BDoyle Thesis

Rupture Behaviour of Abdominal Aortic Aneurysms:
A Computational and Experimental Investigation
AUTHOR
Barry Joseph Doyle
A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

AT THE COLLEGE OF ENGINEERING, UNIVERSITY OF LIMERICK,
IRELAND.
SUPERVISOR
Prof. Timothy M. McGloughlin
DEPARTMENT OF MECHANICAL AND AERONAUTICAL ENGINEERING,
UNIVERSITY OF LIMERICK, IRELAND.
SUBMITTED TO THE UNIVERSITY OF LIMERICK, APRIL 2009.

UNIVERSITY of LIMERICK
O L L S C O I L L U I M N I G H
DECLARATION
April, 2009
The substance of this thesis is the original work of the author and due reference and
acknowledgement has been made, when necessary, to the work of others. No part of this
thesis has been accepted for any degree and is not being concurrently submitted for any
other award.
______________________ ________________________
Barry Doyle Prof. Timothy McGloughlin
(Candidate) (Supervisor)
University of Limerick. Tel: +353-61-202700 Fax: +353-61-330316

Web: http://www.ul.ie
Dedication
This thesis is dedicated to my loving parents for their continued

support and encouragement
Also in loving memory of
Jonathan Kiely
29 August 1979 – 11th November 2008
th
R.I.P.
You’ll never be forgotten…
Table of Contents
TABLE OF CONTENTS
Abstract i
List of Figures ii
List of Tables vi
Presentation of Thesis viii
Glossary of Terms ix
Introduction 1
PART I Computational Investigation

Chapter I Improved assessment and treatment of abdominal aortic 13
aneurysms: The use of 3D reconstructions as a surgical
guidance tool in endovascular repair.
Irish Journal of Medical Science (in press)
Chapter II A comparison of modelling techniques for computing wall 28

stress in abdominal aortic aneurysms.
Biomedical Engineering Online 2007;6:38
Chapter III A finite element analysis rupture index (FEARI) as an 51

additional tool for abdominal aortic aneurysm rupture
prediction.
Vascular Disease Prevention 2009;6:114-121
Chapter IV Vessel asymmetry as an additional diagnostic tool in the 67

assessment of abdominal aortic aneurysms.
Journal of Vascular Surgery 2009;49:443-454
PART II Experimental Investigation

Chapter V 3D reconstruction and manufacture of real abdominal 89
aortic aneurysms: From CT scan to silicone model.
Journal of Biomechanical Engineering 2008;130:034501-5
Table of Contents
Chapter VI An experimental and numerical comparison of the rupture 103

locations of an abdominal aortic aneurysm.
Journal of Endovascular Therapy (in press)
Chapter VII Experimental modeling of aortic aneurysms: Novel 127

applications of silicone rubbers.
Medical Engineering and Physics (submitted)
Chapter VIII Identification of the rupture locations in patient-specific 155

abdominal aortic aneurysms using experimental and
computational techniques.
Journal of Biomechanics (submitted)
Summary and Conclusions 191

Recommendations for Future Work 198
APPENDICES
Appendix A Journal Publications
Appendix B Conference Publications
Appendix C AAA Moulds
Appendix C Honours and Awards
Abstract
ABSTRACT
Abdominal aortic aneurysm (AAAs) is a permanent and irreversible dilation of the
infrarenal section of the aorta. These aneurysms are typically defined as having a
diameter 50% greater than the normal aortic diameter. If left untreated, AAAs will
continually expand until rupture, with rupture usually resulting in death. Currently, the
decision to surgically repair an AAA is determined by the maximum transverse
diameter. If this diameter exceeds 5cm the AAA is deemed a high rupture-risk and
subsequently repaired, by either open repair or by the minimally invasive technique of
endovascular aneurysm repair (EVAR). There is continued debate as to the
appropriateness of the maximum-diameter criterion as a “one-size fits all” approach to
AAA management, as AAAs are known to be patient-specific in terms of geometry, and
ultimately wall stress and strength. This thesis aims to explore alternative approaches to
AAA severity assessment using both computational and experimental methodologies
and attempt to provide new insights into AAA rupture, thereby improving the overall
clinical assessment. In this study it was found that 80% of realistic experimental AAAs
rupture at areas experiencing peak wall stress and not at regions of maximum diameter.
These regions of elevated wall stress can be identified by the finite element method or
the surface curvature of the AAA. Surface defects affect rupture location as they shift
the site of rupture from regions of high wall stress to the areas of localised irregularities.
These experimental surface defects are comparable to calcifications, blebs and localised
hypoxia in the AAA wall in vivo. Idealised AAA models were also shown to rupture at
regions of complex surface curvature and not at the maximum diameter. A range of
silicone rubbers were created that allow improved AAA analogues to be produced.
These silicones have varying wall strengths which can be determined non-destructively
using a colour analysis technique, resulting in experimental AAAs more akin to the in
vivo situation. Bench-top models were manufactured using an injection-moulding lost-
wax technique that create anatomically-correct AAA replications and can be used in a
variety of experimental methods. A direct correlation between numerically predicted
posterior wall stress and vessel asymmetry was identified, and the study showed that
asymmetry may be as significant as diameter in determining the severity of a particular
AAA. A new computational approach to AAA rupture assessment was also created.
This index, described as the Finite Element Analysis Rupture Index (FEARI), indicates
that not all AAAs may be at a high risk of rupture, and is independent to diameter.
Results suggest that similar diameter AAAs may have contrasting rupture potentials.
All computational AAA analyses stem from 3D reconstructions of medical images.
These virtual 3D models not only allow further investigation with techniques such as
the finite element method, but also allow pre-operative planning once surgery is decided
upon. Clinical guidance in the form of stent-graft sizing is possible and anatomical
complications can be foreseen and subsequently overcome. The results and conclusions
presented throughout this thesis may further contribute to the understanding of AAA
biomechanics and rupture potential, and in the future may help provide improved
clinical guidance on the timing of surgery for AAA treatment.
i
List of Figures
List of Figures
PART I
Chapter I
Figure 1 (A) Typical CT scan showing aneurysmal aorta in centre of image, (B) 16
algorithm defines regions of interest and (C) software generates 3D
reconstruction of desired regions.
Figure 2 Full 3D reconstruction of CT scan data set. 17
Figure 3 Meshed reconstruction for use in FEA. 18
Figure 4 Example reconstruction showing measurements used to size the stent-graft. 19
Figure 5 3D reconstruction of extremely tortuous proximal neck. 20
Figure 6 3D reconstruction showing thoracic aortic aneurysm (TAA), abdominal 20
aortic aneurysm (AAA) and iliac aortic aneurysm (IAA) from various
viewpoints.
Figure 7 Resulting wall stress distribution for case examined using FEA. 21
Figure 8 Post-operative EVAR illustration showing deployed TAA stent-graft 24
excluding TAA sac.
Chapter II
Figure 1 Typical CT scan and 3D reconstruction of patient suffering from an AAA. 32
Figure 2 Effect of axial smoothing on the polylines constructed from the CT scans. 33
Figure 3 Resulting wall stress distributions of the various degrees of smoothing 34
examined.
Figure 4 Effect of smoothing on the reconstructed AAA. 34
Figure 5 von Mises wall stress distributions for both the AAA(SIMP)L and 38
AAA(SIMP)NL models at an internal pressure of 120 mmHg.
Figure 6 von Mises wall stress distributions for the both the AAA(MOD)L and 38
AAA(MOD)NL models at an internal pressure of 120mmHg.
Figure 7 von Mises wall stress distributions for both the AAA(COMP)L and 39
AAA(COMP)NL models at an internal pressure of 120mmHg.
Figure 8 Normalised wall stress results and locations of peak wall stress in all the 41
linearly elastic models examined.
Figure 9 Normalised wall stress results and locations of peak wall stress in all the non- 41
linearly elastic models examined.
Chapter III
Figure 1 Illustration showing a representative AAA stress distribution for Patient 9 56
and how each AAA model was segmented in order to determine the
corresponding UTS value for the particular region of peak stress.
Figure 2 Graph displaying FEARI results for all ten cases. 59
Figure 3 FEARI results for the “worst case scenario” of extremely low failure 61
strengths of the AAA tissue.
Chapter IV
Figure 1 Schematic of representative AAA showing how dimensions are obtained 69
from each AAA model.
Figure 2 Original asymmetric AAA of Patient 2 (left) and the modified axisymmetric 70
AAA (right).
Figure 3 Simple illustration showing method of obtaining asymmetry measurements. 71
Figure 4 Example CT scans showing the creation of polylines on each slice, together 72
with the centre point of each polyline.
Figure 5 Diagram depicting how measurements of asymmetry are determined once 72
AAA centreline is created.
Figure 6 between posterior wall stress and anterior asymmetry (left column) and 75
posterior wall stress and diameter (right column) for Patients 1-5.
Figure 7 Relationships between posterior wall stress and anterior asymmetry (left 76
column) and posterior wall stress and diameter (right column) for Patients 6-
10.
ii
List of Figures
Figure 8 Relationships between posterior wall stress and anterior asymmetry (left 77
column) and posterior wall stress and diameter (right column) for Patients
11-15.
Figure 9 Comparison of the posterior wall stress for the symmetric and asymmetric 78
AAA case for Patient 2.
PART II
Chapter V
Figure 1 Segmentation and polyline generation of CT scan. Figure 1(A) shows the full 91
CT scan, whilst Fig. 1(B) is a close-up of the region of interest.
Figure 2 Example mould designs of patient-specific AAAs. Figure 1(A) is a mould 93
design including iliac arteries and Figure 1(B) without iliac arteries.
Figure 3 Example machined mould-piece of inner AAA model. 93
Figure 4 Example FEA von Mises wall stress distribution for Patient A showing 96
region of peak stress on anterior wall. The corresponding mould piece and
resulting silicone model of the same patient can be seen on the right of this
figure.
Chapter VI
Figure 1 (A) Type 2 dumb-bell specimen from BS ISO 37. (B) Modified trouser test 106
piece from BS ISO 34-1
Figure 2 Illustration showing the loading directions that are applied to the trouser test 107
specimen using the tensile test machine.
Figure 3 Mesh independence results for Type 2 dumb-bell model. 108
Figure 4 Experimental set-up of rupture test. 109
Figure 5 Cross-sections of example ideal AAA rubber model. Cross-sections (left) 110
correspond to slices shown on model (right).
Figure 6 Representative numerical ideal AAA, uniform wall. (A) The finite element 111
mesh and (B) loading conditions and constraints.
Figure 7 Engineering stress and engineering strain experimental data (black lines) with 112
4th order polynomial curve applied to the results (red line).
Figure 8 Comparison of results for the experimental tensile tests and the numerical 113
simulation of Sylgard 184.
Figure 9 Sequence of events leading to model rupture of Test 1. (A) The inflated 114
model, (B) the initial point of rupture, (C) propagation of the failure zone and
(D) complete failure of the silicone model.
Figure 10 Box and whisker plot showing wall thickness categorised by AAA region. 115
Figure 11 Effect of varying wall thickness on wall stress results. 116
Figure 12 In vitro rupture locations and FEA stress patterns. High stress occurs at 117
inflection points which correlate with rupture locations in experimental
models.
Chapter VII
Figure 1 (A) The CIE colour space which defines the location of a particular colour 130
with regards to L*, a* and b*, and (B) the corresponding ellipse sizes
throughout L*a*b* colour space.
Figure 2 Ideal AAA models created (A) Example Sylgard 160 model, (B) Example 133
Sylgard 170 model and (C) Mixed AAA model.
Figure 3 Ideal AAA mixed model created using both Sylgard 160 and Sylgard 170. 134
Figure 4 Numerical ideal AAA model used in the analysis. (A) Shows the meshed 135
symmetrical model and (B) illustrates the boundary conditions used.
Figure 5 Variation in 1st order Ogden μ coefficient with differing silicone mixtures. 138
Figure 6 Stress-strain curves for experimental and numerical results for Sylgard 160 138
and 170.
Figure 7 True stress-strain curves for the experimental and numerical results for the 139
new silicone materials.
Figure 8 UTS calibration curve for silicones ranging from Sylgard 170 to Sylgard 160. 139
Figure 9 Tear strength calibration curve for silicones ranging from Sylgard 170 to 140
Sylgard 160.
Figure 10 Predicted UTS results compared with those found experimentally for each 141
silicone mixture.
iii
List of Figures
Figure 11 Predicted TS results compared with those found experimentally for each 141
silicone mixture.
Figure 12 Relationship between tear strength and UTS for each silicone rubber. 142
Figure 13 ∆E variation along the longitudinal distance of the ideal AAA mixed model. 143
Figure 14 Pressure-diameter results for the Sylgard 160, Sylgard 170 and mixed ideal 144
AAA models.
Figure 15 Comparison of results for maximum diameter change with increased pressure 145
loading for both the experimental (exp) and FEA results.
Chapter VIII
Figure 1 Idealised AAA models created with Sylgard 160 and Sylgard 170. 159
Figure 2 3D reconstructions showing various AAA geometries used. 160
Figure 3 View through high-speed camera of test rig. 161
Figure 4 Representative meshed AAA models. 162
Figure 5 Method of measuring wall thickness. 163
Figure 6 Cross-sections which were assessed for wall thickness. 163
Figure 7 Grid system used to compare results between experimental and 164
computational results.
Figure 8 Burst pressures for the various AAA models examined. 165
Figure 9 Typical sequence of events of rupture test. 166
Figure 10 Stress distributions on the outer surface of the undeformed ideal AAA with 167
increasing pressure loading.
Figure 11 Resulting stress contours on the outer surface of the Sylgard 170 AAA of 167
Patient 1.
Figure 12 Box and whisker plot showing peak wall stress categorised by silicone type. 168
Figure 13 Box and whisker plot showing wall thickness categorised by AAA case. 170
Figure 14 Comparison of rupture sites with regions of high stress for Sylgard 170 171
(Black) idealised models.
(Grey) idealised models.
Figure 16 Contour plot of Gaussian curvature for the idealised AAA. 172
(Black) models of Patient 1.
(Grey) models of Patient 1.
Figure 19 Contour plot of Gaussian curvature for Patient 1. 174
Figure 24 CT scan at rupture site of Model 1 revealed a small tear in the model prior to 177
experimental rupture.
Figure 28 (A) Stereomicroscope image of typical surface (B) Cross-section at rupture 179
site revealing entrapment of microscopic air bubbles.
Figure 31 (A) Wall stress contour plot compared to (B) Gaussian surface curvature 183
contour plot.
iv
List of Figures
FUTURE WORK
Figure 1 Example 3D reconstruction with centreline shown in red. 199
Figure 2 Example asymmetry plot automatically generated from MATLAB. 199
Figure 3 Example process of determining asymmetry from 2D images. 200
Figure 4 Example plot of asymmetry determined from 2D CT images. 201
Figure 5 Realistic AAA created using a mixture of Sylgard 160 and Sylgard 170. 202
Figure 6 Patient-specific AAA created using PL-3 epoxy resin for use with the 203
photoelastic method.
Figure 7 Illustration showing DIC process of measurement. 204
Figure 8 TAA with centreline for use in asymmetry assessment. 204
Figure 9 Preliminary method of thresholding and segmenting ultrasound images. 205
v
List of Tables
List of Tables
PART I
Chapter II
Table 1 Comparison of peak wall stress, location of peak wall stress and FEA 40
computational time for the AAA(SIMP), AAA(MOD) and the AAA(COMP)
models.
Chapter III
Table 1 Pre-operative patient details for each study subject. 54
Table 2 Regional UTS values obtained by combining and averaging previous 56
experimental data.
Table 3 Maximum diameter, FEA computed peak wall stress, location of peak wall 57
stress, and UTS of peak stress region in all ten cases examined.
Table 4 Resulting FEARI values compared to the corresponding maximum diameters 58
for each AAA studied.
Chapter IV
Table 1 Patient details for each study subject. 73
Table 2 Maximum diameter, peak wall stress, location of peak stress, and diameter at 74
peak wall stress for each patient studied.
Table 3 Statistical analysis of patient-specific parameters and peak wall stress. 79
Table 4 Statistical analysis of patient-specific parameters and posterior wall stress. 80
Table 5 Correlation coefficients for posterior wall stress and both asymmetry and 80
diameter.
Table 6 Correlation between asymmetry and diameter for each case examined. 81
PART II
Chapter V
Table 1 Averaged wall thickness measurements at four locations on the AAA wall. 95
Table 2 Averaged wall thickness measurements for each patient-specific AAA model 95
Chapter VI
Table 1 Material coefficients determined for the 3rd order Ogden model for Sylgard 113
184.
Table 2 Summary of experimental rupture results. 114
Table 3 Regional variations in wall thickness. Thickness values were applied to 116
numerical models to determine the effect of wall thickness on stress results.
Table 4 Summary of FEA wall stress results. 118
Table 5 Summary of FEA wall tension results. 118
Chapter VII
Table 1 Results of the uniaxial tensile testing for each mixture of silicone. 136
Table 2 Results of the tear testing for each mixture of silicone. 136
Table 3 1st order Ogden SEF coefficients for each silicone mixture. 137
Table 4 ∆E measurements at various locations along the length of the ideal AAA 142
mixed model.
Table 5 Pressure-diameter results for the Sylgard 160 ideal AAA models. 143
Table 6 Pressure-diameter results for the Sylgard 170 ideal AAA models. 144
Table 7 Pressure-diameter results for the randomly-mixed ideal AAA model. 144
Table 8 Summarised results comparing diameter changes between those found 145
experimentally and numerically.
Chapter VIII
Table 1 Case details of AAA geometries used in study. 157
Table 2 Material coefficients for the 1st order Ogden SEF for Sylgard 160 and 158
Sylgard 170.
vi
List of Tables
Table 3 Percentage increase in peak wall stress with increased internal pressure 166
loading.
Table 4 Peak wall stress results for each AAA case. 168
Table 5 Maximum and minimum Gaussian surface curvature for each AAA 169
geometry.
Table 6 Results of the wall thickness study. 169
Table 7 Wall thickness measurements at the site of experimental rupture. 170
vii
Presentation of Thesis
Presentation of Thesis
This thesis has been prepared in the style “Thesis by Publication.” Therefore
referencing styles vary throughout the thesis as required by the specific journal of
publication. Table 1 below outlines the chapters and the journals in which that chapter
has been published, accepted for publication, or submitted for publication. Target
audience for each journal is also indicated and the corresponding current ISI Impact
Factor of the respective journal. The research undertaken and described in this thesis
highlights a biomedical engineering approach aimed at addressing clinical problems.
Table 1: Journal of publication, target audience and ISI Impact Factor for each chapter
presented in this thesis.
Chapter Journal Audience Impact Factor
1 Irish Journal of Medical Science Medical 0.29
2 Biomedical Engineering Online Biomedical 1.58

Engineering
3 Vascular Disease Prevention Medical N/A*
4 Journal of Vascular Surgery Medical 3.272
5 Journal of Biomechanical Engineering Biomedical 1.591

Engineering
6 Journal of Endovascular Therapy Medical 2.392
7 Medical Engineering & Physics Biomedical 1.471

Engineering
8 Journal of Biomechanics Biomedical 2.897

Engineering
* ISI Impact Factor not available at time of print.
viii
Glossary of Terms
Glossary of Terms
Aneurysm A permanent swelling of a blood vessel caused by disease or

weakening of the vessel wall.
Aneurysmal Relating to or affected by an aneurysm.
Endovascular A minimally-invasive technique of inserting a medical device
into a blood vessel.
In vivo Latin for within the living.
In vitro Latin for within the glass.
Endoleaks Term that describes the presence of persistent flow of blood
into the aneurysm sac after the placement of a medical device.
Endotension Persistent or recurring pressurisation of the aneurysm sac after
placement of a medical device.
Proximal Nearer to a point of reference such as an origin, a point of
attachment, or the midline of the body.
Distal Anatomically located far from a point of reference, such as an
origin or a point of attachment.
Anterior Located on or near the front of the body.
Posterior Located behind a part or toward the rear of a structure.
Hypertension Abnormally elevated blood pressure.
Hyperlipidaemia Presence of excess lipids in the blood.
Pulmonary Of, relating to, or affecting the lungs.
Intraluminal Within the lumen of any tubular structure or organ.
Lumen The cavity or channel within a tube or tubular organ.
Thrombus A stationary blood clot along the wall of a blood vessel.
Calcification A calcified substance or part.
Bleb A small bladder or sac containing liquid, similar to a blister.
ix
INTRODUCTION
Introduction
Introduction
Cardiovascular disease is the leading cause of morbidity and premature death of modern
era medicine. It is estimated that approximately 81 million people in the United States
(US) currently have one or more of the many forms of cardiovascular disease, resulting
in 1 in every 2.8 deaths, or 900,000 deaths per year. 40% of all deaths in Europe are a
result of cardiovascular disease in people under the age of 75, with mortality rates in
Ireland reported to be as high as 43% in 1997 (Creagh et al., 2002). Aneurysms form a
significant portion of these cardiovascular related deaths and are defined as a permanent
and irreversible localised dilation of a blood vessel greater than 50% of its normal
diameter (Sakalihasan et al., 2005). Although aneurysms can form in any blood vessel,
the more lethal aneurysms develop in the cranial arteries, thoracic aorta and abdominal
aorta, which left untreated will eventually expand until rupture. The vast majority of
these aneurysms form in the abdominal aorta and are termed abdominal aortic aneurysm
(AAA). The actual mechanisms resulting in AAA formation are still not fully
understood. It is believed that these aneurysms form due to alterations of the connective
tissue in the aortic wall. The degradation of the aortic wall can be attributed to risk
factors such as tobacco smoking, sex, age, hypertension, chronic obstructive pulmonary
disease, hyperlipidaemia and family history of the disorder (Sakalihasan et al., 2005).
With recent advancements in medical imaging technology, the recorded incidence of

AAA is on the increase. It has been reported that over the last 30 years, AAA diagnosis
has tripled in the Western world and this is likely to increase over the coming years as
the average population age is on the rise (Bosch et al., 2001). There are approximately
150,000 - 200,000 new cases in the US and 500,000 patients worldwide diagnosed each
year. It is estimated that 1 million people worldwide are currently living with an
undiagnosed AAA and that 95% of these cases could be successfully treated if detected
prior to rupture. Mortality figures are also high, with a 95% mortality rate. These
statistics contribute to rank AAAs as the 13th leading cause of death in the US and the
10th leading cause of death in men over the age of 55. Annually there are
approximately 15,000 - 20,000 deaths in the US from AAA, and 8,000 deaths in the
United Kingdom. The numbers of cases and subsequent deaths for Ireland is similar to
international figures (Brosnan et al., 2009). 30 - 50% of patients with a ruptured AAA
die before they ever reach a hospital and even with surgical repair a 50 - 70% mortality
1
Introduction
rate remains. The condition is also more common in men than women, however, the
disease is more lethal in women.
AAAs are typically asymptomatic and are usually detected as a result of unrelated
health problems or screening programs of the elderly. Screening involves the use of
ultrasonography to detect AAA and the implementation of these programs is becoming
increasingly common. From a theoretical screening model of a population of 100,000 it
has been estimated that 1500 lives could be saved, at a cost of $78,000 per life (Quill et
al., 1989; Ernst, 1993). Measurements of AAA size determined using ultrasonography
are accurate up to 6 mm (Ernst, 1993), and therefore has become the most cost-effective
method of AAA detection. It has been recommended that people over the age of 60-65
years old, in particular men, should be screened for AAA, with the recommended age
reducing to 50 - 55 when there is a history of aneurysmal disease in the family as this
may increase the likelihood of AAA by up 6 times. AAA screening programs are
becoming more widespread in the UK with many private institutions providing
screening. The UK National Health Service (NHS) recently announced that a full
screening program will be made available throughout the UK, but is unlikely to become
widely available until 2013 (National Heath Service, 2009). Recently, it was suggested
that AAA screening may be beneficial in Irish males aged 65 - 75 years (Brosnan et al.,
2009). According to the US Preventative Services Task Force (USPSTF, 2005), the
potential benefit of screening for AAA among women over the age of 65 is low because
of the number of age-related deaths in this population. The majority of AAA related
deaths occur in women over the age of 80, and as there are many competing health risks
at this age, any benefit of screening would be minimal (USPSTF, 2005). Various
lifestyle considerations can also affect the prevalence of AAA, such as smoking, diet
and obesity; with people who fall into these categories may be more likely to either
have or develop an AAA.
Upon detection of AAA, there is currently much debate as how to best assess the
severity and rupture risk of the patient and it is this ongoing debate that has formed the
basis of this thesis. Currently, the trend in determining the severity of an AAA is to use
the maximum diameter criteria (Cronenwett et al., 1985; Glimaker et al., 1991).
Patients with an AAA that has a maximum diameter greater than 5 - 5.5 cm are deemed
a high rupture risk and are usually recommended for surgical repair (Lederle et al.,
2002). In the case of smaller AAAs where the diameter is <5 cm, the preferred
2
Introduction
approach is often careful and frequent observation using either ultrasonography or

computed tomography (CT) scanning. Recent research however, has cast doubt over
the suitability of surgical repair based on a maximum diameter criteria alone (Raghavan
et al., 2000; Sayers et al., 2002; Fillinger et al., 2002, 2003; Vande Geest et al., 2006;
Kleinstreuer and Li, 2006; Leung et al., 2006). Although the diameter-criterion can be
justified, as the rupture risk for an AAA is clearly related to its maximum diameter
(Conway et al., 2001; Fillinger et al., 2002), surgical decision-making using the
diameter-criterion alone may in fact lead to both unnecessary AAA repairs and also
exclude certain cases (AAA <5 cm) from surgical repair (Darling et al., 1977;
Cronenwett et al., 1985; Nicholls et al., 1998; Fillinger et al., 2002). Nicholls et al.
(1998) reported that 10 - 24% of ruptured AAAs were less than 5cm in diameter.
Darling et al. (1977) also reported that of 473 non-repaired AAAs examined from
autopsy reports, there were 118 cases of rupture, 13% of which were less than 5 cm in
diameter. They also showed that 60% of the AAAs greater than 5 cm (including 54% of
those AAAs between 7.1 and 10 cm) never experienced rupture. Vorp et al. (2008) later
deduced from the findings of Darling et al. (1977) that if the maximum diameter
criterion were followed for the 473 subjects, only 7% (34/473) of cases would have
succumbed to rupture prior to surgical intervention as the diameter was less than 5 cm,
with 25% (116/473) of cases possibly undergoing unnecessary surgery since these
AAAs may never have ruptured.
Alternative methods of rupture assessment have been recently reported. The majority of
these approaches involve the numerical analysis of AAAs using the common
engineering technique of the finite element method (FEM) to determine the wall stress
distributions. Recent reports have shown that these stress distributions have been
shown to correlate to the overall geometry of the AAA rather than solely to the
maximum diameter (Vorp et al., 1998; Sacks et al., 1999; Doyle et al., 2009a). It is also
known that wall stress alone does not govern failure as an AAA will rupture when the
local wall stress exceeds the local wall strength. Therefore, rupture assessment is most
suitable when both the patient-specific wall stress is coupled with patient-specific wall
strength. A non-invasive method of determining patient-dependent wall strength was
recently reported by Vande Geest et al. (2006), with more traditional approaches to
strength determination via tensile testing performed by others (Raghavan et al., 1996,
2006; Thubrikar et al., 2001). Newly proposed AAA assessment methods include:
AAA wall stress (Fillinger et al., 2002, 2003; Venkatasubramaniam et al., 2004); AAA
3
Introduction
expansion rate (Hirose et al., 1998); degree of asymmetry (Doyle et al., 2009a);
presence of intraluminal thrombus (ILT) 1 (Wang et al., 2002); a rupture potential index
(RPI) (Vorp et al. 2005; Vande Geest et al., 2006); a finite element analysis rupture
index (FEARI) (Doyle et al., 2009b); biomechanical factors coupled with computer
analysis (Kleinstreuer and Li, 2006); growth of ILT (Stenbaek et al., 2006); geometrical
parameters of the AAA (Giannoglu et al., 2006); and also a method of determining
AAA growth and rupture based on mathematical models (Watton et al., 2004; Volokh
and Vorp, 2008). Based on these hypotheses, it is believed that an improved predictor
of AAA rupture is desirable and may have clinical importance (Vorp et al., 1998;
Raghavan et al., 2000a, 2000b; Sayers et al., 2002; Wang et al., 2002; Fillinger et al.,
2002, 2003; Vande Geest et al., 2006; Kleinstreuer and Li, 2006; Leung et al., 2006;
Doyle et al., 2009a, 2009b).
Detection of AAAs leads to the clinical question: should the AAA be surgically repaired
or monitored? Monitoring of AAAs, particularly smaller aneurysms, has been shown in
the past to be effective, with the usual recommendations being repeat ultrasound every
six months for AAAs 4 – 5 cm in diameter and every three months for larger AAAs
(Ernst, 1998). During this monitoring phase, expansion rates are also observed. If the
expansion rate exceeds 0.5 cm per year, surgical intervention is often recommended.
Further to aneurysm size deciding the fate of the AAA, other contraindications to
elective repair are also commonly considered. These include, but are not limited to:
myocardial infarction within the past six months; intractable congestive heart failure;
intractable angina pectoris; severe chronic renal insufficiency, incapacitating effects
from stroke, and a life expectancy of less than two years.
If the clinical outcome is to surgically repair the AAA, two approaches may be
considered, both utilising a graft to exclude blood flow and pressure from the
aneurysmal sac and return it to a relatively normal state. These options are either the
traditional approach of open repair or the minimally-invasive technique of endovascular
aneurysm repair (EVAR). The gold standard in AAA repair is the open surgical
method. The long-term durability of the procedure is excellent, with little need for post-
operative intervention (Garcia-Madrid et al., 2004; Goueffic et al., 2005; Corbett et al.,
2008). However, this procedure is highly invasive and poses serious risk of
1
Intraluminal thrombus (ILT) consists of a fibrin structure incorporated with blood cells, platelets, blood
proteins and cellular debris, and are found in most AAAs
4
Introduction
complications during the operation (Sakalihasan et al., 2005; Vogel et al., 2005; Corbett
et al., 2008). EVAR, on the other hand, is the preferred treatment in older patients and
patients unfit for open repair. This procedure uses minimally-invasive techniques,
entering the aorta via the femoral artery, resulting in minimal scarring, reduced blood-
loss and a significant reduction in operation time. Questions have, however, been raised
recently about the effectiveness of EVAR compared to open repair, in particular with
regards to the biomechanical performance of the stent-graft resulting in post-operative
complications such as endoleak, graft migration, and fabric tears (EVAR Trial
Participants, 2005a; EVAR Trial Participants, 2005b; Corbett et al., 2008).
The main aim of this thesis is to examine alternative methods of AAA assessment,
rather than the traditional diameter-based approach. This work is divided into two
sections: Part I utilises numerical methodology to determine the severity of patient-
specific AAAs and Part II investigates the use of experimental methods to create and
test AAA analogues, and determine the rupture locations of these aneurysms.
In Part I, Chapter I addresses the issue of 3D reconstructions of computed tomography

(CT) datasets. Traditionally, clinicians examine the 2D CT scans prior to surgical
repair. Vital dimensions are obtained from these images with regards to the correct
sizing of the grafts used to repair the aneurysm. The use of even basic 3D
reconstructions at this stage can improve the sizing of the graft, as measurements can be
taken from exact 3D models, rather than from 2D images. Also, further to graft sizing,
the actual AAA can be examined. Although reconstructions are becoming increasingly
common prior to surgery, it is still not the standard. 3D reconstructions allow the AAA
to be “visualised” and rigorously examined by the clinician. This increases confidence
in the surgical approach by the clinician, as they can now determine tortuosity of the
proximal neck and the iliac arteries, crucial for effective EVAR. Also, irregularities that
may cause concern during the operation can be highlighted, negating any unforeseen
problems during the surgical procedure due to the shape and morphology of the AAA.
Although numerical modelling of AAAs to determine wall stress has increased over
recent years (Mower et al., 1997; Vorp et al., 1998; Raghavan et al., 2000; Di Martino
et al., 2001; Thubrikar et al. 2001; Hua et al., 2001; Wang et al., 2002, Fillinger et al.,
2002, 2003; Venkatasubramaniam et al. 2004; Giannoglu et al., 2006; Leung et al.,
2006; Papaharilaou et al., 2007; Speelman et al., 2007; Scotti et al., 2005, 2007;
5
Introduction
Truijuers et al., 2007; Doyle et al., 2007, 2009a, 2009b, 2009c; Rissland et al., 2009),
there appears to be little in the way of consistency with approaches. Many reports differ
in the methods of 3D reconstruction, levels and degrees of smoothing, mesh generation
for the FEM, biomechanical material properties and incorporation of ILT. Chapter II
compares some of the more common approaches among researchers and draws
conclusions based on the results observed.
It is known from the simple definition of material failure that an AAA will rupture when
the local wall stress exceeds the local wall strength. Based on this hypothesis, Chapter
III reports a novel tool to predict the propensity of a patient-specific AAA to rupture.
Based on previously published work (Raghavan et al., 1996, 2006; Thubrikar et al.,
2001) by two leading groups in AAA research, ultimate tensile strength (UTS) was
determined from a total set of 149 AAA tissue samples. Combining these previous UTS
results, average regional strengths could be determined for the main areas of the AAA.
Numerical analyses then generated approximate results for the peak stress in the patient-
specific AAA. Combining this peak stress along with the UTS of the region allows a
rupture index to be determined. This novel tool for assessing AAA rupture potential is
the Finite Element Analysis Rupture Index (FEARI) and could be readily incorporated
into the decision-making process of the clinician.
The asymmetry of AAAs has been reported to affect wall stress distributions in
idealised AAA models (Vorp et al., 1998; Scotti et al., 2005). Chapter IV develops
this biomechanical insight and moves it towards patient-specific AAA cases. Darling et
al., (1977) performed a comprehensive study over the course of 23 years, from which
they concluded that 82% of ruptures occurred along the posterior wall of the aneurysm.
Based on this clinical fact, the relationship between posterior wall stress and various
geometrical factors was investigated. Peak stress is still regarded as the most clinically
relevant factor obtained from numerical studies, but as most ruptures occur on the
posterior wall, posterior wall stress may be equally important. This section identifies a
direct correlation between posterior stress and anterior asymmetry and outlines a novel
method of measuring asymmetry. This new technique of asymmetry assessment is
easily implemented and could also be applied to a clinical setting.
Within Part II of this thesis, experimental methods are reported. Chapter V discusses
an injection-moulding lost-wax technique used to manufacture realistic AAA
6
Introduction
experimental models. 3D reconstruction of AAA cases allows further computer-aided

design (CAD) to be performed, enabling the design of computer numerically controlled
(CNC) machined moulds (Appendix C). AAA models can then be manufactured using
injectable liquid materials. In this study silicone was chosen as the AAA wall analogue
as silicone lends itself well to the study of arterial mechanics and has been used in
previous studies (O’Brien et al., 2005; Marins et al., 2006; Doyle et al., 2008a, 2008b,
2009d). The effectiveness of this method to produce realistic AAA models was
investigated.
The lost-wax technique outlined in Chapter V was used to create idealised AAA
models. This ideal AAA geometry has been previously used in both experimental
(O’Brien et al., 2005; Morris et al., 2004) and numerical studies (Callanan et al., 2004).
Chapter VI describes a method of utilising high-speed photography to capture the
rupture locations of these rubber models. This section also reports the mechanical
characterisation of the material used, generating material coefficients for an optimum
strain energy function (SEF) that accurately describes the stress-strain behaviour of the
material. Rupture locations were found to resemble those observed in previous a study
(Morris et al., 2004). The FEM was used to predict the regions of high stress within
these models, with good agreement found. Idealised AAAs experience elevated wall
stresses at regions of inflection on the AAA surface. Elevated stress at inflection
regions and not at areas of maximum diameter in idealised models has been reported in
recent work using both numerical methods (Vorp et al., 1998; Callanan et al., 2004;
Scotti et al., 2005) and experimentally using the photoelastic method (Morris et al.,
2004). Elevated stress at inflection regions was also reported in realistic AAA cases
(Doyle et al., 2007, 2009a, 2009b). Wall thickness does not appear to significantly
effect the wall stress distributions, with the proximal inflection region experiencing high
stress regardless of variations in wall thickness. Of the five idealised models examined
in this chapter, four experienced rupture at inflection points.
In order to successfully model aneurysms in vitro, material analogues are required.

Chapter VII reports the design and development of a novel range of silicone rubber
compositions used in experimental testing. An important aspect of this section was the
ability to non-destructively assess material properties. This was achieved by
incorporating colour analysis techniques into the study. A range of silicone rubbers
were created, each with differing colour intensities. Mechanical testing then determined
7
Introduction
the failure properties of each material, which were correlated to colour intensity.
Calibration curves were created and assessed for accuracy. Results showed good
correlation between the predicted and measured UTS and tear strength (TS) for all
compositions within the range. Mechanical characterisation was performed for all new
materials in order to determine material coefficients for SEFs that accurately model the
materials. These silicones were then implemented into experimental tests. The lost-
wax process was again utilised to create idealised AAA models, which were then used
with videoextensometry. Deformations of the idealised AAA wall were measured at
incrementally increasing pressure loadings. The FEM was used to numerically validate
these deformations.
Chapter VIII links the work of Chapters V, VI and VII and progresses it from the
work of idealised AAA cases to patient-specific models. Realistic AAA rubber models
were created using the process outlined in Chapter V. Each experimental model was
imaged using CT in order to reconstruct experimental models into exact numerical
models. Rupture testing was carried out in a similar manner to that of Chapter VI,
with the FEM used to numerically predict rupture zones. Experimental and numerical
techniques correlated well with surface curvature indicating regions that may experience
elevated wall stress. It was observed that 80% of AAA models ruptured at regions of
peak wall stress and not at regions of maximum diameter. Wall anomalies were shown
to effect rupture location.
A comprehensive overview of our results with conclusive remarks and the implication
of our findings on the clinical community complete this thesis.
References
Department of Health and Children: Dublin. (1999) Building Healthier Hearts: The report of the
cardiovascular health strategy group.
Brosnan, M., Collins, C.G., Moneley, D.S., Kelly, C.J., Leahy, A.L. (2009) “Making the case for
cardiovascular screening in Irish males: Detection of abdominal aortic aneurysms, and assessment of
cardiovascular risk factors.” European Journal of Vascular and Endovascular Surgery, 37, 300-304.
Bosch, J.L., Lester, J.S., McMahon, P.M., Beinfeld, M.T. Halpern, E.F., Kaufman, J.A., Brewster, D.C.,
Gazelle, G.S. (2001) “Hospital costs for elective endovascular and surgical repairs of infrarenal
abdominal aortic aneurysms.” Radiology, 220(2), 492-497.
Callanan, A., Morris, L.G., McGloughlin, T.M. (2004) “Numerical and experimental analysis of an
idealised abdominal aortic aneurysm.” European Society of Biomechanics, S-Hertogenbosch,
Netherlands.
8
Introduction
Conway, K.P., Byrne, J., Townsend, M., Lane, I.F. (2001) “Prognosis of patients turned down for
conventional abdominal aortic aneurysm repair in the endovascular and sonographic era: Szilagyi
revisted?” Journal of Vascular Surgery, 33, 752-757.
Corbett, T.J., Callanan, A., Morris, L.G., Doyle, B.J., Grace, P.A., Kavanagh, E.G. McGloughlin, T.M.
(2008) “A review of the in vivo and in vitro biomechanical behaviour and performance of postoperative
abdominal aortic aneurysms and implanted stent-grafts.” Journal of Endovascular Therapy, 15, 468-464.
Creagh, D., Neilson, S., Collins, A., Colwell, N., Hinchion, R., Drew, C., O’Halloran, D., Perry, I.J.
(2002) Established cardiovascular disease and CVD risk factors in a primary care population of middle-
aged Irish men and women. Irish Medical Journal, 95(10), 298-301.
Cronenwett, J.L., Murphy, T.F., Zelenock, G.B., Whitehouse Jr., W.M., Lindenauer, S.M., Graham, L.M.,
Quint, L.E., Silver, T.M., Stanley, J.C. (1985) “Actuarial analysis of variables associated with rupture of
small abdominal aortic aneurysms.” Surgery, 98, 472-483.
Darling, R.C., Messina, C.R., Brewster, D.C., Ottinger, L.W. (1977) “Autopsy study of unoperated
abdominal aortic aneurysms. The case for early resection.” Circulation, 56(2), 161-164.
DiMartino, E.S., Guadagni, G., Fumero, A., Ballerini, G., Spirito, R., Biglioli, P, Redaelli, A. (2001)
“Fluid-structure interaction within realistic 3D models of aneurysmatic aorta as a guidance to assess the
risk of rupture of the aneurysm.” Medical Engineering and Physics, 23, 647-655.
Doyle, B.J., Callanan, A., Burke, P.E., Grace, P.A., Walsh, M.T., Vorp, D.A., McGloughlin, T.M.
(2009a) “Vessel asymmetry as an additional tool in the assessment of abdominal aortic aneurysms.”
Journal of Vascular Surgery, 49:443-454.
Doyle, B.J., Morris, L.G., Callanan, A., Kelly, P., Vorp, D.A., McGloughlin, T.M. (2008a) “3D
reconstruction and manufacture of real abdominal aortic aneurysms: From CT scan to silicone model.”
Journal of Biomechanical Engineering, 130, 034501-5.
Doyle, B.J., Callanan, A., Corbett, T.J., Cloonan, A.J., O’Donnell, M.R., Vorp, D.A., McGloughlin, T.M.
(2008b) “The use of silicone to model abdominal aortic aneurysm behaviour.” Society of Plastics
Engineers, SPE European Conference on Medical Polymers, pp.115-120.
Doyle, B.J., Callanan, A., McGloughlin, T.M. (2007) “A comparison of modelling techniques for
computing wall stress in abdominal aortic aneurysms.” Biomedical Engineering Online, 6, 38.
Doyle, B.J., Callanan, A., Walsh, M.T., Grace, P.A., McGloughlin, T.M. (2009b) “A finite element
analysis rupture index (FEARI) as an additional tool for abdominal aortic aneurysm rupture prediction.”
Vascular Disease Prevention, 6, 75-82.
Doyle, B.J., Grace, P.A. Kavanagh, E.G., Burke, P.E. Wallis, F., Walsh, M.T., McGloughlin, T.M.
(2009c) “Improved assessment and treatment of abdominal aortic aneurysms: The use of 3D
reconstructions as a surgical guidance tool in endovascular repair.” Irish Journal of Medical Science, in
press.
Doyle, B.J., Corbett, T.J., Callanan, A., Walsh, M.T., Vorp, D.A., McGloughlin, T.M. (2009d) “An
experimental and numerical comparison of the rupture locations of an abdominal aortic aneurysm.”
Journal of Endovascular Therapy, in press.
Ernst, C.B. (1993) “Abdominal aortic aneurysm.” The New England Journal of Medicine, 328(16), 1167-
1172.
EVAR Trial Participants (2005a) “Endovascular aneurysm repair versus open repair in patients with
abdominal aortic aneurysm (EVAR Trial 1): randomised controlled trial.” Lancet, 365, 2179-2186.
EVAR Trial Participants (2005b) “Endovascular aneurysm repair and outcome in patients unfit for open
repair of abdominal aortic aneurysm (EVAR Trail 2): randomised controlled trial.” Lancet, 365(9578),
2187-2192.
9
Introduction
Fillinger, M.F., Marra, S.P., Raghavan, M.L., Kennedy, F.E. (2003) “Prediction of rupture risk in
abdominal aortic aneurysm during observation: wall stress versus diameter.” Journal of Vascular Surgery,
37, 724-732.
Fillinger, M.F., Raghavan, M.L., Marra, S.P., Cronenwett, J.L., Kennedy, F.E. (2002) “In vivo analysis of
mechanical wall stress and abdominal aortic aneurysm rupture risk.” Journal of Vascular Surgery, 36,
589-597.
Garcia-Madrid, C., josa, M., Riambau, V., Mestresa, C.A., Muntanab, J., Muleta, J. (2004) “Endovascular
versus open surgical repair of abdominal aortic aneurysm: a comparison of early and intermediate results
in patients suitable for both techniques.” Journal of Endovascular Therapy, 28(4), 365-372.
Giannoglu, G., Giannakoulas, G., Soulis, J., Chatzizisis, Y., Perdikides, T., Melas, N., Parcharidis, G.,
Louridas, G. (2006) “Predicting the risk of rupture of abdominal aortic aneurysms by utilizing various
geometrical parameters: Revisiting the diameter criterion.” Angiology, 57(4), 487-494.
Glimaker, H., Holmberg, L., Elvin, A., Nybacka, O., Almgren, B., Bjorck, C.G., Eriksson, I. (1991)
“Natural history of patients with abdominal aortic aneurysm.” European Journal of Vascular Surgery, 5,
125-130.
Goueffic, Y., Becquemin, J.P., Desgranges, P., Kobeiter, H. (2005) “Midterm survival after endovascular
versus open repair of infrarenal aortic aneurysms.” Journal of Endovascular Therapy, 12(1), 47-57.
Hirose, Y., Takamiya, M. (1998) “Growth curve of ruptured aortic aneurysm.” Journal of Cardiovascular
Surgery, 39, 9-13.
Hua, J., Mower, W.R. (2001) “Simple geometric characteristics fail to reliably predict abdominal aortic
aneurysm wall stress.” Journal of Vascular Surgery, 34, 308-315.
Kleinstreuer, C., Li, Z. (2006) “Analysis and computer program for rupture-risk prediction of abdominal
aortic aneurysms.” Biomedical Engineering Online, 5, 19.
Lederle, F.A., Johnson, G.R., Wilson, S.E., Ballard, D.J., Jordan Jr., W.D., Blebea, J., Littooy, F.N.,
Freischlag, J.A., Bandyk, D., Rapp, J.H., Salam, A.A. (2002) “Rupture rate of large abdominal aortic
aneurysms in patients refusing or unfit for elective repair.” Journal of the American Medical Association,
287, 2968-2972.
Leung, J.H., Wright, A.R., Cheshire, N., Crane, J., Thom, S.A., Hughes, A.D., Xu, Y. (2006) “Fluid
structure interaction of patient specific abdominal aortic aneurysms: a comparison with solid stress
models.” Biomedical Engineering Online, 5, 33.
Marins, P.A.L.S., Natal Jorge, R.M., Ferreira, A.J.M. (2006) “A comparative study of several material
models for prediction of hyperelastic properties: Application to silicone-rubber and soft tissues.” Strain,
42, 135-147.
Morris, L., O’Donnell, P., Delassus, P., McGloughlin, T. (2004) “Experimental assessment of stress
patterns in abdominal aortic aneurysms using the photoelastic method.” Strain, 40, 165-172.
Mower, W.R., Quinones, W.J., Gambhir, S.S. (1997) “Effect of intraluminal thrombus on abdominal
aortic aneurysm wall stress.” Journal of Vascular Surgery, 26, 602-608.
National Health Service (2009) National Screening Program for Abdominal Aortic Aneurysm [online]
available: http://aaa.screening.nhs.uk [accessed 9 Feb 2009].
Nicholls, S.C., Gardner, J.B., Meissner, M.H., Johansen, H.K. (1998) “Rupture in small abdominal aortic
aneurysms.” Journal of Vascular Surgery, 28, 884-888.
O’Brien, T., Morris, L., O’Donnell, M., Walsh, M. and McGloughlin, T. (2005) “Injection-moulded
models of major and minor arteries: The variability of model wall thickness owing to casting technique.”
Proceedings of the International Mechanical Engineers, 219, Part H: Journal of Engineering in Medicine.
10
Introduction
Papaharilaou, Y., Ekaterinaris, J.A., Manoussaki, E., Katsamouris, A.N. (2007) “A decoupled fluid
structure approach for estimating wall stress in abdominal aortic aneurysms.” Journal of Biomechanics,
40, 367-377.
Quill, D.S., Colgan, M.P., Sumner, D.S. (1989) “Ultrasonic screening for the detection of abdominal
aortic aneurysms.” Surgical Clinics of North America, 69, 713-720.
Raghavan, M.L., Kratzberg, J., de Tolosa, E.M.C., Hanaoka, M.M., Walter, P., da Silva, E.S. (2006)
“Regional distribution of wall thickness and failure properties of human abdominal aortic aneurysm.”
Journal of Biomechanics, 39(16), 3010-6.
Raghavan, M.L., Vorp, D.A. (2000) “Toward a biomechanical tool to evaluate rupture potential of
abdominal aortic aneurysm: identification of a finite strain constitutive model and evaluation of its
applicability.” Journal of Vascular Surgery, 33, 475-482.
Raghavan, M.L., Vorp, D.A., Federle, M.P., Makaroun, M.S., Webster, M.W. (2000) “Wall stress
distribution on three-dimensionally reconstructed models of human abdominal aortic aneurysm.” Journal
of Vascular Surgery, 31, 760-769.
Raghavan, M.L., Webster, M.W., Vorp, D.A. (1996) “Ex vivo biomechanical behaviour of abdominal
aortic aneurysm: assessment using a new mathematical model.” Annals of Biomedical Engineering, 24,
573-582.
Rissland, P., Alemu, Y., Einav, S., Ricotta, J., Bluestein, D. (2009) “Abdominal aortic aneurysm risk of
rupture: Patient-specific FSI simulations using anisotropic model.” Journal of Biomechanical
Engineering, 131, 031001-0310010.
Sacks, M.S., Vorp, D.A., Raghavan, M.L., Federle, M.P., Webster, M.W. (1999) “In vivo three-
dimensional surface geometry of abdominal aortic aneurysms.” Annals of Biomedical Engineering, 27(4),
469-479.
Sakalihasan, N., Limet, R., Defawe, O.D. (2005) “Abdominal aortic aneurysm.” Lancet, 365(9470),
1577-89.
Sayers, R.D. (2002) “Aortic aneurysms, inflammatory pathways and nitric oxide.” Annals of the Royal
College of Surgeons England, 84(4), 239-246.
Scotti, C.M., Finol, E.A. (2007) “Compliant biomechanics of abdominal aortic aneurysms: a fluid-
structure interaction study.” Computers and Structures, 85, 1097-1113.
Scotti, C.M., Shkolnik, A.D., Muluk, S.C., Finol, E. (2005) “Fluid-structure interaction in abdominal
aortic aneurysms: effect of asymmetry and wall thickness.” Biomedical Engineering Online, 4, 64.
Speelman, L., Bohra, A., Bosboom, E.M.H., Schurink, G.W.H., van de Vosse, F.N., Makaroun, M.S.,
Vorp, D.A. (2007) “Effects of wall calcifications in patient-specific wall stress analyses of abdominal
aortic aneurysms.” Journal of Biomechanical Engineering, 129, 1-5.
Stenbaek, J., Kalin, B., Swedenborg, J. (2000) “Growth of thrombus may be a better predictor of rupture
than diameter in patients with abdominal aortic aneurysms.” European Journal of Vascular and
Endovascular Surgery, 20, 466-469.
Thubrikar, M.J., Labrosse, M., Robicsek, F., Al-Soudi, J., Fowler, B. (2001) “Mechanical properties of
abdominal aortic aneurysm wall.” Journal of Medical Engineering and Technology, 25(4), 133-142.
Truijers, M., Pol, J.A., SchultzeKool, L.J., van Sterkenburg, S.M., Fillinger, M.F., Blankensteijn, J.D.
(2007) “Wall stress analysis in small asymptomatic, symptomatic and ruptured abdominal aortic
aneurysms.” European Journal of Vascular and Endovascular Surgery, 33, 401-407.
United States Preventative Services Task Force. (2005) “Screening for abdominal aortic aneurysm:
Recommendation statement.” Annals of internal Medicine, 142(3), 198-202.
11
Introduction
Vande Geest, J.P., Di Martino, E.S., Bohra, A., Makaroun, M.S., Vorp, D.A. (2006) “A biomechanics-
based rupture potential index for abdominal aortic aneurysm risk assessment.” Annals of the New York
Academy of Science, 1085, 11-21.
Vande Geest, J.P., Wang, D.H.J., Wisniewski, S.R., Makaroun, M.S., Vorp, D.A (2006) “Towards a non-
invasive method for determination of patient-specific wall strength distribution in abdominal aortic
aneurysms.” Annals of Biomedical Engineering, 34(7), 1098-1106.
Venkatasubramaniam, A.K., Fagan, M.J., Mehta, T., Mylankal, K.J., Ray, B., Kuhan, G., Chetter, I.C.,
McCollum, P.T. (2004) “A comparative study of aortic wall stress using finite element analysis for
ruptured and non-ruptured abdominal aortic aneurysms.” European Journal of Vascular and Endovascular
Surgery, 28, 168-176.
Vogel, T.R., Nackman, G.B., Crowley, J.G., Bueno, M.M., Banavage, A., Odroniec, K., Brevetti, L.S.,
Ciocca, R.G., Graham, A.M. (2005) “Factors impacting functional health and resource utilisation
following abdominal aortic aneurysm repair by open and endovascular techniques.” Annals of Vascular
Surgery, 19, 641-647.
Volokh, K.Y., Vorp, D.A. (2008) “A model of growth and rupture of abdominal aortic aneurysm.”
Vorp, D.A. (2008) “Biomechanics of abdominal aortic aneurysm.” Journal of Biomechanics, 40, 1887-
1902.
Vorp, D.A., Raghavan, M.L., Webster, M.W. (1998) “Mechanical wall stress in abdominal aortic
aneurysm: influence of diameter and asymmetry.” Journal of Vascular Surgery, 27(4), 632-639.
Vorp, D.A., Vande Geest, J. (2005) Biomechanical determinants of abdominal aortic aneurysm rupture.”
Artheriosclerosis, Thrombosis and Vascular Biology, 25, 1558-1566.
Wang, D.H.J., Makaroun, M.S., Webster, M.W., Vorp, D.A. (2002) “Effect of intraluminal thrombus on
wall stress in patient-specific models of abdominal aortic aneurysm.” Journal of Vascular Surgery, 36,
598-604.
Watton, P., Hill, N., Heil, M. (2004) A mathematical model for the growth of the abdominal aortic
aneurysm.” Biomechanics and Modelling in Mechanobiology, 3(2), 98-113.
12
PART I
COMPUTATIONAL INVESTIGATION
Chapter I
Improved assessment and treatment of abdominal
aortic aneurysms: The use of 3D reconstructions as a
surgical guidance tool in endovascular repair.
Barry J. Doyle,1 Pierce A. Grace,2 Eamon G. Kavanagh,2 Paul E.

Burke,2 Fintan Wallis,3 Michael T. Walsh1 and Timothy M.
McGloughlin1
1. Centre for Applied Biomedical Engineering Research (CABER), Department of Mechanical
and Aeronautical Engineering, and the Materials and Surface Science Institute, University of
Limerick, Ireland.
2. Department of Vascular Surgery, HSE Midwestern Regional Hospital, Limerick, Ireland.
3. Department of Radiology, HSE Midwestern Regional Hospital, Limerick, Ireland
Irish Journal of Medical Science (in press)

Conception and design: TM, BD
Analysis and interpretation: BD, EK
Data collection: PG, EK, PB, FW
Writing the article: BD
Critical revision of article: PG, EK, PB, MW, TM
Final approval: TM, MW, PG
Obtained funding: TM, BD
Chapter One 3D Reconstructions for EVAR
ABSTRACT
Background
Endovascular repair is fast becoming the treatment of choice for abdominal aortic
aneurysms in anatomically suitable patients. 3D reconstructions not only aid
conventional 2D measurements but also allow further analyses of the vessel anatomy.
Methods
CT scan data was obtained for four male patients that were either awaiting or had
undergone endovascular repair. 3D reconstructions were performed to determine
measurements. Wall stress was assessed on one particular case using finite element
analysis.
Results
3D reconstruction allows measurements to be obtained that can be difficult to measure
using 2D images. This method compliments traditional 2D approaches.
Reconstructions also provided imaging of potential anatomical problems. Wall stress
results showed key regions that may be possible rupture sites.
Conclusions
3D reconstructions greatly aid surgical planning. As stent-graft devices evolve,
anatomical difficulties previously considered contraindications to endovascular repair
can now be overcome with careful planning. 3D reconstruction is a useful adjunct to
assessment and planning of endovascular repair.
13
INTRODUCTION
An abdominal aortic aneurysm (AAA) can be defined as a permanent and irreversible
localised dilatation of the aorta [1]. Aneurysms form due to alterations of the
connective tissue in the aortic wall. This degradation of the aortic wall is attributed to
risk factors such as tobacco smoking, sex, age, hypertension, chronic obstructive
pulmonary disease, hyperlipidaemia, and family history of the disorder [1]. With recent
advancements in non-invasive diagnostic imaging more AAAs are being detected.
Approximately 150,000 new cases are diagnosed each year in the US [2-4], and 15,000
deaths per year are attributed to AAA rupture [5]. The normal diameter of the
abdominal aorta can vary with increasing age, sex and bodyweight [1,2], and decreases
progressively from its entry into the abdominal cavity to the iliac bifurcation. The
infrarenal abdominal aortic diameter ranges from 15 mm to 24 mm for most elderly men
[1] and it was proposed [6] that an abdominal aortic aneurysm is an aorta with an
infrarenal diameter greater than 30 mm. As the diameter of the infrarenal aorta is
dependent on age and sex [7], a size of at least 1.5 times greater than the expected
normal diameter of the infrarenal aorta [8], has been suggested as an appropriate
definition of an aneurysm.
Currently, the indication for surgical intervention is based on the maximum diameter of
the abdominal aortic aneurysm in the context of the patient’s general health status, with
most repairs performed when the aneurysm exceeds 50 to 60 mm [9-12]. The aim of
AAA treatment is to totally exclude the aneurysm, isolating the aneurysm sac from
systemic blood pressure and flow so as to minimise the risk of rupture. As a result of
AAA treatment, most aneurysms should stabilise or shrink [13], although, it has been
reported that aneurysms can often enlarge after AAA treatment due to endotension [14].
Surgical repair can be either traditional open repair or endovascular repair (EVAR).
With traditional surgical repair, the abdomen is entered via a long midline or a wide
transverse incision. Clamping of the infrarenal, or rarely the suprarenal, aorta is
necessary while suturing either a Dacron or expanded polytetrafluoroethylene (ePTFE)
graft to the neck of the aorta proximal to the aneurysm and either the distal aorta (tube
graft) or to both iliac arteries (bifurcated graft) distal to the aneurysm which is
effectively removed from the circulation. Endovascular repair consists of the placement
of a stent-graft via an intraluminal introducer across the aneurysm and its fixation to the
normal aortic and iliac walls with self-expanding stents at both ends [1]. This treatment
involves the surgical exposure of the common femoral arteries where the endovascular
14
graft can be inserted by an over-the-wire technique. The early benefit of the minimally
invasive EVAR approach has been proven in several randomised trials in reducing the
perioperative mortality compared with open repair [14-17]. It is thought that this
benefit is a result of the avoidance of the large abdominal incision and consequent
systemic inflammatory response, respiratory and wound complications as well as the
avoidance of aortic clamping with resultant cardiovascular complications.
The conversion of computed tomography (CT) scans into 3D reconstructions of patient-

specific regions is of great use to surgeons and is becoming widely used [18]. For many
years regions of the human body have been reconstructed in order to give clinicians a
further insight into the particular problem region or condition. The focus of this paper is
on the use of 3D reconstruction of the human abdominal aorta in patients suffering from
aneurysmal disease. Also included in this study is the methodology and results of a
finite element analysis (FEA) of a particularly extreme case of aneurysmal disease.
FEA is a commonly used engineering technique to determine stress distributions
experienced in complex structures and is used widely in biomedical engineering
applications. The use of FEA in assessing the propensity of an aneurysm to rupture is
becoming more widespread among researchers [11,12,19-22], and helps provide a
useful insight into the biomechanical behaviour of the aorta. Ultimately, these
reconstructions provide the clinician with valuable data regarding dimensions,
geometry, and possible problem areas that may arise during endovascular surgical
repair.
METHODS
3D Reconstructions
Computed tomography (CT) scan data was obtained from the Midwestern Regional
Hospital, Limerick, and St. James’s Hospital, Dublin, for 4 patients who were either
awaiting or had undergone surgical repair of an AAA. All patients were male with a
mean age of 77.5 yrs (range 59 - 87 yrs). CT scans were obtained using a Somotom
Plus 4 (Siemens AG, D-91052 Erlangen, Germany). The mean pixel size of the CT
scans was 0.675 mm with all scans taken using a 3 mm slice increment. The CT scans
were imported into the commercially available software Mimics v12 (Materialise,
Belguim) for reconstruction. The full reconstruction technique has been reported
previously by our group [23]. Briefly, a thresholding technique is applied to each scan
in the series. This process assigns a pixel intensity value measured in Hounsfield units
15
(HU) to each pixel in the image. From this, the HU value can be controlled so that only
the regions of interest are thresholded. Following this thresholding, segmentation of the
image is possible. This assigns a certain colour to a certain region of interest, in this
case the diseased aorta. By applying this algorithm to the series of CT scans, a
complete 3D reconstruction can be generated. This process can be seen in Figure 1.
Figure 1: (A) Typical CT scan showing aneurysmal aorta in centre of image, (B)
algorithm defines regions of interest and (C) software generates 3D reconstruction of
desired regions.
Assigning regions into different layers allows certain areas to be highlighted or removed
depending on the desired region of interest. An example of this layered approach to
reconstruction can be seen in Figure 2. In this particular reconstruction, all major
internal structures were generated for illustration purposes.
For the particular reconstructions used in this study, only the lumen regions were of
clinical importance as this is the region to which the stent-graft is deployed and fixated.
For the purposes of FEA, the full aneurysmal sac of the extremely diseased aorta was
reconstructed.
16
Figure 2: Full 3D reconstruction of CT scan data set. Image shows the use of layers to
highlight/remove particular regions of interest depending on desired areas.
Finite Element Analysis (FEA)

The process of determining wall stresses experienced within the aortic wall from CT
scan data sets has been previously reported by this group [21,22,24]. The FEA
technique works by subdividing the structure into smaller regions called elements, with
each element connected via nodes. The resulting subdivision is known as a mesh. The
stress at each integration point is then mathematically computed using the appropriate
software, in this case ABAQUS v6.7 (Dassault Systemes, SIMULIA, R.I., USA).
Realistic boundary conditions were then applied to the aorta. In this study, FEA was
used to determine the resulting wall stress distributions for the extremely diseased case
shown in Figure 3. The model was meshed using shell elements. The intraluminal
thrombus (ILT) was omitted from the analysis. The ILT has been shown to reduce wall
stress by acting like a mechanical buffer between the internal blood pressure and the
aortic wall [24]. An internal peak systolic blood pressure of 120 mmHg (16 kPa) was
applied to the internal region of the aorta, with the reconstruction constrained from
movement at the entrance to the aortic arch and the iliac arteries to represent tethering to
the remainder of the ascending aorta and the iliac arteries. The aortic wall was assumed
to have a uniform thickness of 1.9 mm [10,20,25], and was modelled as a non-linear
17
hyperelastic material [10]. These boundary conditions have been implemented in many
previous studies [10,19,21,22,24,26-28].
Figure 3: Meshed reconstruction for use in FEA. Case shown is that of an extremely
diseased aorta with thoracic aortic aneurysm (TAA), abdominal aortic aneurysm (AAA)
and also iliac aortic aneurysm (IAA). Case shown here is from the anterior viewpoint.
RESULTS
3D Reconstructions
Three patients awaiting surgical repair and one who had received repair were
reconstructed into virtual 3D models for examination. Reconstruction times can range
from as little as 2 minutes for a basic model where minor details are ignored, to 1 hour
where all details are included such as ILT and surface indentations. The critical
dimensions when concerned with EVAR stent-grafts are the neck diameter of the stent-
graft, the iliac diameter, and also the length of the device. Clinician’s can usually
determine the neck diameter without difficulty using 2D CT scans, but difficulties can
arise when determining length. 3D reconstruction allows the overall length to be easily
18
measured, thus allowing the exact sizing of the medical device. An example of this
AAA length measurement can be seen in Figure 4. The measurements are taken from
below the lowest renal artery to a point in the iliac arteries where the clinician feels
comfortable will maintain an adequate fixation of the device.
Figure 4: Example reconstruction showing measurements used to size the stent-graft.

All measurements are in millimeters (mm). Total length of stent-graft was determined to
be 159 mm.
Concerns arose about one particular patient when CT scans revealed an extremely
tortuous proximal neck that may cause difficulties during the surgical procedure. 3D
reconstruction highlighted the degree of curvature of the neck and allowed the clinician
to decide on the most suitable approach before EVAR. As shown in Figure 5, the
proximal neck of the AAA required special attention during the operation. As the upper
portion of the stent-graft is embedded into the proximal neck, the tortuous nature of this
could complicate placement and fixation of the device.
The patient-specific reconstruction shown in Figure 6 was a particularly extreme case.

In this reconstruction, the full extent of the aneurysmal aorta was revealed. This patient
had a descending thoracic abdominal aneurysm (TAA), abdominal aortic aneurysm
(AAA) and also an iliac aortic aneurysm (IAA).
19
Figure 5: 3D reconstruction revealed extremely tortuous proximal neck which can be

identified using 2D CT scans. The 3D reconstruction however, showed exactly the
degree of tortuosity, thus allowing the clinician to decide on the most suitable means of
approach during EVAR. Reconstruction shown from the left viewpoint.
Figure 6: 3D reconstruction showing thoracic aortic aneurysm (TAA), abdominal aortic

aneurysm (AAA) and iliac aortic aneurysm (IAA) from various viewpoints.
20
Finite Element Analysis (FEA)

The resulting wall stress distributions revealed that the TAA of this case experienced
higher stresses than that of both the AAA and IAA. In this case, the peak wall stress
was 1.356 MPa and was located at an inflection point at the proximal region of the
TAA, as shown in Figure 7. Inflection points have been previously reported to act as
stress raisers on the surface of aneurysms [19,21,22,24], with inflection points defined
as regions where the curvature changes from concave to convex. The peak wall stress
was computed to be 1.356 MPa, and the failure strength of thoracic aortic aneurysmal
tissue has been reported to be approximately 1.19 MPa [25]. Since it is known that
rupture will occur when local wall stress exceeds local wall strength, these findings
provided further support for the clinical intervention.
Figure 7: Resulting wall stress distribution for case examined using FEA. Peak stress
was located at the proximal inflection point of the TAA. High stress is shown in red in
the figure. Lower stresses were observed in both the AAA and the IAA. Model shown
from the posterior viewpoint.
21
DISCUSSION
The focus of this paper was to highlight the use of 3D reconstruction as a surgical
guidance tool for vascular surgeons. Currently, many clinicians use 2D CT scans to
determine the correct sizing of stent-grafts and also to visualise the aneurysm prior to
surgery. 3D reconstructions provide a useful addition to 2D measurements for stent-
graft sizing, and therefore can possibly help improve the outcome of EVAR. This
increases surgical confidence in the operation, as the clinician will not encounter any
unforeseen geometrical obstacles to the already difficult procedure. Measurements
obtained from 2D CT scans are often accurate, in particular when measuring diameter,
as the measurement must be taken in a plane perpendicular to the true lumen of the
vessel. Many clinicians measure stent-graft length from 2D images as the tortuous
nature of the aneurysmatic aorta can often be reduced with the introduction of the stiff
guide wire. A combination of the methods may provide a better approximation of the
exact length, thus allowing more precise stent-graft lengths to be obtained.
Measurements from the virtual 3D model however, offer further insight into the
morphology of the diseased aorta. Iliac bifurcation angles can now be accurately
measured in order to determine the correct stent-graft for the particular application.
Also, when fenestrated stent-grafts are to be utilised, the distances between the
mesenteric arteries and the renal arteries can be accurately obtained. These exact
measurements can be difficult to determine from 2D images. It is also common practice
to use a graduated measuring catheter placed over the stiff guide wire prior to
deployment of the device, in order to ensure the correct length stent-graft is to be
introduced. The case presented in Figure 4 was repaired using the Talent™ stent-graft
by Medtronic with no complications 2 years post-operation. Reconstruction times can
vary depending on the complexity of the case. Most reconstructions can be performed
under 1 hour, which should be adequate as the majority of CT scans are taken a
considerable time prior to the actual operation. Therefore, the clinician can often afford
to examine a 3D reconstruction without sacrificing the health status of the patient.
The reconstruction shown in Figure 5 revealed a tortuous proximal neck and iliac
aneurysm, both of which increase the difficulty of EVAR. Fixation of the distal end of
the device in the iliac arteries is hampered by the aneurysmal or tortuous artery, and
therefore, the device may dislodge and ultimately fail. Calcified or tortuous iliac
arteries may also rupture upon insertion or withdrawal of the introducer. The problems
associated with irregular iliac arteries were highlighted in 2003 when Guidant were
22
forced to withdraw the Ancure™ stent-graft from the market [29]. The diameter of the
iliac arteries is of obvious importance when sizing stent-grafts, but so too is the
tortuosity, as the introducer can place large stresses on the artery wall whilst being
navigated through the vessel. Nowadays clinicians are tackling more challenging
anatomy with EVAR compared to recent years, and therefore, the ability to view the
exact morphology of the iliac arteries has clinical relevance, and may not always be
possible with 2D CT scans.
Reconstruction also gives the clinician valuable information about the aneurysm before
surgery. This ensures that the surgeon does not encounter any unforeseen problems
upon surgical repair. This was particularly useful for cases where the aneurysm
displays extreme tortuosity and curvature, like the cases presented in Figures 5 and 6.
Enabling the clinician to visualise the AAA prior to surgery increases confidence in the
procedure as the degree of curvature of the neck or iliac arteries can be readily
determined. The case presented in Figure 5 was repaired using the Aorfix™ stent-graft
by Lombard due to the tortuous nature of the anatomy. This stent-graft showed no
complications 1 year post-operation. This patient later died due to unrelated causes.
Extreme cases, such as that of Figure 6, highlight the advantages of 3D reconstruction
as a surgical tool. From the reconstruction, precise measurements can be determined,
such as length, degree of curvature, angles, and exact distances over surfaces, therefore
aiding in the choice of a patient-specific stent-graft from the range currently available.
Cases such as that presented in Figure 6 may often require tailor-made stent-grafts, and
the use of reconstructions aid this task. Tailor-made stent-grafts can be designed using
the centrelines generated from the reconstructions, and the resulting flow patterns can
be identified [30]. Clinicians will usually opt to have a range of device sizes at hand
during the operation so that the correct device can be obtained in the event of
unforeseen circumstances. Full examinations using 3D reconstructions can help to
alleviate this worry.
3D reconstructions also allow the clinical problem to be examined using various

software applications. Diseased aortas can now be modelled using complex
mathematical software to determine stresses and strains experienced in the wall, and
also the complex flow patterns of the blood through the diseased artery allowing for
stent-graft design [30]. The results of the FEA studied here show that the aorta is
experiencing a peak stress of 1.356 MPa at the proximal inflection point of the TAA,
23
which exceeds the average failure strength of tissue in this area as reported previously
[25]. FEA also identifies hotspots on the diseased aorta that may indicate possible
rupture sites in vivo. The reconstruction of the stent-graft after placement within the
TAA can be seen in Figure 8.
Figure 8: Post-operative EVAR illustration showing deployed TAA stent-graft

excluding TAA sac. Stent-graft now returns blood flow in the descending aorta to a
relatively normal state. Transparency of the aorta in the image allows easy visualisation
of the device. Model shown from the posterior viewpoint.
The device now excludes the region of high stress and thus aids to return the aorta to a
lower stress state that is relatively safe from rupture. This reconstruction was generated
from post-operative CT scans that help to monitor the diseased aorta, and highlight any
additional aneurysm growth from endotension [14], or further progression of the AAA.
Post-operative 3D reconstructions can also aid the clinician to observe the outcome of
the EVAR procedure at regular intervals. Stent-graft limbs can often become kinked,
twisted or occluded by thrombosis [17], which can complicate the flow of blood and
24
may require further intervention by the clinician. Although these complications can be
identified using 2D CT scans, 3D reconstructions allow exact visualisation of the
problem and aid towards a better understanding of the overall situation of the patient.
Currently, there is no institution-wide standard regarding stent-graft sizing, with many
clinicians opting for their own methods of measurements, particularly lengths. There is
still a degree of learning among the EVAR community when concerned with AAA
morphology and stent-graft behaviour. The latest generation of the Medtronic
Endurant™ graft appears to shorten more in length on deployment than its predecessor,
Talent™, due to design modifications. This consequently affects the recommended
measurements and sizing by the manufacturer. Ultimately, better imaging and 3D
reconstructions can only add to the quality of the endovascular repair.
CONCLUSION
3D reconstruction of AAAs is a powerful and useful surgical guidance tool.
Reconstruction allows the clinician to obtain measurements useful for the sizing of
stent-grafts, and also allows the clinician to visualise the aneurysm prior to surgery.
Reconstructions are also necessary for further use with FEA which has been shown to
be a good method of determining wall stress in the diseased aorta. FEA provides an
additional source of information to the clinician and helps towards a greater
understanding of the biomechanical behaviour of the anatomy prior to surgery. 3D
reconstruction is quick to perform and could not only help towards standardising the
measurement and sizing of stent-grafts, but could also aid surgeons in improving the
treatment of AAAs.
ACKNOWLEDGMENTS
The authors would like to thank (i) the Irish Research Council for Science, Engineering
and Technology (IRCSET) Grant RS/2005/340 (ii) Grant #R01-HL-060670 from the
US National Heart Lung and Blood Institute (iii) the Department of Vascular Surgery in
the HSE Midwestern Regional Hospital, Ireland (iv) the Department of Radiology in the
HSE Midwestern Regional Hospital, Ireland (v) Prof. David A. Vorp from the
Departments of Surgery and Bioengineering, Centre for Vascular Remodelling and
Regeneration, McGowan Institute of Regenerative Medicine, University of Pittsburgh,
USA and (vi) Mr. Prakash Madhavan, Consultant Vascular Surgeon, St. James Hospital,
Dublin, Ireland.
25
REFERENCES
1. Sakalihasan, N., Limet, R., Defawe, O.D. (2005) “Abdominal aortic aneurysm.”
Lancet, 365, 1577-1589.
2. Bengtsson, H., Nilsson, P., Bergqvist, D. (1998) “Natural history of abdominal
aortic aneurysm detected by screening.” British Journal of Surgery, 80, 718-720.
3. Vorp, D.A. (2007) “Biomechanics of abdominal aortic aneurysm.” Journal of
Biomechanics, 40, 1887-1902.
4. Ouriel, K., Green, R.M., Donayre, C., Shortell, C.K., Elliott, J., DeWeese, J.A.
(1992) “An evaluation of new methods of expressing aortic aneurysm size:
relationship to rupture.” Journal of Vascular Surgery, 15, 12-20.
5. Kleinstreuer, C., Li, Z. (2006) “Analysis and computer program for rupture-risk
prediction of abdominal aortic aneurysms.” Biomedical Engineering Online, 5, 19.
6. McGregor, J.C., Pollock, J.G., Anton, H.C. (1975) “The value of ultrasonography
in the diagnosis of abdominal aortic aneurysm.” Scottish Medical Journal, 20, 133-
137.
7. Grimshaw, G.M., Thompson, J.M. (1997) “Changes in diameter of the abdominal
aorta with age: An epidemiological study.” Journal of Clinical Ultrasound, 25, 7-
13.
8. Johnston, K.W., Rutherford, R.B., Tilson, M.D., Shah, D.M., Hollier, L., Stanley,
J.C. (1991) “Suggested standards for reporting on arterial aneurysms.
Subcommittee on reporting standards for arterial aneurysms, Ad hoc committee on
reporting standards, Society for Vascular Surgery and North American Chapter,
International Society for Cardiovascular Surgery.” Journal of Vascular Surgery, 13,
452-458.
9. Sayers, R.D. (2002) “Aortic aneurysms, inflammatory pathways and nitric oxide.”
Annals of the Royal College of Surgeons of England, 84(4), 239-246.
10. Raghavan, M.L., Vorp, D.A. (2000) “Toward a biomechanical tool to evaluate
rupture potential of abdominal aortic aneurysm: identification of a finite strain
constitutive model and evaluation of its applicability.” Journal of Vascular
Surgery, 33, 475-482.
11. Fillinger, M.F., Marra, S.P., Raghavan, M.L., Kennedy, F.E. (2003) “Prediction of
rupture risk in abdominal aortic aneurysm during observation: wall stress versus
diameter.” Journal of Vascular Surgery, 37, 724-732.
12. Fillinger, M.F., Raghavan, M.L., Marra, S.P., Cronenwett, J.L., Kennedy, F.E.
(2002) “In vivo analysis of mechanical wall stress and abdominal aortic aneurysm
rupture risk.” Journal of Vascular Surgery, 36, 589-597.
13. Morris, L.G. (2004) “Numerical and experimental investigation of mechanical
factors in the treatment of abdominal aortic aneurysms.” PhD Thesis, University of
Limerick.
14. EVAR Trial Participants. (2004) “Comparison of endovascular aneurysm repair
with open repair in patients with abdominal aortic aneurysm (EVAR trial 1), 30-
day operative mortality results: randomised controlled trial.” Lancet, 364, 843-848.
15. EVAR trial participants. (2005) “Endovascular aneurysm repair versus open repair
in patients with abdominal aortic aneurysm (EVAR trial 1): Randomised control
trial.” Lancet, 365, 2179-2186.
16. Schermerhorn, M., O’Malley, J., Jhaveri, A., Cotterill, P., Pomposelli, F., Landon,
B.E. (2008) “Endovascular vs. open repair of abdominal aortic aneurysms in the
Medicare population.” New England Journal of Medicine, 358, 464-474.
17. Corbett, T.J., Callanan, A., Morris, L.G., Doyle, B.J., Grace, P.A., Kavanagh, E.G.,
McGloughlin, T.M. (2008) “In vivo and in vitro biomechanical behaviour and
performance of post-operative abdominal aortic aneurysms and implanted stent-
grafts: A review.” Journal of Endovascular Therapy, 15(4), 468-484.
26
18. Resch, T.A. “TeraRecon Clinical Case Studies: Endograft planning for complex
aortic aneurysms. Available via TeraRecon.”
http://www.terarecon.com/downloads/news/clinical_case_studies_v2/casestudy_Re
sch-EndograftPlanning.pdf: Accessed 10 March 2008.
19. Vorp, D.A., Raghavan, M.L., Webster, M.W. (1998) “Mechanical wall stress in
abdominal aortic aneurysm: influence of diameter and asymmetry.” Journal of
Vascular Surgery, 27(4), 632-639.
20. Raghavan, M.L., Vorp, D.A., Federle, M.P., Makaroun, M.S., Webster, M.W.
(2000) “Wall stress distribution on three-dimensionally reconstructed models of
human abdominal aortic aneurysm.” Journal of Vascular Surgery, 31, 760-769.
21. Doyle, B.J., Callanan, A., Walsh, M.T., Grace, P.A., McGloughlin, T.M. (2009)
“A finite element analysis rupture index (FEARI) as an additional tool for
abdominal aortic aneurysm burst prediction.” Vascular Disease Prevention, 6, 114-
121.
22. Doyle, B.J., Callanan, A., Burke, P.E., Grace, P.A., Walsh, M.T., Vorp, D.A.,
McGloughlin, T.M. (2009) “Vessel asymmetry as an additional diagnostic tool for
the assessment of abdominal aortic aneurysms.” Journal of Vascular Surgery, 49,
443-454.
23. Doyle, B.J., Morris, L.G., Callanan, A., Kelly, P., Vorp, D.A., McGloughlin, T.M.
(2008) “3D reconstruction and manufacture of real abdominal aortic aneurysms:
From CT scan to silicone model.” Journal of Biomechanical Engineering, 130,
034501-5.
24. Doyle BJ, Callanan A and McGloughlin TM (2007) A comparison of modelling
techniques for computing wall stress in abdominal aortic aneurysms. Biomedical
Engineering Online 6:38.
25. Vorp, D.A., Shiro, B.J., Ehrlich, M.P., Juvonen, T.S., Ergin, M.A., Griffith, B.P.
(2003) “Effect of aneurysm on the tensile strength and biomechanical behaviour of
the ascending thoracic aorta.” Annals of Thoracic Surgery, 75, 1210-1214.
26. Thubrikar, M.J., Al-Soudi, J., Robicsek, F. (2001) “Wall stress studies of
abdominal aortic aneurysm in a clinical model.” Annals of Vascular Surgery, 15,
355-366.
27. Truijers, M., Pol, J.A., SchultzeKool, L.J., van Sterkenburg, S.M., Fillinger, M.F.,
Blankensteijn, J.D. (2007) “Wall stress analysis in small asymptomatic,
symptomatic and ruptured abdominal aortic aneurysms.” European Journal of
Vascular and Endovascular Surgery, 33, 401-407.
28. Speelman, L., Bohra, A., Bosboom, E.M.H., Schurink, G.W.H., van de Vosse,
F.N., Makaroun, M.S., Vorp, D.A. (2007) “Effects of wall calcifications in patient-
specific wall stress analyses of abdominal aortic aneurysms.” Journal of
Biomechanical Engineering, 129, 1-5.
29. Faries, P.L., Dayal, R., Rhee, J., Trocciola, S., Kent, C.K. (2004) “Stent graft
treatment for abdominal aortic aneurysm repair: recent developments in therapy.
Diseases of the aorta, pulmonary, and peripheral vessels.” Current Opinion in
Cardiology, 19(6), 551-557.
30. Molony, D.S., Callanan, A., Morris, L.G., Doyle, B.J., Walsh, M.T., McGloughlin,
T.M. (2008) “Geometrical enhancements for abdominal aortic stent grafts.” Journal
of Endovascular Therapy, 15(5), 518-529.
27
Chapter II
A comparison of modelling techniques for computing
wall stress in abdominal aortic aneurysms.
Barry J. Doyle, Anthony Callanan and Timothy M. McGloughlin
Centre for Applied Biomedical Engineering Research (CABER), Department of Mechanical and
Aeronautical Engineering, and the Materials and Surface Science Institute, University of
Limerick, Ireland.
Biomedical Engineering Online 2007; 6: 38

Conception and design: BD
Analysis and interpretation: BD, AC
Data collection: BD
Critical revision of article: AC, TM
Final approval: TM
Chapter Two Comparison of FEA Techniques
ABSTRACT
Background
Aneurysms, in particular abdominal aortic aneurysms (AAA), form a significant portion
of cardiovascular related deaths. There is much debate as to the most suitable tool for
rupture prediction and interventional surgery of AAAs, and currently, maximum
diameter is used clinically as the determining factor for surgical intervention. Stress
analysis techniques, such as, finite element analysis (FEA) to compute the wall stress in
patient-specific AAAs have been regarded by some authors to be more clinically
important than the use of a “one-size-fits-all” maximum diameter criterion, since some
small AAAs have been shown to have higher wall stress than larger AAAs and have
been known to rupture.
Methods
A patient-specific AAA was selected from our AAA database and 3D reconstruction
was performed. The AAA was then modelled in this study using three different
approaches, namely, AAA(SIMP), AAA(MOD) and AAA(COMP), with each model
examined using linear and non-linear material properties. All models were analysed
using the finite element method for wall stress distributions.
Results
Wall stress results show marked differences in peak wall stress results between the three
methods. Peak wall stress was shown to reduce when more realistic parameters were
utilised. It was also noted that wall stress was shown to reduce by 59% when modelled
using the most accurate non-linear complex approach, compared to the same model
without intraluminal thrombus.
Conclusions
The results here show that using more realistic parameters affect resulting wall stress.
The use of simplified computational modelling methods can lead to inaccurate stress
distributions. Care should be taken when examining stress results found using
simplified techniques, in particular, if the wall stress results are to have clinical
importance.
28
BACKGROUND
Cardiovascular disease is the leading cause of morbidity and premature deaths of
modern era medicine and aneurysms are a major contributor to the poor clinical
outcomes. Aneurysm, meaning widening, can be defined as a permanent and
irreversible localised dilatation of a vessel [1]. Although this disease can form in any
blood vessel, artery or vein, the more serious aneurysms occur in the abdominal aorta,
the brain arteries, and the heart and thoracic aorta. The vast majority of these
aneurysms occur in the abdominal aorta, and are termed Abdominal Aortic Aneurysm
(AAA). There is much debate amongst researchers and surgeons as to the most
appropriate criteria for surgical intervention of unruptured AAAs [2-6]. Currently, the
maximum diameter of the aneurysm is the most commonly used determinant of surgical
intervention, with clinicians opting for surgery when the AAA exceeds 5 cm in
transverse diameter [4,5,7]. Although this maximum diameter criterion can be justified,
as the rupture risk for an AAA is clearly related to its maximal diameter [5,8], it is also
known that small AAAs can rupture [5,9-11]. Nicholls et al. [11] reported in one study
that 10% to 24% of ruptured aneurysms were 5 cm or less in diameter, and therefore
there is clearly a need for an improved predictor of rupture.
It is believed by many researchers that stress analyses are more accurate indicators than
diameter for predicting rupture [2,5-7,12-14], and it is clear that a reliable method of
predicting AAA rupture has definite clinical importance [2,3,6,7,13].
It has been previously documented how AAA formation is accompanied by an increase

in wall stress [2,7], and a decrease in wall strength [15-17]. It has been identified that
aneurysm wall stress does not follow the traditional Law of Laplace, in that AAAs with
equivalent diameters and pressures could have largely different stress distributions
[2,3,7,13]. Researchers have been using stress analysis techniques, such as finite
element analysis (FEA), to compute stresses within AAAs for some time now
[2,5,7,13], but this work has yet to be validated experimentally using realistic AAA
models. Morris et al. [18] showed experimentally the stress contours in an idealised
AAA using the photoelastic method, which was later validated numerically by Callanan
et al. [19]. Previous numerical studies have shown how asymmetry and diameter affect
wall stress [2], and also how this wall stress is distributed within a realistic AAA
[5,7,13,20]. Dobrin [21] suggested that the presence of intraluminal thrombus (ILT)
neither reduces the luminal pressure exerted on the wall nor offers a retractive force and
29
thus has no effect on the wall stress. This ILT was later reported to significantly reduce
the stress acting on the AAA wall [13] and to act as a mechanical cushion [13,17],
although some researchers dispute this [22].
The purpose of this study is to analyse and compare the differences associated with
finite element analysis (FEA) modelling of AAAs. Many researchers have used
simplified AAA models in their studies [2,23-25], compared to others who have used
much more complex, realistic methods [2,3,7,13,20,26-28]. The more common
simplifying assumptions include treating the biomechanical material properties of the
vessel as linearly elastic and excluding the intraluminal thrombus (ILT) from the model.
One patient-specific AAA was examined in this study. The model was examined from
various aspects of complexity, starting with a simplified approach, AAA(SIMP),
followed by a moderate method, AAA(MOD), and then a complex technique,
AAA(COMP), with results and differences contrasted and compared. In this study,
simplified refers to the use of common simplifying assumptions. Moderate refers to the
use of a limited amount of simplifying assumptions, with complex inferring the
inclusion of ILT. All models were examined using both linear and non-linear material
properties.
METHODS
A patient suffering from an AAA was selected from our AAA database which included
modest levels of intraluminal thrombus (ILT). The study subject was male, and the
AAA maximum transverse diameter was 5.1 cm, and so had exceeded the 5 cm
threshold for surgical intervention. The total length of the AAA was 13.2 cm, had a
total volume of 176.7 cm3, of which 67.5 cm3 was ILT. For clinically meaningful stress
analysis of AAAs, patient-specific computed tomography (CT) scans were obtained for
the patient. This CT data was then reconstructed using commercially available software
(Mimics v10.0, Materialise, Belgium). In total, six models were created and analysed.
Three different methods of reconstruction were utilised to create the AAA, with each
model examined using both linear and non-linear material properties. For the purpose
of this study, the AAAs modelled using the first set of simplifying assumptions will be
referred to as AAA(SIMP)L and AAA(SIMP)NL, with the subscripts L and NL referring to
linear and non-linear material properties, respectively. The second set of AAAs will be
referred to as AAA(MOD)L and AAA(MOD)NL, due to the use of moderate simplifying
assumptions. The third set of AAAs, which used a more complex method of analysis
30
will be referred to as AAA(COMP)L and AAA(COMP)NL. All stress analyses were

performed using the finite element analysis software ABAQUS v6.6-2 (Dassault
Systemes, SIMULIA, Rhode Island, USA) on a standard desktop with a 3.2 GHz
processor and 4 GB of RAM.
3D Reconstruction
Spiral CT data was then used to reconstruct the infrarenal section of the aorta. As CT
scanning is routinely performed on AAA patients scheduled for repair, collection of this
information involved no extra participation by the study subject. Digital files in Digital
Imaging and Communications in Medicine (DICOM) file format, containing cross-
sectional information was then imported to Mimics for reconstruction. A Hounsfield
unit (HU) thresholding technique was then applied to each CT slice in order to identify
the region of interest. A HU value of 226 is sufficient to identify the lumen region of
the AAA due to the non-ionic contrast. This contrast dye is an organic solution that
makes internal bodily structures visible. The ILT regions of the AAA must be assigned
a lower HU value, as the material has a pixel intensity that is closer to fat than bone. A
HU value of 0 is sufficient to distinguish the ILT from the surrounding tissue. For all
models reconstructed, the iliac arteries have been omitted as with previous research [5].
Figure 1 shows the conversion from CT scan to 3D model. On the left the thresholding
and segmentation process highlights both the ILT and lumen of the AAA, with the
software generating 3D reconstructions, shown on the right. For the AAA(SIMP) and
AAA(MOD) models, the ILT was omitted, and so the reconstruction consisted of a
single layer wall of uniform thickness.
AAA(SIMP)
For this set of AAAs, the ILT was omitted from the reconstruction, and was instead
incorporated into the full AAA model as lumen. Previous work [5,6] has also taken this
approach to AAA modelling. The thickness of the aortic wall is not easily identifiable
from CT scans. Previous research [2,5-7,18,19,25,26,29] has used wall thickness
varying from 1-2 mm. As no information was available from the CT scans about the
wall thickness of each AAA, in this study the wall was assumed to be of uniform
thickness throughout the model and was set to 1.5 mm [2,26,29].
31
Figure 1: Typical CT scan and 3D reconstruction of patient suffering from an AAA. On

the left shows a typical CT scan after segmentation using Mimics v10.0, with the ILT
(red) and lumen (yellow) clearly distinguishable. 3D reconstruction of examined AAA
is shown on right.
AAA(MOD)
When reconstructing the AAA(MOD) models, the surface polylines created using
Mimics v10.0, were exported to ProEngineer Wildfire 3.0 (Parametric Technology
Corporation, Needham, M.A., USA) in order to create a 3D wall. The original polylines
were offset by 1.5 mm [2,26,29] so as to create a second artificial surface offset from
the original surface. These surfaces were then connected so as to form a realistic
uniform 3D aortic wall.
AAA(COMP)
For this AAA, the ILT region was included into the models. Therefore, the ILT is a
completely different entity to that of the AAA wall, with the two sections tied together
using realistic constraints. This is a more realistic representation of the actual AAA
with the ILT. An artificial wall was created to represent the AAA wall and, like earlier,
was set to be a uniform 1.5 mm in thickness [2,26,29].
Model Smoothing
Each of these 3D models were initially quite rough containing sharp edges and surface
artefacts, and required smoothing to ensure that the model could be readily meshed and
that stress analyses could be performed. The method of reconstruction and smoothing
was validated from previous work by our group, [30,31] and it was shown that there are
negligible differences between the two methods. In order to determine the optimum
32
level of smoothing for these reconstructions, four degrees of smoothing were examined.
These four smoothing levels were based on axial smoothing of individual polyline slices
created from the CT scans. Four models were then reconstructed. Model 1 contained 7
control points per polyline slice, model 2 had 20 control points, model 3 had 50 control
points, and model 4 had 70 control points per slice. The more control points per
polyline decreases the surface smoothness of the resulting AAA. Figure 2 shows the
difference in polylines between the various degrees of smoothing. Non-linear stress
analyses were then performed on each of the four models. Resulting wall stress
distributions are shown in Figure 3, with stress results normalised to the peak stress
experienced in the roughest model, namely model 4. Mesh independence was
performed on all models. From these results, model 2 was deemed to be the optimum
level of smoothing as unwanted surface detail is removed without causing unnecessary
over-smoothing. Figure 4 illustrates the difference between the resulting rough and
smooth models. These AAA surfaces were then exported in IGES file format for
further work.
Figure 2: Effect of axial smoothing on the polylines constructed from the CT scans.
Model 1 consists of 7 control points, Model 2 has 20 control points, Model 3 has 50
control points, and Model 4 has 70 control points per polylines slice. Green line is the
original polyline from the CT scan slice, with the red line being the new smoothed
polyline.
33
Figure 3: Resulting wall stress distributions of the various degrees of smoothing

examined. Wall stress results are normalised to the peak stress found in Model 4. Black
mark indicates abnormal locations of elevated wall stress in Models 3 and 4.
Figure 4: Effect of smoothing on the reconstructed AAA. The optimum smoothing

factor removes unwanted surface detail without over-smoothing the AAA. (A) Shows
the rough model without smoothing and (B) shows the effect of a 20 control point axial
smoothing factor. Model shown in the lateral view.
Biomechanical Material Properties

AAA(SIMP)
Firstly, the AAA material was assumed to be homogenous and isotropic with linear
elastic material properties. This model is referred to as AAA(SIMP)L. Although human
arterial tissue acts like a non-linear material, at pressures above 80 mmHg (10.67 kPa)
34
the aorta behaves more like a linearly elastic material [23,24,32,33]. The modulus of
elasticity applied to the AAA wall was 4.66 MPa, as used in previous research [24,27].
Many researchers have modelled the aorta as linearly elastic in order to examine the
influence of geometry and various other parameters on wall stress [2,4,23-25,34].
The same model was then analysed using non-linear elastic material properties, and is
referred to as AAA(SIMP)NL. The AAA wall was modelled as a homogenous isotropic
hyperelastic material using the finite strain constitutive model proposed by Raghavan
and Vorp [3] in Eqn. 1. These material properties have been utilised for many stress
analysis studies [5-7,13,20,26-28].
W = C1(IB − 3) + C2 (IB − 3)2 Eqn. (1)
Where, W is the strain energy and IB is the first invariant of the left Cauchy-Green
tensor B (IB = tr B). The constants were set to the population mean values C1=174,000
Pa (0.174 MPa) and C2=1,881,000 Pa (1.881 MPa). Aortic tissue is also known to be
nearly incompressible with a Poisson’s ratio of approximately 0.49 [2,24,35].
AAA(MOD)
For this set of AAAs, the wall was again assigned both linear and non-linear material
properties, referred to as AAA(MOD)L and AAA(MOD)NL, respectively. AAA(MOD)L
was assigned an Young’s modulus of 4.66 MPa, like earlier, and AAA(MOD)NL utilized
the non-linear model proposed by Raghavan and Vorp [3].
AAA(COMP)
In the AAA(COMP) models, the ILT was included as a separate entity in the AAA
model, and therefore required its own material property. Previous work has used the
theory of linear elasticity to model the ILT [4,28], and so for the AAA(COMP)L the ILT
was assigned a Young’s modulus of 0.11 MPa, and a Poisson’s ratio of 0.45. The AAA
wall was also assumed to be linearly elastic with the same material properties as
described earlier. For the AAA(COMP)NL model, the ILT was modelled as a
hyperelastic material using the material constants derived from 50 ILT specimens from
14 patients performed by Wang et al. [36]. The AAA wall was again analysed using the
non-linear model proposed by Raghavan and Vorp [3]. Therefore, the model consisted
of two non-linear materials, both with realistic material properties.
35
Mesh Generation
AAA(SIMP)
Once the AAA was imported into ABAQUS v.6.6-2 for stress analysis, a mesh was
generated on the AAA model. The elements chosen for the application were linear shell
elements [2,3,23]. The surface was partitioned, by dividing the model into different
regions in order to optimise the mesh creation. Each shell element was assigned a
uniform thickness of 1.5 mm [2,26,29]. Mesh independence was performed in order to
determine the optimum number of elements. In order to gain confidence in the mesh
size of each AAA model, the number of elements was incrementally increased and the
peak wall stress computed. The optimum mesh size was determined once the peak
stress did not increase by more than 2%. This method of convergence has been used in
previous studies [37,38]. It was noted that the locations of the peak stress did not
significantly alter when examining each model for mesh convergence. Both the
AAA(SIMP)L and the AAA(SIMP)NL models were meshed using shell elements, with
element numbers for each model in the order of 17000 and 30000, respectively.
AAA(MOD)
To mesh the 3D wall of the AAA(MOD) models, 3D stress elements were required in
order to mesh throughout the thickness of the AAA wall. The elements used for this
part of the study were quadratic, 10-noded, tetrahedral elements. Mesh convergence
was achieved using the same approach as described earlier. In this case, the
AAA(MOD)L and the AAA(MOD)NL models required 29000 elements. The location of
peak stress did not significantly alter in location during the convergence study.
AAA(COMP)
As with the AAA(MOD) models, the elements employed here were quadratic
tetrahedral 3D stress elements, with 10-nodes per element. The AAA wall and ILT
regions were meshed using a master-slave contact scenario, where the slave region
(ILT) has a larger number of elements than the master region (AAA wall). Both the
AAA(COMP)L and the AAA(COMP)NL models converged at 13000 elements using the
same convergence criteria as described previously.
Forces and Boundary Conditions

The blood pressure within the AAA acts on the AAA inner wall, therefore, the pressure
was applied to the inner surface of all the virtual AAA models studied. A static peak
36
pressure of 120 mmHg (16 kPa) was used. Ideally, patient-specific blood pressures
would be applied to each particular AAA, by measuring blood pressure at the time of
CT scan, but for this study the standard peak pressure of 120 mmHg was felt to be
sufficient. The shear stress induced by blood flow was neglected in this study [7,24],
although the effects of blood flow have been shown to reduce wall stress by 10% in
uniformly thick walled ideal models and by up to 30% in variable wall thickness models
[34]. In order to simulate the attachment of the AAA to the aorta at the renal junction
and iliac bifurcation, the AAA model was fully constrained in the proximal and distal
regions. Residual stresses that may exist within the aortic wall in vivo and tethering
forces on the posterior surface caused by the lumbar arteries were neglected. Bulging is
on the anterior surface due to the constraint the spinal cord places on posterior dilation.
RESULTS
In order to easily observe and visualise the resulting wall stress of each AAA model,
contours of the von Mises stress were plotted on the surface of each AAA model. The
von Mises stress is a stress index especially suited for failure analysis, as stress is a
tensor quantity with nine components, with the von Mises stress being a combination of
these components [7]. The normalised computed wall stress results for the AAA(SIMP),
AAA(MOD) and AAA(COMP) models can be seen in Figures 5, 6 and 7. In these
figures, wall stress results were normalised by using the peak stress experienced in the
linearly elastic model of each case. In each case, the linearly elastic model returned a
higher peak stress than the corresponding non-linear model. The location of peak stress
remained the same in all models except the AAA(SIMP) models, where the location of
peak stress shifted from a centred anterior location to a left lateral location. For both
linearly elastic and non-linearly elastic AAA(MOD) and AAA(COMP) models, the peak
stress region was located on the inner surface of the AAA wall. When using 3D stress
elements the stress is not interpolated through the wall thickness, as with shell elements,
but rather at the individual integration points of the element. For the AAA(COMP)
models the peak stress was located at regions between the intersection of the ILT region
and the AAA wall.
37
Figure 5: von Mises wall stress distributions for both the AAA(SIMP)L and
AAA(SIMP)NL models at an internal pressure of 120 mmHg. Wall stress results are
normalised to the peak stress found AAA(SIMP)L. The black mark indicates the region
of peak wall stress. Models are shown in the anterior view.
Figure 6: von Mises wall stress distributions for the both the AAA(MOD)L and
AAA(MOD)NL models at an internal pressure of 120 mmHg. Wall stress results are
normalised to the peak stress found AAA(MOD)L. The black mark indicates the region
38
Figure 7: von Mises wall stress distributions for both the AAA(COMP)L and
AAA(COMP)NL models at an internal pressure of 120 mmHg. Wall stress results are
normalised to the peak stress found AAA(COMP)L. The black mark indicates the region
In both the AAA(SIMP) models, the peak stress occurred at regions of relatively
maximum diameter, rather than at inflection points along the surface of the AAA sac.
For all AAA(MOD) and AAA(COMP) cases studied, the peak stress occurred at an
inflection point on the inner surface of the AAA. An inflection point is defined as
points on the AAA surface at which the local AAA wall shape changes from convex to
concave [2]. Peak stress located at inflection points has been previously reported in
idealised models, both numerically [2,19,34] and experimentally [18]. In general, wall
stress reduces across the entire surface of the AAA when the material properties become
non-linear. The recorded peak stresses for each case are also noticeably different. The
peak stress results, along with location of peak stress and overall computational time
can be seen in Table 1. Computational time here is rounded to the nearest minute. Peak
stresses varied considerably depending on the modelling approach. By comparing the
first approach of AAA reconstruction consisting of a single layer of shell elements,
AAA(SIMP), the peak stress reduced from 0.8833 MPa to 0.5814 MPa when more
realistic non-linear material properties were implemented. This drop in peak stress
when modelled from a non-linear approach followed through to the second case also,
that is, AAA(MOD). In this comparison of AAAs with 3D 1.5 mm thick uniform walls,
the peak stress reduced from 2.232 MPa to 1.034 MPa. In the third case, AAA(COMP),
39
again peak stress was reduced when realistic material properties were utilised. In this
case the peak stress reduced from 0.8036 MPa to 0.4219 MPa. These large reductions
in peak stress for each comparison can be attributed to the use of realistic material
properties, as the geometry of the models are identical. The peak stress in the
AAA(SIMP), AAA(MOD) and AAA(COMP) models reduced by 34%, 55% and 47%,
respectively, when accounting for material non-linearity. This results in an average
reduction in peak stress of 45% in all models.
The effect of geometrical modelling parameters can be seen in Figure 8 for the linearly
elastic models, and in Figure 9 for the non-linearly elastic models. The location of peak
stress is also indicated in these figures. The wall stresses were normalised using the
peak wall stress experienced in the linear elastic study (Figure 8) and the peak stress
obtained in the non-linear elastic study (Figure 9). By comparing the three models
examined from a linearly elastic approach, the AAA(MOD)L returned the highest peak
stress of 2.282 MPa. Both the AAA(SIMP)L and AAA(COMP)L returned similar peak
stresses of 0.8833 MPa and 0.8036 MPa, respectively.
Table 1: Comparison of peak wall stress, location of peak wall stress and FEA
computational time for the AAA(SIMP), AAA(MOD) and the AAA(COMP) models.
Peak Wall Stress CPU Time
Location of Peak Stress
(MPa) (min)
AAA(SIMP)L 0.8833 Centre of anterior region 1
AAA(SIMP)NL 0.5814 Left lateral region 9

Inflection point on inner surface
AAA(MOD)L 2.282 4
of anterior region
AAA(MOD)NL 1.034 19
of anterior region
AAA(COMP)L 0.8036 8
of anterior region
AAA(COMP)NL 0.4291 82
of anterior region
40
Figure 8: Normalised wall stress results and locations of peak wall stress in all the
linearly elastic models examined. Wall stress results are normalised to the peak stress
found in AAA(MOD)L. The figure shows the effect of modelling parameters and
inclusion of the ILT on peak wall stress. The black mark indicates the region of peak
wall stress. Location of peak stress for the AAA(MOD)L and AAA(COMP)L models are
on the inner surface of the AAA wall. All models are shown in the anterior view.
Figure 9: Normalised wall stress results and locations of peak wall stress in all the non-
linearly elastic models examined. Wall stress results are normalised to the peak stress
found AAA(MOD)NL. The figure shows the effect of modelling parameters and inclusion
of the ILT on peak wall stress. The black mark indicates the region of peak wall stress.
Location of peak stress for the AAA(MOD)NL and AAA(COMP)NL models are on the
inner surface of the AAA wall. All models are shown in the anterior view.
41
The “cushioning effect” of the ILT which has been previous documented by other
researchers [13,17,35-40] can be observed by comparing the linear and non-linear cases
of both the AAA(MOD) models and the AAA(COMP) models. The AAA(MOD)L and the
AAA(COMP)L models are similar in material properties except for the inclusion of
thrombus. There is a 65% reduction in stress when including the ILT in the linear case.
This “cushioning effect” is also evident in the non-linear cases, AAA(MOD)NL and
AAA(COMP)NL. Here, inclusion of ILT reduces peak stress by 59%.
DISCUSSION
In this study, one patient-specific AAA was selected for rigorous finite element
analysis. The 3D reconstruction and stress analysis was performed using three different
techniques, with each technique studied using both linearly elastic and non-linearly
elastic material properties. This study is believed to be the first paper to contrast and
compare established and widely used techniques of computing AAA wall stress
distributions. The first technique is classed as AAA(SIMP), the second as AAA(MOD),
and the third as AAA(COMP). Each model was then thoroughly analysed and compared
to determine any differences in both peak wall stress and also wall stress distribution.
The FEA technique used in this study has been validated both experimentally by Morris
et al. [18] and numerically by Callanan et al. [19] in idealised AAA models in previous
research by our group. Briefly, these previous studies used an idealised AAA model
designed by our group using realistic dimensions from a large population study [41].
Aluminium moulds were then created using a CAD/CAM procedure, and experimental
models manufactured using a technique described in previous research by both Morris
et al. [18] and Doyle et al. [31]. The photoelastic method of determining stress
distributions was then employed. A numerical model of the same AAA was then
analysed using the finite element method, with good agreement between the two. Peak
stresses were found to occur at regions of inflection on the surface of the AAA rather
than at regions of maximum diameter. This result was also reported by other authors
using both FEA [2] and fluid-structure interaction (FSI) [34]. 3D solid stress elements
were used in the numerical aspect of this previous work as shell elements resulted in
erroneous stress distributions.
3D Reconstruction
The reconstruction process used here compared favourably to previous methods used by
our group [30,31]. For the AAA(SIMP) and AAA(MOD) models, reconstruction
42
involved the omission of the ILT, and so the AAA was modelled as a uniform walled
vessel, consisting of one structure. For the AAA(SIMP) models, the single layer
surfaces of these 3D reconstructions were then exported for stress analyses. The
AAA(MOD) required the generation of a uniform artificial wall 1.5 mm thick in
ProEngineer Wildfire 3.0, before being exported to the FEA solver. For the
AAA(COMP), the ILT was incorporated into the vessel wall and so the complex AAA
had an artificial uniform wall and ILT region of varying thickness. The full model was
then exported for stress analyses. The iliac arteries were omitted from these models, as
in previous studies [5]. The effect of smoothing on the 3D reconstruction was also
examined. Four degrees of smoothing were studied in order to determine the effect
smoothing plays on wall stress distributions. Normalised wall stress for the four models
is shown in Figure 3. For model 1, the AAA was over-smoothed, and resulted in low
wall stress. Smoothing was then reduced for model 2, and resulted in higher stress and
less gradual changes in stress along the AAA surface. Smoothing was then reduced
further for models 3 and 4, which both resulted in elevated stresses and also more
surface detail. This surface detail results in rapid changes from regions of higher and
lower stresses, showing jagged wall stress distributions. Also, if the stress distributions
of models 3 and 4 were used in the study, they would indicate possible failure sites in
abnormal regions of the AAA, as elevated stresses are shown in the healthy proximal
neck. This observation is shown in Figure 3 by the black marks indicating these
irregular locations of high stress. From this study the optimum level of smoothing was
determined to be that of model 2, 20 control points per axial slice, and was used in all
further reconstructions of this study.

In each of the three geometrically different models, a linearly elastic model was
examined. It has been argued in previous AAA wall stress studies [23,24] that the aorta
behaves more like a linearly elastic material at pressures above 80 mmHg (10.67 kPa).
Therefore, many researchers utilise the much simpler technique of linear elasticity in
their studies [2,4,23-25]. Other researchers however, have reported that the use of
simplified linearly elastic material properties or other inappropriate tissue constitutive
models can lead to erroneous stress distributions [2,36-40]. For each of the non-linear
models, the material properties of 69 AAA specimens form the basis of the finite strain
constitutive model for AAA wall tissue proposed by Raghavan and Vorp [3] and has
been utilised by many previous researchers [5-7,13,20,26-29]. These material
43
properties are believed to resemble the behaviour of the actual AAA wall more closely
than those of a linearly elastic material. When utilising the AAA(COMP) method, the
ILT is an entirely different structure than that of the AAA wall, and therefore required
its own material properties. The ILT was modelled as linearly elastic in the
AAA(COMP)L using previously published data [4,28,35]. The non-linear material
properties derived by Wang et al. [36] have been implemented to model the ILT region
of the AAA(COMP)NL in a more realistic manner.
Computational Effort
The computational time for each stress analysis increases with the inclusion of non-
linear material properties and complexity of the model, and so a compromise is usually
desirable between computational time and accuracy. For this study the computational
time for each analysis was recorded and compared. These results can be seen in Table
1. The differences in CPU times can be clearly seen, and highlight how the inclusion of
non-linearity increases computational effort. Inclusion of the non-linear behaviour of
the AAA wall resulted in a 9-fold increase in computational time for the AAA(SIMP)
models, 5-fold increase for AAA(MOD) models, and a 10-fold increase for the
AAA(COMP) models. Accounting for the “nonslip” contact between the AAA wall and
the ILT region of the AAA(COMP) cases also increased the computational effort.
Wall Stress
The modelling techniques used in this study all return significantly different wall stress
results. Peak wall stress was shown to reduce when taking into account the realistic
non-linear behaviour of AAA tissue in all three modelling approaches. All wall stress
results can be seen in Table 1. When modelled using non-linear material properties,
wall stress reduced by 34% in the AAA(SIMP) case, 55% in the AAA(MOD) case, and
by 47% in the AAA(COMP) case. These reductions in wall stress can be observed in
Figures 5 - 7 for each of the three modelling techniques. From these figures it can be
seen how the change in material properties does not significantly alter the stress
distribution patterns on the surface of the AAA, instead it simply reduces the magnitude
of these acting stresses.
The location of peak stress only differed when modelled using the AAA(SIMP)
approach. This shift in peak stress location in the AAA(SIMP) models may indicate that
the use of shell elements results in inaccurate stress distributions. Previous research
44
undertaken by our group [19] also noted that shell elements can lead to erroneous stress
distributions, in particular when using a linearly elastic material. In this previous
numerical work [19], the ideal AAA examined showed peak stresses along the region of
maximum diameter, rather than at the proximal and distal inflection points, as observed
experimentally [18]. Callanan et al. [19] then used 3D stress elements to numerically
analyse the ideal AAA model, and found that the stress distributions matched those
found experimentally, with good correlation between the two sets of results. This
background may indicate that the use of shell elements could possibly lead to inaccurate
results in realistic AAA models also. For both the AAA(MOD) and AAA(COMP)
models, 3D stress elements were used, and the location of peak stress did not
significantly shift when the model accounted for non-linearity of the material. As the
location of peak stress did not significantly alter in these models, it is believed that the
use of 3D elements is more accurate than the use of shell elements in FE analyses of
realistic AAAs. As mentioned earlier, this observation was reported in previous
research by our group [19].
Figure 8 compares the wall stress results of the three linearly elastic models examined.
Both the AAA(MOD) and AAA(COMP) models showed peak stress regions at inflection
points on the inner surface at the proximal region of the AAA sac. These findings are
consistent with previous experimental [18] and numerical work [19]. The AAA(MOD)L
model returned the highest peak stress in all three cases. The peak stress in
AAA(MOD)L is over 2.5 times the peak stress in both AAA(SIMP)L and AAA(COMP)L.
Both the AAA(SIMP)L and AAA(COMP)L models returned similar peak stresses of
0.8833 MPa and 0.8036 MPa, respectively. In comparison, the wall stress results for
the three non-linear cases are presented in Figure 9. Again, the AAA(MOD)NL model
experienced the highest wall stress, with a peak stress 44% higher than the
AAA(SIMP)NL and 59% higher than that experienced in the AAA(COMP)NL model. By
including the ILT into the models, peak wall stress was reduced by 65% in the linear
case and 59% in the non-linear case. This results in an average reduction of 62% in
peak wall stress simply by including the ILT. The ILT absorbs the internal pressure
induced by blood flow, and transfers the load to the AAA wall in a reduced amount.
From the stress results found for all the various models analysed, the AAA(COMP)NL
model is deemed to be the most accurate method of predicting wall stress in AAAs.
This model incorporates the realistic non-linear behaviour of both the AAA wall and the
ILT, and therefore produces the most accurate wall stress results.
45
Significance of Results
In order to determine the significance of the results reported here, other factors must be
considered. It is known that AAA rupture occurs when the locally acting wall stress
exceeds the locally acting wall strength. To make valid conclusions from the wall stress
results presented here, wall stress was examined with respect to AAA wall strength.
The AAA case examined in this study underwent routine CT scanning, thus detecting
the aneurysm. Previous work has identified that AAA wall strength has an average
failure stress of 0.942 MPa [16]. If the AAA was analysed for wall stress without
inclusion of the ILT, such as cases AAA(MOD)L and AAA(MOD)NL, results would
suggest that rupture has already occurred as local stress had exceeded local strength. By
implementing the ILT into the model, peak stress was reduced to a level believed to
more accurately represent that in vivo. The reported failure strength of 0.942 MPa [16]
is higher than the predicted wall stress of 0.4291 MPa, indicating that the rupture
potential of the AAA may be equalised. If wall stress results are to be incorporated into
the clinical decision-making process, inclusion of non-linear material properties and the
ILT may be important to the accuracy of the results.
Limitations
The study presented here is not without limitations. Firstly, it is known that
calcifications occur in almost all AAAs. These calcium deposits are believed to act as
stress raisers within the wall and so incorporation into AAA stress studies may have
significance [29]. Also, patient-specific blood pressures should be recorded at the time
of CT scan. This would allow the accurate stress analysis of patient-specific AAAs, and
as most patients exhibit elevated blood pressures, this could have an impact on peak
wall stress. It is known that the AAA wall thickness is non-uniform [42], and therefore,
incorporating a non-uniform wall into the modelling process may also alter stress
distributions and peak stresses. The use of pulsatile pressures, possibly leading to
fatigue testing, could also be incorporated into the modelling process. The relatively
new computational field of fluid-structure interaction (FSI) could also lead to more
accurate wall stress results [26-28,34]. Material properties could be improved. It is
known that AAA tissue is non-homogenous and anisotropic, factors of which have been
ignored for this study. The inclusion of more accurate material properties could lead to
more precise wall stress distributions. This comparison also only accounts for one
AAA model. Each individual AAA is unique, and a larger number of cases may give a
better understanding of the role of certain modelling parameters and assumptions.
46
Another limitation to this work, and also other AAA stress analyses, is that stress results
are not intuitive to the clinician [5]. Use of the maximum diameter criterion is easy and
clear to the surgeon, and therefore, wall stress results should be converted into a more
clinician-friendly index. There is ongoing research on the rupture prediction of AAAs
with the view to develop an easy to use tool that the clinician can implement into the
decision-making process.
CONCLUSIONS
This paper has examined the effect of modelling on the resulting wall stress
distributions of a realistic AAA. Three different modelling techniques were
implemented and each studied using both linear and non-linear material properties. It
was found from the resulting stress distributions that inclusion of the ILT and
implementing non-linear material properties may be important if accurate stress
distributions are to be obtained. Peak stresses were shown to significantly vary
depending on the modelling technique, with the most accurate model, namely
AAA(COMP)NL, returning the lowest peak stress. Shell elements were shown to yield
poor stress results, as there was a significant shift in peak stress location when non-
linear material properties were implemented into the model. Also, in these AAA(SIMP)
models the location of peak stress was shown to occur at regions of maximum diameter,
which previous research [18,19] has shown not to be the case. Simplifying assumptions
used in predicting wall stress with FEA should be limited.
The results reported here suggest that even patients with very modest levels of ILT,
should be analysed using the more complex, time-consuming methods, such as
including the ILT into FEA models, and the use of realistic, non-linear material
properties. The effect of omitting these important parameters may lead to erroneous
peak stress results, and ultimately to incorrect surgical decision-making. The principal
idea behind reconstruction and stress analysis of AAAs is to determine the wall stress
distributions of the AAA, and the potential of rupture. As rupture will occur when the
local wall stress exceeds the local wall strength, and AAA have been shown to have
variations in local wall strength [42], the need for improved prediction of peak wall
stress location may also have clinical importance.
To conclude, the modelling technique employed in patient-specific FEA of AAAs, may

be a valuable additional tool for clinical use.
47
ACKNOWLEDGEMENTS
We would like to thank (i) the Irish Research Council for Science, Engineering and
Technology (IRCSET) Grant no. RS/2005/340 (ii) NIH Grant R01-HL-060670-05A1
(iii) Prof. David A. Vorp of the Centre for Vascular Remodelling and Regeneration,
McGowan Institute for Regenerative Medicine, University of Pittsburgh (iv) Mr. Eamon
Kavanagh, a Consultant Vascular Surgeon at Midwestern Regional Hospital, Limerick,
for his help in collecting patient data and background information.
REFERENCES
Lancet, 365(9470), 1577-89.
Surgery, 33, 475-482.
4. Di Martino, E.S., Guadagni, G., Fumero, A., Ballerini, G., Spirito, R., Biglioli,
P., Redaelli, A. “Fluid-structure interaction within realistic 3D models of the
aneurysmatic aorta as a guidance to assess the risk of rupture of the aneurysm.”
Medical Engineering and Physics, 23, 647-655.
5. Fillinger, M.F., Marra, S.P., Raghavan, M.L., Kennedy, F.E. (2003) “Prediction
of rupture risk in abdominal aortic aneurysm during observation: wall stress
versus diameter.” Journal of Vascular Surgery, 37, 724-732.
(2002) “In vivo analysis of mechanical wall stress and abdominal aortic
aneurysm rupture risk.” Journal of Vascular Surgery, 36, 589-597.
8. Conway, K.P., Byrne, J., Townsend, M., Lane, I.F. (2001) “Prognosis of patients
turned down for conventional abdominal aortic aneurysm repair in the
endovascular and sonographic era: Szilagyi revisted?” Journal of Vascular
Surgery, 33, 752-757.
9. Darling, R.C., Messina, C.R., Brewster, D.C., Ottinger, L.W. (1977) “Autopsy
study of unoperated abdominal aortic aneurysms. The case for early resection.”
Circulation, 56(2), 161-164.
10. Cronenwett JL, Murphy TF, Zelenock GB, Whitehouse WM Jr, Lindenauer SM,
Graham LM: Actuarial analysis of variables associated with rupture of small
abdominal aortic aneurysms. Surgery 1985; 98: 472-483.
11. Nicholls, S.C., Gardner, J.B., Meissner, M.H., Johansen, H.K. (1998) “Rupture
in small abdominal aortic aneurysms.” Journal of Vascular Surgery, 28, 884-
888.
12. Vande Geest, J.P., Wang, D.H.J., Wisniewski, S.R., Makaroun, M.S., Vorp, D.A
(2006) “Towards a non-invasive method for determination of patient-specific
wall strength distribution in abdominal aortic aneurysms.” Annals of Biomedical
Engineering, 34(7), 1098-1106.
48
13. Wang, D.H.J., Makaroun, M.S., Webster, M.W., Vorp, D.A. (2002) “Effect of
intraluminal thrombus on wall stress in patient-specific models of abdominal
aortic aneurysm.” Journal of Vascular Surgery, 36, 598-604.
14. Sacks, M.S., Vorp, D.A., Raghavan, M.L., Federle, M.P., Webster, M.W. (1999)
“In vivo three-dimensional surface geometry of abdominal aortic aneurysms.”
Annals of Biomedical Engineering, 27(4), 469-479.
15. Vorp, D.A., Lee, P.C., Wang, D.H.J., Makaroun, M.S., Nemoto, E.M., Ogawa,
S., Webster, M.W. (2001) “Association of intraluminal thrombus in abdominal
aortic aneurysm with local hypoxia and wall weakening.” Journal of Vascular
Surgery, 34, 291-9.
16. Raghavan, M.L., Webster, M.W., Vorp, D.A. (1996) “Ex vivo biomechanical
behaviour of abdominal aortic aneurysm: assessment using a new mathematical
model.” Annals of Biomedical Engineering, 24, 573-582.
17. Vorp, D.A., Mandarino, W.A., Webster, M.W., Gorcsan III, J. (1996) “Potential
influence of intraluminal thrombus on abdominal aortic aneurysm as assessed by
a new non-invasive method.” Cardiovascular Surgery, 4, 732-739.
18. Morris, L., O’Donnell, P., Delassus, P., McGloughlin, T. (2004) “Experimental
assessment of stress patterns in abdominal aortic aneurysms using the
photoelastic method.” Strain, 40, 165-172.
19. Callanan, A., Morris, L.G., McGloughlin, T.M. (2004) “Numerical and
experimental analysis of an idealised abdominal aortic aneurysm.” European
Society of Biomechanics, S-Hertogenbosch, Netherlands.
20. Venkatasubramaniam, A.K., Fagan, M.J., Mehta, T., Mylankal, K.J., Ray, B.,
Kuhan, G., Chetter, I.C., McCollum, P.T. (2004) “A comparative study of aortic
wall stress using finite element analysis for ruptured and non-ruptured
abdominal aortic aneurysms.” European Journal of Vascular and Endovascular
Surgery, 28, 168-176.
21. Dobrin, P.B. (1989) “Pathophysiology and pathogensis of aortic aneurysms.
Current concepts.” Surgical Clinics of North America, 69, 687-703.
22. Schurink, G.W., van Baalen, J.M., Visser, M.J., van Bockel, J.H. (2000)
“Thrombus within an aortic aneurysm does not reduce pressure on the
aneurysmal wall.” Journal of Vascular Surgery, 31, 501-506.
23. Giannoglu, G., Giannakoulas, G., Soulis, J., Chatzizisis, Y., Perdikides, T.,
Melas, N., Parcharidis, G., Louridas, G. (2006) “Predicting the risk of rupture of
abdominal aortic aneurysms by utilizing various geometrical parameters:
Revisiting the diameter criterion.” Angiology, 57(4), 487-494.
355-366.
25. Hua, J., Mower, W.R. (2001) “Simple geometric characteristics fail to reliably
predict abdominal aortic aneurysm wall stress.” Journal of Vascular Surgery, 34,
308-315.
26. Leung, J.H., Wright, A.R., Cheshire, N., Crane, J., Thom, S.A., Hughes, A.D.,
Xu, Y. (2006) “Fluid structure interaction of patient specific abdominal aortic
aneurysms: a comparison with solid stress models.” Biomedical Engineering
Online, 5, 33.
27. Li, Z., Kleinstreuer, C. (2005) “Blood flow and structure interactions in a stented
abdominal aortic aneurysm model.” Medical Engineering and Physics, 27, 369-
382.
28. Papaharilaou, Y., Ekaterinaris, J.A., Manoussaki, E., Katsamouris, A.N. (2007)
“A decoupled fluid structure approach for estimating wall stress in abdominal
aortic aneurysms.” Journal of Biomechanics, 40, 367-377.
49
F.N., Makaroun, M.S., Vorp, D.A. (2007) “Effects of wall calcifications in
patient-specific wall stress analyses of abdominal aortic aneurysms.” Journal of
30. Morris, L, Delassus, P., Callanan, A., Walsh, M., Wallis, F., Grace, P.,
McGloughlin, T. (2005) “3D numerical simulation of blood flow through
models of the human aorta.” Journal of Biomechanical Engineering, 127, 767-
775.
31. Doyle, B.J., Morris, L.G., Callanan, A., Kelly, P., Vorp, D.A., McGloughlin,
T.M. (2008) “3D reconstruction and manufacture of real abdominal aortic
aneurysms: From CT scan to silicone model.” Journal of Biomechanical
Engineering, 130, 034501-5.
32. Langewouters, G.J., Wesseling, K.H., Goedhard, W.J.A. (1984) “The static
elastic properties of 45 human thoracic and 20 abdominal aortas in vitro and the
parameters of a new model.” Journal of Biomechanics, 17, 425-435.
33. McDonald, D.A. (1974) “The elastic properties of the arterial wall”, Blood flow
in arteries.” Baltimore, Williams and Wilkins, p238-282.
34. Scotti, C.M., Shkolnik, A.D., Muluk, S.C., Finol, E. (2005) “Fluid-structure
interaction in abdominal aortic aneurysms: effect of asymmetry and wall
thickness.” Biomedical Engineering Online, 4, 64.
35. DiMartino, E., Mantero, S., Inzoli, F., Melissano, G., Astore, D., Chiesa, R.,
Fumero, R. (1998) “Biomechanics of abdominal aortic aneurysm in the presence
of endoluminal thrombus: experimental characterisation and structural static
computational analysis.” European Journal of Vascular and Endovascular
Surgery, 15, 290-299.
36. Wang, D.H.J., Makaroun, M., Webster, M.W., Vorp, D.A. (2001) “Mechanical
properties and microstructure of intraluminal thrombus from abdominal aortic
aneurysm.” Journal of Biomechanical Engineering, 123, 536-539.
38. Truijers, M., Pol, J.A., SchultzeKool, L.J., van Sterkenburg, S.M., Fillinger,
M.F., Blankensteijn, J.D. (2007) “Wall stress analysis in small asymptomatic,
39. Mower, W.R., Quinones, W.J., Gambhir, S.S. (1997) “Effect of intraluminal
thrombus on abdominal aortic aneurysm wall stress.” Journal of Vascular
Surgery, 26, 602-608.
40. Vorp, D.A., Vande Geest, J. (2005) Biomechanical determinants of abdominal
aortic aneurysm rupture.” Artheriosclerosis, Thrombosis and Vascular Biology,
25, 1558-1566.
41. EUROSTAR Data Registry Centre. (2001) Laheij, R., van Marrewijk, C., Buth,
J. “Progress report on the procedural and follow up results of 3413 patients who
received stent graft treatment for infrarenal aortic aneurysms for a period of 6
years.”
42. Raghavan, M.L., Kratzberg, J., de Tolosa, E.M.C., Hanaoka, M.M., Walter, P.,
da Silva, E.S. (2006) “Regional distribution of wall thickness and failure
properties of human abdominal aortic aneurysm.” Journal of Biomechanics,
39(16), 3010-6.
50
Chapter III
A finite element analysis rupture index (FEARI) as
an additional tool for abdominal aortic aneurysm
rupture prediction.
Barry J. Doyle,1 Anthony Callanan,1 Michael T. Walsh,1 Pierce A.

Grace2 and Timothy M. McGloughlin1
Limerick, Ireland.
Vascular Disease Prevention 2009; 6: 114-121

Data collection: PG
Critical revision of article: AC, MW, PG, TM
Final approval: TM
Chapter Three FEARI
ABSTRACT
Background
Currently, abdominal aortic aneurysms (AAAs), which are a permanent dilation of the
aorta, are treated surgically when the maximum transverse diameter surpasses 5 cm.
AAA rupture occurs when the locally acting wall stress exceeds the locally acting wall
strength. There is a need to review the current diameter-based criterion, and so it may
be clinically useful to develop an additional tool to aid the surgical decision-making
process. A Finite Element Analysis Rupture Index (FEARI) was developed.
Methods
Ten patient-specific AAAs were reconstructed, and the corresponding wall stress
computed. Previous experimental work on determination of ultimate tensile strengths
(UTS) from AAA tissue samples was implemented in this study. By combining peak
wall stress along with average regional UTS, a new approach to the estimation of
patient-specific rupture risk has been developed.
Results
Ten cases were studied, all of which had previously undergone surgical AAA repair. A
detailed examination of these ten cases utilising the FEARI analysis suggested that there
was a possibility that some of the AAAs may have been less prone to rupture than
previously considered.
Conclusions
It is proposed that FEARI, used alongside other rupture risk factors, may improve the
current surgical decision-making process. The use of FEARI as an additional tool for
rupture prediction may provide a useful adjunct to the diameter-based approach in
surgical decision-making.
51
Chapter Three FEARI
INTRODUCTION
Cardiovascular disease is a leading cause of death in the Western world. Aneurysms,
which are dilations or widening of blood vessels, form a significant portion of these
deaths. The healthy infrarenal abdominal aorta diameter ranges from 15 mm to 24 mm
for most elderly men [1], with abdominal aortic aneurysms (AAAs) defined as an aorta
with an infrarenal diameter 1.5 times that of the normal aortic diameter [2]. The actual
mechanisms resulting in AAA formation are still not clearly understood. It is believed
that these aneurysms form due to alterations of the connective tissue in the aortic wall.
This degradation of the aortic wall can be attributed to risk factors such as tobacco
smoking, sex, age, hypertension, chronic obstructive pulmonary disease,
hyperlipidaemia, and family history of the disorder [1]. Increased screening of subjects
and improved imaging technologies, have given rise to increased detection of AAAs.
Approximately 150,000 new cases are diagnosed each year in the USA [3-5], resulting
in 15,000 deaths per year due to AAA rupture [6].
Currently, the rupture risk of AAAs is regarded as a continuous function of aneurysm

size. Traditionally, surgeons operate once the AAA exceeds a transverse diameter of 50
mm [7-10], but previous work has shown that AAAs smaller than 50 mm can also
rupture [11,12]. It is believed by many researchers that there is a need to review the
determination of the timing of surgical intervention based solely on aneurysm diameter,
and consider the inclusion of other relevant risk factors.
It is known that AAA rupture occurs when the locally acting wall stress exceeds the
locally acting wall strength. Therefore, the AAA tissue strength must play an equal role
to AAA wall stress in determining failure. A region of AAA wall that is under elevated
wall stress may also have high wall strength, thus equalising its rupture potential. Many
researchers have proposed alternative methods of determining AAA rupture potential.
These additional risk factors could, for example, include: AAA wall stress [9,10]; AAA
expansion rate [6,13,14]; degree of asymmetry [15-17]; presence of intraluminal
thrombus (ILT) [18] which is a fibrin structure incorporated with blood cells, platelets,
blood proteins, and cellular debris, and are found in most AAAs; a rupture potential
index (RPI) [19,20]; biomechanical factors and computer analysis [6]; growth of
thrombus [21]; geometrical parameters [22]; and also a method of determining AAA
growth and rupture based on a mathematical model [23].
52
Chapter Three FEARI
The purpose of this study is to examine the use of a new additional tool to assist in the
assessment of AAA rupture risk. This new approach focuses on a combination of the
finite element method (FEM) coupled with published ultimate tensile strength (UTS)
data from AAA tissue reported by previous researchers [13,24,25]. The FEM is a
numerical technique widely used in engineering that divides complex 3D structures into
small areas called elements, for which the stress distribution can be more easily studied.
This new approach which we have described as the Finite Element Analysis Rupture
Index (FEARI), may be clinically useful in aiding surgeons as to the most appropriate
time to surgically intervene, and may serve as a useful adjunct to maximum diameter.
MATERIALS AND METHODS

Study Subjects and Reconstructions
Computed tomography (CT) scan data was obtained for 10 patients (male, n=6; female,
n=4). These patient scans were obtained from the Midwestern Regional Hospital,
Limerick, Ireland, and the University of Pittsburgh Medical Centre, Pittsburgh, PA,
USA. All ten patients were awaiting AAA repair at the time of CT scan, as AAA
diameters had reached or exceeded the current 5.0 cm threshold for repair. 3D
reconstructions were performed using the commercial software Mimics v10.0
(Materialise, Belgium). This reconstruction technique was validated with previous
work performed by our group [26,27]. The iliac arteries were omitted from the
reconstruction as they have been shown not to significantly alter the resulting wall stress
distributions [10]. Patient-specific wall thickness was obtained using the equation
proposed by Li and Kleinstreuer [28] shown in Eqn. 1, where t is wall thickness and
Diamax is the maximum diameter. The ILT was also included in the reconstructions as
this structure has been shown to reduce wall stress [18]. The models developed by this
method formed the basis for the FEM stress analysis. The pre-operative details of the
ten cases examined can be seen in Table 1.
−0.2892
⎛ Diamax ⎞
t = 3.9 ⎜ ⎟ Eqn. (1)
⎝ 2 ⎠
53
Chapter Three FEARI
Table 1: Pre-operative patient details for each study subject. Note that Ø is diameter.
Max Ø Wall Thickness AAA Length AAA Volume ILT Volume
Patient Sex
(cm) (mm) (cm) (cm3) (cm3)
1 Male 5.1 1.53 13.2 176.7 54.1
2 Male 5.75 1.48 11.2 192.4 122.8
3 Male 5.85 1.47 11.6 194.9 96.8
4 Female 5.0 1.54 9.3 136.9 38.4
5 Male 5.9 1.47 12.8 220.9 116.8
6 Male 7.4 1.37 14.3 311.6 147.6
7 Female 5.3 1.51 10.5 137.9 82.5
8 Female 6.2 1.48 9.7 61.6 29.1
9 Female 6.5 1.43 10.5 217.3 150.2
10 Male 9.0 1.29 11.7 445.5 124.5
Wall Stress Analysis

The ten 3D reconstructions were imported into the commercial finite element solver
ABAQUS v6.7 (Dassault Systemes, SIMULIA, R.I., USA) for stress analysis. In order
to simulate in vivo wall stress in the AAA wall, realistic boundary conditions were
applied to each model. The AAA wall was modelled as a homogenous isotropic
hyperelastic material using the finite strain constitutive model proposed by Raghavan
and Vorp [8]. These material properties have been utilised in many previous stress
analysis studies [9,10,17,18,29,30-35]. The aorta is also known to be nearly
incompressible with a Poisson’s ratio of 0.49. The ILT was modelled as a hyperelastic
material using the material characterisation derived from 50 ILT specimens from 14
patients performed by Wang et al. [36]. Each AAA was constrained in the proximal and
distal regions to simulate tethering to the aorta at the renal and iliac bifurcations. The
blood pressure within the AAA acts on the luminal contour of the ILT and therefore,
pressure was applied to the inner surface of the computational AAA model. A static
peak systolic pressure of 120 mmHg (16 kPa) was used, as employed in most AAA
stress analyses [15,22,33,37,38]. It is known that patient-specific blood pressures may
be higher than 120 mmHg, but for the purpose of this study a standard value was more
appropriate so as to eliminate some of the unknown variables in the analysis. The shear
stress induced by blood flow was neglected in this study, although the effects of blood
flow have been shown to reduce wall stress in idealised AAA models [16]. Residual
stresses that may exist within the aortic wall in vivo and tethering forces on the posterior
surface caused by the lumbar arteries were also neglected.
Once adequate boundary conditions were applied to each model, a mesh was generated
on each AAA. Mesh independence was performed for all AAA models in order to
54
Chapter Three FEARI
determine the optimum number of elements, and therefore, the optimum mesh. Mesh
independence was performed by increasing the number of elements in the mesh until the
difference in peak stress was less than 2% of the previous mesh [17,18,34,38]. This
method of determining wall stress distributions has been previously shown to be the
most effective in computing accurate results compared with other approaches [34].
FEARI
The Finite Element Analysis Rupture Index (FEARI) is defined by Eqn. 2. In this
equation, the peak wall stress is computed using the FEM, whereas, the wall strength
values are obtained from previous research on experimental testing of AAA wall
specimens [13,24,25].
FEA Wall Stress Eqn. (2)

FEARI =
Experimental Wall Strength
This equation is based on the simple engineering definition of material failure, that is,
failure will occur when the stress acting on the material exceeds the strength of the
material. This index then returns a value ranging from 0 to 1, where 0 indicates a very
low rupture potential, and a value close to 1 indicates a very high rupture potential.
In order to determine strength values for the AAA wall, the previous research of both
Raghavan et al. [24,25] and Thubrikar et al. [13] were analysed. These publications are
the most detailed reports of experimental uniaxial testing of AAA tissue. Raghavan et
al. [24] tensile tested 52 specimens of AAA tissue and found that the average UTS of
this aneurysmal tissue was 0.942 MPa. Thubrikar et al. [13] later segmented AAAs into
posterior, anterior and lateral regions and tensile tested 49 tissue specimens. Average
regional UTS values were shown to be 0.46 MPa, 0.45 MPa and 0.62 MPa,
respectively. Raghavan et al. [25] subsequently furthered their work, and also divided
each AAA into regional sections and tested 48 samples. They showed that the UTS of
AAA tissue can range from 0.336 – 2.35 MPa. They also reported that the regional
variations in UTS for the anterior, posterior, left and right regions, were 1.099 MPa,
1.272 MPa, 1.217 MPa and 1.224 MPa, respectively.
By combining all the previously published experimental data [13,24,25], average

regional UTS values for the four main regions of the AAA could be obtained. By
55
Chapter Three FEARI
subdividing the AAA into a further four sections, eight in total, more regionally
accurate strength estimates were obtained, thus allowing FEARI values to be calculated.
The method of dividing each AAA into regions is shown in Figure 1, with the resulting
UTS values for the varying regions shown in Table 2.
Table 2: Regional UTS values obtained by combining and averaging previous

experimental data [13,24,25].
AAA Region UTS (MPa)
Anterior 0.7744
Posterior 0.8658
Left 0.9221
Right 0.9187
Anterior/Left 0.8482
Anterior/Right 0.8465
Posterior/Left 0.8939
Posterior/Right 0.8922
Once the combined UTS values of AAA tissue were calculated, these values could be
coupled with the computed wall stress results from the FEM. By recording the location
of peak stress in each AAA model, the region can also be assigned a regional UTS
value. FEARI was then computed for each of the ten cases. Statistical analysis was
performed on all results using a Pearson’s correlation, with P<0.05 accepted as
significant.
Figure 1: Illustration showing a representative AAA stress distribution for Patient 9

(ellipse passes through region of peak stress, indicated by black circle) and how each
AAA model was segmented in order to determine the corresponding UTS value for the
particular region of peak stress. In this case, peak stress occurs on the anterior wall of
the AAA. AAA model shown in the anterior view.
56
Chapter Three FEARI
RESULTS
Wall Stress Results
The finite element analyses using ABAQUS v6.7 produced detailed stress distributions
on each of the models under the pressure loading [19]. The von Mises stress is a stress
index especially suited for failure analysis, as stress is a tensor quantity with nine
components, with the von Mises stress being a combination of these components. By
observing the stress distributions, it was noted that the regions of elevated and peak wall
stresses occurred at inflection points on the AAA surface, and not at regions of
maximum diameter. This observation was also observed by previous researchers in
idealised models, both experimentally [39] and numerically [15,40], and also in realistic
models [17,34]. Inflection points are defined as points on the AAA surface at which the
local AAA wall shape changes from concave outward to concave inward [15]. The
peak wall stresses found in this study ranged from 0.3167 – 1.282 MPa, with a mean ±
standard deviation of 0.6201 ± 0.2836 MPa. The results of the computational stress
analysis, along with the maximum diameter and UTS for each case can be seen in Table
3. There was no statistical significance between peak wall stress and any geometrical
parameters analysed here. There was however a significant relationship between both
FEARI and maximum diameter (P=0.043), and FEARI and AAA volume (P=0.036).
Table 3: Maximum diameter, FEA computed peak wall stress, location of peak wall
stress, and UTS of peak stress region in all ten cases examined. Wall thickness was
incorporated in peak wall stress calculations.
Max Diameter Peak Wall Stress
Patient Location UTS (MPa)
(cm) (MPa)
1 5.1 0.4291 Anterior 0.7744
2 5.75 0.3167 Left 0.9221
3 5.85 0.4346 Anterior/Right 0.8465
4 5.0 0.6641 Anterior/Right 0.8465
5 5.9 0.5866 Anterior/Right 0.8465
6 7.4 0.707 Right 0.9187
7 5.3 0.4 Anterior/Right 0.8465
8 6.2 1.282 Anterior 0.7744
9 6.5 0.5263 Anterior 0.7744
10 9.0 0.855 Posterior 0.8658
FEARI Results
Once all peak wall stresses, locations of peak wall stress, and corresponding regional
UTS values were obtained and grouped, FEARI results using Eqn. 2 could be
calculated. Table 4 and Figure 2 show the resulting FEARI values with the
corresponding maximum diameters of each patient. In this figure, an indication of the
57
Chapter Three FEARI
possible rupture risk using the FEARI analysis of each patient is shown. An example
calculation of this index for Patient 9 can be seen below.
FEA Wall Stress Eqn. (2)

FEARI =
Where, peak FEA wall stress is 0.5263MPa, and this peak stress occurred in the anterior
region of the AAA. Therefore, the UTS of the peak stress region is 0.7744 MPa using
Table 2. Eqn. 2 now becomes Eqn. 3.
0.5263
FEARI = = 0.6796 Eqn. (3)
0.7744
The resulting FEARI for Patient 9 is 0.6796, representing a possible 68% chance of
rupture.
Table 4: Resulting FEARI values compared to the corresponding maximum diameters

for each AAA studied.
Patient Maximum Diameter (cm) FEARI
1 5.1 0.5541
2 5.75 0.3435
3 5.85 0.5133
4 5.0 0.7845
5 5.9 0.6929
6 7.4 0.7696
7 5.3 0.4725
8 6.2 1.6554
9 6.5 0.6796
10 9.0 0.9876
58
Chapter Three FEARI
Figure 2: Graph displaying FEARI results for all ten cases. Horizontal line indicates
possible AAA rupture based on the FEARI model. Diameters of each case are presented
also, indicating how diameter is not related to FEARI.
DISCUSSION
Considering the high level of mortality associated with AAA rupture, improved surgical
decision tools may lead to better clinical outcomes. At present, the timing of
intervention is determined on the basis of maximum diameter alone. However, it was
reported that AAAs with diameters <5.0 cm may be safe from rupture, and should be
rigorously monitored [1]. The results presented here suggest that the inclusion of the
FEARI monitoring approach to AAA management may be useful in the assessment of
all size AAAs, rather than repairing the aneurysm based on maximum diameter alone.
It is known that AAA rupture occurs when the locally acting wall stress exceeds the
locally acting wall strength, and thus the inclusion of wall strength as a rupture
parameter may also be useful. The FEARI approach presented here utilises previously
published experimental data from tissue specimens [13,24,25] which were applied to
establish FEARI results. The wall stress was computed using the commercial FEA
solver, ABAQUS v6.7, from which, peak wall stress values and the location of peak
stress could be recorded for each case. Convergence studies were also performed on
each model to establish confidence in the finite element mesh size and accuracy of the
results [17,18,34,38]. Patient-specific wall thickness values were applied to each case
using the equation proposed by Li and Kleinstreuer [28] to provide more patient-
specific results. Peak stresses varied throughout the ten cases, and were independent of
59
Chapter Three FEARI
the maximum diameter of the AAA. There was no statistical significance between peak
wall stress and any of the measured parameters (maximum diameter, P=0.227; sex,
P=0.404; AAA volume, P=0.936; AAA length, P=0.501; ILT volume, P=0.247; wall
thickness, P=0.375; location of peak stress, P=0.54). FEARI results were also
statistically compared with the same parameters. Of these comparisons, FEARI was
statistically significant with maximum diameter (P=0.043) and AAA volume
(P=0.036), whereas, there was no significance with other parameters (sex, P=0.961;
AAA length, P=0.804; ILT volume, P=0.881; wall thickness, P=0.056; location of
peak stress, P=0.755). Significance was obtained when P<0.05. These results alone
suggest that other rupture indicating parameters should be included in the surgical
decision-making process. From the ten cases examined, all AAAs experienced peak
wall stress at areas where the ILT was regionally thinner, therefore supporting the
hypothesis that ILT reduces wall stress by acting like a “mechanical cushion”
[18,19,38] for the AAA wall. Another interesting observation relates to Patient 8. In
this case, the AAA geometry had a rapid change in diameter from 33.4 mm to 61.4 mm
over 12.5 mm. This sudden change in geometry resulted in a large peak stress of 1.282
MPa, which is approximately 60% higher than the similarly sized-diameter AAA of
Patient 9.
A FEARI value was determined for each case from Eqn. 2. A FEARI value close to or
above 1, suggests that the AAA may be a high rupture-risk AAA, and should be
repaired immediately. FEARI values closer to 0, on the other hand, would indicate that
the risk of rupture is low, and should be monitored closely through regular ultrasound or
CT scanning. All ten cases examined had reached or exceeded the current 5.0 cm
threshold for surgical repair. At the time of this study, all cases had undergone either
traditional open repair or endovascular aneurysm repair (EVAR). The peak stresses
show that smaller AAAs can have higher peak wall stresses than larger AAAs (see
Table 3). Wall stress is closely related to geometrical parameters, with highly stressed
areas occurring at regions of inflection on the surface on the AAA. Tortuous and
irregularly shaped, or asymmetric, AAAs can experience many regions of inflection,
and thus, many regions acting as stress raisers on the surface.
Although FEARI results presented here are preliminary, in that they currently use
experimental data from other researchers [13,24,25], the approach may be clinically
useful. In order to gauge the effect of the UTS on the resulting FEARI, the “worst case
60
Chapter Three FEARI
scenario” in terms of UTS was examined. Raghavan et al. [25] reported that the UTS of
AAA tissue can range from 0.336 – 2.35 MPa, and therefore, this minimum strength of
0.336 MPa was applied to our FEARI equation. The results can be seen in Figure 3.
This resulted in much higher rupture potentials for all ten cases, and suggested that all
patients were suitable for immediate repair. Although this minimum UTS value was
recorded, only 8 of the 48 specimens tested by Raghavan et al. [25] resulted in UTS
values below 0.5 MPa, with 0.336 MPa being the absolute minimum displayed by any
specimen. Therefore, to assume that all AAAs, in all regions experience this low failure
stress would appear to be overly conservative.
Figure 3: FEARI results for the “worst case scenario” of extremely low failure strengths
of the AAA tissue. Horizontal line indicates possible AAA rupture using the FEARI
model.
It is proposed that FEARI could serve as a useful adjunct to diameter-based surgical

decision making. Diameter, and ultimately size, of the AAA is an obvious concern for
the clinician, and must remain a consideration. The overall geometry of the aneurysm
should also play a role. Asymmetry has been shown to affect wall stress in idealised
AAA models, both numerically [15,16,40] and experimentally [39], and may also affect
realistic cases [17]. Other researchers have proposed the Rupture Potential Index (RPI)
[20] which uses a statistical modelling technique with the inclusion of factors such as
age, gender, family history, smoking status, among others, to deduce patient-specific
wall strength [41]. This RPI approach uses a more theoretical method of calculating
wall strength, compared to the experimental approach of FEARI, and combining the
61
Chapter Three FEARI
two approaches may lead to improved predictions. Ultimately, the decision to

surgically intervene may include a combination of factors including diameter,
asymmetry, RPI and FEARI, along with clinical experience, and may determine the
most suitable approach to a particular AAA.
The use of fluid-structure interaction (FSI) software, as opposed to the current static
solid-mechanics approach may reduce peak stresses [16]. The effect of varying blood
pressure may also alter results. Blood pressure can change over time, often as a result
of stress or exercise. Increasing or decreasing the applied pressure to the stress analyses
would increase or decrease the FEARI results respectively. Monitoring a patients’
blood pressure over the course of a day may allow a more accurate loading condition to
be developed, and thus provide more accurate patient-specific FEARI results. It is
known that the realistic AAA has a non-uniform wall thickness [25] and also non-
uniform material properties due to regions of calcifications [15] which can lead to
alterations in stress distributions [33]. The refinement of the technique used to compute
wall stress may therefore produce more realistic FEARI results even when considering
the “worst case scenario” regarding tissue strength. Along with these limitations in
computing the peak wall stress, further experimental testing of AAA tissue is necessary.
Both Raghavan et al. [24,25] and Thubrikar et al. [13] recorded varying UTS values for
the differing regions of the AAA, and therefore, a more widespread study on the UTS of
AAA tissue is required. Collation of data from other centres worldwide, would lead to a
much larger database of tissue strength, which could then be averaged over the
population. Biaxial tensile testing of tissue is also necessary. It has been reported that
AAA tissue is anisotropic, and that 3D multi-axial mechanical evaluation would allow
for more appropriate modelling of aneurysmal tissue [42]. There have been recent
reports on the use of anisotropic material data in wall stress analyses which may also
yield better results [43]. By actually measuring AAA tissue harvested from unruptured
AAAs during surgery, tensile strength data can be gathered. By expanding the database
to many institutions, larger populations can be covered, thus increasing the effectiveness
of the AAA strength database. This may lead to a more accurate FEARI measure in
order to improve the assessment of patients with AAAs.
CONCLUSIONS
FEARI may be clinically useful due to the simplicity of the approach. Rupture occurs
when the AAA tissue cannot withstand the locally acting wall stress exerted, and
62
Chapter Three FEARI
therefore, tissue strength must be considered when assessing AAA rupture potential.
Peak wall stresses were computed, along with the location, and therefore UTS, of peak
wall stress region. FEARI results indicate that surgery may not be necessary for all
cases, but rather continued monitoring may suit particular patients. It is proposed to
couple FEARI together with diameter and other important factors in AAA assessment to
allow the clinician a greater understanding of the severity of an individual AAA before
deciding on surgical intervention. This preliminary study of a FEARI suggests that
further work into this approach may yield more accurate results, and may provide a
useful adjunct to the diameter-based approach in surgical decision-making.
ACKNOWLEDGEMENTS
the Midwestern Regional Hospital, Ireland, in particular, Mr. Eamon Kavanagh, Mr.
Paul Burke and Prof. Pierce Grace (iv) Aidan Cloonan for his assistance with the
numerical analyses (iv) Prof. David A. Vorp from the Centre for Vascular Remodelling
and Regeneration, University of Pittsburgh and (v) Michel S. Makaroun, MD,
Department of Surgery, University of Pittsburgh.
REFERENCES
Lancet, 365(9470), 1577-89.
2. Johnston, K.W., Rutherford, R.B., Tilson, M.D., Shah, D.M., Hollier, L.,
Stanley, J.C. (1991) “Suggested standards for reporting on arterial aneurysms.
Subcommittee on reporting standards for arterial aneurysms, Ad hoc committee
on reporting standards, Society for Vascular Surgery and North American
Chapter, International Society for Cardiovascular Surgery.” Journal of Vascular
Surgery, 13, 452-458.
4. Ouriel, K., Green, R.M., Donayre, C., Shortell, C.K., Elliott, J., DeWeese, J.A.
(1992) “An evaluation of new methods of expressing aortic aneurysm size:
relationship to rupture.” Journal of Vascular Surgery, 15, 12-20.
5. Bengtsson, H., Sonesson, B., Bergqvist, D. (1996) “Incidence and prevalence of
abdominal aortic aneurysms, estimated by necroscopy studies and population
screening by ultrasound.” Proceedings Volume 800 of the Annals of the New
York Academy of Science, United States.
prediction of abdominal aortic aneurysms.” Biomedical Engineering Online, 5,
19.
63
Chapter Three FEARI
7. Sayers, R.D. (2002) “Aortic aneurysms, inflammatory pathways and nitric

oxide.” Annals of the Royal College of Surgeons of England, 84(4), 239-246.
Surgery, 33, 475-482.
11. Nicholls, S.C., Gardner, J.B., Meissner, M.H., Johansen, H.K. (1998) “Rupture
in small abdominal aortic aneurysms.” Journal of Vascular Surgery, 28, 884-
888.
Circulation, 56(2), 161-164.
13. Thubrikar, M.J., Labrosse, M., Robicsek, F., Al-Soudi, J., Fowler, B. (2001)
“Mechanical properties of abdominal aortic aneurysm wall.” Journal of Medical
Engineering and Technology, 25(4), 133-142.
14. Hirose, Y., Takamiya, M. (1998) “Growth curve of ruptured aortic aneurysm.”
Journal of Cardiovascular Surgery, 39, 9-13.
17. Doyle, B.J., Callanan, A., Burke, P.E., Grace, P.A., Walsh, M.T., Vorp, D.A.,
McGloughlin, T.M. (2009) “Vessel asymmetry as an additional tool in the
assessment of abdominal aortic aneurysms.” Journal of Vascular Surgery, 49,
443-454.
25, 1558-1566.
20. Vande Geest, J.P., Di Martino, E.S., Bohra, A., Makaroun, M.S., Vorp, D.A.
(2006) “A biomechanics-based rupture potential index for abdominal aortic
aneurysm risk assessment.” Annals of the New York Acadamy of Science, 1085,
11-21.
21. Stenbaek, J., Kalin, B., Swedenborg, J. (2000) “Growth of thrombus may be a
better predictor of rupture than diameter in patients with abdominal aortic
aneurysms.” European Journal of Vascular and Endovascular Surgery, 20, 466-
469.
64
Chapter Three FEARI
23. Volokh, K.Y., Vorp, D.A. (2008) “A model of growth and rupture of abdominal
aortic aneurysm.” Journal of Biomechanics, 41(5), 1015-21.
24. Raghavan, M.L., Webster, M.W., Vorp, D.A. (1996) “Ex vivo biomechanical
behaviour of abdominal aortic aneurysm: assessment using a new mathematical
model.” Annals of Biomedical Engineering, 24, 573-582.
25. Raghavan, M.L., Kratzberg, J., de Tolosa, E.M.C., Hanaoka, M.M., Walter, P.,
39(16), 3010-6.
775.
28. Li, Z., Kleinstreur, C. (2005) “A new wall stress equation for aneurysm-
rupture.” Annals of Biomedical Engineering, 33(2), 209-213.
wall stress using finite element analysis for ruptured and non-ruptured
abdominal aortic aneurysms.” European Journal of Vascular and Endovascular
Surgery, 28, 168-176.
Online, 5, 33.
34. Doyle, B.J., Callanan, A., McGloughlin, T.M. (2007) “A comparison of
modelling techniques for computing wall stress in abdominal aortic aneurysms.”
Biomedical Engineering Online, 6, 38.
35. Raghavan, M.L., Fillinger, M.F., Marra, S.P., Naegelein, B.P., Kennedy, F.E.
(2005) “Automated methodology for determination of stress distribution in
human abdominal aortic aneurysm.” Journal of Biomechanical Engineering,
127, 868-871.
36. Wang, D.H.J., Makaroun, M., Webster, M.W., Vorp, D.A. (2001) “Mechanical
properties and microstructure of intraluminal thrombus from abdominal aortic
aneurysm.” Journal of Biomechanical Engineering, 123, 536-539.
37. Inzoli, F., Boschetti, F., Zappa, M., Longo, T., Fumero, R. (1993)
“Biomechanical factors in abdominal aortic aneurysm rupture.” European
Journal of Vascular Surgery, 7, 667-674.
65
Chapter Three FEARI
38. Truijers, M., Pol, J.A., SchultzeKool, L.J., van Sterkenburg, S.M., Fillinger,
M.F., Blankensteijn, J.D. (2007) “Wall stress analysis in small asymptomatic,
(2006) “Towards a non-invasive method for determination of patient-specific
wall strength distribution in abdominal aortic aneurysms.” Annals of Biomedical
Engineering, 34(7), 1098-1106.
42. Vande Geest, J.P., Sacks, M.S., Vorp, D.A. (2006) “The effects of aneurysm on
the biaxial mechanical behaviour of human abdominal aorta.” Journal of
43. Rodriguez, J.F., Ruiz, C., Doblare, M., Holzapfel, G.A. (2008) “Mechanical
stresses in abdominal aortic aneurysms: influence of diameter, asymmetry, and
material anisotropy.” Journal of Biomechanical Engineering, 130, 0210231-10.
66
Chapter IV
Vessel asymmetry as an additional diagnostic tool in
the assessment of abdominal aortic aneurysms.
Barry J. Doyle,1 Anthony Callanan,1 Paul E. Burke,2 Pierce A.

Grace,2 Michael T. Walsh,1 David A. Vorp3 and Timothy M.
McGloughlin1
Limerick, Ireland.
3. Departments of Surgery and Bioengineering, Centre for Vascular Remodelling and

Regeneration, McGowan Institute for Regenerative Medicine, University of Pittsburgh, PA,
USA.
Journal of Vascular Surgery 2009; 49: 443-454

Analysis and interpretation: BD, AC, TM
Data collection: PG, DV, PB
Critical revision of article: BD, AC, MW, PB, PG, TM, DV
Final approval: TM
Obtained funding: TM, BD, DV
Chapter Four Asymmetry of AAAs
ABSTRACT
Background
Abdominal aortic aneurysm rupture (AAA) is believed to occur when the local
mechanical stress exceeds the local mechanical strength of the wall tissue. Based on
this hypothesis, the knowledge of the stress acting on the wall of an unruptured
aneurysm could be useful in determining the risk of rupture. The role of asymmetry has
previously been identified in idealised AAA models, and is now studied using realistic
AAAs in the current work.
Methods
Fifteen patient-specific AAAs were studied to estimate the relationship between wall
stress and geometrical parameters. 3D AAA models were reconstructed from CT scan
data. The stress distribution on the AAA wall was evaluated by the finite element
method, and peak wall stress was compared to both diameter and centreline asymmetry.
A simple method of determining asymmetry was adapted and developed. Statistical
analyses were also performed to determine potential significance of results.
Results
Mean von Mises peak wall stress ± standard deviation was shown to be 0.4505 ± 0.14
MPa, with a range of 0.3157 – 0.9048 MPa. Posterior wall stress increases with anterior
centreline asymmetry. Peak stress increased by 48% and posterior wall stress increased
by 38% when asymmetry was introduced into a realistic AAA model.
Conclusions
The relationship between posterior wall stress and AAA asymmetry showed that
excessive bulging of one surface results in elevated wall stress on the opposite surface.
Assessing the degree of bulging and asymmetry that is experienced in an individual
AAA may be of benefit to surgeons in the decision making process, and may provide a
useful adjunct to diameter as a surgical intervention guide.
67
INTRODUCTION
There is currently much debate as to the most appropriate time to surgically intervene
and repair an abdominal aortic aneurysm (AAA) [1-7]. Surgery is often performed
when the detected AAA exceeds 5.0 - 5.5 cm in maximum diameter. Previous research
[8,9] has shown how AAAs smaller than 5.0 cm in maximum diameter can also rupture.
The reliability of the maximum diameter as the main criterion for rupture has been
questioned recently, and a need for a more reliable clinical predictor of AAA rupture
has been identified [1-7,10-12]. Previous work [12,13] has identified the importance of
asymmetry in idealised AAA models, and also indicated the need to investigate this
aspect in realistic models. In this study, we have examined the role of asymmetry and
resulting wall stress in realistic patient-specific AAA cases.
METHODS
Computed tomography (CT) scan data was obtained for 22 patients. For this study,
AAAs that were asymmetric in the anterior-posterior plane were deemed applicable.
Using this criteria, seven of the 22 cases were excluded from the analysis as these case
were asymmetric in other directions. The resulting cohort of 15 cases comprised of 10
males and 5 females. Mean age ± standard deviation of the case subjects was 73.2 ± 6.7
years. These patient scans were obtained from the Midwestern Regional Hospital,
Limerick, Ireland, and the University of Pittsburgh Medical Centre, Pittsburgh, PA,
USA. All 15 patients were awaiting AAA repair, as AAA diameters had reached or
exceeded the current 5 cm threshold for repair. CT scans were acquired using both the
Somatom Plus 4 (Siemens AG, D-91052 Erlangen, Germany) and LightSpeed Plus (GE
Medical Systems, General Electric Company) range of imaging equipment. All scans
were single CT slices with a standard width x height of 512 x 512 pixels. Mean pixel
size of scans was 0.742 ± 0.072 mm. The bodily structures of each subject were made
visible using the non-ionic contrast dye, Optiray® (Mallinckrodt Inc., Convidien, MO,
USA). This CT data was then reconstructed using the commercially available software,
Mimics v10.0 (Materialise, Belgium), and these reconstructions allowed the
computation of stress distributions within the geometries. The patient-specific details
for the cases studied can be seen in Table I, which were obtained using the schematic of
Figure 1.
68
Figure 1: Schematic of representative AAA showing how dimensions are obtained from
each AAA model. Surface area and volume both encompass the total surface area or
volume of the AAA from immediately below the renal arteries to immediately prior to
the iliac bifurcation. Dianorm is the infrarenal aortic diameter of the particular case.
3D Reconstruction Procedure
Spiral CT data was used to reconstruct the infrarenal section of the aorta. As CT
scanning is routinely performed on AAA patients scheduled for repair, collection of this
information involved no extra participation by the study subjects. Digital files in
Digital Imaging and Communications in Medicine (DICOM) file format, containing
cross-sectional information was imported to the software, Mimics v10.0 for
reconstruction. All reconstructions were developed from scan positions immediately
distal to the lowest renal artery to immediately proximal to the iliac bifurcation. The
intraluminal thrombus (ILT) was neglected in this study as with previous approaches
[6,7,11-13,15]. The thickness of the aorta wall is not easily identifiable from CT scans,
therefore the wall was assumed to be uniform throughout the model and set to 2 mm
[16]. Once regions of interest were identified, 3D reconstructions were generated. The
reconstruction method employed here was validated and reported in previous work
performed by our group [17-19], along with the effect of geometry smoothing on
resulting wall stress [20]. All AAAs underwent the same degree of smoothing. The
iliac bifurcation was omitted from this study, as in previous stress analysis work as it
has been shown to not significantly affect the wall stress results of the AAA [7]. The
influence of asymmetry compared to a symmetric AAA was also examined. The
reconstruction of Patient 2 was modified using ProEngineer Wildfire 3.0 (Parametric
Technology Corporation, Needham, MA, USA) so that the AAA now formed along a
straight central axis, becoming an axisymmetric fusiform aneurysm. This symmetric
69
AAA was created using the same diameter information as the original case. The two
forms of this AAA can be seen in Figure 2.
Figure 2: Original asymmetric AAA of Patient 2 (left) and the modified axisymmetric
AAA (right).

The AAA material was assumed to be homogenous and isotropic with non-linear
realistic material properties [5] that have been implemented in many previous
publications [3,6,7,10,11,20-22]. The aorta is also known to be nearly incompressible
with a Poisson’s ratio of 0.49.
FEA Mesh Generation

Once the AAA surfaces were imported into ABAQUS v.6.6-2 (Dassault Systemes,
SIMULIA, Rhode Island, USA) for stress analysis, a mesh was generated on the AAA
model. As wall thickness cannot be fully determined from the AAA scan data, each
shell element was assigned a uniform thickness of 2 mm [16]. Mesh independence was
performed by increasing the number of elements in the mesh until the difference in peak
stress was less than 2% of the previous mesh [10,20,23].
Forces and Boundary Conditions

The blood pressure within the AAA acts on the AAA inner wall and therefore, pressure
was applied to the inner surface of the computational AAA model. A static peak
systolic pressure of 120 mmHg (16 kPa) was used. In order to simulate the tethering of
the AAA to the aorta at the renal junction and iliac bifurcation, the AAA model was
fully constrained in the proximal and distal regions.
70
Asymmetry Definition
Most AAAs are constrained from radial expansion in the posterior direction due to the
spinal column, therefore, AAAs predominantly dilate in the anterior plane. All cases
studied in this analysis were naturally asymmetric in the anterior-posterior direction. In
order to examine the effect asymmetry has on wall stress, the centreline of each AAA
was automatically found using the software Mimics v10.0. The centreline passes
through the centroid of each polyline slice in the series. Asymmetry is defined, in this
case, as the perpendicular distance from the proximal and distal points of the centreline
to a defined point on the centreline. Figures 3-5 show how these asymmetry measures
are obtained. This method of determining asymmetry was adapted from previous work
by Young Suh et al. [24] Starting with the 3D AAA model in Figure 3A, a centreline is
automatically created through the polyline centroids of Figure 3B, thus creating Figure
3C. Then these polylines are exported from Mimics v10.0 to ProEngineer Wildfire 3.0.
Next, using the software, the end points of the centreline are connected with a straight
line (Figure 3D) and a perpendicular line is extended from this connecting line, to
predetermined points along the centreline (Figure 3E and 3F). The asymmetry at a
specified distance along the AAA model is regarded as this perpendicular distance, and
is measured in millimetres (mm). This method of determining AAA asymmetry is
shown for Patient 3 in Figures 4 and 5. Figure 4 shows the creation of the polylines on
the CT scan after the thresholding and segmentation process in Mimics v10.0, and also
the resulting model of polylines and AAA centreline. Figure 5 shows the measurement
process of asymmetry for Patient 3, and the resulting asymmetry plot. Maximum
asymmetry for this case was 24.5 mm.
Figure 3: Simple illustration showing method of obtaining asymmetry measurements.
71
Figure 4: Example CT scans showing the creation of polylines on each slice, together
with the centre point of each polyline. Once polylines are created on each scan in the
series, the slices are stacked together to form the model on the right. Example shown is
for Patient 3.
Figure 5: Diagram depicting how measurements of asymmetry are determined once

AAA centreline is created. Example shown is for Patient 3.
72
Statistical Analysis
In order to evaluate the statistical significance of the wall stress results, the software
SPSS 14.0 (SPSS Inc., Chicago, IL, USA) was utilised. This allowed any significant
correlations within the results to be identified. Correlations between various
geometrical parameters with both peak wall stress and posterior wall stress were
assessed for significance.
RESULTS
The results from the geometrical examination of each case are shown in Table I. The
finite element analysis using ABAQUS v6.6-2 produced a detailed stress pattern on
each of the aneurysmal models under the pressure loading [25]. From these stress
results, factors affecting wall stress could be examined, in particular, the role of
asymmetry.
Table I: Patient details for each study subject.

Max Total Total
Total S.A. Dia/
Patient Sex Age Diameter Length Volume ROD
(cm2) Length
(cm) (cm) (cm3)
1 Male 66 5.6 13.2 176.7 19.8 0.425 1.533
2 Male 78 6.1 11.2 192.4 18.5 0.544 2.071
3 Male 70 5.7 13 194.9 19.1 0.439 1.752
4 Female 65 5.6 9.3 136.9 14.8 0.606 2.474
5 Male 81 5.9 12.8 220.9 21.8 0.463 1.772
6 Female 68 5.7 10 148.2 16.3 0.570 2.953
7 Female 67 5.3 10.5 137.9 15.5 0.505 2.000
8 Male 70 6.0 11.1 216.7 20.4 0.541 1.508
9 Female 77 5.8 8.8 94.2 12.6 0.661 2.535
10 Male 87 9.0 11.7 445.5 32.4 0.769 1.768
11 Male 66 6.5 10.5 207.6 19.2 0.617 1.952
12 Male 81 6.8 14 267.6 25.2 0.484 1.994
13 Male 77 6.2 16.8 320.8 27 0.371 1.757
14 Female 72 5.7 11.4 143.8 16.7 0.497 2.675
15 Male 73 7.9 11.8 286.5 24.3 0.671 2.865
Note that S.A. is total AAA surface area, Dia/Length is the ratio of maximum diameter
to total AAA length, and ROD is the ratio of maximum AAA diameter to infrarenal
diameter of that patient [14].
Peak Wall Stress

It was noted from the computed stress results that the regions of peak wall stress
occurred at regions of inflection on the surface of the AAA models. Inflection points
are defined as points on the AAA surface at which the local AAA wall shape changes
from concave outward to concave inward [12]. The peak stress occurring at regions of
inflection was also observed by previous researchers in idealised models [12,13,26,27].
The von Mises peak wall stress, diameter at peak stress, and location were recorded for
each case and were compared to the maximum diameter in Table II. All peak stress
73
values are recorded at the peak systolic pressure of 120 mmHg (16 KPa). Shown also is
the diameter of the region through which the peak wall stress occurred. Mean von
Mises peak wall stress ± standard deviation was shown to be 0.4505 ± 0.14 MPa, with a
range of 0.3157 – 0.9048 MPa. The circumferential stress was also recorded for each
case. Mean circumferential stress ± standard deviation was 0.1176 ± 0.061 MPa, with a
range of 0.07 – 0.3271 MPa.
Table II: Maximum diameter, peak wall stress, location of peak stress, and diameter at
peak wall stress for each patient studied.
Max Diameter Peak Wall Location of Peak Diameter at Peak
Patient
(cm) Stress (MPa) Wall Stress Wall Stress (cm)
1 5.6 0.4018 Anterior-Right 3.8
2 6.1 0.4213 Anterior-Right 4.1
3 5.7 0.5524 Posterior 5.3
4 5.6 0.3157 Posterior 4.9
5 5.9 0.3822 Posterior 5.3
6 5.7 0.3823 Left 5.8
7 5.3 0.3621 Left 4.8
8 6.0 0.3872 Posterior 4.6
9 5.8 0.4093 Posterior-Right 5.5
10 9.0 0.9048 Posterior 6.2
11 6.5 0.4608 Left 5.8
12 6.8 0.4991 Left 5.6
13 6.2 0.4523 Left 6
14 5.7 0.3703 Posterior 5.4
15 7.9 0.4747 Left 7.6
Wall Stress – Asymmetry Relationship

Figures 6-8 show how the von Mises wall stress varies with respect to the asymmetry of
the AAA centreline. It was noted that regions of elevated centreline asymmetry
experienced a region of elevated posterior wall stress.
74
Figure 6: Relationships between posterior wall stress and anterior asymmetry (left
column) and posterior wall stress and diameter (right column) for Patients 1-5.
75
76
77
Effect of Asymmetry on Wall Stress

In order to gauge the effect of asymmetry on resulting wall stress, the AAA of Patient 2
was modified into a symmetric aneurysm as described earlier. The symmetric wall
stress can be seen compared to that of the posterior wall stress in the asymmetric case in
Figure 9. Peak stress increased from 0.2186 MPa to 0.4213 MPa when asymmetry was
introduced into the AAA. This resulted in a 93% increase in peak wall stress in the
asymmetric model. There was also a noticeable increase of 61% from 0.2186 MPa to
0.3527 MPa in posterior wall stress between the two models.
0.4
Asymmetric Symmetric
0.35
0.3
Stress (MPa)
0.25
0.2
0.15
0.1
0.05
0
0 20 40 60 80 100 120 140
Distance (mm)
Figure 9: Comparison of the posterior wall stress for the symmetric and asymmetric
AAA case for Patient 2. Peak stress for this asymmetric case was located on the
anterior-right wall.
A Spearman’s Rho correlation test was considered in order to assess any relationships
evident between both peak and posterior wall stress and various patient-specific
measurable parameters. Correlations are deemed significant when P<0.05. The P
values of this study can be seen in Tables III and IV. There was no significant
correlation between peak circumferential stress and either maximum asymmetry
(P=0.0708) or maximum diameter (P=0.5197). The relationship between posterior wall
stress with both asymmetry and diameter was also examined using a Spearman’s Rho
correlation test. Coefficients were found using a bivariate correlation test to compare
posterior wall stress with both asymmetry and diameter at 10 mm intervals along the
78
longitudinal distance of each patient. These results can be seen in Table V. The
significance of the relationship between asymmetry and diameter was also assessed
using a nonparametric correlation test. The results are shown in Table VI. Eleven of
the fifteen cases revealed that there was a significant correlation between asymmetry
and diameter. Patient age also correlated well with both maximum diameter (P=0.009)
and peak posterior wall stress (P=0.028).
Table III: Statistical analysis of patient-specific parameters and peak wall stress.
P
Max Diameter 0.0003
Peak Posterior Stress 0.0021
Asymmetry at Peak Circ Stress 0.0036
AAA Volume 0.0043
Sex 0.0061
Max CSA 0.0126
AAA Surface Area 0.0130
ILT Volume 0.0232
AAA Length 0.0961
Peak Circ Stress 0.1580
Lumen Volume 0.1658
ROD 0.3307
Asymmetry at Peak VM Stress 0.4384
AAA Diameter/AAA Length 0.5409
Peak Stress Location 0.5814
Peak Asymmetry 0.6384
AAA Volume/ILT Volume 0.8994
Average Asymmetry 0.9345
79
Table IV: Statistical analysis of patient-specific parameters and posterior wall stress.
P
AAA Volume 0.0002
AAA Surface Area 0.0008
Lumen Volume 0.0012
Max CSA 0.0013
Peak Stress 0.0021
Max Diameter 0.0028
Sex 0.0081
Asymmetry at Peak Circ Stress 0.0136
AAA Length 0.0144
Peak Circ Stress 0.0378
AAA Volume/ILT Volume 0.0983
ILT Volume 0.1728
ROD 0.2597
Peak Stress Location 0.5399
Peak Asymmetry 0.6025
Asymmetry at Peak VM Stress 0.7466
AAA Diameter/AAA Length 0.9295
Average Asymmetry 0.9496
Note that CSA is the cross-sectional area at the region of maximum diameter; Circ
Stress is the circumferential stress; ROD is the ratio of the maximum diameter to the
infrarenal diameter; and VM Stress is the von Mises stress. This refers to both Table III
and IV.
Table V: Correlation coefficients for posterior wall stress and both asymmetry and
diameter.
Patient Asymmetry P Diameter P
1 0.574 0.032 0.420 0.135
2 0.781 0.003 0.895 0.000
3 0.175 0.587 0.755 0.005
4 0.37 0.293 0.733 0.016
5 0.862 0.000 0.820 0.000
6 0.82 0.002 0.609 0.047
7 0.464 0.151 0.573 0.066
8 0.834 0.001 0.573 0.051
9 0.474 0.166 0.529 0.116
10 0.443 0.130 0.505 0.078
11 0.683 0.014 0.811 0.001
12 0.411 0.128 0.593 0.020
13 0.411 0.128 0.593 0.020
14 0.834 0.000 0.709 0.007
15 0.667 0.013 -0.132 0.668
80
Table VI: Correlation between asymmetry and diameter for each case examined.
Patient Coefficient P
1 0.801 0.001
2 0.823 0.001
3 0.755 0.005
4 0.612 0.060
5 0.935 0.000
6 0.752 0.008
7 0.9 0.000
8 0.806 0.002
9 0.799 0.006
10 0.627 0.022
11 0.746 0.005
12 0.391 0.150
13 0.391 0.150
14 0.889 0.000
15 0.066 0.830
The statistical significance between peak stress and other relevant parameters were also
assessed. The rate of change of both asymmetry and diameter along the length of the
AAA were not statistically significant (P=0.089 and P=0.501, respectively). The
diameter at which peak stress occurred proved to be significant with a P value of 0.039.
DISCUSSION
In this study, fifteen patient-specific AAAs were reconstructed, and wall stress
distributions in each aneurysm were estimated using the finite element method. From
the von Mises wall stress distributions, the peak stress was found to occur at regions of
inflection. This finding is consistent with previous research, both numerical [12,26] and
experimental [27]. The peak wall stresses found in this study ranged from 0.3157 –
0.9584 MPa, with a mean ± standard deviation value of 0.482 ± 0.197 MPa. The AAA
with the lowest peak wall stress (0.3157 MPa), Patient 4, also had the second smallest
maximum AAA diameter (5.6 cm), whereas, the AAA with the highest peak stress
(Patient 10) of 0.9048 MPa had the largest AAA diameter (9 cm). These finding may
suggest that the maximum diameter criterion may be a good predictor of AAA rupture.
From the previous research into this hypothesis [9] it is known that this may not always
be the case, as small AAAs can also rupture. Fillinger et al. [7] performed stress
analysis on an AAA which had a smaller diameter than the current 5.5 cm threshold
(diameter = 4.8 cm). This particular AAA experienced a peak wall stress of 0.335 MPa,
which is within 10% of the peak stress found in Patient’s 4, 7 and 14 of this study.
81
It is important to note that ruptures do not necessarily occur at the region of peak wall
stress, but in fact occur where the locally acting wall stress exceeds the locally acting
wall strength. This study examined the role of realistic asymmetry on posterior wall
stress in patient-specific cases. In order to determine the affect the asymmetry of the
AAA has on posterior wall stress, a simple method of calculating asymmetry was
established. Figures 3-5 show the approach used to plot the degree of asymmetry in
each case, which were then coupled with the posterior wall stress results. These stress-
asymmetry relationships are shown in Figures 6-8, which also show the relationship
between diameter and posterior wall stress. It was stated by Raghavan et al. [28] that
the posterior wall tends to be the higher stressed region, and also the rupture site, even
though the bulge is predominantly anterior. Vorp et al. [12] identified the link between
asymmetry and wall stress in idealised AAA models using FEA, with Scotti et al. [13]
using a fluid-structure interaction (FSI) approach to highlight the relationship in
idealised models. It was concluded in the work of Scotti et al. [13] that AAAs
experiencing asymmetry may be exposed to higher mechanical stresses and increased
risk of rupture than more fusiform AAAs. This current work agrees with this
hypothesis, and has furthered the work of idealised AAAs to that of realistic AAA
geometries with results suggesting that there may be an intrinsic relationship between
asymmetry and posterior wall stress.
Darling et al. [9] determined from 118 AAA autopsies that 82% of ruptures occur on the
posterior wall, indicating that these ruptures may have been as a result of elevated
posterior wall stress. Approximately 68% of cases (15 of 22 patients) examined for this
study experienced posterior-anterior bowing, resulting in elevated posterior wall stress.
It has been recently reported [29] that the posterior and right regions of AAAs are
regionally thinner than the anterior and left regions. Raghavan et al. [29] also reported
that failure tension may be a better indicator of rupture rather than failure stress, with
failure tension described as; peak wall tension = peak wall stress x wall thickness.
Applying this failure tension to this study results in failure tensions ranging from 0.6314
– 1.8096 N/mm, compared to the range of 0.42 – 1.48 N/mm observed by Fillinger et al.
[6]. Actual failure stress of AAA tissue has been shown to range from 0.336 – 2.351
MPa (median = 1.266 MPa) [29].
Giannoglou et al. [30] also determined that mean AAA curvature may be a better
predictor of AAA rupture risk, although, this previous study implemented linearly
82
elastic material properties. AAA centreline curvature was also analysed as part of this
present study using the curvature analysis function in ProEngineer Wildfire 3.0. The
resulting centreline curvature readings are difficult to interpret as minor changes in the
centreline result in large spikes of curvature. Gaussian surface curvature was also
examined using ProEngineer Wildfire 3.0, as previously reported [22]. It has been
shown that rapid changes in surface curvature may indicate regions of high wall stress
[22]. As with centreline curvature, the model geometries are too complex to achieve
surface curvature results with any quantifiable meaning. The definition of asymmetry
in this study is a measurement that is easy to interpret and calculate, and shows good
agreement with wall stress results.
As the anterior region of the AAA bulges outwards, the posterior region is often
constrained from radial expansion by the spinal column, and results in elevated
posterior wall stress. AAAs may also rupture at regions experiencing a wall stress less
than that of the peak wall stress as AAA are known to rupture when the local stress
exceeds the local wall strength, with AAAs experiencing regional variations in wall
strength [29]. In this study, two cases experienced peak wall stress on the anterior-right
wall. Peak stress can occur at any region along the AAA surface, but is predominantly
found at regions where there is high local surface curvature or asymmetry. Therefore,
when analysing patient-specific AAA cases, it is difficult to pre-determine the location
of peak stress. Six cases resulted in peak stress on the left wall of the AAA. Again,
these locations of elevated stress are due to the local topology of the surface. Even
though peak stress does not necessarily occur along the posterior wall, in all cases
examined there was an increase in posterior stress along the length of the particular
AAA in relation to anterior asymmetry.
From the results of the statistical analysis shown in Tables III and IV, one can
determine that the significant parameters that relate to peak wall stress are maximum
diameter (P=0.0003), peak posterior wall stress (P=0.0021), asymmetry at the region of
peak circumferential stress (P=0.0036), AAA volume (P=0.0043), sex (P=0.061),
maximum cross-sectional area (P=0.0126), AAA surface area (P=0.013), and also the
ILT volume (P=0.0232). In comparison, the significant relationships with posterior
wall stress are AAA volume (P=0.0002), AAA surface area (P=0.0008), lumen volume
(P=0.0012), maximum cross-sectional area (P=0.0013), peak wall stress (P=0.0021),
maximum diameter (P=0.0028), sex (P=0.0081), asymmetry at the region of peak
83
circumferential stress (P=0.0136), AAA length (P=0.0144), and also peak

circumferential stress (P=0.0378). From these results, maximum diameter appears to be
significantly related to peak stress, but if the sample size of 15 used in this study was
increased to much larger numbers this relationship may not be as strong. The majority
of these parameters are also based on diameter, and therefore, if diameter returns a
strong correlation with peak stress, it may be obvious that similar parameters will also
score highly. There was no statistical significance between peak wall stress and the rate
of change of diameter (P=0.501) or rate of change of asymmetry (P=0.089).
Closer examination of the relationship between both asymmetry and diameter with
posterior wall stress involved analyses using a nonparametric Spearman Rho
correlation. These results are presented in Table V and show how both asymmetry and
diameter are both comparable in their significance towards posterior wall stress. From
the resulting correlations, 8/15 cases show asymmetry is significant and 9/15 cases
show that diameter is significant. These results suggest that if posterior wall stress is to
gain clinical acceptance as a possible high-risk rupture indicator, that both asymmetry
and diameter may be as important in determining the posterior wall stress, and therefore
may both equally contribute to AAA rupture. It has been previously postulated by Vorp
et al. [25] and also by Fillinger et al. [6,7] that the biomechanics of the AAA may
provide useful clinical guidance over the maximum diameter criteria. This work
supports this biomechanics-based approach and in particular suggests that posterior wall
stress may be clinically important. The results presented also suggest that if peak wall
stress is to remain the primary purpose of AAA stress analyses, then diameter remains a
significant factor.
Although, ideally, stress analysis should be carried out on every AAA detected, the
reality is that the decision to repair lies with the surgeon. The use of the maximum
diameter criterion is very easy to implement for the surgeon, in that they must simply
measure the maximum diameter from CT scans. The asymmetry condition described in
this study could also be readily incorporated into the surgeon’s decision making. This
dilation, and ultimately asymmetry, can be identified by the clinician from basic 3D
reconstruction, and could greatly aid in their decision to surgically intervene. A method
of determining asymmetry from 2D CT scans is also currently under development
within our group. Also, our group is developing an approach that accounts for
asymmetry in all directions, and therefore the relationship between asymmetry and wall
84
stress can be assessed in all AAAs, regardless of orientation. Once detected, the degree
of bulging could be incorporated into the decision making process of the surgeon, and
may refine and improve the current system of deciding on surgical intervention solely
on the basis of maximum diameter. It is suggested that, to include AAA asymmetry as
another means of assessing the rupture potential of AAAs could serve as a useful
adjunct to the maximum diameter criterion, and may ultimately lead to improved
surgical decision making.
Limitations
Like previous work [6,7,11-13,15], this study did not include ILT in the AAA 3D
reconstructions. The ILT has been shown to reduce wall stress by up to 30% [10,30]
and can act as a “mechanical cushion” [31] for the AAA wall. It is known that the
realistic AAA has a non-uniform wall thickness [29] varying regionally from 0.23 mm
at a rupture site to 4.26mm at a calcified site (median = 1.48 mm), and also non-uniform
material properties due to regions of calcifications [32], which can lead to alterations in
stress distributions [32,33]. This study examined AAA wall stress using a static
analysis. It is possible that a dynamic loading, such as a realistic infrarenal aortic pulse
may influence the stress distributions. Researchers have shown that the use of fluid-
structure interaction methods to determine wall stress can give more accurate results
[3,13]. Computational time is increased by as much as 2500-fold [13] from that of a
static pressure finite element analysis, with maximum stress locations found to be the
same using both methods. Variations in maximum wall stress in realistic models have
been reported to range from 1-25%, according to prior studies [3,21,33]. In order to
establish the suitability of this method for clinical applications, a larger cohort of patient
data is required. The authors are investigating the possibility of applying this method to
a database of previously screened patients, with a view to enhancing confidence in the
asymmetry approach. Applying this study to a larger cohort may also significantly alter
the statistical results as only 15 patients are studied here.
CONCLUSIONS
Most AAA ruptures occur on the posterior wall. It has been shown here how the
posterior wall stress can be related to anterior asymmetry in patient-specific cases.
Results suggest that an increase in asymmetry may cause increases in posterior wall
stress. Statistical analyses revealed that maximum diameter still significantly influences
wall stress, particularly peak wall stress, but that asymmetry may also have a significant
85
role in posterior wall stress. This study suggests that AAA asymmetry may be an
important criteria in AAA assessment, and could possibly be included as a factor in the
clinicians’ decision to surgically intervene. Further evaluation is needed to determine
clinical applicability.
ACKNOWLEDGEMENTS
the Midwestern Regional Hospital, Ireland (iv) Dr. Liam Morris in the Galway Medical
Technology Centre, Galway and Mayo institute of Technology, Ireland (v) Samarth
Shah from the Centre for Vascular Remodelling and Regeneration and (vi) Michel S.
Makaroun, MD, University of Pittsburgh, Department of Surgery.
REFERENCES
(2006) “Towards a non-invasive method for determination of patient-specific wall
strength distribution in abdominal aortic aneurysms.” Annals of Biomedical
Engineering, 34(7), 1098-1106.
prediction of abdominal aortic aneurysms.” Biomedical Engineering Online, 5,
19.
Online, 5, 33.
4. Sayers, R.D. (2002) “Aortic aneurysms, inflammatory pathways and nitric
oxide.” Annals of the Royal College of Surgeons of England, 84(4), 239-246.
Surgery, 33, 475-482.
(2002) “In vivo analysis of mechanical wall stress and abdominal aortic aneurysm
rupture risk.” Journal of Vascular Surgery, 36, 589-597.
8. Nicholls, S.C., Gardner, J.B., Meissner, M.H., Johansen, H.K. (1998) “Rupture in
small abdominal aortic aneurysms.” Journal of Vascular Surgery, 28, 884-888.
Circulation, 56(2), 161-164.
86
14. Hatakeyama, T., Shigematsu, H., Muto, T. (2001) “Risk factors for rupture of
abdominal aortic aneurysm based on a three-dimensional study.” Journal of
Vascular Surgery, 33, 453-461.
355-366.
wall stress using finite element analysis for ruptured and non-ruptured abdominal
aortic aneurysms.” European Journal of Vascular and Endovascular Surgery, 28,
168-176.
McGloughlin, T. (2005) “3D numerical simulation of blood flow through models
of the human aorta.” Journal of Biomechanical Engineering, 127, 767-775.
18. Morris, L.G. (2004) “Numerical and experimental investigation of mechanical
factors in the treatment of abdominal aortic aneurysms.” PhD Thesis, University
of Limerick.
20. Doyle, B.J., Callanan, A., McGloughlin, T.M. (2007) “A comparison of
modelling techniques for computing wall stress in abdominal aortic aneurysms.”
Biomedical Engineering Online, 6, 38.
22. Sacks, M.S., Vorp, D.A., Raghavan, M.L., Federle, M.P., Webster, M.W. (1999)
“In vivo three-dimensional surface geometry of abdominal aortic aneurysms.”
Annals of Biomedical Engineering, 27(4), 469-479.
23. Truijers, M., Pol, J.A., SchultzeKool, L.J., van Sterkenburg, S.M., Fillinger, M.F.,
Blankensteijn, J.D. (2007) “Wall stress analysis in small asymptomatic,
24. Young Suh, G., Choi, G., Draney Blomme, M., Taylor, C. (2007) “Quantification
of three-dimensional motion of the renal arteries using image-based modelling
techniques.” Proceedings of the American Society of Mechanical Engineers,
Summer Bioengineering Conference 2007, United States.
87
28. Raghavan, M.L., Kratzberg, J.A., Golzarian, J. (2005) “Introduction to
biomechanics related to endovascular repair of abdominal aortic aneurysm.”
Techniques in Vascular and Interventional Radiology, 8, 50-55.
29. Raghavan, M.L., Kratzberg, J., de Tolosa, E.M.C., Hanaoka, M.M., Walter, P., da
Silva, E.S. (2006) “Regional distribution of wall thickness and failure properties
of human abdominal aortic aneurysm.” Journal of Biomechanics, 39(16), 3010-6.
25, 1558-1566.
32. Vorp, D.A., Mandarino, W.A., Webster, M.W., Gorcsan III, J. (1996) “Potential
influence of intraluminal thrombus on abdominal aortic aneurysm as assessed by
a new non-invasive method.” Cardiovascular Surgery, 4, 732-739.
34. Scotti, C.M., Finol, E.A. (2007) “Compliant biomechanics of abdominal aortic
aneurysms: a fluid-structure interaction study.” Computers and Structures, 85,
1097-1113.
88
PART II
EXPERIMENTAL INVESTIGATION
Chapter V
3D reconstruction and manufacture of real abdominal
aortic aneurysms: From CT scan to silicone model.
Barry J. Doyle,1 Liam G. Morris,2 Anthony Callanan,1 Paddy Kelly,1

David A. Vorp3 and Timothy M. McGloughlin1
Limerick, Ireland.
2. Galway Medical Technology Centre, Galway Mayo Institute of Technology, Galway, UK.

USA.
Journal of Biomechanical Engineering 2008; 130: 034501-5

Analysis and interpretation: BD
Data collection: LM, DV, PK
Critical revision of article: BD, AC, LM, DV, TM
Final approval: DV, TM
Chapter Five From CT Scan to Silicone Model
ABSTRACT
Background
Abdominal aortic aneurysm (AAA) can be defined as a permanent and irreversible
dilation of the infrarenal aorta. AAAs are often considered to be an aorta with a
diameter 1.5 times the normal infrarenal aorta diameter. This paper describes a
technique to manufacture realistic silicone AAA models for use with experimental
studies.
Methods
This paper is concerned with the reconstruction and manufacturing process of patient-
specific AAAs. 3D reconstruction from computed topography (CT) scan data allows
the AAA to be created. Mould sets are then designed for these AAA models utilising
CAD/CAM techniques to be used with the injection-moulding technique. Silicone
rubber forms the basis of the resulting AAA model.
Results
Assessment of wall thickness and overall percentage difference from the final silicone
model to that of the computer-generated model was performed. In these realistic AAA
models, wall thickness was found to vary by an average of 9.21%.
Conclusions
The percentage difference in wall thickness recorded can be attributed to the contraction
of the casting wax and the expansion of the silicone during model manufacture. This
method may be used in conjunction with wall stress studies using the photoelastic
method, or in fluid dynamic studies using a laser Doppler anemometry (LDA). In
conclusion, these patient-specific rubber AAA models can be used in experimental
investigations, but should be assessed for wall thickness variability once manufactured.
89
INTRODUCTION
An abdominal aortic aneurysm (AAA) can be defined as a permanent and irreversible
localised dilation of the infrarenal aorta [1]. It has been proposed that an AAA is an
aorta with a diameter 1.5 times that of the normal infrarenal aorta [2]. Currently, the
timing of surgical intervention is determined by the maximum diameter of the AAA,
with an AAA diameter greater than 5 cm deemed to be at high risk of rupture. Much
work has been aimed at the rupture prediction of these aneurysms, in particular the use
of finite element analysis (FEA) to determine wall stress [3-8]. Although, the use of
numerical studies to gain an insight into the stress acting on the AAA wall is of obvious
benefit to the particular AAA case, validation of these techniques is of equal
importance. The ability to manufacture patient-specific AAA models for experimental
studies could extend the use of pre-operative wall stress techniques. These realistic
silicone models could be employed, not only for stress analyses, similar to the
photoelastic work of Morris et al. [9], but also for fluid dynamics studies and post-
operative experimental testing, such as stent graft distraction testing. The models are
created by first reconstructing a virtual AAA model, leading to mould design, and then
to manufacturing via the injection-moulding technique. Previous research has examined
the use of rapid-prototyping as a method of producing elastomeric replicas of arterial
vessels [10]. This method, although quick and effective, does not produce the surface
finish that can be achieved using the injection-moulding process. Surface finish is of
paramount importance when using arterial models for experimental testing using the
photoelastic method, such as that previously conducted at our laboratory [9]. Other
techniques have been employed in order to make models for use in laser-Doppler
anemometry (LDA) [11] and particle image velocimetry (PIV) flow studies, where
surface finish was of lesser importance then when conducting wall stress studies.
The principle objective of this study is to describe the modeling and manufacturing
process used and to determine the effectiveness of the technique. This process of
converting a standard computed tomography (CT) scan to a patient-specific silicone
model is of value to many researchers in this field.
METHODS
3D Reconstruction
Four patients were chosen from our AAA database. The CT scans of each patient were
then imported into the software package Mimics v10.0 (Materialise, Belgium). This
90
software allows the transformation of 2D CT scans into realistic 3D models of exact

geometry. This software uses a marching squares algorithm to threshold and segment
the regions of interest of the CT scan according to a predetermined grayscale value.
Once segmented, the software generates polylines around the segmented regions, to a
user-controlled level of smoothing. An example of this image segmentation and
polyline generation can be seen in Figure 1. In this study, polylines were created with
approximately 20 control points per scan, allowing optimum smoothing without the loss
of model accuracy. These polylines can then be exported as initial graphics exchange
specification (IGES) format. Previous work has utilized various other forms of
reconstruction software, such as Scion Image (Release Beta 4.0.2, Scion Corporation,
Frederick, MD) [12]. Validation of Mimics against this work has been performed, with
a percentage difference of 1.2% between the reconstruction methods.
Figure 1: Segmentation and polyline generation of CT scan. (A) The full CT scan and
(B) is a close-up of the region of interest. For the design of moulds the AAA was
regarded to be the full volume of the lumen and ILT.
Computer Aided Design (CAD)

Polylines created in Mimics are imported into ProEngineer Wildfire 2.0 (PTC,
Parametric Technology). Surfaces are then recreated along these polylines. These
surfaces are then split exactly into two halves, thus creating a two-piece mould set used
in the manufacturing technique. Each patient-specific mould design consists of two sets
of moulds. The first mould is designed to produce the casting wax model of the AAA,
and the second set to produce the outer silicone model. The outer mould is
approximately 2 mm larger in all regions than that of the wax mould, so as to produce a
silicone model with a 2 mm thick wall. As the wall of an AAA can range in thickness
from 0.23 to 4.33 mm [13], a wall thickness of 2 mm is a reasonable assumption and
has been used in previous studies [14].
91
Example mould designs can be seen in Figure 2. Each outer mould design includes
supports for the inner wax cast to ensure location of the wax model inside the larger
outer mould. Of the four AAAs used in this study, three AAAs were modelled without
the iliac arteries (Patients A, B, and C), and one AAA with the iliac arteries included
(Patient D). For experimental studies involving stress analyses, the iliac arteries are
believed to be unimportant, whereas, for fluid dynamic and stent graft testing, the iliac
arteries are of paramount importance. Moulds designed without the iliac arteries have
cylindrical sections included both in the proximal and distal regions of the AAA, to
allow attachment to experimental test rigs.
Computer Aided Manufacture (CAM)

Once the mould sets have been designed in ProEngineer, the designs are exported again
in IGES format. These files are imported into the software package AlphaCAM in
order to generate the toolpath commands used to control the milling machine. Each
mould is set-up with the same reference points so as that each mould piece fits together
exactly, ensuring the resulting model has an almost negligible seamline.
Machining is performed by a three-axis computer numerical control (CNC) milling

machine. Moulds are machined from solid aluminium blocks and are finished by hand
to remove any unwanted burrs that results from the milling process. Figure 3 shows an
example machined mould piece. This illustration shows regions where extensions have
been included into the design in the proximal and distal regions of the AAA, and also
the inlet through which the wax is poured. Necessary holes and vents were added to
each model after machining.
92
Figure 2: Example mould designs of patient-specific AAAs. (A) Mould design including
iliac arteries and (B) without iliac arteries.
Figure 3: Example machined mould-piece of inner AAA model.
93
Model Manufacture
All mould pieces were cleaned using acetone prior to use. The wax moulds were
preheated to 40ºC to minimise the contraction of the wax upon pouring. A casting wax,
(Castylene B581, REMET Corporation) was used to make the lumen casts. The lumen
casts were then placed into the outer moulds, which were coated with releasing agent
(Ambersil Formula 8, Chemcraft Industries Ltd, Dublin, Ireland) and then clamped.
The silicone rubber (Wacker RT601) was then prepared, and slowly injected into the
preheated outer mould. Silicone-rubber was employed as the material due to its non-
linear behaviour when subjected to large strains [15] and is believed to be a good
arterial analogue. The mould is then placed into an oven at a temperature of 50ºC and
cured for 24 hours. Once cured, the model is removed and the temperature is increased
to 100ºC in order to melt the wax from the mould. The resulting silicone model is then
thoroughly cleaned, dried and inspected for defects. The full procedure can be seen in
the Appendix.
RESULTS
Sectioning the Model
Each model was sectioned at regular intervals to assess the dimensional accuracy of the
resulting silicone model compared to the CAD model. Each silicone AAA model was
carefully split using a scalpel longitudinally along the left and right sides, thus leaving
each model in two halves. Each half-model was then sliced axially at 10 mm intervals
along the length of the model, leaving a series of cross-sectional slices for each patient-
specific model.
Wall Thickness Measurements

For each cross-sectional slice of the silicone model, the wall thickness was measured at
four 90º equispaced locations around the edge. Therefore, wall thickness was measured
along the left, right, anterior and posterior walls of the full AAA model. Measurements
were obtained using a digital micrometer. Measurement readings ranged from 40 to 60
readings per AAA model, depending on the patient. Measurement results were then
averaged for each patient-tailored silicone model with the percentage difference
between the actual silicone model and the 2 mm wall CAD model noted. The standard
deviation was also included in the results. The measurement results can be seen in
Table 1 and are summarised as a total model wall thickness for each patient in Table 2.
Percentage difference refers to the difference between the silicone model wall thickness
94
and the original 2 mm wall thickness in the mould design. The AAA model of Patient D
included the iliac arteries.
Table 1: Averaged wall thickness measurements at four locations on the AAA wall.
Axial Position
Anterior Posterior Right Left
Average Wall Thickness (mm) 1.87 1.97 2.09 2.55
Patient A Standard Deviation 0.276 0.314 0.173 0.327
Percentage Difference 6.95 1.78 4.23 21.47

Patient B Standard Deviation 0.207 0.418 0.235 0.248

Patient C Standard Deviation 0.223 0.293 0.352 0.306

Patient D Standard Deviation 0.653 0.282 0.414 0.281
Table 2: Averaged wall thickness measurements for each patient-specific AAA model
Average Wall Thickness (mm) Average Percentage Difference
Patient A 2.12 4.24 %
Patient B 2.20 9.01 %
Patient C 2.29 12.51 %
Patient D 2.22 11.09 %
Wall Stress Distribution

Figure 4 shows the resulting von Mises wall stress distribution for Patient A and the
region of resulting peak stress. The results show how the peak stress in the AAA model
was 0.533 MPa, and was located on the anterior wall of the AAA. The combination of
FEA results with validated experimental wall stress studies could further the use of
numerical studies in the field of AAA rupture prediction. A more detailed study of the
wall stress experienced in AAAs is currently in progress, and will expand on the
preliminary FEA results presented here.
95
Figure 4: Example FEA von Mises wall stress distribution for Patient A showing region
of peak stress on anterior wall. The corresponding mould piece and resulting silicone
model of the same patient can be seen on the right of this figure.
DISCUSSION
This study describes a procedure for manufacturing patient-specific rubber models of
abdominal aortic aneurysms both with and without the iliac arteries. A 3D
reconstruction technique using commercially available software is coupled with a
CAD/CAM technique to achieve the desired mould designs capable of forming realistic
AAAs using the injection moulding method. Previous studies [14,16] have used similar
techniques to produce rubber models of vascular vessels. The models developed in this
study are of greater complexity. These reproducible silicone models can be utilised for
experimental testing of vessel hemodynamics, wall stress analyses and stent graft
studies, all of which may contribute to experimental validation of numerical work. This
technique may allow other researchers to begin manufacturing realistic AAA silicone
models for use in their experimental work. In recent years, emphasis has been placed
on the use of numerical studies to attempt to predict AAA rupture. The use of
experimental research into this area is also of importance. Not only could these silicone
AAA models help in validation purposes, but may also become an important asset in
AAA rupture prediction.
For each of the patient-specific models created, wall thickness is the most variable
factor. It has been reported [13] how the AAA wall can range in thickness from 0.23 –
4.33 mm, with aortic wall thickness reported to range from 1.1 to 3.4 mm [7,17,18].
Average wall thickness over the four models was noted to be 2.26 ± 0.39 mm. In this
study, wall thickness lies within the range recorded by previous researchers [7,17,18],
and so can be deemed as acceptable. Wall thickness results also compare favourably
96
with those found for a realistic aorta by O’Brien et al. [14], who recorded a wall
thickness in realistic aortas of 2.39 ± 0.32 mm. Although, wall thickness appears to be
within an acceptable range, the moulds were designed to have a wall thickness of 2 mm.
The resulting silicone models differ from the mould designs by an average of 9.21% in
wall thickness, which can be attributed to the contraction of the casting wax during
solidification and the thermal expansion of the silicone during the curing process.
These limitations of the technique were also observed by O’Brien et al [14]. This
previous work recorded percentage differences in wall thickness to mould design
ranging from 20% for a realistic straight section of an aorta, to 58% for a section of a
saphenous vein. The results observed in this study are deemed acceptable in that
percentage differences are considerably less than those previously reported [14]. It
should also be mentioned that the models produced here are full AAA models and not
straight vessel sections and, therefore, one would expect the percentage differences to
be higher than those previously reported [14]. Therefore, confidence has been
established in the method used to produce these models.
The issue of uniformity in wall thickness should also be addressed. In the mould
designs, the wall thickness was set at 2 mm, and therefore, the resulting silicone models
should also have a uniform wall thickness. Due to reasons mentioned above, that is,
wax contraction and silicone expansion, the wall thickness varies in each AAA model.
These differences in wall thickness can be attributed to the complex and tortuous
geometry of these patient-specific AAA models. This limitation was also noted by
O’Brien et al. [14]. Both O’Brien et al. [14] and Chong et al. [16] produced idealised
vascular models and in these much simpler models the wall uniformity issue was easily
overcome, thus highlighting the fact that realistic geometries increase the difficulty of
not only the CAD/CAM process, but also the model manufacture itself.
The use of a uniform wall in these experimental models may be considered in

appropriate, since it is known that the actual AAA wall tissue can include various forms
of arterial tissue (calcified deposits and thrombus), and thus is usually non-uniform.
Whilst these regions of both calcified tissue and thrombus can be detected from the CT
scan, their incorporation into the wall of the model introduces additional complexity.
Firstly, incorporating regions of varying material properties into the AAA wall, results
in greatly increased computations, as the numerical equations used to solve for wall
stress at these locations become extremely complex. Also, the primary purpose of
97
producing realistic silicone models of known uniform wall is to experimentally validate

numerical studies of the same AAA model. Most previous work regarding numerical
stress analyses of AAAs [3-8,19-22] has conducted testing using uniform walls. Our
earlier work [9] on idealised AAA models of known uniform wall thickness proved
quite successful and has paved the way for the introduction of realistic AAA models to
be tested using the same technique. This experimental photoelastic work was later
validated numerically on an idealised AAA model by our group [23], which confirmed
the locations of peak stress on the model. It is planned to revisit the concept of non-
uniformity of the silicone models using this described procedure. Work has also begun
in this laboratory on the inclusion of thrombus in the AAA models.
It has been suggested [6] that the use of a fluid-structure interaction (FSI) approach to
stress analysis may yield more accurate wall stress results than FEA alone. Some
researchers have shown that the wall stress is increased by amounts <1% [21], whereas
others have reported increases ranging from 12.5% [6] to 20.5% [20]. As there are
conflicting results in the benefits of FSI in wall stress studies, these uniform-walled
AAA models can help towards validating both methods of stress studies. Work has
begun within our laboratory on the use of FSI, employing Mesh-based Parallel Code
Coupling Interface (MpCCI 3.0.6, Fraunhofer SCAI, Germany) software, which couples
both ABAQUS and Fluent together to obtain realistic wall stress values. Notably, the
use of a uniform wall is widespread among researchers conducting FSI studies [5-6,19-
22], and so this future work will allow the numerical and experimental validation of
wall stress in realistic AAA models based on the uniform wall thickness.
CONCLUSION
The procedure for the manufacture of patient-specific AAA models has been described.
Confidence has been established in the reproducibility of the rubber models, and the
limitations noted. In general, rubber models of good geometrical accuracy can be
produced by sensible mould design and use of controlled parameters in the silicone
production. Models showed a maximum percentage difference of 9.21% between the
designed moulds and the resulting silicone models. Uniformity of wall thickness
proved to be the most difficult parameter to control, with finished silicone models
always inspected and assessed before commencement of experimental testing. This
technique may aid in the validation of numerical methods through photoelastic
98
validation [9] or also through experimental testing such as laser Doppler anemometry
(LDA) or particle image velocimetry (PIV).
In conclusion, 3D reconstruction and CAD/CAM techniques proved to be successful in

the replication of patient-specific rubber AAA models, and may help contribute to the
use of patient-specific AAA models in experimental testing. Therefore, the use of
silicone AAA models with uniformly thick walls will help towards the validation of
numerical work, for both wall stress studies and flow dynamics.
ACKNOWLEDGEMENTS
US National Heart Lung and Blood Institute (iii) Mr. Eamon Kavanagh, an
Endovascular surgeon at Midwestern Regional Hospital, Limerick, for his help in
collecting patient data and background information, and (iv) Samarth Shah from the
Centre for Vascular Remodelling and Regeneration, University of Pittsburgh, USA.
APPENDIX
Realistic Model Manufacture
In creating the male wax models, the following steps are given.
1. The first sets of moulds are used to create the wax models.
2. Clean the surface of the moulds with acetone; make sure it is free of loose
debris and coat with mould release (Ambersil Formula 8, Chemcraft
Industries Ltd, Dublin, Ireland).
3. Bolt the two mould pieces together tightly.
4. Melt the casting wax (Castylene B581, REMET Corporation) on a hotplate
at a temperature of around 150ºC.
5. Pre-heat moulds in oven to 40ºC.
6. Position the mould at an angle of 45° to aid the liquid wax to flow into the
mould and minimise the risk of trapping air, thus creating voids and bubbles.
7. Pour the wax into the mould as slowly as possible to prevent splashing,
which can also create voids. As the cavity is filled the mould is returned to
the vertical position to finish pouring the wax.
99
8. The wax is then left to solidify at room temperature for up to four hours.
During this cooling period, the mould is gently tapped with a mallet to allow
any trapped air to rise to the surface.
9. As the wax cools and solidifies, additional wax is added to the mould to
ensure a complete wax model.
10. Open the moulds and carefully remove the wax model from the mould.
In creating the silicone model, the following steps are given.

11. Mix the silicone in the required ratio of silicone and curing agent (9:1).
12. Hand mixing the material components for a period of 2 minutes is sufficient.
This liquid silicone will represent the aorta wall. The liquid silicone
contains air bubbles once hand mixed which must be removed. The working
time for the mixed components at room temperature is around 90min.
13. In order to remove the trapped bubbles, place the container of liquid silicone
into a freezer until all bubbles have been naturally removed. Duration of
time in the freezer depends on the viscosity of the liquid silicone and can
vary from 1-3 hours.
14. Once all the air has been removed, suck all the liquid silicone into a 60ml
syringe.
15. Clean the second set of aluminium moulds with acetone.
16. Spray on a coat of silicone mould release, (Ambersil Formula 8) onto both
the aluminium moulds.
17. Carefully remove any excess material and flash from the wax model.
18. Position the inner wax model inside the outer aluminium mould ensuring
that the wax model is sitting in a way that will allow uniform wall thickness.
19. Bolt the two aluminium moulds together tightly.
20. Seal around the edges of the mould with sealant to avoid any unwanted
leakage from the mould.
21. Slowly inject the liquid silicone into the aluminium cast using the filled
60ml syringe.
22. Once the liquid silicone is injected, place the mould in an oven at a
temperature of 50ºC for a period of 24 hours.
23. Once cured, open the cast and carefully remove the silicone and wax model.
24. Place the silicone and wax model into the oven at a temperature of 100ºC to
melt the wax from the mould.
100
REFERENCES
1. Sakalihasan, N., Limet, R.., Defawe, O.D. (2005) “Abdominal aortic aneurysm.”
Lancet, 365(9470), 1577-89.
2. Johnston, K.W., Rutherford, R.B., Tilson, M.D., Shah, D.M, Hollier, L., Stanley,
J.C. (1991) “Suggested standards for reporting on arterial aneurysms.
Subcommittee on reporting standards for arterial aneurysms, Ad hoc committee
on reporting standards, Society for Vascular Surgery and North American
Chapter, International Society for Cardiovascular Surgery.” Journal of Vascular
Surgery, 13, 452-458.
of rupture risk in abdominal aortic aneurysm during observation: Wall stress
5. Di Martino, E.S., Guadagni, G., Fumero, A., Ballerini, G., Spirito, R., Biglioli,
P., Redaelli, A. (2001) “Fluid-structure interaction within realistic 3D models of
the aneurysmatic aorta as a guidance to assess the risk of rupture of the
aneurysm.” Medical Engineering and Physics, 23, 647-655.
6. Papaharilaou, Y., Ekaterinaris, J.A., Manousaki, E., Katsamouris, A.N. (2007)
aortic aneurysms.” Journal of Biomechanics, 40(2), 367-377.
9. Morris, L, O’Donnell, P., Delassus, P., McGloughlin, T. (2004) “Experimental
10. Seong, J., Sadasivan, C., Onizuka, M., Gounis, M.J., Christian, F., Miskolczi, L.,
Wakhloo, A.K., Lieber B.B. (2005) “Morphology of elastase-induced cerebral
aneurysm model in rabbit and rapid prototyping of elastomeric transparent
replicas.” Biorheology, 42(5), 345-361.
11. Loth, F., Jones, S.A., Giddens, D.P., Bassiouny, H.S., Glagov, S., Zarins, C.K.
(1997) “Measurements of velocity and wall shear stress inside a ptfe vascular
graft model under steady flow conditions.” Journal of Biomechanical
Engineering, 119(2), 187-194.
12. Morris, L., Delassus, P., Callanan, A., Walsh, M., Wallis, F., Grace, P.,
775.
13. Raghavan, M.L., Kratzberg, J., de Tolosa, E.M.C., Hanaoka, M.M., Walker, P.,
39(16), 3010-6
14. O’Brien, T, Morris, L., O’Donnell, M., Walsh, M., McGloughlin, T. (2005)
“Injection-moulded models of major and minor arteries: the variability of model
wall thickness owing to casting technique.” Proceedings of the International
Mechanical Engineers, 219, Part H: Journal of Engineering in Medicine.
101
15. Marins, P.A.L.S., Natal Jorge, R.M., Ferreira, A.J.M. (2006) “A comparative
study of several material models for prediction of hyperelastic properties:
application to silicone-rubber and soft tissues.” Strain, 42, 135-147.
16. Chong, C.K., Rowe, C.S., Sivanesan, S., Rattray, A., Black, R.A., Shortland,
A.P., How, T.V. (1999) “Computer aided design and fabrication of models for in
vitro studies of vascular fluid dynamics.” Proceedings of the International
Mechanical Engineers, 213, Part H: Journal of Engineering Medicine.
17. Nicholls, S.C., Gardner, J.B., Meissner, M.H., Johansen, H. K. (1998) “Rupture
in small abdominal aortic aneurysms.” Journal of vascular Surgery, 28, 884-888.
18. Gullace, G., Ruffa, F. (1995) “Effect of pravastatin on abdominal aorta and
carotid wall thickness in dyslipidemia.” Journal of the American College of
Cardiology, 25(2), 369A-370A.
19. Li, Z., Kleinstreuer, C. (2005) “blood flow and structure interactions in a stented
abdominal aortic aneurysm model.” Medical Engineering and Physics, 27, 369-
382.
20. Scotti, C.M., Finol, E.A. (2007) “Compliant biomechanics of abdominal aortic
aneurysms: a fluid-structure interaction study,” Computers and Structures, 85,
1097-1113.
21. Leung, J.H., Wright, A.R., Cheshire, N., Crane, J. Thom, S.A., Hughes, A.D.,
aneurysms: A comparison with solid stress models,” Biomedical Engineering
Online, 5, 33.
22. Gao, F., Watanabe, M., Matsuzawa, T. (2006) “Stress analysis in a layered aortic
arch model under pulsatile blood flow.” Biomedical Engineering Online, 5, 25.
23. Callanan, A., Morris, L., McGloughlin, T. (2004) “Numerical and experimental
analysis of an idealised abdominal aortic aneurysm.” European Society of
Biomechanics, S-Hertogenbosch, Netherlands.
102
Chapter VI
An experimental and numerical comparison of the
rupture locations of an abdominal aortic aneurysm.
Barry J. Doyle,1 Timothy J. Corbett,1 Anthony Callanan,1 Michael T.

Walsh,1 David A. Vorp2 and Timothy M. McGloughlin1
Limerick, Ireland.

USA.
Journal of Endovascular Therapy (in press)

Analysis and interpretation: BD, TC, AC
Data collection: BD, DV
Critical revision of article: BD, TC, AC, MW, DV, TM
Chapter Six Ideal AAA Rupture Study
ABSTRACT
Background
Abdominal aortic aneurysms are a permanent and irreversible dilation of the infrarenal
aorta, which, if left untreated may rupture usually leading to death. Recently, the
effectiveness of the maximum diameter criterion for determining elective repair has
been questioned, with many believing that other parameters should be considered. This
paper aims to identify the rupture locations of idealised physical models of aneurysms
using an in vitro set-up, and compare the findings to those predicted numerically.
Methods
Dow Corning Sylgard 184 was mechanically characterised from tensile tests, tear tests
and finite element analysis. Five idealised abdominal aortic aneurysms were
manufactured using this silicone rubber and then inflated to the point of rupture.
Images were recorded using a high-speed camera. Numerical modelling attempted to
confirm these rupture locations. Regional variations in wall thickness of the silicone
models was also quantified and applied to numerical models.
Results
Four of the five models tested ruptured at inflection points in the proximal and distal
regions of the aneurysm sac, and not at regions of maximum diameter. These findings
agree with high stress regions computed numerically. Wall stress appears to be
independent of wall thickness with high stress occurring at regions of inflection
regardless of wall thickness variations.
Conclusions
According to the experimental and numerical findings presented here, abdominal aortic
aneurysms experience higher stresses at regions of inflection compared to regions of
maximum diameter. Ruptures of the idealised silicone models occurred predominantly
at the inflection points, as numerically predicted. Regions of inflection can be easily
identified from basic 3D reconstruction, and as ruptures appear to occur at inflection
points, these findings may provide a useful insight into the clinical significance of
inflection regions. This approach will be applied to patient-specific models in a future
study.
103
INTRODUCTION
Aneurysms form a significant portion of cardiovascular related deaths in the Western
world. Aneurysms are permanent and irreversible localised dilations1 and although they
can form in any blood vessel, artery or vein, the more serious aneurysms occur in the
abdominal aorta, the arteries of the brain, and the thoracic aorta. The majority of these
serious aneurysms develop in the infrarenal section of the abdominal aorta, and are
termed Abdominal Aortic Aneurysm (AAA). If left untreated, AAAs can eventually
rupture, yet these aneurysms are frequently asymptomatic. Due to the advancements in
medical imaging technology over recent years, larger numbers of AAAs are being
detected through computed tomography (CT) scanning and ultrasound, either as part of
screening programs for the elderly, or by routine scanning at medical check-ups. AAA
screening programs involve the use of ultrasound to detect any abnormalities in the
aorta, and also allow the maximum diameter to be measured. Approximately 500,000
AAAs are diagnosed worldwide each year2 resulting in 15,000 deaths per year in the
USA alone.3
Typically, an AAA is surgically repaired once shown to have reached or exceeded a

diameter of 5-5.5 cm. There have been reports that this threshold may lead to
inaccurate surgical-decision making as not only can smaller AAAs rupture4-6 but also
large AAAs can remain stable.6 It is known that by applying the definition of material
failure to AAAs, the AAA will rupture when the locally acting wall stress exceeds the
locally acting wall strength. Currently, much research is aimed at examining the wall
stress distributions within the AAA wall with a view to improving rupture prediction.7-16
Researchers have also examined methods to determine the strength of the AAA wall,
both invasively from excised tissue17-19 and non-invasively.20 While there is much
focus on attempting to numerically understand the mechanics of AAA rupture, only
limited work has focused on the development of experimental methods of determining
rupture potential.
Previous work by Morris et al.21 reported the use of the photoelastic method in
determining wall stress distributions in an idealised AAA case that used dimensions
taken from a comprehensive study.22 Morris et al.21 observed rupture of the photoelastic
model at the proximal and distal inflection points. This is the only publication in which
rupture of an experimental AAA model has been reported. Callanan et al.23 later
validated this photoelastic work numerically. Other researchers24 have also employed
104
an idealised AAA in experimental studies, this time using strain gauges attached to a
latex model to determine the stress in the AAA wall. The technique of creating rubber
models of AAAs has been reported previously by Doyle et al.25 which was adapted from
related techniques developed by O’Brien et al.26 The primary purpose of this study is to
identify the sites of rupture in idealised AAA models by performing in vitro rupture
studies and compare these findings with those predicted using numerical modelling.
Mechanical characterisation of the material used was also performed, along with the
examination of silicone model wall thickness.
METHODS
Material Characterisation and Validation
Material Selection
Dow Corning Sylgard 184 silicone was used as the material for this study. This silicone
has an ultimate tensile strength (UTS) of 7.1 MPa and tear strength of 2.6 N/mm,
according to the Dow Corning specification sheet. This silicone is an elastomeric
poly(dimethylsiloxane) (PDMS) kit, which is supplied in two parts. Sylgard 184 is
mixed in a 10:1 ratio of base:curing agent and is transparent in appearance. Although
silicone rubber and arterial tissue are not identical in their stress/strain behaviour,
silicone remains the most suitable analogue.
Sample Preparation
Aluminium moulds were designed and manufactured to be used with the injection-
moulding method to create the silicone tensile testing samples. The mould cavity for
the tensile test samples conforms to a Type 2 tensile test specimen as outlined in BS
ISO 37 standards, with a second set of moulds designed for creating trouser test samples
to be used for tear testing, as per BS ISO 34-1 standards. The dimensions of these
samples are shown in Figure 1. The samples prepared for tear strength testing are a
modification of those outlined in BS ISO 34-1, in that the dimensions differ somewhat.
Trouser test specimens are named due to their trouser-like appearance, with the “legs”
extended in opposite directions to determine the tear strength of the material. Figure 2
illustrates this test method. The protocol for preparing samples for testing is outlined in
Appendix I. Briefly, the moulds are thoroughly cleaned and bolted together, and the
necessary amounts of silicone are manually mixed. Degassing is required at this stage
as air bubbles can get trapped in the silicone during the mixing process. Once all air is
removed, the silicone is transferred from the mixing container to a syringe, where it is
105
then injected into the mould cavity. Further degassing in a vacuum is required after the
injection process. The mould is then placed into an oven at 45ºC for 18 hours to ensure
adequate curing of the samples. Once cured, the samples are removed from the mould
and left at room temperature for 3 hours, before further preparation for both tensile and
tear testing.
Figure 1: (A) Type 2 dumb-bell specimen from BS ISO 37. (B) Modified trouser test
piece from BS ISO 34-1, a indicates the direction of the cut. All dimensions are in
millimetres (mm) and the thickness of each specimen is 2 mm.
Uniaxial Tensile Testing

The procedure for tensile testing each specimen is described in Appendix II. The test
apparatus used was the Tinius Olsen H25KS (Surrey RH1 5DZ, England) with a 1 kN
load cell. For the tensile testing, an extension rate of 500 mm/min was applied, as
recommended in BS ISO 37 for Type 2 specimens. Preconditioning of each sample was
also performed. An initial stretch-relax program of ten cycles to an extension of 20%
was carried out in order for the stress-strain function of the material to become
stabilised and overcome the Mullin’s effect.27 The sample was then stretched to failure.
A sample size of 9 was used. It is recommended in BS ISO 37 that at least 3 samples be
used to determine material properties.
Tear Testing
The procedure for the tear test is outlined in Appendix III. As with the tensile test, the
Tinius Olsen test apparatus was used. An extension rate of 100 mm/min was applied to
each sample, as per BS ISO 34-1. 12 samples were used to determine tear strength.
Figure 2 shows the load directions that were applied to the trouser test specimen. This
test involves extension of the legs of the sample in opposite directions until the point of
complete material failure.
106
Figure 2: Illustration showing the loading directions that are applied to the trouser test
specimen using the tensile test machine.
Strain Energy Function

In order to mechanically characterise the material, the experimental data from the
tensile tests was converted to engineering stress and engineering strain, and then a 4th
order polynomial curve was fitted to the data to obtain a representative experimental
data curve. The polynomial curve applied to the experimental data was then used as the
material model for the finite element analysis (FEA). The commercial FEA solver
ABAQUS v6.7 (Dassault Systemes, SIMULIA, Rhode Island, USA) was used. An
identical dumb-bell model to that of the tensile tests was replicated numerically, with
the 4th order polynomial curve applied to represent the experimental data and used to
determine the most suitable strain energy function (SEF). Various SEF’s were
examined such as the Neo-Hookean, Ogden, Yeoh, and Mooney-Rivilin, and the most
applicable SEF determined. Material constants could then be determined for any
chosen SEF, thus allowing the material to be readily modelled in numerical studies.
Finite Element Analysis

By replicating the tensile test experiments using a numerical simulation, the accuracy of
the material coefficients could be examined. For this, the numerical model was
simulated in the same manner as that of the tensile tests, with a half-model analysed due
to symmetry. The model was constrained from all movement at one end, with a
displacement constraint identical to that of the tensile test applied to the opposite end.
Mesh independence was determined by examining the stress and strain experienced in a
particular node at the centre of the gauge length over the duration of the loading. This
was performed for mesh densities of 260 elements, 1752 elements, and 5000 elements.
A mesh of 1752 elements was deemed adequate as there was no significant difference in
results when the mesh density was increased to 5000 elements as shown in Figure 3.
107
Figure 3: Mesh independence results for Type 2 dumb-bell model.
Rupture Modelling
Experimental AAA Model
In order to study the rupture of the aneurysm in vitro, a previously reported21,23,26
idealised AAA model was used. This ideal AAA was developed using realistic
dimensions obtained from the EUROSTAR data registry.22 In brief, the ideal AAA has
a maximum internal diameter of 50 mm, a total length of 260 mm, a uniform wall
thickness of 2 mm and is geometrically symmetric. Experimental ideal AAA silicone
models were manufactured using the lost-wax process previously reported by our
group.25,26 Using this technique, male and female aluminium moulds are designed using
computer-aided design (CAD) and machined using a computer numerical control (CNC)
3-axis milling machine. The female mould allows the creation of a wax model of the
AAA which upon solidification is then placed into the outer male mould. This male
mould has a uniform wall cavity of 2 mm that facilitates the injection of silicone around
the wax model. Once the silicone rubber is cured, the complete model can be removed
and the wax melted out from the model. To ensure the optimum uniform wall thickness
of the resulting silicone rubber model, the female mould is heated to slow down the
solidification process of the wax model. This reduces the amount of shrinkage involved
with the cooling process, resulting in more accurate wax models. An accurate wax
model allows a uniform 2 mm cavity to surround the wax upon placement in the male
outer mould. Five AAA models to be used in the rupture study were manufactured
using this technique, with a further four models created to determine wall thickness.
108
Experimental Set-Up
The experimental test rig consisted of a series of mirrors, a pneumatic airline, a pressure
regulator, a pressure manometer, a high-speed camera and various clamps and silicone
tubing. The use of correctly positioned and angled mirrors allows 360º of the model to
be viewed, which is of paramount importance to the success of the experiments (Fig. 4).
A high-speed camera (Olympus i-Speed, Olympus Corporation) was used to capture the
point and location of rupture. This camera is capable of recording images at rates up to
33,000 frames per second (fps). A pixel resolution of 800 x 600 at 1000 fps was
deemed adequate for this application, and images were recorded using a monochrome
image sensor. Once the AAA model was attached to the test rig, the high-speed camera
was adjusted to ensure optimum focus and angle. After a satisfactory set-up was
established, the air pressure was increased and the camera set to record. Air pressure
was incrementally increased by 20 mmHg every 30 seconds until rupture occurred. The
pressure readings were also recorded in the video image, and so could be examined
post-rupture.
Figure 4: Experimental set-up of rupture test. (1) Retort stand, (2) plug, (3) clamp, (4)
silicone rubber AAA model, (5) plug, (6) pressure gauge, (7) mirror-walls, (8) pressure
regulator, (9) pneumatic air-line. Mirror walls allow 360º of AAA model to be
examined.
109
Measurement of Wall Thickness

In order to determine the variations in wall thickness throughout the AAA models, four
additional silicone rubber ideal AAA models were created. These models were
sectioned at 20 mm increments along the longitudinal distance of each model as shown
in Figure 5. This measurement technique is similar to that previously reported.25,26
Cross-sections were then measured for wall thickness at four 90º equidistant locations.
Results for each region were averaged across the four models so that regional variations
in thickness could be compared. Wall thickness measurements were then analysed
using SPSS 15 (SPSS Inc., Chicago, Ill, USA) to determine the 25th and 75th percentiles,
maximum and minimum wall thickness, and also the mean and standard deviation wall
thickness.
Figure 5: Cross-sections of example ideal AAA rubber model. Cross-sections (left)

correspond to slices shown on model (right). Slices were taken at 20 mm longitudinal
increments. Image shows relative uniformity of wall thickness throughout rubber model.
Numerical Modelling
To correlate experimental and numerical results, the evaluated material constants were
applied to the ideal AAA model and implemented in the FEA software. The pressure
loadings observed experimentally were then applied numerically. Boundary conditions
110
similar to the experimental set-up were implemented in the virtual AAA model, so that
the model was constrained from all movement at the proximal neck and iliac legs. Due
to the symmetrical nature of the model, only half the model was numerically analysed
using symmetry constraints to reduce computational time. Each model was meshed
using 3D stress elements with mesh independence performed as previously reported.13
This mesh independence test involves increasing the number of elements of the mesh
until the peak stress is <2% of that recorded with the previous mesh size. Initially, a
uniform 2 mm thick model was subjected to an internal pressure of 364.6 mmHg, which
is an average of the five rupture pressures observed during experiments. The model was
also tested using all five of the actual experimental rupture pressures. Peak von Mises
wall stress was analysed along with peak wall tension.19 In order to gauge the effect of
wall thickness on the resulting wall stress, AAA geometries were created using varying
wall thickness corresponding to the 25th and 75th percentiles, maximum and minimum
wall thickness, mean wall thickness, a thick walled inflection region, and a thin walled
inflection region. This required the creation of an additional seven numerical models.
Each of the models was subjected to an internal pressure of 120 mmHg which is
commonly regarded as the average peak systolic pressure of the cardiac cycle. By
applying a common internal pressure the stress on the AAA wall could be compared
across each model. An example numerical model can be seen in Figure 6 which shows
the finite element mesh and boundary conditions used in the study.
Figure 6: Representative numerical ideal AAA, uniform wall. (A) The finite element
mesh and (B) loading conditions and constraints.
111
RESULTS
Material Characterisation and Validation
Uniaxial Tensile Test
Tensile testing revealed that the average UTS ± standard deviation of Sylgard 184 is
7.7361 ± 1.6597 MPa (range = 4.74 – 9.39 MPa), compared to 7.1 MPa on the Dow
Corning specification sheet. The stress/strain curves for each sample, along with the 4th
order polynomial curve fit is shown in Figure 7.
Figure 7: Engineering stress and engineering strain experimental data (black lines) with
4th order polynomial curve applied to the results (red line). Engineering stress is defined
as σE=F/A, where σE is engineering stress, F is force, and A is cross-sectional area.
Engineering strain is defined as εE=∆l/l, where ε is engineering strain, ∆l is the change
in gauge length, and l is the original gauge length.
Tear Test
Tear testing showed that the average tear strength ± standard deviation of the material is
0.419 ± 0.094 N/mm. Dow Corning report a tear strength (TS) of 2.6 N/mm on the
specification sheet, but it is known that results can depend heavily on the type of tear
test sample used,28 therefore it is difficult to compare the results found in this study with
those of Dow Corning. For this study the average tear strength of 0.419 N/mm was
taken as the maximum wall tension the material can resist prior to rupture.
Strain Energy Function

For this particular material, the 3rd order Ogden SEF proved to be the most suitable
constitutive equation as not only did it provide a good curve fit to the data (R2=0.9812),
112
but was also stable at all stresses and strains. The general form of the Ogden SEF is
shown (Eqn.1), with the resulting material coefficients presented (Table 1).
N 2μ p
w (λ1, λ2 , λ3 ) = ∑
α αp αp
(λ1 p + λ2 + λ3 − 3) (1)
p =1 αp 2
Where W is the strain energy density per undeformed unit volume, (λ1, λ2, λ3) are the
principal stretch ratios, α is a strain hardening exponent, and μ has the interpretation of
the shear modulus under infinitesimal straining.
Table 1: Material coefficients determined for the 3rd order Ogden model for Sylgard
184.
μ α
1 -304.235 1.2667
2 148.232 1.5962
3 157.156 0.9075

Converting the engineering stress and strain presented in Figure 7 to true stress and
strain allowed the numerical and experimental methods to be compared. The results of
this comparison are shown in Figure 8 and agree favourably. Therefore, confidence was
established in the material characterisation of Sylgard 184.
Figure 8: Comparison of results for the experimental tensile tests and the numerical
simulation of Sylgard 184. True stress is defined as σT= σE*(1+εE), where σT is true
stress, σE is engineering stress, and εE is the engineering strain. True strain is defined as
εT=ln(1+εE), where εT is true strain and εE is the engineering strain.
113
Rupture Modelling
Experimental Rupture Tests
Of the five silicone models ruptured in vitro, four models experienced rupture at a
region of inflection on the surface of the model. An inflection point is defined as a
point on the AAA surface at which the local AAA wall shape changes from concave
outward to concave inward.29 This finding is consistent with previous numerical and
experimental reports by our group13,21,23 and others29 that noted peak stresses occur at
these regions instead of at the maximum diameter of the AAA. A summary of the
rupture locations along with corresponding burst pressures is shown (Table 2). One
silicone model ruptured at the iliac bifurcation. The sequence of events leading to
rupture for Test 1 can be seen in Figure 9. This illustration shows the frame where the
material failure initiates, leading to rupture of the model, and ultimately complete
failure of the silicone model.
Table 2: Summary of experimental rupture results.

Test Rupture Location Rupture Pressure (mmHg)
1 Proximal Inflection Point 254.7
4 Distal Inflection Point 278.7
5 Iliac Bifurcation 544.6
Figure 9: Sequence of events leading to model rupture of Test 1. (A) The inflated model,
(B) the initial point of rupture, (C) propagation of the failure zone, and (D) complete
failure of the silicone model. A similar sequence was observed for all four models that
ruptured at regions of inflection.
114
Wall Thickness Results

By slicing the four ideal AAA models into cross-sections, wall thickness could be
assessed. These four models were made using the same process as the previous AAA
models. The measured wall thickness for all four models is shown in Figure 10, with
the average regional wall thickness shown in Table 3. The locations in Table 3
correspond to the locations in Figure 10. This table also shows the wall thickness used
to create numerical models. The minimum wall thickness recorded in any model was
1.17 mm and was located at the iliac bifurcation region. The maximum wall thickness
was 2.53 mm and was also found at the iliac bifurcation region. The effect of wall
thickness on stress distributions within the AAA sac was examined and is shown in
Figure 11. The numerical models used to generate the results presented in Figure 11
were created using the data in Table 3. The region of maximum diameter is located at a
distance of 40 mm. The uniform 2 mm wall model, with the exception of the minimum
wall thickness model, appears to represent the overall behaviour of the AAA model
regardless of variations in wall thickness. This study highlighted that the inflection
region, in particular the proximal inflection region, experiences higher stresses than the
region of maximum diameter.
Figure 10: Box and whisker plot showing wall thickness categorised by AAA region.
The boxes indicate the median (centreline of box), 25th and 75th percentiles of the group,
and the whiskers show maximum and minimum measurements. Shown also are the
regions on the AAA model for each specific area.
115
Table 3: Regional variations in wall thickness. Thickness values were applied to

numerical models to determine the effect of wall thickness on stress results. AAA
regions correspond to Figure 10. All results are presented in millimetres (mm).
AAA Region
Proximal Max Distal
Model Neck Bifurcation
Inflection Diameter Inflection
Uniform 2 mm 2 2 2 2 2
25th Percentile 1.94 1.89 1.63 1.7 1.55
75th Percentile 2.16 2.29 1.87 2.18 2.17
Min Thickness 1.72 1.65 1.34 1.26 1.17
Max Thickness 2.49 2.51 2.04 2.44 2.53
Mean Thickness 2.03 2.08 1.78 1.91 1.87
Median Thickness 2.005 2.075 1.8 1.915 1.905
Thick Inflection 2 2.5 1.5 2.5 2
Thin Inflection 2 1.5 2.5 1.5 2
Figure 11: Effect of varying wall thickness on wall stress results at an internal pressure
of 120 mmHg. X-axis refers to the longitudinal distance from above the AAA sac to
immediately above the iliac bifurcation. Legend refers to numerical models created
using the corresponding data of Table 3.
Numerical Modelling
Stress distributions on the surfaces of the virtual AAA model reveal that high stresses
occur at the regions of inflection and not at regions of maximum diameter. This has
been previously shown by our group for this particular idealised AAA model both
numerically using FEA,23 and experimentally using the photoelastic method.21 The
resulting von Mises stress distribution on the surface of the ideal AAA at a mean
116
experimental pressure of 364.5 mmHg, along with the image captured at the point of
rupture, is shown (Fig. 12). Applying the same mean internal experimental pressure to
the FEA model (364.5 mmHg) returned a peak stress of 1.16 MPa located at the iliac
bifurcation. Also shown (Fig. 12) is the qualitative agreement in stress distributions at
this pressure loading with the rupture location found experimentally. Inflection zones
experience higher stress than regions of maximum diameter. Wall stress results can also
be expressed as wall tension results, as the failure tension of the wall is believed to be a
better indicator of a regions susceptibility to failure,19 with peak wall tension = peak
wall stress x wall thickness. The numerical results from the five ranges of pressures
applied during the experimental tests, resulted in varied wall stresses and wall tensions
(Tables 4 and 5).
Figure 12: In vitro rupture locations and FEA stress patterns. High stress occurs at
inflection points which correlate with rupture locations in experimental models. The
undeformed image shows stress contours displayed on the AAA model prior to loading,
with the deformed illustration showing the stress contours displayed on the model when
an internal loading of 364.5 mmHg is applied. There is a noticeable change in geometry
with increased internal pressure, particularly at the inflection regions.
117
Table 4: Summary of FEA wall stress results. Results are also shown as percentages of
the UTS of the material as determined from tensile testing (UTS = 7.7 MPa).
Rupture Pressure َ σmax σinflection َ σmax σinflection
Test (mmHg) (MPa) (MPa) % of UTS % of UTS
1 254.7 0.6808 0.4156 9% 5%
2 278.6 0.7375 0.4586 10% 6%
3 466.2 1.337 0.8668 17% 11%
4 278.7 0.7375 0.4586 10% 6%
5 544.6 1.733 1.1033 23% 14%
σ = stress.
Table 5: Summary of FEA wall tension results. Results are also shown as percentages of
the TS of the material as determined from tear testing (TS = 0.419 MPa).
Rupture Pressure َ Ωmax Ωinflection َ Ωmax Ωinflection
Test (mmHg) (N/mm) (N/mm) % of TS % of TS
1 254.7 1.3616 0.8312 325% 198%
2 278.6 1.475 0.9172 352% 219%
3 466.2 2.674 1.7336 638% 414%
4 278.7 1.475 0.9172 352% 219%
5 544.6 3.466 2.2066 827% 527%
Ω = wall tension
DISCUSSION
The focus of this paper was to experimentally rupture idealised silicone models of AAA
and compare these locations to numerical predictions. There has been little reported on
the in vitro rupture of abdominal aortic aneurysms, with much focus on the
computational analysis of these aneurysms.7-16 Morris et al.21 observed rupture
locations at the inflection regions during the use of the photoelastic method. These
rupture locations correlate with the sites of rupture experienced in this study. Nine
models were manufactured using the injection-moulding technique25,26 of which five
were experimentally ruptured and four used to assess wall thickness. The commercially
available silicone rubber Sylgard 184 from Dow Corning was used to represent the
AAA wall. Although silicone rubber does not behave identically to arterial tissue when
subjected to large strains, it is currently the most suitable analogue and has been widely
used in previous studies.25,26,30-32 Until a more suitable material is available, one that
possibly incorporates the layered structure of the arterial wall and the presence of a
fibrinous network (collagen and elastin), silicone rubber will remain the most feasible
arterial mimic.
118
Material Characterisation
The silicone was mechanically characterised in order to determine the most suitable
strain-energy function. Tear testing resulted in maximum tear strength (mean ±
standard deviation) of 0.419 ± 0.094 N/mm, indicating that wall tensions above this
value may cause rupture of the material. It was observed that the tear in each sample
was smooth, possibly indicating that the bond within the PDMS chains is not strong
enough to prevent rupture under an applied force.32 Tensile testing revealed that the
UTS of the material (mean ± standard deviation) is 7.7361 ± 1.6597 MPa (range = 4.74
– 9.39 MPa). This correlates well with the UTS of 7.1 MPa stated by Dow Corning on
the specification sheet. By fitting a 4th order polynomial curve to the experimental data
(Fig. 7) and evaluating the material using ABAQUS v6.7, material coefficients were
determined. An exact replica of the Type 2 dumb-bell sample used for the tensile tests
was then created numerically. Identical boundary conditions to those applied
experimentally were implemented, with mesh independence performed. Good
agreement was observed between the results of the experimental and numerical tests
(Fig. 8). The optimum strain-energy function for this particular material is a 3rd order
Ogden model and allows the material to remain stable at all stresses and strains within
the ranges presented. As the SEF is stable up to a strain of 2, whereas the maximum
strain experienced in the FEA ideal AAA models does not exceed 0.5, the 3rd order
Ogden SEF accurately describes this application. For the purpose of this paper, we are
interested in rupture regions of the models, and not entirely with the numerical
quantities.
Rupture Modelling
The use of a high-speed camera to record the point of rupture proved to be a very
powerful experimental tool. A frame-rate of 1000 fps captured the exact point of
rupture without sacrificing picture detail. When concerned with high-speed video
recording, the higher the recording frame rate, the lower the resulting picture quality.
As image quality is of paramount concern with this work, determining the optimum
resolution and frame rate was crucial to the success of the testing. The mirror-wall
arrangement (Fig. 4) also proved to be very effective as this set-up allowed all 360º of
the model to be viewed. Upon pressurisation, it was observed that the AAA models did
not inflate uniformly, resulting in asymmetry. This may be due to localised variations
in wall thickness. However, it was shown that variations in wall thickness do not
significantly affect the overall stress distributions. Rupture pressures varied
119
significantly for two of the five models (Tests 3 and 5), as shown (Table 2). Rupture
pressure (mean ± standard deviation) was 364.56 ± 131.89 mmHg. Three models
experienced rupture at the proximal inflection regions, one model at the distal inflection
region, and one model ruptured at the iliac bifurcation. It is known that in a clinical
setting real AAAs rarely experience rupture at the iliac bifurcation, with AAAs
predominately rupturing at the posterior wall.5 Test 5 in this study may therefore have
ruptured at the iliac bifurcation region due to a localised flaw in the model. Elevated
stresses at regions of inflection have been previously reported in experimental models,
both using the photoelastic method21 and using rosette strain gauge arrangements.24
Flora et al.24 observed stresses at the inflection region nearly twice those experienced at
the region of maximum diameter.
Wall Thickness
The results recorded in the wall thickness study indicate that although slight deviations
in the wall thickness exist, the mean ± standard deviation thickness is 1.954 ± 0.266 mm
which is comparable to the 2 mm original design of the aluminium moulds. Therefore it
may be acceptable to use a uniform 2 mm wall to numerically represent the ideal AAA.
The effect of local changes in wall thickness was assessed using a statistical method to
determine the 25th and 75th percentiles of the wall thickness. This allowed numerical
AAA models to be created and analysed that had realistic wall thickness variations. The
maximum and minimum wall thickness values were also used to create “worst case
scenario” AAA models with quite large variations in wall thickness. Also, two models
were generated that had extreme changes in wall thickness (referred to as Thick and
Thin Inflection in Table 3). Computations of the wall stress revealed that the inflection
regions, particularly the proximal inflection zone, experiences higher stress than the
region of maximum diameter, regardless of wall thickness variations (Fig. 11).
Numerical Modelling
Initial results using a mean internal pressure of 364.6 mmHg resulted in a peak wall
stress of 1.16 MPa located immediately prior to the iliac bifurcation. This wall stress is
much lower than the UTS of the material (7.7 MPa), and therefore one would not expect
rupture of the model. Peak stresses at the iliac bifurcation are common in numerical
computations of AAA wall stress. It is known that stress artefacts arise due to the
inclusion of this bifurcation and a method of evaluating and refining this localised stress
region is yet to be developed.33 The peak stress obtained numerically was on average
120
only 13.8% (range 9-23%, Table 4) of the materials’ UTS, with the stress at the
inflection regions only 8.4% (range 5-14%, Table 4) of the UTS. It is also important to
note that Dow Corning state on the specification sheet that UTS values can have a very
large range of results. The stated UTS of 7.1 MPa can actually be as low as 3.5 MPa
according to Dow Corning.
By comparing the wall tension calculated numerically in the experimental models to the
tear strength of the material, one can see that the resulting wall tension exceeds the tear
strength of the material (Table 5). These results may suggest that the model has
experienced rupture in all cases examined, if indeed tear strength is to be considered a
possible indicator of material failure strength. At elevated pressures, silicone AAA
models may rupture due to the pressure applied to the inner surface causing an increase
in wall tension. This results in the propagation of a nick or tear, and ultimately,
complete failure of the silicone wall (Fig. 9). In realistic situations, calcified deposits
embedded in the AAA wall may act as stress raisers, therefore acting as rupture
initiators. Ruptures may occur at these areas, in particular if calcifications occur at
regions of inflection. Analogue models to replicate calcified deposits can be included in
future rubber models using available techniques at our laboratory. Good agreement was
observed when comparing the experimental rupture locations with the high stress
regions found using FEA (Fig. 12). It is believed that the geometry of the idealised
AAA model may play a role in rupture behaviour. The 3D nature of the model, along
with slight deviations in wall thickness25,26 may increase the likelihood of rupture.
Another interesting point to note is that when the experimental model is inflated, the
radius of curvature of the inflection region changes, whilst the numerical model does
not experience such a significant change. This may also contribute to the difference in
the wall stress results. Also, it is known that no silicone model is absolutely free from
defects and imperfections. The manufacturing process results in a silicone model that
may contain microscopic flaws. Tensile testing determined the UTS of the material
using flat, planar dog-bone shapes, whereas, the AAA model has regions of inflection,
along with a bifurcation branching into the iliac arteries. These factors may all
contribute to the propensity of the silicone model to rupture, and hence the FEA results
would differ somewhat. It has also been reported31 using the Law of Laplace, that
actual AAAs can withstand up to 982 mmHg before rupture with much variability
between intrasac rupture pressures.34
121
Work has begun by our group on the rupture of realistic AAA models and the
correlation with both FEA-predicted and photoelastic-predicted high stress regions.
Also, our group has developed a successful method of creating silicone AAA models
complete with a region of intraluminal thrombus (ILT) and representative calcified
deposits. This thrombus has been shown to reduce wall stress in numerical
models,13,15,35 but has also shown to have a negligible effect on wall pressures in vivo,36
hence there is also a need to examine the role of ILT on in vitro rupture tests.
Calcifications can affect numerically computed wall stress and also decrease the
biomechanical stability of the AAA,37,38 and therefore should be included in future
work.
Limitations
This research is not without limitations. A more suitable AAA wall analogue could be
used. Use of silicones with lower tensile strengths would allow lower, more realistic,
pressures to cause rupture. Work has begun in our group on the use of more suitable
arterial vessel analogues.39 Also, the use of a realistic cardiac pressure wave instead of
static air pressure may alter the results. These factors are to be examined in future
work.
CONCLUSIONS
According to the experimental and numerical findings presented here, abdominal aortic
aneurysms experience higher stresses at regions of inflection compared to regions of
maximum diameter. Rupture of the idealised silicone models occurred predominantly at
the inflection points, as numerically predicted. The use of a high-speed camera is a
useful experimental tool in observing rubber AAA rupture locations. To improve the
method described, more suitable arterial analogues that mimic arterial properties more
closely are required, thus possibly leading to an improved understanding of AAA
rupture. Regions of inflection can be easily identified from basic 3D reconstruction,
and as ruptures appear to occur at inflection points, these findings, along with future
work, may provide a useful insight into the clinical significance of inflection regions.
122
APPENDIX I
Sample Preparation
1. Clean moulds thoroughly using acetone.
2. Bolt moulds tightly together and seal edges to prevent leaks.
3. Weigh appropriate amounts of components depending on the quantity of silicone
material desired, i.e. silicone and catalyst. Silicone:catalyst ratio is 10:1.
4. Combine components in mixing container and thoroughly mix for 3 minutes.
5. Degas mixture at a pressure of 100 mBar for 15 minutes.
6. Remove from vacuum and suck mixture into a 60 ml syringe.
7. Slowly inject mixture into mould, ensuring no leaks.
8. Place mould into vacuum to further degas for 15 minutes at a pressure of 350
mBar.
9. Remove and cure in oven for 18 hours at 45˚C.
10. Upon curing, remove samples from mould and leave at room temperature for 3
hours (BS ISO 37).
APPENDIX II
Tensile Testing
1. Set up tensile test machine apparatus as recommended by manufacturer.
2. Measure thickness and width of each sample before testing.
3. Place specimen in clamps and ensure proper location, i.e. specimen is clamped
in the same position each time.
4. Set up software with pre-determined loading and failure criteria.
a. Dimensions = Input thickness and width of sample (mm)
b. Extension rate = 500 mm/min (BS ISO 37)
c. Preconditioning = 10 cycles at 20% extension
d. End criteria = Stretch sample to failure
5. On failure of sample, machine resets to pre-determined setting.
6. Software automatically generates UTS (MPa) and other user-defined variables.
7. Record data such as UTS and repeat test for next sample.
APPENDIX III
Tear Testing
1. Set up tensile machine apparatus as recommended by manufacturer.
2. Measure thickness of each sample before testing.
123
3. Place specimen in clamps and ensure proper location, i.e. specimen is clamped
in the same position each time.
4. Set up software with pre-determined loading and failure criteria.
a. Dimensions = Input thickness of sample (mm)
b. Extension rate = 100 mm/min (BS ISO 34-1)
5. On failure of sample, machine resets to pre-determined setting.
6. Software automatically generates tear strength (N/mm) and other user-defined
variables.
7. Record data and repeat test for next sample.
ACKNOWLEDGEMENTS
US National Heart Lung and Blood Institute (iii) Dr. Liam Morris from the Galway
Medical Technology Centre, GMIT, Ireland (iv) Kevin O’Flanagan from the
Department of Manufacturing and Operations Engineering, University of Limerick,
Ireland, for his assistance with the high-speed imaging (v) Maria Ryan for her help with
the experimental rupture modelling (vi) the Department of Vascular Surgery in the
Midwestern Regional Hospital, Ireland, in particular, Mr. Eamon Kavanagh, Mr. Paul
Burke and Prof. Pierce Grace and (vii) Michel S. Makaroun, MD, Department of
Surgery, University of Pittsburgh, USA.
REFERENCES
1. Sakalihasan N, Limet R, Defawe OD. Abdominal aortic aneurysm. Lancet.
2005;365(9470):1577-89.
2. Vande Geest JP, Sacks MS, Vorp DA. A planar biaxial constitutive relation for
the luminal layer of intra-luminal thrombus in abdominal aortic aneurysms. J
Biomech. 2006;39:2347-2354.
3. Kleinstreuer C, Li Z. Analysis and computer program for rupture-risk prediction
of abdominal aortic aneurysms. Biomed Eng Online. 2006;5:19.
4. Nicholls SC, Gardner JB, Meissner MH et al. Rupture in small abdominal aortic
aneurysms. J Vasc Surg. 1998;28:884-888.
5. Darling RC, Messina CR, Brewster DC, et al. Autopsy study of unoperated
abdominal aortic aneurysms. The case for early resection. Circ. 1977;56(II):161-
164.
6. Vorp DA. Biomechanics of abdominal aortic aneurysm. J Biomech.
2007;40:1887-1902.
7. Raghavan ML, Vorp DA, Federle MP, et al. Wall stress distribution on three-
dimensionally reconstructed models of human abdominal aortic aneurysm. J
Vasc Surg. 2000;31:760-769.
124
8. Venkatasubramaniam AK, Fagan MJ, Mehta T, et al. A comparative study of

aortic wall stress using finite element analysis for ruptured and non-ruptured
abdominal aortic aneurysms. Eur J Vasc Endovasc Surg. 2004;28:168-176.
9. Fillinger MF, Marra SP, Raghavan ML, et al. Prediction of rupture risk in
abdominal aortic aneurysm during observation: wall stress versus diameter. J
Vasc Surg. 2003;37:724-732.
10. Fillinger MF, Raghavan ML, Marra SP, et al. In vivo analysis of mechanical
wall stress and abdominal aortic aneurysm rupture risk. J Vasc Surg.
2002;36:589-597.
11. Truijers M, Pol JA, SchultzeKool LJ, et al. Wall stress analysis in small
asymptomatic, symptomatic and ruptured abdominal aortic aneurysms. Eur J
Vasc Endovasc Surg. 2007;33:401-407.
12. Mower WR, Quinones WJ, Gambhir SS. Effect of intraluminal thrombus on
abdominal aortic aneurysm wall stress. J Vasc Surg. 1997;26:602-608.
13. Doyle BJ, Callanan A, McGloughlin TM. A comparison of modelling
techniques for computing wall stress in abdominal aortic aneurysms. Biomed
Eng Online. 2007;6:38.
14. Rodriguez JF, Ruiz C, Doblare M, et al. Mechanical stresses in abdominal aortic
aneurysms: Influence of diameter, asymmetry, and material anisotropy. J
Biomech Eng. 2008;130:021023-1-10.
15. Doyle BJ, Callanan A, Walsh MT, et al. A finite element analysis rupture index
(FEARI) as an additional tool for abdominal aortic aneurysm burst prediction.
Vasc Dis Prev. 2009;6:75-82.
16. Doyle BJ, Callanan A, Burke PE, et al. Vessel asymmetry as an additional
diagnostic tool in the assessment of abdominal aortic aneurysms. J Vasc Surg.
2009;49:443-54.
17. Thubrikar MJ, Labrosse M, Robicsek F, et al. Mechanical properties of
abdominal aortic aneurysm wall. J Med Eng Tech. 2001;25(4):133-142.
18. Raghavan ML, Webster MW, Vorp DA. Ex vivo biomechanical behaviour of
abdominal aortic aneurysm: assessment using a new mathematical model. Ann
Biomed Eng. 1996;24:573-582.
19. Raghavan ML, Kratzberg J, de Tolosa EMC, et al. Regional distribution of wall
thickness and failure properties of human abdominal aortic aneurysm. J
Biomech. 2006;39(16):3010-6.
20. Vande Geest JP, Wang DHJ, Wisniewski SR, et al. Towards a non-invasive
method for determination of patient-specific wall strength distribution in
abdominal aortic aneurysms. Ann Biomed Eng. 2006;34:7:1098-1106.
21. Morris L, O’Donnell P, Delassus P, et al. Experimental assessment of stress
patterns in abdominal aortic aneurysms using the photoelastic method. Strain.
2005;40:165-172.
22. Laheij R, van Marrewijk C, Buth J. Progress report on the procedural and follow
up results of 3413 patients who received stent graft treatment for infrarenal
aortic aneurysms for a period of 6 years. EUROSTAR Data Registry Centre.
2001.
23. Callanan A, Morris LG, McGloughlin TM. Numerical and experimental analysis
of an idealised abdominal aortic aneurysm. European Society of Biomechanics
2004, S-Hertogenbosch, Netherlands.
24. Flora HS, Talie-Faz B, Ansdell L, et al. Aneurysm wall stress and tendency to
rupture are features of physical wall properties: An experimental study. J
Endovasc Ther. 2002;9:665-675.
125
25. Doyle BJ, Morris LG, Callanan A, et al. 3D reconstruction and manufacture of
real abdominal aortic aneurysms: From CT scan to silicone model. J Biomech
Eng. 2008;130:034501-1-5.
26. O’Brien T, Morris L, O’Donnell M, et al. Injection-Moulded Models of Major
and Minor Arteries: The Variability of Model Wall Thickness Owing to Casting
Technique. Proceedings of the International Mechanical Engineers, Part H:
Journal of Engineering in Medicine. 2005;219.
27. Mullins L. Softening of rubber by deformation. Rubber Chem Tech.
1969;42:339-362.
28. Brown R. Physical testing of rubber, 4th Ed. 2006. Springer
Science+Business+Media, Inc., NY, USA, pp. 159-167.
29. Vorp DA, Raghavan ML, Webster MW. Mechanical wall stress in abdominal
aortic aneurysm: influence of diameter and asymmetry. J Vasc Surg.
1998;27(4):632-639.
30. Marins PALS, Natal Jorge RM, Ferreira AJM. A Comparative Study of Several
Material Models for Prediction of Hyperelastic Properties: Application to
Silicone-Rubber and Soft Tissues. Strain. 2006;42:135-147.
31. Shergold OA, Fleck NA, Radford D. The uniaxial stress versus strain response
of pig skin and silicone rubber at low and high strain rates. Int J Imp Eng.
2006;32:1384-1402.
32. Aziz T, Waters M, Jagger R. Analysis of the properties of silicone rubber
maxillofacial prosthetic materials. J Dent. 2003;31:67-74.
33. Fillnger M. The long-term relationship of wall stress to the natural history of
abdominal aortic aneurysms (finite element analysis and other methods). Ann NY
Acad Sci. 2006;1085:22-28.
34. Vallabhaneni SR, Gilling-Smith GL, How TV, et al. Heterogeneity of tensile
strength and matrix metalloproteinase activity in the wall of abdominal aortic
aneurysms. J Endovasc Ther. 2004;11:494-502.
35. Wang DHJ, Makaroun MS, Webster MW, et al. Effect of intraluminal thrombus
on wall stress in patient-specific models of abdominal aortic aneurysm. J Vasc
Surg. 2002;36:598-604.
36. Schurink GW, van Baalen JM, Visser MJ, et al. Thrombus within an aortic
aneurysm does not reduce pressure on the aneurysmal wall. J Vasc Surg.
2000;31:501-506.
37. Speelman L, Bohra A, Bosboom EMH, et al. Effects of wall calcifications in
patient-specific wall stress analyses of abdominal aortic aneurysms. J Biomech
Eng. 2007;129:105-109.
38. Li Z-Y, U-King-Im J, Tang TY, et al. Impact of calcification and intraluminal
thrombus on the computed wall stresses of abdominal aortic aneurysm. J Vasc
Surg. 2008;47:928-935.
39. Doyle BJ, Corbett TJ, O’Donnell MR, et al. Design and development of a range
of silicone elastomers for use in experimental studies. Proceedings of the
Fifteenth Annual of the Section of Bioengineering of the Royal Academy of
Medicine in Ireland. 2009;112.
126
Chapter VII
Experimental modelling of aortic aneurysms: Novel
applications of silicone rubbers.
Barry J. Doyle,1 Timothy J. Corbett,1 Michael R. O’Donnell,1 Aidan

J. Cloonan,1 Michael T. Walsh,1 David A. Vorp2 and Timothy M.
McGloughlin1
Limerick, Ireland.

USA.
Medical Engineering & Physics (submitted)

Analysis and interpretation: BD, TC, AC, MOD
Critical revision of article: BD, MW, DV, TM
Chapter Seven Novel Application of Silicone Rubber
ABSTRACT
Background
A range of silicone rubbers were created based on existing commercially available
materials. These silicones were designed to be visually different from one another and
have distinct material properties, in particular, ultimate tensile strengths and tear
strengths.
Methods
In total, eleven silicone rubbers were manufactured, with the materials designed to have
a range of increasing tensile strengths from approximately 2-4 MPa, and increasing tear
strengths from approximately 0.45-0.7 N/mm. The variations in silicones were detected
using a standard colour analysis technique. Calibration curves were then created relating
colour intensity to individual material properties. All eleven materials were
characterised and a 1st order Ogden strain energy function applied. Material coefficients
were determined and examined for effectiveness. Six idealised abdominal aortic
aneurysm models were also created using the two base materials of the study, with a
further model created using a new mixing technique to create a rubber model with
randomly assigned material properties (seven models in total). These models were then
examined using videoextensometry and compared to numerical results.
Results
Colour analysis revealed a statistically significant linear relationship (P<0.0009) with
both tensile strength (R2=0.8082) and tear strength (R2=0.7205), allowing material
strength to be determined using a non-destructive experimental technique. The
effectiveness of this technique was assessed by comparing predicted material properties
to experimentally measured methods, with good agreement in the results.
Videoextensometry and numerical modelling revealed minor percentage differences,
with all results achieving significance (P<0.0009).
Conclusions
This study has successfully designed and developed a range of silicone rubbers that
have unique colour intensities and material strengths. Strengths can be readily
determined using a non-destructive analysis technique with proven effectiveness. These
silicones may further aid towards an improved understanding of the biomechanical
behaviour of aneurysms using experimental techniques.
127
INTRODUCTION
Silicone is a synthetic polymer consisting of a linear backbone of repeating, alternating
silicone and oxygen atoms. Each silicone atom has two groups attached to it, referred to
as R groups, representing any organic group that may be attached to the backbone. This
structure forms a polymer called polydimethylsiloxane (PDMS), and is the most
commonly used silicone. Rubber-like materials are comprised of very long-polymeric
chains united by vulcanisation into a network structure. These rubbers can therefore
undergo large recoverable deformations, hence the wide use of silicone rubbers as
material analogues in the study of arterial vessels1-4 and other soft tissues.5 At high
strain rates, similar to those naturally found within the cardiac cycle, polymeric chain
deformation is usually restricted to bending and stretching of the chemical bonds within
the network. As a result the storage modulus of the rubber can increase by up to three
orders of magnitude.5 Abdominal aortic aneurysms (AAA) are permanent irreversible
dilations of the infrarenal section of the aorta, and will eventually expanded to the point
of rupture if left untreated. Many previous studies have focused on the numerical
prediction of wall stresses and ultimately rupture prediction of AAAs.6-16 However,
there has been limited reports as to how these aneurysms behave experimentally,2,17,18 in
particular, how these models react to increased pressure loadings above the average
peak systolic pressure of 120 mmHg. Also, previous experimental work has employed
the use of a single material to represent the AAA wall, even though it is known that a
realistic AAA may have differing materials properties at various locations throughout
the aneurysm.19
The primary aim of this study was to develop a range of silicones of known colour and
known material properties. Commercially available silicone was used in conjunction
with a method of colour analysis in order to develop calibration curves. These
calibration curves consist of a direct relationship between material colour and both
tensile strength and tear strength. Material characterisation was determined from
uniaxial tensile tests and also from tear strength tests. These rubbers were then used to
form the arterial wall analogue for experimental testing of idealised abdominal aortic
aneurysms. These rubber models can be created using previously reported techniques1,4
for use with videoextensometry. These novel materials could be used to create more
physiologically realistic in vitro arterial models. The use of a combination of silicones
to create a diseased vessel wall could serve as a useful tool in future experimental work.
128
In particular, these materials could be incorporated into experimental rupture studies to

provide more accurate material analogues than those used in previous reports.2

Material Selection
The commercially available Sylgard silicone from Dow Corning was chosen as the base
material for this study, in particular, Sylgard 160 and Sylgard 170. Both Sylgards are
supplied as a two-part silicone elastomer with Sylgard 160 appearing grey and Sylgard
170 appearing black. These two rubbers are prepared in a 50:50 by weight arrangement,
which facilitates mixing and preparation. These silicones were identified as appropriate
materials as each material is easily identifiable due to its colour, and importantly, they
have dissimilar material properties.
Material Development
Sylgard 160 is naturally grey in appearance with an ultimate tensile strength (UTS) of 4
MPa, whereas, Sylgard 170 is naturally black in colour with a UTS value of 2 MPa.
These UTS values were obtained from the Dow Corning specification sheets. These
two materials were mixed together in various ratios in order to create a range of new
silicones, with gradually increasing colour intensity from grey to black and gradually
decreasing failure properties from 4-2 MPa. The ratios of each mix were increased by
10% for each new silicone, resulting in 11 complete materials, including the original
Sylgard 160 and 170, as shown, for example, in Column I of Table 1.
Colour Analysis
The colour intensity of each silicone was analysed using a ColorLite sph850
Spectrophotometer (ColorLite GmbH). This device allows each silicone mix to be
assigned an individual colour intensity value. Colour measurements are given in as a
variation of ∆E, where pure black has a ∆E value of zero. This mathematical model for
colour measurement was developed by the Commission International de l'Eclairage
(CIE) and is often referred to as the CIELAB formula. The numerical value (ΔE) can
be used to differentiate between colours and is given by Equation 1.
ΔE = [ ΔL2 + Δa 2 + Δb 2 ]2
1
(1)
129
This equation is based on L*a*b* values where the lightness value, or luminosity, (ΔL)
indicates how light or dark the colour is, Δa represents the position on the red-green
axis, and Δb shows the position on the yellow - blue axis (Δb). L*a*b* values are
calculated from the tristimulus values (X,Y,Z) which are the backbone of all colour
mathematical models. The location of a colour is defined by a three dimensional
Cartesian coordinate system which determines the numerical values of L*, a* and b*,
and is shown in Figure 1(A). Once the location of L*a*b* is determined, a modification
of the CIELAB tolerancing is performed, known as CMC tolerancing (Colour
Measurement Committee of the society of Dyers and Colourists of Great Britain). This
technique mathematically defines an ellipsoid around the standard colour with semi-axis
corresponding to hue, chroma, and lightness. Figure 1(B) shows the variation of the
ellipse sizes throughout the L*a*b* colour space. In order to actually determine the ΔE
value of the particular material, firstly the device is calibrated using a standard colour,
in this case the standard reference colour was black. The device is then placed against
the material to be tested and allowed to operate. Average colour intensity was recorded
from 3 measurements taken for each sample.
Figure 1: (A) The CIE colour space which defines the location of a particular colour
with regards to L*, a* and b*, and (B) the corresponding ellipse sizes throughout
L*a*b* colour space.
Uniaxial Tensile Testing

Uniaxial tensile testing was performed using a Tinius Olsen H25KS (Tinius Olsen, Ltd.,
Surrey RH1 5DZ, England) with a 1 kN load cell. Each material was formed into Type
2 dumb-bell specimens conforming to BS ISO 37. Samples were injected into the
dumb-bell shape to eliminate the cutting process, which has to been reported to possibly
130
lead to poor results.20 Each sample was subjected to a extension rate of 500 mm/min
(BS ISO 37), with a preconditioning of 10 cycles to 20% of the gauge length.
Preconditioning helps to increase repeatability of the tests by stabilising the stress-strain
function of the material. The structural properties of elastomers change significantly
during the first several times that the material experiences straining. This behaviour is
commonly referred to as the Mullin’s effect.21 The primary purpose of the tensile
testing in this application was to generate force-extension data, which can be converted
to stress-strain data, and determine the ultimate tensile strength (UTS) of each silicone
mixture.
Tear Testing
Tear testing was performed using a modification of the trouser test pieces outlined in
BS ISO 34-1. The Tinius Olsen was again used for this analysis. A strain rate of 100
mm/min was applied to each sample as per BS ISO 34-1. Tear testing results in tear
strength (TS) values, which provide an indication of the resistance of the material to
tearing, with results given in force per unit length. On failure of the specimen, the tear
resistance, or tear strength, is calculated by Equation 2.
F
TS = (2)
t
Where: TS is the tear strength (N/mm); F is the maximum load (N); and t is the
specimen thickness (mm).
In order to mechanically characterise each material, the experimental force-extension
data from the tensile tests were converted to engineering stress and engineering strain.
A 2nd order polynomial curve was applied to the data to obtain a mean experimental
data curve. This mean data was then applied to the commercial finite element analysis
(FEA) solver ABAQUS v.6.7 (Dassault Systemes, SIMULIA, RI, USA) in order to find
the most applicable strain energy function (SEF), and allow the determination of
material coefficients. Material coefficients were then assessed using a Type 2 dumb-
bell numerical model. The model was examined using identical boundary conditions to
those applied experimentally. The stress and strain at a central node was then mapped
throughout the course of the analysis, and compared to the results found experimentally.
131
Calibration Curves
Once data was compiled from tensile tests, tear tests, and colour analysis, calibration
curves could be created. These curves directly relate the colour intensity of the material
to a particular UTS and TS, and ultimately, to material properties and material
coefficients. For each calibration curve, the ∆E value of the material was plotted
against both the UTS and TS, with relevant trendlines applied to the data. Error bars
represent standard deviation. These linear curves then allow the prediction of material
properties depending on the ∆E value determined through colour analysis.
Effectiveness of Calibration Curves

In order to examine the accuracy of the calibration curves, a repeated batch of samples
were created at each silicone mixture. Each sample was probed with the
spectrophotometer to measure the ∆E value. Each ∆E value relates to both a UTS and
TS, thus allowing the UTS and TS to be predicted prior to testing. These samples were
then tensile tested and tear tested so as to record both the UTS and TS. The predicted
UTS and TS of each material could then be compared to the actual measured value.
Idealised AAA Experimental Models

Idealised AAA silicone models were manufactured. These idealised models were
designed to have realistic dimensions based on population averages obtained from the
EUROSTAR data registry.22 The maximum outer diameter of this model is 54 mm.
This ideal AAA model has been utilised in previous research by our group.2,17,23,24 The
technique used to create these models has been previously reported. 1,4 Briefly, each
silicone model is made using the lost-wax process. An inner wax model is first created
using an aluminium mould, which is then placed into a larger mould, consisting of a
uniform 2 mm cavity surrounding the wax model. The liquid silicone rubber is then
injected into this cavity, and cured. Three models were created using Sylgard 160,
three using Sylgard 170, and one model was created by randomly injecting both Sylgard
160 and Sylgard 170 through a Y-tubing connection into the wall cavity. This allowed
both silicones to randomly mix via the injection process, resulting in a model of random
material properties and various shades of black and grey. These models can be seen in
Figure 2.
132
Figure 2: Ideal AAA models created (A) Example Sylgard 160 model, (B) Example
Sylgard 170 model and (C) Mixed AAA model.
Videoextensometry
The purpose of this aspect of the study was to examine how the maximum diameter of
these silicone models react to increasing pressure loadings, and compare with
numerically predicted results using the material coefficients determined earlier. Each
model was set-up in the same manner, that is, the model was connected to a pneumatic
air source via a pressure regulator and pressure manometer. The iliac legs of each
model were blocked and constrained from movement. Applied air pressures were
incrementally increased by 20 mmHg from 0-160 mmHg. The diametrical change of
each model was recorded with the results then averaged. These measurements were
performed with the Messphysik Materials Testing Videoextensometer 1362CA
(Messphysik Materials Testing, Austria) in conjunction with the Messphysik Dot
Measurement for Windows software.
Mixed Model Colour Analysis

The spectrophotometer was also used to measure the colour intensity of the mixed ideal
AAA model of Sylgard 160 and 170. Measurements were taken at 20 mm intervals
along the length of the model, at the front, back, left, and right sides. These
measurement locations are shown in Figure 3. The readings were taken longitudinally
from the start of the proximal neck to the end of the iliac legs.
133
Figure 3: Ideal AAA mixed model created using both Sylgard 160 and Sylgard 170.
Red dots indicate locations where colour intensity (∆E) readings were taken. Distance
between dots is 20 mm.
Significant correlations between results were assessed using a non-parametric
Spearman’s Rho bivariate correlation test using SPSS 15.0 (SPSS Inc., Chicago, Ill,
USA). Correlation coefficient (CC) is shown where significant relationships were
observed.
Numerical Modelling
In order to assess the effectiveness of the material coefficients derived in this study, the
videoextensometry experiment was replicated using the finite element method. An
identical idealised AAA model was used, which has been used in many previous
studies, both experimentally2,4 and numerically.23 The numerical model has a uniform
wall of 2 mm, and was constrained from movement at the proximal neck and iliac legs,
reproducing the experimental set-up. A half model was examined due to the
symmetrical nature of the idealised AAA. The model was meshed using 11,443
quadratic tetrahedral 3D stress elements. Mesh independence was achieved by
increasing the mesh size until the peak stress was <2% of the previous mesh.6,16,25 The
same loading conditions were applied to the inner surface of the model and the
134
displacement of the maximum diameter region was measured at each loading. Figure 4
shows the meshed model and also the boundary conditions used. This FEA replication
was first modelled as a Sylgard 160 material, then as a Sylgard 170 material, and lastly,
with the region of maximum diameter corresponding to the material coefficients
determined from the colour analysis. For this mixed material numerical model, a
circumferential 10 mm band was partitioned at the region of maximum diameter and
assigned material properties corresponding to the ∆E value of the region (~Sylgard
160). As the regions from which diameter measurements were to be obtained were all
approximately the same ∆E value, a circumferential band was deemed to adequately
represent the region.
Figure 4: Numerical ideal AAA model used in the analysis. (A) Shows the meshed
symmetrical model and (B) illustrates the boundary conditions used. Symmetry
constraints were placed along the Y-axis, with the model constrained in all directions at
the proximal neck and iliac legs. The pressure loading was applied to the inner surface
of the model.
135
RESULTS
Uniaxial Tensile Testing, Tear Testing and Colour Analysis
The results of the experimental uniaxial tensile tests for each silicone mixture can be
seen in Table 1, with the results of the tear testing shown in Table 2. These results
show the silicone mix ratio, sample size (n), average ∆E value, average UTS and
average TS. Dow Corning reported the UTS of Sylgard 160 and Sylgard 170 to be 4
MPa and 2 MPa, respectively, although it is known that the upper and lower limits of
these values can vary significantly. Good agreement with these reported values were
found, with each of the mixes shown to have UTS values within this range. Results
from the tear testing show that the TS decreases with respect to the UTS of the material.
It was also noted that there was jagged tearing along the direction of the cut in all
samples examined for tear strength.
Table 1: Results of the uniaxial tensile testing for each mixture of silicone. ∆E and UTS
results are mean values of the sample size. Standard deviation (SD) is also shown.
Silicone Type n ∆E SD UTS (MPa) SD (MPa)
160 18 47.72 2.57 3.822 0.51
10:90 6 40.16 0.97 3.537 0.498
20:80 6 36.91 0.7 3.599 0.635
30:70 5 31.92 0.95 3.289 0.357
40:60 6 32.52 0.45 2.611 0.33
50:50 12 29.88 0.7 3.206 0.377
60:40 5 27.33 0.39 2.473 0.093
70:30 10 26.41 1.01 2.445 0279
80:20 5 25.13 0.77 2.199 0.243
90:10 5 24.55 1.21 2.401 0.391
170 20 23.86 1.85 2.077 0.375
Silicone mixes are in ratios of Sylgard 170:Sylgard 160, therefore, 10:90 refers to 1 part
Sylgard 170 to 9 parts Sylgard 160
Table 2: Results of the tear testing for each mixture of silicone. ∆E and tear strength
(TS) results are mean values of the sample size. Standard deviation (SD) is also shown.
Silicone Type n ∆E SD TS (N/mm) SD (N/mm)
160 6 45.73 0.33 0.697 0.094
10:90 6 40.74 0.22 0.68 0.088
20:80 5 37.23 0.38 0.754 0.079
30:70 6 34.52 0.52 0.692 0.045
40:60 6 31.32 0.58 0.688 0.095
50:50 5 28.54 0.45 0.573 0.034
60:40 6 28.76 0.48 0.537 0.054
70:30 3 26.39 0.14 0.479 0.051
80:20 6 25.40 0.24 0.519 0.069
90:10 6 23.79 0.32 0.433 0.047
170 8 23.84 0.63 0.452 0.049
136
The stress-strain data generated from tensile testing allowed a SEF to be applied to each
material, and therefore, material coefficients to be calculated. A 1st order Ogden model
was deemed to be the optimum SEF as it provided the most accurate curve fit to the data
and also the material remains stable at all stresses and strains. The general Ogden
SEF26 takes the form of Equation 3.
N 2μ p
W (λ1, λ2 , λ3 ) = ∑
α αp αp
(λ1 p + λ2 + λ3 − 3) (3)
p =1 αp 2
Where W is the strain energy density per undeformed unit volume, (λ1, λ2, λ3) are the
principal stretch ratios, α is a strain hardening exponent, and μ has the interpretation of
the shear modulus under infinitesimal straining. By applying this SEF to the
experimental data obtained from the tensile tests, material coefficients could be
determined. Table 3 shows the μ and α coefficients for each silicone mix examined,
with Figure 5 illustrating the effect the silicone mixture has on the resulting μ
coefficients. The coefficients of each material were then assessed using a comparative
Type 2 dumb-bell modelled in ABAQUS v6.7. Results, shown in Figure 6, highlight
the agreement between the stress and strain found experimentally from the tensile tests
compared to those determined using FEA for Sylgard 160 and Sylgard 170. Figure 7
shows the good correlation between experimental and numerical results for the new
silicone materials.
Table 3: 1st order Ogden SEF coefficients for each silicone mixture
Silicone Type ∆E μ α
160 46.63 1.6525 3.2395
10:90 40.45 0.7885 3.0366
20:80 37.07 0.7485 2.9704
30:70 33.22 0.6854 2.9807
40:60 31.92 0.5769 2.9243
50:50 29.59 0.5768 2.9278
60:40 28.04 0.5589 2.9286
70:30 26.40 0.5248 2.8296
80:20 25.27 0.4154 3.0707
90:10 24.17 0.3632 3.0265
170 23.82 0.6988 2.9741
137
Figure 5: Variation in 1st order Ogden μ coefficient with differing silicone mixtures.
Figure 6: Stress-strain curves for experimental and numerical results for Sylgard 160
and 170.
138
Figure 7: True stress-strain curves for the experimental and numerical results for the
new silicone materials.
Calibration Curves
From the experimental testing and colour intensity measurement, calibration curves
could be generated. First, a calibration curve relating ∆E to UTS was created, and is
shown in Figure 8. Then, a calibration curve relating ∆E to TS was generated. This
second curve can be seen in Figure 9. The relationship between ∆E and both UTS
(CC=0.955, P<0.0009) and TS (CC=0.909, P<0.0009) were both observed to be
significant. Each data point in Figure 8 and 9 is labelled with the corresponding
silicone mix ratio of Sylgard 170:Sylgard 160. Good linear relationships were shown
between both ∆E and UTS (R2=0.8082), and ∆E and TS (R2=0.7205).
Figure 8: UTS calibration curve for silicones ranging from Sylgard 170 to Sylgard 160.
139
Figure 9: Tear strength calibration curve for silicones ranging from Sylgard 170 to
Sylgard 160.
Effectiveness of Calibration Curves

The accuracy of Figures 8 and 9 were also examined. This was achieved by preparing
both dumb-bell and trouser test samples of each silicone mixture and measuring the
colour intensity of each sample. This ∆E value was then related to UTS using Figure 8
and TS using Figure 9. Dumb-bell samples were then tensile tested, and trouser
samples measured for tear strength. The measured UTS and TS of the samples were
then compared to the predicted UTS and TS. The results of this test can be seen in
Figures 10 and 11. The predicted and measured TS results showed to be significant for
the 30:70 (CC=0.955, P=0.003) and 70:30 mixtures (CC=0.851, P=0.032).
Significance was not observed between the predicted and measured results for other
mixes. Good linear agreement was observed between the predicted values and the
measured values for both the UTS (R2=0.8556) and the TS (R2=0.8061). Overall, the
predicted UTS and measured UTS results were significant (CC=0.945, P<0.0009), as
were the predicted TS and measured TS results (CC=0.891, P<0.0009).
140
Figure 10: Predicted UTS results compared with those found experimentally for each
silicone mixture.
Figure 11: Predicted TS results compared with those found experimentally for each
silicone mixture.
Relationship Between UTS and Tear Strength

The UTS results and tear strength results were also compared resulting in Figure 12,
which showed a good linear trend (R2=0.7159) and significant relationship (CC=0.9,
P<0.0009) between the two material properties. This allows tear strength to be related
to UTS for each varying silicone material thus giving a further insight into the
relationship between the two properties.
141
Figure 12: Relationship between tear strength and UTS for each silicone rubber.
Colour Analysis: Mixed AAA Model

The spectrophotometer was used to determine the ∆E value at 20 mm intervals along
the longitudinal distance of the model, as shown in Figure 3. The results show how the
∆E value differs depending on the concentration of either Sylgard 170 or Sylgard 160.
This resulted in a series of colour intensity values corresponding to a specific location
on the model, with the values shown in Table 4 and graphically in Figure 13. It can be
clearly seen how the colour intensity of the material changes at various locations on the
model.
Table 4: ∆E measurements at various locations along the length of the ideal AAA mixed
model. Measurement locations correspond with Figure 3.
Front Back Left Right
Distance (mm) ∆E ∆E ∆E ∆E
0 30.38 24.77 31.4 31.58
20 26.02 23.27 28.43 27.25
40 23.94 23.26 29.38 28.07
60 24.07 25.5 28.13 42.98
80 35.09 24.29 27.56 29.91
100 27.45 27.21 45.41 45.31
120 20.12 25.22 39.84 24.42
140 23.16 27.56 44.04 33.41
160 - - 45.13 45.41
180 - - 26.99 26.85
200 - - 27.52 36.93
220 - - 30.11 30.94
Note: Front and back do not have ∆E values past 140mm due to the geometry of the
model.
142
Figure 13: ∆E variation along the longitudinal distance of the ideal AAA mixed model.
Horizontal lines indicate mean ∆E values for both Sylgard 170 (dashed) and Sylgard
160 (solid).
Videoextensometry
As the air pressure was incrementally increased, the diametrical change of each model
was measured and recorded. The initial maximum diameter of each model was 54 mm,
and so any displacement past 54 mm at each pressure loading is recorded with the
software. Tables 5, 6 and 7 show the results of the experimental pressure-diameter
testing, with the results summarised graphically in Figure 14.
Table 5: Pressure-diameter results for the Sylgard 160 ideal AAA models. Each set of
results for each model is a combination of measurements taken from both the front and
side
Pressure (mmHg) Model 1 Model 2 Model 3 Average (mm)
0 54 54 54 54
20 54.01 54.01 54.02 54.01
40 54.02 54.02 54.05 54.03
60 54.04 54.03 54.07 54.05
80 54.08 54.06 54.09 54.07
100 54.13 54.09 54.11 54.11
120 54.18 54.13 54.15 54.15
140 54.23 54.17 54.19 54.19
160 54.27 54.22 54.23 54.24
143
Table 6: Pressure-diameter results for the Sylgard 170 ideal AAA models. Each set of
results for each model is a combination of measurements taken from both the front and
side
Pressure (mmHg) Model 1 Model 2 Model 3 Average (mm)
0 54 54 54 54
20 54.07 54.04 54.04 54.05
40 54.14 54.09 54.12 54.12
60 54.23 54.16 54.19 54.19
80 54.39 54.30 54.32 54.33
100 54.54 54.41 54.47 54.47
120 54.71 54.66 54.66 54.67
140 54.92 54.86 54.86 54.88
160 55.10 55.05 55.06 55.07
Table 7: Pressure-diameter results for the randomly-mixed ideal AAA model.

Pressure (mmHg) Diameter (mm)
0 54
20 54.00
40 54.01
60 54.06
80 54.20
100 54.32
120 54.51
140 54.67
160 54.83
Figure 14: Pressure-diameter results for the Sylgard 160, Sylgard 170 and mixed ideal
AAA models.
Numerical Modelling
By measuring the displacement at the region of maximum diameter, the deformation of
the experimental models could be compared to those observed numerically. The
144
experimental and numerical results were statistically significant for the Sylgard 160
model (CC=1.0, P<0.0009), Sylgard 170 model (CC=1.0, P<0.0009) and also the mixed
silicone model (CC=0.971, P<0.0009). Overall, there was a percentage difference in
the diameter change of 0.24% (range 0.01-1.78%) and 0.38% (range 0.3-5.27%) for the
Sylgard 160 and Sylgard 170 models, respectively. For the mixed ideal AAA model,
deformations were only measured from the front of the model, as with the experimental
model, with an average percentage difference of 0.76% (range 0.25-5.03%). The results
can be seen tabulated in Table 8 and graphically in Figure 15.
Table 8: Summarised results comparing diameter changes between those found

experimentally and numerically. All dimensions are maximum diameter and are given
in millimetres (mm).
Sylgard 160 Sylgard 170 Mixed Model
Pressure
Experimental FEA Experimental FEA Experimental FEA
(mmHg)
0 54 54 54 54 54 54
20 54.012 54.030 54.050 54.040 54.000 54.035
40 54.028 54.039 54.115 54.084 54.010 54.070
60 54.045 54.040 54.192 54.189 54.060 54.085
80 54.073 54.059 54.333 54.325 54.200 54.188
100 54.108 54.095 54.468 54.487 54.320 54.314
120 54.150 54.138 54.670 54.673 54.510 54.460
140 54.192 54.186 54.875 54.885 54.670 54.629
160 54.240 54.240 55.068 55.121 54.830 54.813
Note: Experimental and FEA results for all three analyses were significant at the 0.05
level
Figure 15: Comparison of results for maximum diameter change with increased pressure
loading for both the experimental (Exp) and FEA results.
145
DISCUSSION
Material Development
Silicone rubbers are widely used as arterial analogues in experimental studies.1-4 When
utilising silicone in experimental testing, a complete understanding of the mechanical
behaviour of the material is important in order to correctly measure and assess results.
The purpose of this study was to design a number of silicones with a distinct range of
material properties and utilise these materials as improved arterial analogues. It was
also necessary to be able to identify and correlate material variations non-destructively,
as it was essential that material properties could be determined without further
experimental testing. The results presented suggest that the range of silicones
developed could be readily implemented in future in vitro studies providing models
with predictable non-uniform material properties that are more representative of the
behaviour of arterial tissue. In order to develop the range of silicone rubbers presented,
base materials were required. The materials chosen were Sylgard 160 and Sylgard 170
from the Dow Corning range of silicone rubbers. These materials were selected due to
their relatively low UTS values of 4 MPa and 2 MPa, respectively, and also due to their
natural appearance, that is, Sylgard 160 is grey and Sylgard 170 is black in colour.
These strength values are slightly outside the range of tensile strengths of AAA tissue
previously reported: 0.38 - 0.73 MPa;27 0.9415 MPa;28 0.336 - 2.351 MPa.19 However,
as AAA tissue does experience a range of tensile strengths, the repeatable nature of this
developed range can be employed to create distributions of strengths, thus making the
range somewhat comparable to the realistic setting.
Colour Analysis
A result of these various mixes was that each new material had a differing appearance,
with the colour slightly changing from grey to black as the mix ratio increased from
pure Sylgard 160 to Sylgard 170. This change in physical appearance of each new
silicone facilitated the analysis of each material using a non-destructive method of
determining colour intensity. Colour intensities were measured using a
spectrophotometer. This device measures the colour intensity based on the lightness,
chroma and hue of the particular material. The ∆E value proved to be a repeatable
method of determining colour intensities between the various materials, with the results
shown in Tables 1 and 2. Slight deviations exist between ∆E values for the same
material, yet overall the method is robust, showing a linear increase in ∆E as the
material changes from black (Sylgard 170) to grey (Sylgard 160).
146
Once the colour analysis was determined, it was necessary to mechanically characterise
the range of materials. This was performed by uniaxially tensile testing and tear testing
each silicone rubber to assess, firstly, the UTS of each material, and secondly, the tear
resistance, or tear strength (TS), of each rubber. Tensile testing and tear testing was
performed according to BS ISO 37 and BS ISO 34-1 standards, respectively. The
tensile testing generated force-extension curves, which were converted to engineering
stress and strain data. The tear testing records the peak force required to tear the sample
and then divides this by the sample thickness. This results in tear strength values given
in N/mm. UTS values for each material ranged from 2.077 - 3.822 MPa, values of
which lie within the reported UTS values for Sylgard 170 and Sylgard 160 (2-4 MPa)
according to the Dow Corning specification sheet. These results show that the UTS
values have a linear relationship with silicone type (see Table 1), which was the desired
outcome. Tear testing was performed on trouser test pieces, modified from the sample
described in BS ISO 34-1. Again, tear testing revealed that the TS has a linear
relationship with silicone type (see Table 2), with values ranging from 0.452-0.697
N/mm. These values are lower than those supplied on the Dow Corning data sheet and
are not comparable, as it is known that tear strengths depended heavily on the type of
specimen used in the testing.29
Applying the data obtained from the experimental testing to the numerical solver
allowed the 1st order Ogden SEF to be deemed the most suitable SEF. ABAQUS v6.7
then generates material coefficients for the 1st order Ogden SEF that accurately
represent each material. These material coefficients can be seen in Table 3. The
resulting change in μ coefficient for each mixture can be seen in Figure 5. As Sylgard
160 and Sylgard 170 were designed by Dow Corning and never intended to be mixed
together, the actual mixing of these material may cause an unusual reaction within the
polymerisation of the materials, and this had to be investigated. As a result of this, the
pure Sylgards 160 and 170 are homogenous materials, whereas the new materials may
be inhomogeneous. The homogenous Sylgards (160 and 170) have coefficients that do
not conform to the linear trend displayed in Figure 5, whereas, the nine other
inhomogeneous mixtures designed in this study are new materials and the coefficients
vary linearly depending on the mix ratio. It was determined that when numerically
modelling any of the developed silicones, the material coefficients can be implemented
with the 1st order Ogden SEF within the FEA software. A numerical model of the
147
dumb-bell was analysed in ABAQUS v6.7 using the coefficients determined earlier and
by applying identical boundary conditions to those employed experimentally. This was
achieved by plotting the stress and strain at a central node in model throughout the
loading. Results of this analysis agreed well with the experimental results for each of
the eleven materials examined, as shown in Figures 6 and 7.
The next stage in the study was to develop calibration curves for the range of silicones.
These curves allow the use of a silicone rubber to be assessed for colour intensity prior
to mechanical testing. The resulting ∆E value can then be related to a UTS and TS,
allowing the determination of material property through non-destructive testing. The
calibration curves for both the ∆E - UTS and ∆E - TS relationships can be seen in
Figures 8 and 9, respectively. The calibration curves for ∆E - UTS (R2=0.8082,
CC=0.955, P<0.0009) and ∆E - TS (R2=0.7205, CC=0.909, P<0.0009) were observed
to have strong linear, significant relationships. The effectiveness of these curves was
also examined. This was achieved by creating a duplicate batch of dumb-bell and
trouser samples, and then measuring the ∆E value of each sample. The material
property of the particular specimen could then be pre-determined using Figures 8 and 9
prior to testing. Tensile and tear testing of the sample then revealed the actual measured
UTS and TS of the sample. Figures 10 and 11 show how these predicted values
compared to those measured experimentally. For both the UTS and TS, the values
predicted using the calibration curves agreed well with those determined from further
testing. The predicted TS and measured TS for the 30:70 mixture (CC=0.955, P=0.003)
and the 70:30 (CC=0.851, P=0.032) mixture were the only materials to exhibit
statistical significance. This may be due to the relatively low sample sizes for each
mixture. It is believed that by increasing the sample sizes may lead to significance of
results. Figures 10 and 11 also show the good linear relationships between both the
predicted and measured UTS values (R2=0.8556, CC=0.945, P<0.0009) and the
predicted and measured TS values (R2=0.8061, CC=0.891, P<0.0009). It was now
possible to relate the UTS with the TS for the range of silicones. These results can be
seen in Figure 12 and also displayed statistical significance (R2=0.7159, CC=0.9,
P<0.0009). As a result, in vitro silicone models with varying material properties can
now be created and the distribution of varying material properties can be determined
using non-destructive testing.
148
Ideal AAA Models

The next aspect of this work focused on applying the methodology developed to more
appropriate geometries by creating silicone models of an idealised abdominal aortic
aneurysm (AAA) model. These models were created using the lost-wax process, and
form an idealised AAA model of silicone rubber with a wall thickness of 2 mm. It has
been reported that slight deviations can exist in the uniformity of the wall but within
acceptable tolerances.1,4 Three models of Sylgard 160 and three of Sylgard 170 were
manufactured, with one further model created using a combination of the two materials
injected through a Y-tubing connection. These models can be seen in Figure 2. The
mixed model therefore consists of a random mixture of Sylgard 160 and 170 throughout
the model, and thus has a random distribution of varying material properties within the
ranges presented. The models were then used to evaluate the methodology (of non-
destructively determining a varying material property distribution) described earlier.
Firstly, the mixed model was examined using the spectrophotometer. As the model had
a random distribution of material mixtures, and therefore a random distribution of
differing shades of black and grey, it was expected that the ∆E value would vary at
different locations on the model. The ∆E value was measured at 20mm intervals along
the longitudinal distance of the model, at the front, back, left and right sides. The
results of this are shown in Table 4 and Figure 13. By comparing Figures 3 and 13, it
can be clearly seen how the ∆E value varies according to the location on the model.
This analysis allows the ∆E value at a specific region to be related to a material
property, and also to a set of coefficients to be used with the 1st order Ogden SEF. For
example, in the case of the mixed model, the ∆E value at the left and right sides of the
maximum diameter region were 45.41 and 45.31, respectively. These values
st
correspond with 1 order Ogden μ and α coefficients of 1.6525 and 3.2395 from Table
3.
Videoextensometry
The results of the videoextensometry work can be seen in Tables 5 - 7 and also in
Figure 14. The change in stiffness between the various materials can be clearly seen.
Sylgard 160 is stiffer than Sylgard 170, and this can be seen by the minor change in
maximum diameter, even at high pressures of 160 mmHg. At this elevated pressure the
maximum diameter of the Sylgard 160 deformed by 0.24 mm, compared with a 1.068
mm diameter change for the Sylgard 170 models. Figure 14 also shows how the mixed
149
model performed. This model was a combination of Sylgard 160 and 170, and
therefore, the pressure-diameter results lay in between those of the two Sylgard
materials. The maximum diameter of this model had a ∆E value of approximately 45,
and so was predominately Sylgard 160, hence the mixed model experienced a maximum
diameter change similar to that of the pure Sylgard 160 model (0.83 mm vs. 1.068 mm).
Numerical Validation
Identical boundary conditions to those used experimentally were applied to the
numerical model, using the material coefficients derived earlier. The numerical model
complete with boundary conditions can be seen in Figure 4. A static pressure was
applied to the internal surface of the model, with the displacements of the maximum
diameter region recorded at each pressure loading. This analysis was performed for a
pure Sylgard 160, pure Sylgard 170, and the mixed model, as with the experimental
testing. Table 8 and Figure 15 show how these experimental and numerical results
compared.
For both the Sylgard 160 and Sylgard 170 models, the percentage difference between
the experimental and numerical results was 0.24% (range 0.01 - 1.78%) and 0.38%
(range 0.3 - 5.27%), respectively. There was also significance observed in the
validation of experimental and numerical results for both the Sylgard 160 models
(CC=1.0, P<0.0009) and the Sylgard 170 models (CC=1.0, P<0.0009). When analysing
the mixed AAA model, the colour analysis of the mixed silicone model revealed that the
left and right side of the maximum diameter region had a ∆E value very close to that of
pure Sylgard 160. In order to replicate the mixed model, which has a random array of
material properties in the AAA wall, a 10mm circumferential band was created at the
region of maximum diameter. The model was analysed with a circumferential band, as
the regions at the maximum diameter from which the experimental results were
obtained had very similar ∆E values (left = 45.41, right = 45.31) and therefore it was
assumed that the region was essentially the same material. Results of this examination
show that the experimental and numerical deformations agree well, with an overall
percentage difference of 0.76% (range 0.25 - 5.03%). Significance was also noted with
these results (CC=0.971, P<0.0009). These percentage differences observed for all
models were calculated from Table 8 and are within acceptable limits, as the maximum
percentage difference of 5.27% recorded in the Sylgard 170 model, only relates to a
150
difference in diameter of 55.07 mm compared to 55.12 mm, that is, a percentage

difference equal to 0.05 mm.
Further to previous reports on rubber model creation,1,4 methods of including

calcification analogues into the AAA wall are possible. Calcifications have been shown
to alter the wall stress distributions in numerical models30 and so the examination of
these calcified deposits within in vitro models may yield interesting results. These
calcification analogues could also be characterised and calibrated to material properties
using the methods described here, thus creating an even more realistic experimental
model. Also, the materials and methods described here are not limited to AAAs. Lethal
aneurysms also form in the cerebral arteries and the thoracic aorta, and as with ongoing
research into AAAs, there is a need to develop more realistic in vitro models of these
aneurysms to further understand the biomechanical behaviour of these diseased vessels.
Therefore, there is a need to develop a method of creating experimental AAA models
with known material properties that can be determined using a non-destructive method.
Limitations
It is known that there are many available techniques to mechanically characterise
silicone rubbers, such as biaxial and equiaxial tensile testing, and compression testing.
It is believed that although these tests were not performed in this study, future work
with these experiments may further enhance the material characterisation presented
here. Deviations still exist within the UTS and TS results for the rubbers. Increasing
the sample sizes used for each variation of silicone may help reduce these errors.
Although, with regards to material UTS, Dow Corning state that the upper and lower
limits of UTS values for silicones can vary significantly as silicone rubbers are not
usually designed to fail. It is often simply the stress-strain response that is of paramount
concern, and not the UTS, depending on the application. Also, the percentage
differences, although quite small, observed in the videoextensometry compared to the
numerical models could be reduced. It is known that slight variations in wall thickness
exist within the silicone models, and the use of computed tomography (CT) scanning of
the rubber model prior to testing may help eliminate this. CT scan data of each model
could be reconstructed using available 3D reconstruction techniques and imported to
FEA software for analysis. This would allow an experimentally accurate model to be
analysed numerically, instead of the perfectly uniform wall model employed in this
study. Also, modelling the mixed silicone model numerically could be improved by
151
creating a “marbled” numerical model similar to that in experimental testing, instead of

the “banded” model examined here. These issues will be addressed in future studies.
CONCLUSIONS
This study has successfully designed and developed a range of silicone rubbers that are
not only visually different and easily detectable using a spectrophotometer, but also
have a range of material properties that correlate to the visual appearance. The
spectrophotometer proved to be effective in differentiating between the various shades
of colour. Extensive experimental testing and colour intensity analyses lead to the
development of calibration curves relating colour to both UTS and TS. Material
characterisation was also performed resulting in a complete set of material coefficients
for each of the new silicone rubbers. Testing also showed that the calibration curves are
effective in predicting the material properties. Experimental AAA models of known
and random material properties were also created. Pressure-diameter experiments and
numerical modelling showed that the material coefficients accurately describe the
behaviour of the silicone rubber, with minimal percentage differences. Future
experimental studies could include these materials, with the material coefficients readily
available to be incorporated into any finite element solver.
ACKNOWLEDGEMENTS
The authors would like to thank our funding sources (i) the Irish Research Council for
Science, Engineering and Technology (IRCSET) Grant RS/2005/340 and (ii) Grant
#R01-HL-060670 from the US National Heart Lung and Blood Institute. The authors
would also like to thank the contribution of Joseph Muthu, Departments of Surgery and
Bioengineering, Centre for Vascular Remodelling and Regeneration, University of
Pittsburgh, USA.
REFERENCES
1. Doyle BJ, Morris LG, Callanan A, Kelly P, Vorp DA, McGloughlin TM. 3D
reconstruction and manufacture of real abdominal aortic aneurysms: From CT
scan to silicone model. J Biomech Eng 2008;130:034501-5.
2. Doyle BJ, Corbett TJ, Callanan A, Walsh, MT, Vorp DA, McGloughlin TM. An
experimental and numerical comparison of the rupture locations of an abdominal
aortic aneurysm. J Endovasc Ther 2009, in press.
3. Marins PALS, Natal Jorge RM, Ferreira AJM. A comparative study of several
material models for prediction of hyperelastic properties: Application to
silicone-rubber and soft tissues. Strain 2006;42:135-47.
152
4. O’Brien T, Morris L, O’Donnell M, Walsh M, McGloughlin TM. Injection-

moulded models of major and minor arteries: The variability of model wall
thickness owing to casting technique. Proc IMechE Part H: J Eng Med
2005;219:381-6.
5. Shergold OA, Fleck NA, Radford D. The uniaxial stress versus strain response
of pig skin and silicone rubber at low and high strain rates. Int J Impact Eng
2006;32:1384-1402.
techniques for computing wall stress in abdominal aortic aneurysms. Biomed
Eng Online 2007;6:38.
7. Doyle BJ, Callanan A, Burke PE, Grace PA, Walsh MT, Vorp DA,
McGloughlin TM. Vessel asymmetry as an additional diagnostic tool for the
assessment of abdominal aortic aneurysms. J Vasc Surg 2009;49:443-454.
8. Doyle BJ, Callanan A, Walsh MT, Grace PA, McGloughlin TM. A finite
element analysis rupture index (FEARI) as an additional tool for abdominal
aortic aneurysm burst prediction. Vasc Dis Prev 2009;6:114-121.
9. Fillinger MF, Raghavan ML, Marra SP, Cronenwett JL, Kennedy FE. In vivo
analysis of mechanical wall stress and abdominal aortic aneurysm rupture risk. J
Vasc Surg 2002;36:589-97.
10. Fillinger MF, Marra SP, Raghavan ML, Kennedy FE. Prediction of rupture risk
in abdominal aortic aneurysm during observation: wall stress versus diameter. J
Vasc Surg 2003;37:724-32.
11. Nicholls SC, Gardner JB, Meissner MH, Johansen HK. Rupture in small
abdominal aortic aneurysms. J Vasc Surg 1998;28:884-8.
12. Raghavan ML, Vorp DA, Federle MP, Makaroun MS, Webster MW. Wall stress
distribution on three-dimensionally reconstructed models of human abdominal
aortic aneurysm. J Vasc Surg 2000;31:760-9.
13. Thubrikar MJ, Al-Soudi J, Robicsek F. Wall stress studies of abdominal aortic
aneurysm in a clinical model. Ann Vasc Surg 2001;15:355-66.
14. Venkatasubramaniam AK, Fagan MJ, Mehta T, Mylankal KJ, Ray B, Kuhan G,
Chetter IC, McCollum PT. A comparative study of aortic wall stress using finite
element analysis for ruptured and non-ruptured abdominal aortic aneurysms. Eur
J Vasc Endovasc Surg 2004;28:168-76.
15. Vorp DA, Raghavan ML, Webster MW. Mechanical wall stress in abdominal
aortic aneurysm: influence of diameter and asymmetry. J Vasc Surg
1998;27(4):632-9.
16. Wang DHJ, Makaroun MS, Webster MW, Vorp DA. Effect of intraluminal
thrombus on wall stress in patient-specific models of abdominal aortic
aneurysm. J Vasc Surg 2002;36:598-604.
17. Morris L, O’Donnell P, Delassus P, McGloughlin T Experimental assessment of
stress patterns in abdominal aortic aneurysms using the photoelastic method.
Strain 2005;40:165-72.
18. Flora HS, Talie-Faz B, Ansdell L, Chaloner EJ, Sweeny A, Grass A, Adiseshiah
M. Aneurysm wall stress and tendency to rupture are features of physical wall
properties: An experimental study. J Endovasc Ther 2002;9:665-75.
19. Raghavan ML, Kratzberg J, de Tolosa EMC, Hanaoka MM, Walter P, da Silva
ES Regional distribution of wall thickness and failure properties of human
abdominal aortic aneurysm. J Biomech 2006;39(16):3010-6.
20. Feger C. Mechanical testing of thin polymer films. In: Proceedings of the 4th
International Symposium on Advanced Packaging Materials. 1998;pp.77-81.
21. Mullins L. Softening of rubber by deformation. Rubber Chem Technol
1969;42:339-62.
153
22. Laheij R, van Marrewijk C, Buth J. Progress report on the procedural and follow
up results of 3413 patients who received stent graft treatment for infrarenal
aortic aneurysms for a period of 6 years. EUROSTAR Data Registry Centre,
2001
23. Callanan A, Morris LG, McGloughlin TM. Numerical and experimental analysis
of an idealised abdominal aortic aneurysm. European Society of Biomechanics,
S-Hertogenbosch, Netherlands 2004.
24. O' Brien T P, Walsh M T, Morris L G, Grace P A, Kavanagh E G, McGloughlin
T M. Numerical and Experimental Techniques for the Study of Biomechanics in
the Arterial System. In: Leondes CT (ed) Biomechanical Systems Technology,
World Scientific Publishing Co., Singapore, 2008;Chap 7:233-70.
25. Truijers M, Pol JA, SchultzeKool LJ, van Sterkenburg SM, Fillinger MF,
Blankensteijn JD. Wall stress analysis in small asymptomatic, symptomatic and
ruptured abdominal aortic aneurysms. Eur J Vasc Endovasc Surg 2007;33:401-7.
26. Ogden RW. Nonlinear elastic deformations. Dover Publication Inc., Mineola,
NY, USA. 1984
27. Thubrikar MJ, Labrosse M, Robicsek F, Al-Soudi J, Fowler B. Mechanical
properties of abdonminal aortic aneurysm wall. J Med Eng Technol
2001;25(4):133-42.
28. Raghavan ML, Webster MW, Vorp DA. Ex vivo biomechanical behaviour of
abdominal aortic aneurysm: assessment using a new mathematical model. Ann
Biomed Eng 1996;24:573-82.
29. Brown R. Physical testing of rubber. 4th Ed. Springer Science+Business+Media,
Inc., NY, USA. 2006;159-67.
30. Speelman L, Bohra A, Bosboom EMH, Schurink GWH, van de Vosse FN,
Makaroun MS, Vorp DA. Effects of wall calcifications in patient-specific wall
stress analyses of abdominal aortic aneurysms. J Biomech Eng 2008;129:105-9.
154
Chapter VIII
Identification of rupture locations in patient-specific
abdominal aortic aneurysms using experimental and
computational techniques.
Barry J. Doyle,1 Aidan J. Cloonan,1 Michael T. Walsh,1 David A.

Vorp2 and Timothy M. McGloughlin1
Limerick, Ireland.

USA.
Journal of Biomechanics (submitted)

Critical revision of article: BD, MW, DV, TM
Chapter Eight Rupture of Realistic AAAs
ABSTRACT
Background
In the event of abdominal aortic aneurysm (AAA) rupture, the outcome is often death.
Traditionally, surgeons determine the severity and likelihood of rupture by the
maximum transverse diameter. It is the general consensus that there is a need to review
this criterion. Computational approaches are predominantly employed to help
understand the biomechanics of AAA rupture, with little focus on experimental
techniques. This paper aims to assess the rupture locations of in vitro AAA models of
both idealised and realistic geometries, and validate these rupture sites using the finite
element method.
Methods
Idealised and patient-specific silicone rubber AAA models were manufactured using
Dow Corning Sylgard 160 and Sylgard 170 and subsequently imaged using computed
tomography. Experimental models were inflated until rupture using an in vitro set-up.
High-speed photography captured the site of rupture. Numerical reconstructions were
generated for each experimental model. Surface curvature and wall stress was
determined for each of the geometries. Wall thickness was assessed in all geometries
and at the specific sites of rupture.
Results
Experimental models ruptured at regions of inflection, not at regions of maximum
diameter. Mean ± SD rupture pressures for the Sylgard 160 and Sylgard 170 models
were 650.6 ± 195.1 mmHg (range = 381.4 – 985 mmHg) and 410.7 ± 159.9 mmHg
(range = 252.2 – 714 mmHg), respectively. Computational models accurately predicted
the locations of rupture. Wall stress distributions do not significantly change with
increasing luminal pressure. Mean ± SD peak wall stress for the Sylgard 160 and
Sylgard 170 models was 2.15 ± 0.26 MPa (range = 1.89 – 2.59 MPa) at an internal
pressure of 650 mmHg and 1.69 ± 0.38 MPa (range = 1.17 – 2.26 MPa) at an internal
pressure of 410 mmHg, respectively. Mean ± SD wall thickness was 2.19 ± 0.40 mm,
with a mean ± SD wall thickness at the location of rupture of 1.85 ± 0.33 mm and 1.71
± 0.29 mm for the Sylgard 160 and Sylgard 170 materials, respectively.
Conclusions
Experimental modelling determined that idealised and realistic AAA geometries rupture
at regions of high wall stress and not at regions of maximum diameter. These rupture
locations were confirmed using the finite element method. Numerical results indicate
several regions of elevated wall stress in AAA geometries, with rupture occurring at the
location of peak stress in 80% (16/20) of cases. 10% (2/20) of models had defects in the
AAA which moved the rupture location from regions of elevated wall stress to the
region of the defect. The results presented may further contribute to the understanding
of AAA biomechanics and ultimately AAA rupture prediction.
155
INTRODUCTION
The incidence of abdominal aortic aneurysm (AAA) is on the increase. It is currently
estimated that approximately 1 million people currently live with an undiagnosed AAA,
and that 95% of these cases could be successfully treated if detected prior to rupture.
15,000 deaths per year in the US alone are related to AAA rupture (Kleinstreuer and Li,
2006). Although the mortality rates associated with AAA are high, there still remains
uncertainty about the correct time to surgically repair these aneurysms. Currently, the
trend in determining the severity of an AAA is to use the maximum diameter criterion
(Cronenwett et al., 1985; Glimaker et al., 1991). Based on this criteria, AAAs that
reach or exceed 5-5.5 cm in maximum diameter are deemed a rupture risk, and
subsequently repaired. However, recent research has cast doubt over the effectiveness
of this method (Raghavan et al., 2000; Sayers, 2002; Fillinger et al., 2002, 2003; Vande
Geest et al., 2006; Kleinstreur and Li, 2006; Leung et al., 2006; Doyle et al., 2009a,
2009b) as it has been reported that smaller AAAs can rupture (Darling et al., 1977;
Nicholls et al., 1998) and also larger AAAs may remain stable for the remainder of the
patients’ life.
The majority of previous reports have focused on computational approaches, in

particular finite element analysis (FEA), to predict regions of elevated wall stress
(Mower et al., 1997; Vorp et al., 1998; Raghavan et al., 2000; Di Martino et al., 2001;
Thubrikar et al., 2001; Hua et al., 2001; Wang et al., 2002, Fillinger et al., 2002, 2003;
Ventkatasubraniam et al., 2004; Giannoglu et al., 2006; Leung et al., 2006; Papaharilaou
et al., 2007; Speelman et al., 2007; Scotti et al., 2005, 2007; Truijers et al., 2007; Doyle
et al., 2007, 2009a, 2009b, 2009c; Rissland et al., 2009), but have neglected
experimental approaches to AAA rupture. Quantitatively assessing the stress within an
AAA wall has many merits and has been shown to be superior over maximum diameter
alone in predicting rupture (Fillinger et al., 2003). However, these high stress regions
need to be assessed using experimental techniques to determine if FEA actually predicts
the correct region of rupture. Recent reports by Doyle et al. (2008a, 2009d) presented
comparative studies suggesting that FEA accurately predicts the high stress regions in
idealised AAA models. This paper aims to examine this hypothesis in anatomically-
correct AAAs. Traditional FEA methods will be used to explore the wall stress
distributions in patient-specific AAAs, with high-speed photography employed to
identify rupture regions in vitro. Experimental AAA models of silicone rubber were
created using a reported technique (Doyle et al., 2008b), and then imaged using
156
computed tomography. 3D reconstruction of these data sets allowed exact numerical

models to be used in computational studies, in order to validate the rupture sites
observed experimentally.
METHODS
Study Subjects
Four computed tomography (CT) datasets for patients with AAA were included in this
study. These CT data sets were obtained from the HSE Midwestern Regional Hospital,
Limerick, Ireland, and the University of Pittsburgh Medical Centre, PA, USA. All four
patients had reached or exceeded the current 5 cm threshold for surgical repair. CT
scans were acquired using both the Siemens Somatom Sensation 64 (Siemens AG, D-
91052 Erlangen, Germany) and LightSpeed Plus (GE Medical Systems, General
Electric Company) range of imaging equipment. All scans were single CT slices with a
standard width x height of 512 x 512 pixels. Mean pixel size of scans was 0.813 ±
0.065 mm. The patient details and geometrical characteristics of each case can be seen
in Table 1.
Table 1: Case details of AAA geometries used in study.

Maximum Total Length Volume Surface Area
Patient Sex Age 3
Diameter (cm) (cm) (cm ) (cm2)
Ideal NA NA 5.0 26* 188.9* 27.5*
1 M 78 6.1 15.2 192.4 18.5
2 M 70 5.7 14.8 194.9 19.1
3 F 65 5.6 13.7 136.9 14.8
4 F 67 5.3 13.6 137.9 15.5
* Length, volume and surface area of idealised AAA includes 90mm section of aorta
and iliac arteries.
In addition to these realistic AAA cases, work was performed using an idealised AAA
model that has been used extensively in previous studies by our group (Callanan et al.,
2004; Morris et al., 2005; Doyle et al., 2008a, 2009d, 2009e). This ideal AAA was
developed using realistic dimensions obtained from the EUROSTAR data registry
(Laheij et al., 2001). In brief, the ideal AAA has a maximum internal diameter of 50
mm, a total length of 260 mm, a uniform wall thickness of 2 mm and is geometrically
symmetric.
3D Reconstruction Procedure
The technique of 3D reconstruction used in this work has been previously reported
(Doyle et al., 2008b, 2009c). Briefly, the commercial software Mimics v12
157
(Materialise, Belgium) allows 3D reconstructions to be created from medical data sets,

such as CT scans. The software utilises user-defined thresholding and segmentation
parameters based on the intensity of the pixels in the original image. All
reconstructions were generated from scan positions immediately distal to the lowest
renal artery to immediately proximal to the iliac bifurcation. The intraluminal thrombus
(ILT) was neglected in this study as with previous approaches (Vorp et al., 1998;
Raghavan et al., 2000; Thubrikar et al., 2001; Fillinger et al., 2002, 2003; Scotti et al.,
2005; Doyle et al., 2009a). The thickness of the aorta wall is not easily identifiable
from CT scans, therefore the wall was assumed to be uniform throughout the model and
for the purpose of mould design was set as 2 mm. The same degree of smoothing was
applied to each reconstruction as previously reported (Doyle et al., 2007, 2008b). This
level of smoothing removes any unwanted surface artefacts, whilst maintaining the
geometrical integrity of the AAA. The iliac bifurcation was omitted from this study as
it is believed its inclusion would not significantly alter the results (Fillinger et al., 2002).
Materials and Characterisation

Sylgard 160 and Sylgard 170 (Dow Corning Corporation) were used as the materials for
this study. These commercially available silicone rubbers have different ultimate tensile
strengths (UTS) according to the Dow Corning specification sheets (Sylgard 160 = 4
MPa; Sylgard 170 = 2 MPa) and are also different in appearance (Sylgard 160 = Grey;
Sylgard 170 = Black). These materials have been mechanically characterised in a
previous study (Doyle et al., 2009e) using uniaxial tensile testing and numerical
methods to determine the optimum strain energy function (SEF), and thus generate
material coefficients. This earlier work established that a 1st order Ogden SEF
accurately describes both materials, with the calculated material coefficients shown in
Table 2. It has been shown that these material coefficients accurately model the
behaviour of the material in both planar dumb-bell geometries and also in more
complex 3D structures (Doyle et al., 2009e). The general Ogden SEF (Ogden, 1984)
takes the form of Equation 1.
N 2μ p
W (λ1, λ2 , λ3 ) = ∑
α αp αp
(λ1 p + λ2 + λ3 − 3) (1)
p =1 αp 2
Table 2: Material coefficients for the 1st order Ogden SEF for Sylgard 160 and Sylgard
170.
μ α
Sylgard 160 1.6525 3.2395
Sylgard 170 0.6988 2.9741
158
AAA Model Creation

An effective method of manufacturing realistic AAA silicone rubber models has been
previously reported (Doyle et al., 2008b). Briefly, the process utilises the injection-
moulding lost-wax procedure. This entails the design and machining of aluminium
moulds using computer-aided design (CAD) and computer-aided manufacturing
(CAM). The moulds consist of a female and male set. The female moulds are used to
create the inner wax model, which once produced is inserted into the male mould set.
The male mould has a 2 mm uniform cavity into which the liquid silicone-rubber is
injected. Upon curing of the silicone, the wax is melted from the rubber model,
resulting in a complete, hollow, silicone model. Each model is then thoroughly
examined for flaws and defects before any experimental testing is performed. For each
of the 5 geometries used in this study, 4 silicone rubber models were created, 2 were
manufactured using Sylgard 160 which are grey in colour, and 2 from Sylgard 170
which are black in colour. In total, 20 silicone rubber AAA models (Sylgard 160, n=10;
Sylgard 170, n=10) were created and tested. This enabled the effect of material strength
on rupture location to be examined. Figure 1 shows the visual difference between the
two materials for the ideal AAA case.
Figure 1: Idealised AAA models created with Sylgard 160 (Grey, UTS = 4 MPa) and
Sylgard 170 (Black, UTS = 2 MPa).
CT Scanning and 3D Reconstruction of Silicone AAA Models

Although the silicone rubber models are designed to have a uniform wall thickness,
minor deviations may still exist (O’Brien et al., 2005; Doyle et al., 2008b). Each model
was imaged using CT on a Siemens Somotom Sensation 64 at the HSE Midwestern
159
Regional Hospital, Ireland. In order to obtain the most accurate image of each model,
the slice increment was reduced to 0.5 mm with a pixel size of 0.287 mm. These
parameters were important in order to accurately capture any variations in wall
thickness from the 2 mm design and ultimately produce exact 3D reconstructions. The
reconstruction of the silicone models was performed in a similar manner to the
reconstruction of the original AAAs using Mimics v12. The 5 AAA geometries can be
seen in Figure 2.
Figure 2: 3D reconstructions showing various AAA geometries used. The ideal AAA is
shown from the anterior and Patients 1-4 are shown from the right. Models are not
shown to scale.
Experimental Modelling
An experimental test rig was designed allowing each model to be connected to a
pneumatic air source and inflated to the point of rupture. A high-speed camera (Photron
Fastcam SA1-1, Photron USA Inc.) was used at a frame rate of 2,000 frames per second
(fps). This frame rate allowed the point of rupture to be accurately captured. The
positioning of a series of mirrors surrounding the experimental model allowed a full
360º view of the model, ensuring that the position of rupture was captured by the high-
speed camera. The test rig consisted of a pressure manometer, series of mirrors,
pneumatic airline, pressure regulator and high-speed camera. When fully connected to
the test rig, the silicone rubber model is constrained from movement at the proximal
neck region. The inside of each model was dusted with calcium carbonate to ease
visualisation of the rupture. The internal pressure was increased so that the rupture of
the model occurred within 240s of testing, in accordance with the standards BS ISO
1402 (British Standards, 1997) for burst pressure tests. Air pressure readings were also
recorded in the video image for analysis post-rupture. The test set-up as viewed through
the high-speed camera is shown in Figure 3. All silicone rubber models were inflated
160
until failure with each test recorded using the high-speed camera. Once rupture had
occurred, the test was terminated, and the video recording exported. Significance of
rupture pressures and geometrical parameters were analysed using a Spearman’s Rho
correlation test in SPSS 15 (SPSS Inc., Chicago, Ill, USA).
Figure 3: View through high-speed camera of test rig. Mirrors allow 360º of model to
be captured.
Computational Modelling
To correlate experimental and numerical results, the experiments were reproduced in the
finite element solver ABAQUS v6.7 (Dassault Systemes, SIMULIA, Rhode Island,
USA). 3D reconstructions of the silicone models were used in these analyses. Each
model was rigidly constrained at the proximal and distal regions, and a uniform static
air pressure applied to the internal surface of each model. A surface mesh was created
using Mimics v12 before being exported to ABAQUS v6.7. Within the finite element
solver, the surface mesh of each model was converted from a triangular surface mesh to
a tetrahedral mesh of quadratic 3D stress elements, with mesh independence performed
as previously published (Wang et al., 2002; Truijers et al., 2007; Doyle et al., 2007,
2009a, 2009b). This mesh independence study involves increasing the number of
elements of the mesh until the peak wall stress and wall stress distribution does not
significantly change (<2%). Representative numerical models are shown in Figure 4.
161
Optimum mesh sizes were determined to be: Ideal AAA = 27,000 elements; Patient 1 =
25,000 elements; Patient 2 = 24,000 elements; Patient 3 = 20,000 elements; Patient 4 =
23,000 elements.
Figure 4: Representative meshed AAA models. Models are not shown to scale and
from the right (except Ideal AAA).
To examine if the location of elevated stresses altered with increasing internal pressure
the idealised model was initially subjected to an internal pressure of 120 mmHg (16
kPa), and the location of elevated stresses recorded. The pressure was then increased in
60 mmHg (8 kPa) increments to 360 mmHg (48 kPa), and the locations of elevated
stresses again recorded. Finally, the internal pressure was increased to the average
rupture pressure of that material.
Assessment of Surface Curvature

As each AAA geometry is a complex 3D structure and it was important to identify
regions of increased inflection on the AAA surface. Inflection regions can be linked to
areas where the surface curvature changes from positive to negative. The Gaussian
surface curvature, which may be associated with corresponding elevations in wall stress
(Sacks et al., 1999), was quantified using ProEngineer Wildfire 4.0 (Parametric
Technology Corporation, Needham, MA, USA). These measurements will highlight
areas of high surface curvature which can be linked to regions of inflection.
Measurement of Wall Thickness

The variation in wall thickness in each of the five AAA geometries was examined using
the CT images in Mimics v12. The wall thickness was measured at eight equidistant
points at five cross-sections along the length of the AAA. This resulted in 40 wall
thickness measurements per model. The process of measurement is shown in Figure 5,
162
with Figure 6 illustrating the cross-sections that were analysed in each geometry. This
methodology is based on previously published wall thickness reports of experimental
rubber models (O’Brien et al., 2005; Doyle et al., 2008b, 2009d), except in this study
the reconstruction software was utilised instead of manual measurement. Results for
each model were then analysed using SPSS 15 to determine the 25th and 75th
percentiles, maximum and minimum wall thickness, and also the mean and standard
deviation (SD) wall thickness.
Figure 5: Method of measuring wall thickness. Example shown is Patient 1. Specific

cross-sections are analysed using Mimics v12.
Figure 6: Cross-sections which were assessed for wall thickness. Distance between each
cross-section was 20 mm for all cases.
163
Qualitative Assessment of Rupture Locations

In order to qualitatively assess the rupture sites observed experimentally and
computationally, a grid was constructed around each model to aid the post-processing of
results. The grid reference of each rupture site, and subsequent high stress regions
could then be compared. Figure 7 shows this grid system with Patient 1 used as an
example. In most cases only one view was necessary for each model to compare
results. In all figures illustrating wall stress results, red indicates high wall stress and
blue indicates low wall stress.
Figure 7: Grid system used to compare results between experimental and computational
results. Example shown is Patient 1.
RESULTS
The results of the experimental rupture tests revealed a large range of burst pressures for
the various AAA models as shown in Figure 8. On average, the AAA models created
using the weaker silicone rubber (Sylgard 170) ruptured at lower pressures than those
made using the stronger silicone type (Sylgard 160). Sylgard 160 models ruptured at a
mean ± SD pressure of 650.6 ± 195.1 mmHg (range = 381.4 – 985 mmHg), whereas the
weaker Sylgard 170 ruptured at a mean ± SD pressure of 410.7 ± 159.9 mmHg (range =
252.2 – 714 mmHg). There was no statistical significance (P>0.05) between the burst
pressure of the Sylgard 160 models and the geometrical parameters of Table 1
(Diameter, P=0.787; Length, P=0.684; Volume, P=0.787; Surface Area, P=0.586).
Correlating the burst pressures of the Sylgard 170 models showed that there was a
significant relationship (P<0.05) between burst pressure and surface area (Surface area,
P=0.01) but not between burst pressure and diameter, length or volume (Diameter,
P=0.088; Length, P=0.229; Volume, P=0.367).
164
Figure 8: Burst pressures for the various AAA models examined.
The ideal AAA models ruptured at the distal regions of inflection and not at regions of
maximum diameter. These rupture locations are consistent with those previously
reported (Doyle et al., 2008a, 2009d). All realistic AAA geometries also ruptured at
regions of inflection. Only the Sylgard 160 models of Patient 1 ruptured at regions near
the maximum diameter, although these rupture locations experienced localised
inflection as there was a sharp change in curvature over a very short length. The
location of rupture in other models varied. A typical sequence of events can be seen in
Figure 9, showing the frames immediately prior to rupture, through to the point of
complete failure. Case shown here is Patient 4 Sylgard 160 Model 1. The model in the
centre of each image in Figure 9 is the original model, with the reflected views to either
side.
165
Figure 9: Typical sequence of events of rupture test. (A) Model is inflected with air, (B)
silicone rubber fails (highlighted in figure), (C) tear develops (highlighted in figure)
until (D) AAA model completely fails.
As it was believed that the experimental models would rupture at areas of high stress, it
was important to examine the effect of pressure on resultant stress distributions. The
ideal AAA model of Sylgard 170 was analysed at increasing pressure loadings from
120-360 mmHg and also at the average experimental rupture pressure for Sylgard 170
models of 410 mmHg. From the resulting stress contours, it was observed that the
distribution of stress on the AAA wall does not alter with increased loading, rather, the
magnitude of the stress increases, as shown in Table 3 and Figure 10. The location of
the elevated stresses did not significantly change, remaining at the regions of inflection.
Table 3: Percentage increase in peak wall stress with increased internal pressure
loading.
Pressure Peak Wall Stress Increase in Peak Stress
(mmHg) (MPa) (%)
120 0.4378 -
180 0.6505 49
240 0.8875 36
300 1.1620 31
360 1.4890 28
166
Figure 10: Stress distributions on the outer surface of the undeformed ideal AAA with
increasing pressure loading.
Stress distributions on the realistic AAA geometries of Patients 1-4 revealed that an
AAA may have several regions of elevated stress in the diseased wall. These elevated
stress regions are due to the morphology of the particular AAA and also minor
variations in wall thickness from the manufacturing process. Figure 11 shows the stress
contours observed in the AAA case of Patient 1. In this particular case, the proximal
anterior region of the AAA experiences the peak wall stress, with elevated stresses
(shown as green areas in Figure 11) also along the midsection of the left, right and
posterior walls.
Figure 11: Resulting stress contours on the outer surface of the Sylgard 170 AAA of
Patient 1. Internal pressure for this case was the average rupture pressure for the Sylgard
170 silicone rubber (410 mmHg). Model shown in the undeformed state.
Figure 12 presents the peak wall stresses categorised by silicone type with the peak wall
stress results for each particular AAA shown in Table 4. All Sylgard 160 models were
167
subjected to an internal pressure of 650 mmHg and all Sylgard 170 models to 410
mmHg. At these mean experimental loadings, the FEA peak wall stress results show
that in all cases for the Sylgard 160 models and 70% of the Sylgard 170 models, the
peak stress does not exceed the UTS of the material. Mean peak wall stress ± SD for
the Sylgard 160 and Sylgard 170 models was 2.15 ± 0.26 MPa (range = 1.89 – 2.59
MPa) and 1.69 ± 0.38 MPa (range = 1.17 – 2.26 MPa), respectively. Table 4 also
presents these peak wall stresses as percentages of the UTS.
Figure 12: Box and whisker plot showing peak wall stress categorised by silicone type.
The boxes indicate the 25th and 75th percentiles of the group, and the whiskers show the
maximum and minimum wall stresses. Horizontal lines indicate UTS of Sylgard 160 (4
MPa) and Sylgard 170 (2 MPa).
Table 4: Peak wall stress results for each AAA case. Shown also is the peak stress as a
percentage of the UTS.
Sylgard 160* Sylgard 170**
Peak Stress Peak Stress
Model (MPa) % of UTS Model (MPa) % of UTS
1 1.902 48 1 1.821 91
Ideal
2 2.142 54 2 1.289 65
1 2.275 57 1 1.830 92
Patient 1
2 2.199 55 2 2.079 104
1 2.100 53 1 1.489 75
Patient 2
2 2.470 62 2 2.259 130
1 1.764 44 1 2.156 108
Patient 3
2 1.890 47 2 1.502 75
1 2.196 55 1 1.174 59
Patient 4
2 2.598 65 2 1.390 70
* Sylgard 160 models: Internal pressure = 650 mmHg; UTS = 4 MPa.
** Sylgard 170 models: Internal pressure = 410 mmHg; UTS = 2 MPa.
168
Assessment of Inflection Regions

Due to the complex nature of realistic AAA geometries, many areas exhibit high
positive and negative surface curvature. The resulting Gaussian surface curvature
contour plots for each case are shown in Figures 16, 19, 22, 26 and 30. These figures
aid the comparison of rupture locations to regions of elevated wall stress. Table 5
presents the maximum and minimum surface curvature for each AAA geometry.
Table 5: Maximum and minimum Gaussian surface curvature for each AAA geometry.
Model Maximum Minimum
Ideal AAA 0.0081 -0.021
Patient 1 0.079 -0.091
Patient 2 0.045 -0.089
Patient 3 0.062 -0.078
Patient 4 0.14 -0.185
Wall Thickness
The results of the wall thickness study can be seen in Table 6 and Figure 13. The mean
± SD wall thickness of all the five models (n=200) examined was 2.19 ± 0.40 mm. The
maximum wall thickness recorded was 3.7 mm and the minimum was 0.91 mm. The
wall thickness was also measured at the exact site of rupture in each experimental
model. These results are presented in Table 7. The mean ± SD wall thickness at the
rupture sites were 1.85 ± 0.33 mm and 1.71 ± 0.29 mm for the Sylgard 160 and Sylgard
170 materials, respectively. There was no statistical significance (P>0.05) between
rupture pressures and wall thickness at rupture site (Sylgard 160 models: P=0.688,
Sylgard 170 models: P=0.881).
Table 6: Results of the wall thickness study. N=40 for each case. All results are in
millimetres.
Ideal AAA Patient 1 Patient 2 Patient 3 Patient 4
Mean 2.08 2.26 2.22 2.19 2.22
Median 2.02 2.25 2.14 2.15 2.05
SD 0.26 0.39 0.50 0.49 0.38
Minimum 1.45 1.53 1.01 0.91 1.81
Maximum 2.60 2.96 3.70 3.15 3.46
Range 1.15 1.43 2.69 2.24 1.65
25th Percentile 2.00 2.01 2.01 1.87 2.00
th
75 Percentile 2.27 2.64 2.33 2.51 2.40
169
Figure 13: Box and whisker plot showing wall thickness categorised by AAA case. The
boxes indicate the 25th and 75th percentiles of the group, and the whiskers show
maximum and minimum measurements.
Table 7: Wall thickness measurements at the site of experimental rupture.

Sylgard 160 Sylgard 170
Rupture Thickness Rupture Thickness
Model Model
(mm) (mm)
1 1.96 1 1.82
Ideal
2 1.75 2 1.72
1 1.84 1 1.51
Patient 1
2 2.03 2 1.7
1 1.41 1 1.41
Patient 2
2 1.24 2 1.97
1 2.09 1 1.21
Patient 3
2 1.78 2 1.57
1 2.34 1 2.1
Patient 4
2 2.09 2 2.11
170
Comparison of Rupture Locations
Idealised AAA
Sylgard 170
The Sylgard 170 idealised AAA models ruptured at regions of inflection, with good
agreement in the experimental and computational results, as shown in Figure 14. Both
models ruptured at the distal region of inflection, with FEA predicting that these regions
experience elevated stress. Figure 16 shows the corresponding surface curvature for
this geometry. From Figure 16 it can be seen that the distal region of the AAA sac
exhibits changes in curvature, resulting in high wall stress.
Figure 14: Comparison of rupture sites with regions of high stress for Sylgard 170
(Black) idealised models. Model 1 is shown from the right and Model 2 from the
anterior. Numerical models shown in undeformed state with an internal pressure of 410
mmHg. Rupture sites: Model 1 = E8; Model 2 = E7.
Sylgard 160
Model 1 in Figure 15 ruptured at a region of elevated stress but not at the region of peak
stress. Regions of maximum diameter exhibit lower wall stress compared to these
inflection zones, and do not experience rupture. High wall stress corresponds with
regions where the surface curvature changes rapidly (Figure 16).
171
(Grey) idealised models. Model 1 is shown from the anterior and Model 2 from the
right. Numerical models shown in undeformed state with an internal pressure of 650
mmHg. Rupture sites: Model 1 = E5; Model 2 = E7.
Figure 16: Contour plot of Gaussian curvature for the idealised AAA.
Patient 1
Sylgard 170
Comparing the results of Patient 1 revealed that for the Sylgard 170 models (Figure 17),
the proximal anterior region experiences peak wall stress, with experimental models
rupturing at these sites. Examining the anterior view in Figure 19, the yellow regions
indicate areas of complex curvature, resulting in elevated wall stress and experimental
rupture.
172
(Black) models of Patient 1. Both models are shown from the anterior. Numerical
models shown in undeformed state with an internal pressure of 410 mmHg. Rupture
sites: Model 1 = E14; Model 2 = F14.
Sylgard 160
Both the Sylgard 160 models (Figure 18) ruptured at a region close to maximum
diameter, with peak wall stress predicted at this region. This area experiences a
localised change in curvature from C9 - E10 in Figure 19, resulting in elevated wall
stress. Examining the posterior surface curvature presented in Figure 19 shows a region
of high curvature close to the rupture site. The rupture site shifts from the anterior wall
of the AAA for Sylgard 170 models to the posterior wall for the higher UTS Sylgard
160 models.
173
(Grey) models of Patient 1. Both models shown from the posterior. Numerical models
shown in undeformed state with an internal pressure of 650 mmHg. Rupture sites:
Model 1 = F11; Model 2 = F11.
Figure 19: Contour plot of Gaussian curvature for Patient 1.
Patient 2
Sylgard 170
Patient 2 also experienced both rupture (experimental) and peak wall stress (FEA) at
inflection regions. Model 1 ruptured at the anterior proximal region and Model 2 at the
anterior distal region, both corresponding to areas of peak wall stress. Comparing these
rupture and high stress zones to the surface curvature of Figure 22 shows that these
regions are areas experiencing shifts in curvature (high curvature shown in red).
174
(Black) models of Patient 2. Both models are shown from the anterior. Model 2 is also
shown distally. Numerical models shown in undeformed state with an internal pressure
of 410 mmHg. Rupture sites: Model 1 = E14; Model 2 = E3.
Sylgard 160
Both Sylgard 160 models ruptured at the posterior proximal region of the AAA sac.
These rupture sites correlate with regions of peak wall stress (Figure 21). Peak wall
stress occurs at regions of inflection (Figure 22, inflection regions shown in yellow)
where the surface curvature changes from positive to negative. It was noticed in this
geometry, as with Patient 1, that the rupture site shifts from the anterior wall to the
posterior wall when the material changes from Sylgard 170 to Sylgard 160.
175
(Grey) models of Patient 2. Both models are shown from the posterior. Numerical
models shown in undeformed state with an internal pressure of 650 mmHg. Rupture
sites: Model 1 = D13; Model 2 = D13.
Patient 3
Sylgard 170
The rupture location and region of peak wall stress correlated well for Model 1 (Figure
23), which also corresponds to changes in surface curvature (Figure 26). The rupture
location and wall stress distribution for Model 2 however did not correlate (Figure 23).
This discrepancy was investigated further. By examining the CT scans of Model 2, a
minute tear on the inner surface of the model at the rupture zone was identified (Figure
24). This tear subsequently influenced the rupture of the model.
176
(Black) models of Patient 3. Model 1 is shown from the left and also shown proximally.
Model 2 is shown proximally and also from the left. Numerical models shown in
undeformed state with an internal pressure of 410 mmHg. Rupture sites: Model 1 =
C14; Model 2 = C10.
Figure 24: CT scan at rupture site of Model 1 revealed a small tear in the model prior to
experimental rupture.
Sylgard 160
Both the models of Sylgard 160 experienced experimental rupture at regions of FEA
predicted peak wall stress (Figure 25), both in the proximal region of the AAA sac.
Again, these regions both experienced rapid changes in the local surface curvature, as
shown in Figure 26.
177
(Grey) models of Patient 3. Model 1 is shown from the left. Model 2 is shown
proximally and also shown from the left. Numerical models shown in undeformed state
with an internal pressure of 650 mmHg. Rupture sites: Model 1 = D14; Model 2 = E12.
Patient 4
Sylgard 170
Comparing the results of Model 1 revealed that the experimental rupture zone did not
correlate with regions of elevated wall stress. Rupture would be expected to occur at
the region of B12 (Figure 26, Model 1). The CT images for this case did not reveal any
inconsistencies in the wall of the model. A stereomicroscope was then used to examine
a typical surface cross-section with that of the rupture site cross-section in this model.
This comparison is shown in Figure 28(A and B). Microscopic air bubbles can be seen
trapped in the model wall which contributed to failure of the model at this location.
178
(Black) models of Patient 4. Model 1 is shown from the left. Model 2 is shown distally
and also shown from the left. Numerical models shown in undeformed state with an
internal pressure of 410mmHg. Rupture sites: Model 1 = F5; Model 2 = G12.
Figure 28: (A) Stereomicroscope image of typical surface (B) Cross-section at rupture
site revealing entrapment of microscopic air bubbles (highlighted in figure).
Sylgard 160
The Sylgard 160 Model 1 results did also differ in rupture and peak stress location
(Figure 29). In this case, the model ruptured at the proximal inflection region, with
FEA predicting peak wall stress at the distal inflection area. There was however, a
localised region of elevated wall stress at F16 (Figure 29, Model 1), with a
corresponding change in surface curvature (anterior view of Figure 30) which may have
caused rupture. The rupture and high stress results of Model 2 agreed favourably, with
rupture occurring at the region of peak wall stress.
179
(Grey) models of Patient 4. Both models are shown from the anterior. Model 1 is also
shown distally. Numerical models shown in undeformed state with an internal pressure
of 650 mmHg. Rupture sites: Model 1 = F16; Model 2 = F5.
DISCUSSION
AAA Models
This study has explored experimental and computational techniques of determining
rupture locations in abdominal aortic aneurysms. Five AAA geometries were studied,
one idealised AAA case based on realistic dimensions (Laheij et al., 2001) and four
realistic AAA cases. These AAAs had a mean diameter of 5.5 cm (range = 5.0 – 6.1
cm) and the study subjects consisted of two males and two females with a mean age of
70 years (range = 65 – 78 years). Anatomically-correct silicone rubber AAA models
were created using a previously reported injection-moulding technique (Doyle et al.,
2008b). Sylgard 160 and Sylgard 170 were selected as the AAA wall analogue for this
180
study, as both materials differ in mechanical properties (Sylgard 160, UTS = 4 MPa;
Sylgard 170, UTS = 2 MPa). The material coefficients (Table 2) were determined in a
previous study (Doyle et al., 2009e). Four experimental models were manufactured for
each AAA case, two created using Sylgard 160 and two using Sylgard 170. This
resulted in a total of twenty experimental models. CT images of each model were
acquired using a Siemens Somotom Sensation 64 and the standard imaging protocol for
the inner ear. This allowed the slice increment of each scan be reduced to 0.5 mm and
the pixel size to be 0.287 mm. Images at this resolution produce exact 3D
reconstructions of the experimental model, capturing any variance in wall thickness
from the original 2 mm design throughout the model. These exact reconstructions were
exported for use with ABAQUS v6.7 and also further analysed using Mimics v12.
The procedure for the experimental rupture tests has been previously reported (Doyle et
al., 2009d) and was adapted in this current work. A recording rate of 2,000 frames per
second (fps) was used. Mirrors allowed all aspects of each in vitro model to be captured
with the camera (Figure 3). Air pressure was applied to the inner wall of each model
until rupture occurred. The pressure was increased at a rate so that rupture occurs
within 240s of testing, complying with BS ISO 1402. Figure 9 shows the rupture of a
realistic AAA case. The inner wall of each model was dusted with calcium carbonate so
that the jet of powder expelled from the model upon failure would aid the visualisation
of the rupture site. Burst pressures were higher for the Sylgard 160 models (650.6 ±
195.1 mmHg) than the Sylgard 170 models (410 ± 159.9 mmHg) possibly due to the
differences in material UTS. Recent reports (Doyle et al., 2008a, 2009d) suggest that
idealised AAA in vitro models will rupture at regions of inflection. It has also been
identified that FEA predicts inflection regions as the predominant zones of elevated wall
stress in realistic AAA geometries (Doyle et al., 2007, 2009a, 2009b, 2009c). The
results presented in this study support this hypothesis as all models, both idealised and
realistic cases, ruptured at regions of inflection which can be determined from surface
curvature. Only two models out of twenty (10%) (Patient 1: Sylgard 160; Models 1 and
2) experienced rupture at a region close to maximum diameter, although the local
surface curvature at this rupture location changed rapidly over a short distance, and
therefore was a region of high inflection, and subsequent wall stress.
181
The 3D reconstructions generated from the detailed CT images for each case provided
good representations of the experimental models. A high quality surface mesh was
created in Mimics v12 on each model before conversion to a volume mesh in ABAQUS
v6.7. Mesh independence revealed the optimum numbers of elements for each case
using a previously reported protocol (Wang et al., 2002; Truijers et al., 2007; Doyle et
al., 2007, 2009a, 2009b).
An initial study was performed using the idealised AAA model. The model was
subjected to increasing internal pressures in order to observe changes in the wall stress
distribution. It was noted that the locations of peak and elevated stress do not
noticeably change with increasing load, but rather increase in magnitude (Table 3 and
Figure 10). Each model was then subjected to the corresponding mean experimental
rupture pressure. Figure 11 illustrates how a particular AAA geometry may have
several regions of elevated stress, indicating possible rupture sites. Figure 12 shows the
computed peak wall stress for both the Sylgard 160 and Sylgard 170 silicone rubbers,
with Table 4 presenting the peak wall stress for each particular AAA model. Mean peak
wall stress ± SD for the Sylgard 160 and Sylgard 170 models was 2.15 ± 0.26 MPa
(range = 1.89 – 2.59 MPa) and 1.69 ± 0.38 MPa (range = 1.17 – 2.26 MPa),
respectively. These results suggest that the FEA predicted peak wall stress does not
exceed the UTS of the material in some cases. This discrepancy may be attributed to
the model-specific experimental rupture pressures. Applying the model-specific rupture
pressure to the corresponding numerical model may increase peak stresses, although the
finite element solver can experience difficulties converging solutions at these large
pressures. This is particularly evident in the cases of the idealised AAA Sylgard 170
models. The high experimental pressures observed for these cases (Model 1 = 673.8
mmHg; Model 2 = 714 mmHg) produce excessive deformation of the FEA elements,
bringing the stresses and strains outside the behaviour of the material model. This
problem was alleviated by applying mean experimental rupture pressures to the
numerical models.
Surface Curvature
Realistic AAAs have complex asymmetric geometries which result in regions where the
signs and magnitudes of the surface curvatures can change rapidly. Table 5 presents the
maximum and minimum surface curvature for AAA geometries and shows how the
182
symmetric idealised AAA has a significantly lower magnitude of curvature compared to

the realistic AAA shapes. Each change in curvature from positive to negative is an
inflection point, and it is at these regions where the stress is believed to elevate.
Gaussian surface curvature results revealed that each AAA exhibits many regions of
inflection. Previous work reported that locations where the curvature changes sign can
be used to identify regions where the stress distribution is likely to be complex (Sacks et
al., 1999). This work supports this hypothesis. In the geometries studied here, the
surface curvature is linked to the resulting wall stress. Figure 31 compares the wall
stress and surface curvature of Patient 1. It can be seen in this figure how the regions of
elevated stress correspond to the regions of curvature where the sign rapidly changes.
The red and yellow spots in Figure 31(B) appear to create a region surrounding D12,
E13 - 14 and F13 - 14, which result in elevated wall stresses in this area. A similar
trend was observed in other geometries.
Figure 31: (A) Wall stress contour plot compared to (B) Gaussian surface curvature
contour plot.
Wall Thickness
The moulds used to create the experimental silicone rubber models are designed to have
a uniform 2 mm cavity to control the resulting wall thickness. Deviations in wall
thickness from the original 2 mm design have been previously reported (Doyle et al.,
2008b). In complex 3D models such as those presented in this study, strict control over
the wall thickness is difficult to achieve due to the contraction of the wax and the
correct positioning of the wax model within the outer wall mould. As high resolution
183
CT images were acquired for each AAA model, wall thickness was accurately measured
in the reconstruction software Mimics v12. Eight measurements were quantified at 5
cross-sections along the length of the different AAA geometries (Figures 5 and 6). This
resulted in 40 measurements per model, as shown in Table 6 and Figure 18. The mean
± SD wall thickness from all measurements (n=200) was 2.19 ± 0.40 mm, with the
minimum and maximum wall thickness observed to be 0.91 mm and 3.7 mm,
respectively. The mean wall thickness may be higher than the original 2 mm design due
to the contraction of the wax model upon solidification, whereas, the large range in wall
thickness may be attributed to the positioning of the wax model inside the outer mould.
The wall thickness at specific rupture sites are presented in Table 7. The relationship
between wall thickness and burst pressure was deemed to be insignificant (Sylgard 160
models: P=0.688, Sylgard 170 models: P=0.881). The mean ± SD wall thickness at the
rupture sites were 1.85 ± 0.33 mm and 1.71 ± 0.29 mm for the Sylgard 160 and Sylgard
170 materials, respectively. These values represent a difference in rupture wall
thickness to the measured mean wall thickness (2.19 mm) of 16% and 22% for the
Sylgard 160 and Sylgard 170 models, respectively. As wall thickness was close to the
mean value reported at the rupture sites, failure of the models may not have been
directly related to wall thickness, but rather due to geometrical factors and surface
anomalies. Previous work has highlighted (Doyle et al., 2009d) that wall thickness may
not significantly affect wall stress distribution, but rather increase or reduce the
magnitudes of these stresses.
Comparison of Rupture Locations

A grid system was used to qualitatively identify and correlate rupture locations with
regions of surface curvature and elevated wall stress. In 90% (18/20) models examined
the rupture locations agreed with the high stress regions predicted using FEA. Of the
models that did correlate with regions of elevated stress, 16/18 (89%) ruptured at
regions of FEA predicted peak wall stress, resulting in peak wall stress accurately
predicting the rupture location in 80% (16/20) of all cases examined. Local wall defects
seen in 10% (2/20) of the models, such as those presented in Figures 24 and 28, can
alter the location of rupture, shifting it from regions of elevated and peak wall stress to
the sites of surface anomalies.
184
Significance of Results
This study has investigated the rupture locations of AAA analogues and the correlation
of the locations with numerically predicted regions of peak and high wall stress.
Computational biomechanical analyses are becoming increasingly common as
researchers attempt to improve the current trend in AAA management. This is the first
attempt to assess the rupture location of anatomically-correct AAAs in vitro. It is
understood that an AAA will fail when the wall stress exceeds the wall strength, with
wall thickness and heterogeneity of the wall contributing to rupture. The results
presented in this report support the hypothesis that many factors influence rupture. Peak
wall stress deviates significantly from case to case, with variations in peak wall stress
witnessed between similar geometries depending on wall thickness. Peak stress is still
regarded as the primary outcome of FEA when concerned with AAAs, although some
have argued that other stresses should be considered, such as posterior wall stress
(Doyle et al., 2009a) or the 99th percentile of peak stress (Speelman et al., 2008). Other
stress parameters may influence rupture and should not be underestimated. The strength
of the AAA wall is believed to be patient-specific with recent reports of methods to
statistically predict strength based on relevant risk factors (Vande Geest et al., 2006).
Rupture of AAAs may also be dependent on wall heterogeneity. As witnessed in this
study, AAAs will fail at regions where the wall is locally damaged or defected. The
situation presented here, where two models failed due to defects in the wall, can be
compared to the in vivo setting of AAAs. Calcifications (Speelman et al., 2007; Li et
al., 2008), blebs, which are small blister-like formations (Hunter et al., 1989) and
localised hypoxia (Vorp et al., 2001) all affect the AAA wall. Wall thickness does
influence wall stress, yet it may not influence rupture to the same extent. The minimum
wall thickness observed in this current study was 0.91 mm, yet this region did not
experience rupture. Thin walled regions may be strong enough to withstand the
pulsatile forces of the cardiac cycle, whereas thicker regions may be weaker due to
conditions such as those mentioned, and vice versa. Therefore, not only is wall strength
a key factor of rupture, but surface anomalies may significantly influence the location of
AAA rupture.
This study has limitations which could be improved to further the results presented in
this work. The number of cases examined is low and therefore statistical significance
was not achieved. Increasing the numbers of models per AAA geometry may enhance
the results and highlight any significant relationships in rupture potential. Static air
185
pressure was used to pressurise the inner surface of the experimental models. In vivo
the AAA is subjected to a cyclic pulse of pressure and fluid force, factors which were
neglected in this study. The arterial wall analogue could be improved to create more
realistic material properties with non-uniform wall strength, more analogous to the in
vivo situation (Raghavan et al., 2006). Also, experimental models could be improved
with the inclusion of the intraluminal thrombus (ILT). Numerical modelling could also
be further developed to include surface defects already present prior to testing and
incorporating these flaws within the finite element model may yield interesting results.
Implementation of damage modelling to represent the actual failure of the silicone
should also be included.
CONCLUSIONS
Experimental modelling determined that AAAs rupture at regions of elevated wall stress
and not at regions of maximum diameter. These rupture locations were confirmed using
the finite element method. Computational results indicate several regions of elevated
wall stress in AAA geometries, with rupture occurring at the location of peak stress in
80% (16/20) of models. Surface defects affect rupture location more so than wall
thickness, moving the location of rupture away from the regions of high wall stress to
the defect region. The results presented may further contribute to the understanding of
AAA biomechanics and ultimately AAA rupture prediction.
ACKNOWLEDGEMENTS
The authors would like to thank our funding sources (i) the Irish Research Council for
Science, Engineering and Technology (IRCSET) Grant RS/2005/340 and (ii) Grant
#R01-HL-060670 from the US National Heart Lung and Blood Institute. The authors
would also like to thank the contribution of (i) Joseph Muthu, Departments of Surgery
and Bioengineering, Centre for Vascular Remodelling and Regeneration, University of
Pittsburgh, USA, and (ii) Dr. Fintan Wallis and Gena Nicholas from the Department of
Radiology, HSE Midwestern Regional Hospital, Limerick, Ireland.
REFERENCES
Callanan, A., Morris, L.G., McGloughlin, T.M., 2004. Numerical and experimental
analysis of an idealised abdominal aortic aneurysm. European Society of Biomechanics,
S-Hertogenbosch, Netherlands.
186
Cronenwett, J.L., Murphy, T.F., Zelenock, G.B., Whitehouse Jr., W.M., Lindenauer,
S.M., Graham, L.M., Quint, L.E., Silver, T.M., Stanley, J.C., 1985. Actuarial analysis of
variables associated with rupture of small abdominal aortic aneurysms. Surgery 98, 472-
483.
Darling, R.C., Messina, C.R., Brewster, D.C., Ottinger, L.W., 1977. Autopsy study of
unoperated abdominal aortic aneurysms. The case for early resection. Circulation 56(2),
161-164.
DiMartino, E.S., Guadagni, G., Fumero, A., Ballerini, G., Spirito, R., Biglioli, P,
Redaelli, A., 2001. Fluid-structure interaction within realistic 3D models of
aneurysmatic aorta as a guidance to assess the risk of rupture of the aneurysm. Medical
Engineering and Physics 23, 647-655.
Doyle, B.J., Callanan, A., McGloughlin, T.M., 2007. A comparison of modelling

techniques for computing wall stress in abdominal aortic aneurysms. Biomedical
Engineering Online 6, 38.
Doyle, B.J., Callanan, A., Corbett, T.J., Cloonan, A.J., O’Donnell, M.R., Vorp, D.A.,
McGloughlin, T.M., 2008a. The use of silicone to model abdominal aortic aneurysm
behaviour. Society of Plastics Engineers, SPE European Conference on Medical
Polymers, pp.115-120.
Doyle, B.J., Morris, L.G., Callanan, A., Kelly, P., Vorp, D.A., McGloughlin, T.M.,
2008b. 3D reconstruction and manufacture of real abdominal aortic aneurysms: From
CT scan to silicone model. Journal of Biomechanical Engineering 130, 034501-5.
Doyle, B.J., Callanan, A., Burke, P.E., Grace, P.A., Walsh, M.T., Vorp, D.A.,
McGloughlin, T.M., 2009a. Vessel asymmetry as an additional tool in the assessment of
abdominal aortic aneurysms. Journal of Vascular Surgery 49:443-454.
Doyle, B.J., Callanan, A., Walsh, M.T., Grace, P.A., McGloughlin, T.M., 2009b. A
finite element analysis rupture index (FEARI) as an additional tool for abdominal aortic
aneurysm rupture prediction. Vascular Disease Prevention 6, 114-121.
Doyle, B.J., Grace, P.A. Kavanagh, E.G., Burke, P.E. Wallis, F., Walsh, M.T.,
McGloughlin, T.M., 2009c. Improved assessment and treatment of abdominal aortic
aneurysms: The use of 3D reconstructions as a surgical guidance tool in endovascular
repair. Irish Journal of Medical Science, in press.
Doyle, B.J., Corbett, T.J., Callanan, A., Walsh, M.T., Vorp, D.A., McGloughlin, T.M.,
2009d. An experimental and numerical comparison of the rupture locations of an
abdominal aortic aneurysm. Journal of Endovascular Therapy, in press.
Doyle, B.J., Corbett, T.J., Cloonan, A.J., O’Donnell, M.R., Walsh, M.T., Vorp, D.A.,
McGloulghin, T.M., 2009e. Experimental modeling of aortic aneurysms: Novel
applications of silicone rubbers. Medical Engineering & Physics, submitted.
Fillinger, M.F., Marra, S.P., Raghavan, M.L., Kennedy, F.E., 2003. Prediction of
rupture risk in abdominal aortic aneurysm during observation: wall stress versus
diameter. Journal of Vascular Surgery 37, 724-732.
187
Fillinger, M.F., Raghavan, M.L., Marra, S.P., Cronenwett, J.L., Kennedy, F.E., 2002. In
vivo analysis of mechanical wall stress and abdominal aortic aneurysm rupture risk.
Journal of Vascular Surgery 36, 589-597.
Giannoglu, G., Giannakoulas, G., Soulis, J., Chatzizisis, Y., Perdikides, T., Melas, N.,
Parcharidis, G., Louridas, G., 2006. Predicting the risk of rupture of abdominal aortic
aneurysms by utilizing various geometrical parameters: Revisiting the diameter
criterion. Angiology 57(4), 487-494.
Glimaker, H., Holmberg, L., Elvin, A., Nybacka, O., Almgren, B., Bjorck, C.G.,
Eriksson, I., 1991. Natural history of patients with abdominal aortic aneurysm.
European Journal of Vascular Surgery 5, 125-130.
Hua, J., Mower, W.R., 2001. Simple geometric characteristics fail to reliably predict
abdominal aortic aneurysm wall stress. Journal of Vascular Surgery 34, 308-315.
Hunter, G.C., Leong, S.C., Yu, G.S., McIntyre, K.E., Bernhard, V.M., 1989. Aortic
blebs: Possible site of aneurysm rupture. Journal of Vascular Surgery 10(1), 93-99.
Kleinstreuer, C., Li, Z., 2006. Analysis and computer program for rupture-risk
prediction of abdominal aortic aneurysms. Biomedical Engineering Online 5, 19.
Laheij R, van Marrewijk C, Buth J., 2001. Progress report on the procedural and follow
up results of 3413 patients who received stent graft treatment for infrarenal aortic
aneurysms for a period of 6 years. EUROSTAR Data Registry Centre.
Leung, J.H., Wright, A.R., Cheshire, N., Crane, J., Thom, S.A., Hughes, A.D., Xu, Y.,
2006. Fluid structure interaction of patient specific abdominal aortic aneurysms: a
comparison with solid stress models. Biomedical Engineering Online 5, 33.
Li, Z.Y., U-King-Im, J., Tang, T.Y., Soh, E., See, T.C. Gillard, J.H., 2008. Impact of
calcification and intraluminal thrombus on the computed wall stresses of abdominal
aortic aneurysm. Journal of Vascular Surgery 47, 928-935.
Morris, L., O’Donnell, P., Delassus, P., McGloughlin, T., 2004. Experimental
assessment of stress patterns in abdominal aortic aneurysms using the photoelastic
method. Strain 40, 165-172.
Mower, W.R., Quinones, W.J., Gambhir, S.S., 1997. Effect of intraluminal thrombus on
abdominal aortic aneurysm wall stress. Journal of Vascular Surgery, 26, 602-608.
Nicholls, S.C., Gardner, J.B., Meissner, M.H., Johansen, H.K., 1998. Rupture in small
abdominal aortic aneurysms. Journal of Vascular Surgery, 28, 884-888.
O’Brien, T., Morris, L., O’Donnell, M., Walsh, M., McGloughlin, T., 2005. Injection-
moulded models of major and minor arteries: The variability of model wall thickness
owing to casting technique. Proceedings of the International Mechanical Engineers 219,
Part H: Journal of Engineering in Medicine.
Ogden, R.W., 1984. Nonlinear elastic deformations. Dover Publication Inc., Mineola,
NY, USA.
188
Papaharilaou, Y., Ekaterinaris, J.A., Manoussaki, E., Katsamouris, A.N., 2007. A

decoupled fluid structure approach for estimating wall stress in abdominal aortic
aneurysms. Journal of Biomechanics 40, 367-377.
Raghavan, M.L., Vorp, D.A., Federle, M.P., Makaroun, M.S., Webster, M.W., 2000.
Wall stress distribution on three-dimensionally reconstructed models of human
abdominal aortic aneurysm. Journal of Vascular Surgery 31, 760-769.
Raghavan, M.L., Kratzberg, J., de Tolosa, E.M.C., Hanaoka, M.M., Walter, P., da Silva,
E.S., 2006. Regional distribution of wall thickness and failure properties of human
abdominal aortic aneurysm. Journal of Biomechanics, 39(16), 3010-6.
Rissland, P., Alemu, Y., Einav, S., Ricotta, J., Bluestein, D., 2009. Abdominal aortic
aneurysm risk of rupture: Patient-specific FSI simulations using anisotropic model.
Journal of Biomechanical Engineering 131, 031001-0310010.
Sacks, M.S., Vorp, D.A., Raghavan, M.L., Federle, M.P., Webster, M.W., 1999. In vivo
three-dimensional surface geometry of abdominal aortic aneurysms. Annals of
Biomedical Engineering, 27, 469-479.
Sayers, R.D., 2002. Aortic aneurysms, inflammatory pathways and nitric oxide. Annals
of the Royal College of Surgeons England 84(4), 239-246.
Scotti, C.M., Shkolnik, A.D., Muluk, S.C., Finol, E., 2005. Fluid-structure interaction in
abdominal aortic aneurysms: effect of asymmetry and wall thickness. Biomedical
Engineering Online 4, 64.
Scotti, C.M., Finol, E.A., 2007. Compliant biomechanics of abdominal aortic

aneurysms: a fluid-structure interaction study. Computers and Structures 85, 1097-1113.
Speelman, L., Bohra, A., Bosboom, E.M.H., Schurink, G.W.H., van de Vosse, F.N.,
Makaroun, M.S., Vorp, D.A., 2007. Effects of wall calcifications in patient-specific wall
stress analyses of abdominal aortic aneurysms. Journal of Biomechanical Engineering
129, 1-5.
Speelman, L., Bosboom, E.M.H., Schurink, G.W.H., Hellenthal, F.A.M.V.I, Buth, J.,
Breeuwer, M., Jacobs, M.J., van de Vosse, F.N., 2008. Patient-specific AAA wall stress
analysis: 99-percentile versus peak stress. European Journal of Vascular and
Endovascular Surgery, 36(6), 668-676.
Thubrikar, M.J., Al-Soudi, J., Robicsek, F., 2001.Wall stress studies of abdominal aortic
aneurysm in a clinical model. Annals of Vascular Surgery, 15,355-366.
Truijers, M., Pol, J.A., SchultzeKool, L.J., van Sterkenburg, S.M., Fillinger, M.F.,
Blankensteijn, J.D., 2007. Wall stress analysis in small asymptomatic, symptomatic and
ruptured abdominal aortic aneurysms. European Journal of Vascular and Endovascular
Surgery 33, 401-407.
Vande Geest, J.P., Di Martino, E.S., Bohra, A., Makaroun, M.S., Vorp, D.A., 2006. A
biomechanics-based rupture potential index for abdominal aortic aneurysm risk
assessment. Annals of the New York Academy of Science 1085, 11-21.
189
Vande Geest, J.P., Wang, D.H.J., Wisniewski, S.R., Makaroun, M.S., Vorp, D.A., 2006.
Towards a non-invasive method for determination of patient-specific wall strength
distribution in abdominal aortic aneurysms. Annals of Biomedical Engineering, 34(7),
1098-1106.
Venkatasubramaniam, A.K., Fagan, M.J., Mehta, T., Mylankal, K.J., Ray, B., Kuhan,
G., Chetter, I.C., McCollum, P.T., 2004. A comparative study of aortic wall stress using
finite element analysis for ruptured and non-ruptured abdominal aortic aneurysms.
European Journal of Vascular and Endovascular Surgery 28, 168-176.
Vorp, D.A., Raghavan, M.L., Webster, M.W., 1998. Mechanical wall stress in
abdominal aortic aneurysm: influence of diameter and asymmetry. Journal of Vascular
Surgery 27(4), 632-639.
Vorp, D.A., Lee, P.C., Wang, D.H.J., Makaroun, M.S., Nemoto, E.M., Ogawa, S.,
Webster, M.W., 2001. Association of intraluminal thrombus in abdominal aortic
aneurysm with local hypoxia and wall weakening. Journal of Vascular Surgery 34, 291-
299.
Wang, D.H.J., Makaroun, M.S., Webster, M.W., Vorp, D.A., 2002. Effect of
intraluminal thrombus on wall stress in patient-specific models of abdominal aortic
aneurysm. Journal of Vascular Surgery 36, 598-604.
190
SUMMARY AND CONCLUSIONS
Summary and Conclusions
The reconstruction of CT scans to 3D images has many benefits for the clinician. Not
only does this process allow the surgeon to visualise the problem prior to surgery, but it
can also provide some useful insights into the particular aneurysm. Complications with
EVAR can arise due to the geometrical nature of the AAA, in particular when there is
an angular or tortuous proximal neck. This can significantly increase the complexity of
the repair and also diminish confidence in the patency and fixation of the device. 3D
reconstructions can also highlight irregularities in the iliac arteries. Access to the AAA
is usually gained via the femoral artery, with the device guided through the iliac arteries
towards the attachment site below the renal arteries. In the event of unsuitable iliac
arteries, either due to tortuosity or occlusion, alternative access may be required. The
EUROSTAR registry reported that 13% of patients experience problems with access
through the iliac arteries. The common carotid artery has been used as an alternative
access site in cases where the iliac arteries are unsuitable, and 3D reconstruction may
also alleviate intraoperative access problems. 3D reconstructions also offer the clinician
the ability to determine exact measurements of the particular AAA. Traditionally,
EVAR stent-grafts are sized using measurements obtained from 2D CT images. These
critical measurements are usually the proximal neck diameter, total length from the
renal arteries to the fixation point in the iliac arteries, and also the iliac artery diameter.
When determining size measurements from 2D images, sizes may be over or under
estimated. Measurements from 3D reconstructions are exact and may help reduce the
need for a range of stent-grafts with varying sizes to be “at-hand” for the surgeon during
the operation. 3D reconstructions also allow effective design for patient-tailored stent-
grafts, such as fenestrated devices. Post-operative examination also benefits from 3D
reconstruction as complications such as type 2 endoleaks can be identified. Recent
advancements in medical imaging allows clinicians to perform reconstructions on-site
immediately after scanning and may provide more useful information than traditional
2D images.
Not only do 3D reconstructions provide useful clinical guidance, but also they allow
engineering techniques to be applied to the problem. Over recent years there has been
much interest in examining the biomechanical aspects of AAAs, in particular, the use of
the finite element method to calculate wall stress, and ultimately, in determining the
propensity of an AAA to rupture. Although there have been many reports indicating
191
wall stress as a possible rupture parameter (Fillinger et al., 2002, 2003;

Venkatasubramaniam et al., 2004), there is considerable variation between reported
methods of calculating this stress. An in-depth examination of the more common
techniques published revealed interesting findings and discrepancies in modelling
approaches. If peak stress is to be used as an indicator of rupture potential, then all
aspects that govern wall stress must be accounted for. Simplifications such as
modelling the AAA wall with linearly elastic material properties may result in
inaccurate wall peak stress values. The intraluminal thrombus (ILT) is often excluded
from analyses as it can significantly increase the computational effort required to
determine the wall stress. In cases where the level of ILT is negligible, this
simplification may not cause much concern, but some AAAs exhibit large amounts of
ILT which can greatly reduce the magnitude of stress and alter the stress distribution. It
is recommended that if peak wall stress is to be considered as a parameter in the
decision-making of the clinician, then simplifications in the modelling approach should
be limited.
With the recent increase in research aimed at highlighting AAA wall stress as a rupture
parameter, in particular peak wall stress, it has become apparent that the wall strength of
AAA tissue does not appear to be receiving the same attention. It is known that failure
will occur when the local wall stress exceeds the local wall strength, and therefore, wall
strength may play an equal role in determining AAA rupture potential. Recent reports
have been published that present the ultimate tensile strength (UTS) of AAA tissue
(Raghavan et al., 1996, 2006; Thubrikar et al., 2001) however these strength values
have not been coupled with patient-specific wall stress calculations. A novel index
(FEARI) described earlier has been developed that used these previous reports on AAA
UTS by expressing this data as regional wall UTS. Together with calculations of
patient-specific peak wall stress and peak wall stress location using the complex
approach outlined in Chapter II, these wall strengths help highlight a new possible
parameter in rupture potential. The use of a simple index that may govern AAA rupture
could have clinical benefits. The index could be readily incorporated into the clinical
decision to surgically repair, as the time taken from CT imaging to calculate the FEARI
value is relatively short (~1 day) and may provide further guidance to the surgeon.
Although many believe that peak wall stress is the most clinically applicable calculation
obtained from the finite element method, alternatives should not be completely
192
excluded. It is known that approximately 82% of AAA ruptures occur on the posterior
wall of the AAA (Darling et al., 1977), and therefore the stress acting on the posterior
wall may play a role in the possibility of rupture. A correlation between the posterior
wall stress and the degree of asymmetry was identified, which may be significant in
determining AAA rupture potential. A novel approach to calculating the degree of
asymmetry of a particular AAA was created, that can be easily determined from 3D
reconstructions. AAA asymmetry is predominantly in the anterior direction due to the
constraint of the spine on posterior dilation. Wall stress was shown to increase when
asymmetry was introduced into an AAA. It was also observed that maximum diameter
still significantly influences wall stress, particularly peak wall stress, but that
asymmetry may also have a significant role in inducing high posterior wall stresses.
Currently, maximum diameter is deemed the most accurate indicator of rupture
potential, as not only is size an obvious factor in the decision-making process, but it is
also easy for the clinician to determine from CT scans. It is suggested that asymmetry
may also be an important criteria in AAA assessment, and as the technique of
quantifying asymmetry can be readily incorporated into the decision-making process, it
could provide a useful adjunct to diameter and other relevant parameters in the severity
assessment of AAAs.
In order to compliment computational methods of AAA assessment, experimental

approaches should be considered to confirm the numerical predictions and to improve
knowledge of AAA burst mechanics. In vitro techniques to determine AAA rupture
potential require accurate models that exactly represent the geometry of the aneurysm.
Following on from previously reported work (O’Brien et al., 2005), a technique was
developed that enables the manufacture of patient-specific silicone rubber AAA models.
This technique utilises the injection-moulding lost-wax process to achieve anatomically-
correct experimental models that exhibit good reproducible uniformity in wall
thickness. These models can be applied to numerous bench-top techniques and can be
created with or without the iliac arteries depending on the desired use. Inclusion of the
iliac arteries is of concern when examining post-operative considerations such as stent-
graft distraction forces and drag forces. Currently, this approach has focused on AAAs,
yet the methodology may be applied to other areas of both diseased and healthy
vasculature, such as the thoracic aorta and the cerebral arteries. These models may help
increase the possibilities of in vitro testing and ultimately provide further information
regarding the biomechanics of diseased vessels.
193
Although there is a need to perform in vitro testing of AAAs, there have been limited
reports to date. This is particularly evident when concerned with the pre-operative
assessment of AAAs. Many studies focus on bench-top techniques to examine AAA
stent-graft related complications. However, the rupture behaviour of AAAs, in
particular the location of rupture, has not been experimentally reported. Using the
described method of manufacturing AAA experimental models, an experimental
approach using high-speed photography was established. This imaging method
effectively captures the location of rupture which can be correlated to numerically
predicted regions of elevated wall stress using the finite element method. Resulting
from this study, it became evident that AAAs may predominantly rupture at regions of
inflection and not at regions of maximum diameter. This hypothesis has been
previously established numerically but never confirmed using experimental techniques.
It was also concluded from this study that the effect of wall thickness on resulting wall
stress may not be as significant as once believed. Although wall stress is governed by
wall thickness, the overall morphology of the aneurysm may be as important. The
inflection regions of aneurysms may experience elevated stress compared to regions of
maximum diameter regardless of wall thickness, and thus these inflection regions may
be significant rupture zones in vivo. Such locations can be easily identified from 3D
reconstructions as demonstrated in Chapter I. This approach may also provide useful
information to the clinician allowing a more informed decision to be reached.
It is known that in vivo AAA tissue is non-uniform in strength (Raghavan et al., 2006).
It is therefore desirable to create experimental AAA models using arterial analogues that
also possess non-uniform strength distributions. A novel range of silicone rubbers was
created that stemmed from commercially available materials. These silicones were
visually distinguishable due to their colour, and thus could be identified. Each silicone
rubber has a distinct ultimate tensile strength (UTS) and tear strength (TS), each of
which could be directly correlated with the colour of the material. These correlations
allowed the generation of calibration curves linking material property to colour
intensity. A novel technique of non-destructively determining both UTS and TS was
developed using spectrophotometry, and proved to be effective in predicting material
strength. These silicone rubbers were then used in experimental studies to monitor the
diameter change with respect to increasing internal pressure, with the finite element
method used to confirm these experiments. As experimental studies may help in the
understanding of AAA biomechanics, this range of materials will allow AAA models
194
with varying wall strengths to be used in further testing. Ultimately, it is hoped that
these materials, used in conjunction with other experimental approaches, may further
the clinical understanding of patient-specific AAA rupture potential.
The work presented on rupture locations using idealised AAA geometries yielded some
interesting findings, and so it was desirable to progress this methodology to realistic
AAA shapes. By implementing previous work, idealised and patient-specific silicone
rubber in vitro models were examined to determine rupture locations, with
computational approaches also employed. Imaging of the experimental models with
computed tomography complimented this study, allowing exact numerical
representations to be used in analyses. In vitro AAA models were shown to rupture at
regions of elevated wall stress and not at regions of maximum diameter regardless of
geometry. Inflection regions were identified by examining the Gaussian surface
curvature of each AAA geometry. Regions of inflection are areas where the surface
curvature changes from positive to negative, and this correlates well with elevated wall
stress. It was observed that rupture primarily (80% of cases) occurs at the region of
numerically predicted peak wall stress, but can also occur at other areas of high wall
stress. Wall thickness was found to vary throughout the experimental models, yet wall
thickness at the rupture sites were found to only deviate from the mean wall thickness
by 16 - 22%. Regions of thin wall thickness may therefore not necessarily be regions of
rupture. This supports previous reports suggesting that regions of thin wall thickness
may be strong enough to withstand the force of the cardiac cycle, and may not act as
sites of rupture in vivo. Surface anomalies were shown to move the site of rupture from
high stress regions to the locality of the wall defect. These defects can be compared to
calcifications, blebs or localised hypoxia in the realistic in vivo setting. The results
presented in this final chapter may further contribute to the understanding of AAA
biomechanics and AAA rupture assessment.
In conclusion based on this work:

♦ 3D reconstructions assist surgeons in preoperative planning. Not only can
valuable dimensional data be obtained, but also, complications with the anatomy
of the problem can be foreseen and overcome.
♦ The numerical techniques used to determine wall stress in AAAs can affect the
subsequent results. Care should be taken with the methodology used.
195
♦ FEARI may be a useful adjunct to diameter in the surgical decision-making

process.
♦ Since posterior wall stress is related to anterior asymmetry and as most AAA
ruptures occur on the posterior wall, the method of determining asymmetry may
be clinically beneficial.
♦ A technique has been developed to produce patient-specific experimental AAA
models that can be applied to several bench-top test methods.
♦ Idealised AAA experimental models rupture at areas of inflection and not at
regions of maximum diameter. Numerical methods predict these regions as
those of elevated wall stress.
♦ A range of silicone rubbers were created. The colour of these materials can be
directly related to material properties. More realistic experimental AAAs can
now be produced.
♦ Realistic AAAs rupture at regions of peak stress and not at regions of maximum
diameter. Surface defects affect rupture locations.
This thesis has explored several existing and novel theories of AAA rupture potential
using both computational and experimental techniques. The results presented
throughout suggest that the diameter criterion alone may not be adequate to
comprehensively determine the severity of patient-specific AAAs. There is a need to
implement alternative factors into the clinical decision-making process and the
methodologies reported in this thesis may be beneficial in future AAA rupture
assessment. There are some limitations to the methodologies and results presented
throughout and are addressed in the Recommendations for Future Work.
References
Darling, R.C., Messina, C.R., Brewster, D.C., Ottinger, L.W. (1977) “Autopsy study of unoperated
abdominal aortic aneurysms. The case for early resection.” Circulation, 56(2), 161-164.
Fillinger, M.F., Marra, S.P., Raghavan, M.L., Kennedy, F.E. (2003) “Prediction of rupture risk in
abdominal aortic aneurysm during observation: wall stress versus diameter.” Journal of Vascular Surgery,
37, 724-732.
Fillinger, M.F., Raghavan, M.L., Marra, S.P., Cronenwett, J.L., Kennedy, F.E. (2002) “In vivo analysis of
mechanical wall stress and abdominal aortic aneurysm rupture risk.” Journal of Vascular Surgery, 36,
589-597.
O’Brien, T., Morris, L., O’Donnell, M., Walsh, M. and McGloughlin, T. (2005) “Injection-moulded
models of major and minor arteries: The variability of model wall thickness owing to casting technique.”
Proceedings of the International Mechanical Engineers, 219, Part H: Journal of Engineering in Medicine.
196
Raghavan, M.L., Kratzberg, J., de Tolosa, E.M.C., Hanaoka, M.M., Walter, P., da Silva, E.S. (2006)
“Regional distribution of wall thickness and failure properties of human abdominal aortic aneurysm.”
Raghavan, M.L., Webster, M.W., Vorp, D.A. (1996) “Ex vivo biomechanical behaviour of abdominal
aortic aneurysm: assessment using a new mathematical model.” Annals of Biomedical Engineering, 24,
573-582.
Thubrikar, M.J., Labrosse, M., Robicsek, F., Al-Soudi, J., Fowler, B. (2001) “Mechanical properties of
abdominal aortic aneurysm wall.” Journal of Medical Engineering and Technology, 25(4), 133-142.
Venkatasubramaniam, A.K., Fagan, M.J., Mehta, T., Mylankal, K.J., Ray, B., Kuhan, G., Chetter, I.C.,
McCollum, P.T. (2004) “A comparative study of aortic wall stress using finite element analysis for
ruptured and non-ruptured abdominal aortic aneurysms.” European Journal of Vascular and Endovascular
Surgery, 28, 168-176.
197
RECOMMENDATIONS FOR
FUTURE WORK
Future Work
Recommendations for Future Work

Whilst this thesis has explored many avenues of AAA assessment, both numerically and
experimentally, many areas of research still remain. The following recommendations
for future work may help alleviate some of the simplifying assumptions and
methodologies that limit this thesis. These recommendations may help build on the
work presented here and further our understanding of the biomechanics attributed to the
rupture of AAAs.
Asymmetry in 3 Dimensions
The tool used to define the degree of asymmetry of AAAs currently assesses asymmetry
in the posterior-anterior plane. Although the majority of AAAs develop in this manner,
it is known that AAAs can experience asymmetry in other directions. This is currently a
limiting factor of the tool described in Chapter IV. A technique has been developed that
measures the degree of asymmetry in all directions. This is achieved by creating the
centreline using the 3D reconstruction software Mimics v12 (Figure 1), and then
exporting the centreline as a text file consisting of coordinates. A script was then
created in the mathematical software MATLAB v7.3 (The Mathworks Inc., MA, USA)
that reads the centreline information. As before, the centreline is connected with an axis
of rotation from which the asymmetry measurements are referenced. This script
automatically generates a plot of the original asymmetry data along the length of the
AAA. This graph can be seen in Figure 2. Future work could include refinement of this
script, in particular to automatically read both the posterior wall stress and wall stress
from other key regions for the corresponding AAA case. Comparison plots would be
automatically generated using the script so as to highlight any regional relationships
between 3D asymmetry and wall stress. Using this tool, the asymmetry in AAAs of all
geometries could be assessed, creating an even more useful tool to aid surgeons with the
decision-making process.
198
Future Work
Figure 1: Example 3D reconstruction with centreline shown in red. Centreline can now
be exported as a text file for further analysis with MATLAB.
Figure 2: Example asymmetry plot automatically generated from MATLAB.
Asymmetry From 2D CT Images

Although asymmetry can be easily determined from 3D reconstructions, there may also
be a need to develop a method of assessing asymmetry from 2D CT scans. This would
allow clinicians to measure AAA asymmetry directly from the 2D images after
completion of CT imaging. This method determines the centrepoints of the AAA using
the coordinates available in most DICOM (Digital Imaging and Communications in
Medicine) viewing software packages, such as the freeware software DicomWorks. In
order to measure the maximum asymmetry, the CT scan immediately below the lowest
199
Future Work
renal artery is examined. The coordinates of the centre point of the AAA are then
recorded. The clinician then follows through the series of CT scans until the scan
revealing the most severe asymmetry is detected. This can be easily identified by
experienced radiologists. The coordinates of the centrepoint, or centroid, of the AAA at
this region is again noted. Figure 3 shows the detection of centre points for the
particular region of interest, that is, the abdominal aortic aneurysm region of the CT
scan. The 2D distance between these two points can then easily be determined by
simply subtracting the Y coordinates of each scan. This results in a measurement of
asymmetry in millimetres (mm). For the example shown in Figure 3, by subtracting the
Y coordinates, the resulting maximum asymmetry is 25 mm. This allows the clinician
to obtain a single parameter to assess the degree of asymmetry of the particular AAA.
The same method can be applied to each CT scan in the series if the full degree of
asymmetry is desired along the length of the AAA, resulting in a plot similar to that
shown in Figure 4. Refinement of this technique and the automation of the process
using MATLAB is possible.
Figure 3: Example process of determining asymmetry from 2D images. (A) CT scan

depicting the proximal neck of the AAA with centre point identified, (B) Close-up view
of region with centroid coordinates recorded, (C) Scan depicting the most severe region
of asymmetry of the particular AAA, again with centroid shown, and (D) Close-up of
scan with centre point coordinates recorded.
200
Future Work
Figure 4: Example plot of asymmetry determined from 2D CT images.
Asymmetry of Ruptured AAAs

Work has begun on the gathering of CT datasets of ruptured AAAs. These datasets will
provide useful information on the geometries of AAAs that ruptured and can be
compared to non-ruptured AAAs, in particular, degrees of asymmetry. As it was
previously shown (Chapter IV) that asymmetry increases wall stress, information from
ruptured AAAs may further strengthen this hypothesis. It has also been shown that
experimental silicone rubber AAA models rupture at regions of inflection (Chapters VI
and VIII), therefore, determining if realistic AAAs actually ruptured at regions of
inflection in vivo may also provide a useful insight into AAA rupture.
Rupture Studies of AAAs

It is known that in vivo AAAs have a non-uniform distribution of wall strength. The
technology is in place to create experimental silicone rubber AAA models with varied
material properties throughout the AAA wall, as outlined in Chapter VII. Figure 5
shows a realistic AAA created using this technique and also the location of rupture
observed during the preliminary rupture study. Preliminary work using these mixed
material models has shown that the AAA will rupture at locations of reduced wall
strength (black region of model, Sylgard 170). Further assessment of the rupture
locations of these models may yield interesting results. Refinement of the FEA
techniques presented throughout this thesis may also provide useful data. Damage
models can be implemented to the FEA simulations that actually represent the
numerical failure of the silicone AAA models, therefore, not only indicating the regions
201
Future Work
of elevated stress, but also highlight where the material of the model ruptures. It is
necessary to increase the number of models used for the rupture studies in order to
determine statistical significance. The numbers of experimental models used in these
previous studies (Chapters VI and VIII) were relatively low and therefore statistical
significance is unrealistic. Also, inclusion of the intraluminal thrombus (ILT) and wall
defects such as calcifications can be manufactured into the silicone model. Previous
work (Chapter VIII) has shown that wall defects affect rupture location, and the ILT
may further change rupture locations.
Figure 5: Realistic AAA created using a mixture of Sylgard 160 and Sylgard 170. This
model has a random distribution of wall strengths which can be related to the UTS
values presented in Chapter VII. Also shown is the location of rupture using the
methodology of Chapter VIII.
Material Characterisation of Silicone Rubbers

The methodology and results presented in Chapter VII are based on a limited series of
mechanical tests to characterise the new range of silicone rubbers. Further refinement
of this characterisation through alternative testing methods, such as biaxial tensile
testing, ball-burst testing, hardness testing, dynamic mechanical thermal analysis
202
Future Work
(DMTA) testing, and tear testing using different test specimen geometries, may lead to
an improved understanding of these materials.
Stress Analyses of AAAs Using the Photoelastic Method

Previous work by our group 1 has reported the methodology and resulting stress
distributions determined by the photoelastic method on the idealised AAA model used
throughout this thesis. Work has begun on the progression of this method to patient-
specific AAA geometries (Figure 6). This recommended future work may help further
validate FEA techniques and provide a further insight into patient-specific AAA wall
stress.
Figure 6: Patient-specific AAA created using PL-3 epoxy resin (PhotoStress® Plus,
Vishay Micromeasurements, Inc.) for use with the photoelastic method.
Digital Image Correlation (DIC) Analyses

The use of Digital Image Correlation (DIC) is also possible with the recently acquired
StrainMaster (LaVisionUK Ltd., UK). This experimental method is a non-invasive
optical tool that measures deformation and strain analyses under loading conditions.
This technique may further compliment the photoelastic method, and ultimately validate
the use of FEA as a feasible and accurate method of determining wall stress in AAAs.
This technique involves the application of a textured coating to the test structure, often
1
Morris, L., O’Donnell, P., Delassus, P., McGloughlin, T. (2004) “Experimental assessment of stress
patterns in abdominal aortic aneurysms using the photoelastic method.” Strain, 40, 165-172.
203
Future Work
in the form of a spray paint. DIC then tracks the movement of the surface pattern
during testing, resulting in vector lengths and direction, and ultimately in strain
measurements. Figure 6 illustrates this process.
Figure 7: Illustration showing DIC process of measurement (LaVisionUK Ltd.).
Application of Tools to Other Aneurysms

The assessment tools presented throughout this thesis, along with those recommended
for future work, can also be applied to other aneurysm types. Other lethal aneurysms
predominantly form in the thoracic aorta and the cerebral arteries, and the assessment of
these diseased vessels may benefit from these techniques. Work has begun on these
aneurysm types such as the thoracic aortic aneurysm (TAA) shown in Figure 8.
Figure 8: TAA with centreline for use in asymmetry assessment.
204
Future Work
3D Reconstructions From Traditional Ultrasound

AAA screening programs are becoming increasingly common in many countries. These
screening programs utilise basic ultrasound technology to detect the presence of AAAs,
and if detected the maximum diameter is recorded. Work has begun on an algorithm to
generate 3D reconstructions of the AAA from these basic ultrasound images. Thus
future work would entail a series of ultrasound images recorded along the diseased aorta
from specific landmark locations, such as the renal arteries and iliac bifurcation.
Automated thresholding and segmentation techniques using either Mimics v12 (Figure
9) or MATLAB could identify the regions of interest, and ultimately create
reconstructions. This procedure may have clinical benefits as 3D reconstructions of the
AAA can be visualised immediately post-ultrasound, and further the effectiveness of
screening programmes worldwide. An on-going collaboration with the UK AAA
Screening Program will help further this technique and apply the tool to a cohort of
patients if successful. This method may also allow smaller aneurysms to be
reconstructed and analysed as currently there is little need to image an AAA <5 cm,
based on the maximum-diameter criteria, using CT unless other complications are
observed.
Figure 9: Preliminary method of thresholding and segmenting ultrasound images. (A)

Shows the ultrasound image, (B) the thresholding and segmentation of the artery and
(C) generation of arterial boundary.
205
APPENDICES
APPENDIX A
Journal Publications
Appendix A
Journal Publications
Corbett, T.J., B.J. Doyle, A. Callanan, M.T. Walsh, and T.M. McGloughlin,
Engineering Silicone Rubbers for In vitro Studies: Creating AAA Models and ILT
Analogues with Physiological Properties, Journal of Biomechanical Engineering,
2009, in press.
Doyle, B.J., T.J. Corbett, A. Callanan, M.T. Walsh, D.A. Vorp and T.M. McGloughlin,
An Experimental and Numerical Comparison of the Rupture Locations of an
Abdominal Aortic Aneurysm, Journal of Endovascular Therapy, 2009, in press, to
appear June 2009.
Doyle, B.J., P.A. Grace, E.G. Kavanagh, P.E. Burke, F. Wallis, M.T. Walsh and T.M.
McGloughlin, Improved Assessment and Treatment of Abdominal Aortic
Aneurysms: The Use of 3D Reconstructions as a Surgical Guidance Tool in
Endovascular Repair, Irish Journal of Medical Sciences, 2009, in press, available
online March 25th 2009.
Doyle, B.J., A. Callanan, E. Kavanagh, P.A. Grace, P.E. Burke, D.A. Vorp and T.M.
McGloughlin, A Finite Element Analysis Rupture Index (FEARI) as an Additional
Tool for Abdominal Aortic Aneurysm Burst Prediction, Vascular Disease
Prevention, 2009; 6: 114-121.
Doyle, B.J., A. Callanan, P.E. Burke, P.A. Grace, M.T. Walsh, D.A. Vorp, and T.M.
McGloughlin, Vessel Asymmetry as an Additional Diagnostic Tool for the
Assessment of Abdominal Aortic Aneurysms, Journal of Vascular Surgery 2009;
49(2): 443-454.
Molony, D., A. Callanan, B.J. Doyle, L.G. Morris, M.T. Walsh and T.M. McGloughlin,
Geometrical Enhancements for Abdominal Aortic Aneurysm Stent-Grafts, Journal
of Endovascular Therapy, 2008; 15: 518-529.
Corbett, T.J., A. Callanan, L.G. Morris, B.J. Doyle, P.A. Grace, E.G. Kavanagh and
T.M. McGloughlin, In Vivo and In Vitro Biomechanical Behaviour and
Appendix A
Performance of Post Operative Abdominal Aortic Aneurysms and Implanted

Stent-Grafts - A Review, Journal of Endovascular Therapy, 2008; 15: 468-484.
Doyle, B.J., L.G. Morris, A. Callanan, P. Kelly, D.A. Vorp and T. McGloughlin, 3D
Reconstruction and Manufacture of Real Abdominal Aortic Aneurysms: From CT
Scan to Silicone Model, Journal of Biomechanical Engineering, 2008; 130: 034501-5.
Doyle, B.J., A. Callanan and T.M. McGloughlin, A Comparison of Modelling

Techniques for Computing Wall Stress in Abdominal Aortic Aneurysms,
Biomedical Engineering Online, 2007; 6: 38.
Engineering Silicone Rubbers for In vitro Studies:
Creating AAA Models and ILT Analogues with
Physiological Properties
T.J. Corbett, B.J. Doyle, A. Callanan, M.T. Walsh and T.M McGloughlin1
Centre for Applied Biomedical Engineering Research (CABER), MSSi, Department of

Mechanical and Aeronautical Engineering, University of Limerick, Limerick, Ireland.
1
Corresponding author: Address:
Professor Tim McGloughlin
MSG-014
MSSi Building
University of Limerick
Limerick
Ireland
Email: tim.mcgloughlin@ul.ie
Corbett et al.
ABSTRACT
Background: In vitro studies of abdominal aortic aneurysm (AAA) have been widely
reported. Frequently mock artery models with intraluminal thrombus (ILT) analogues
are used to mimic the AAA in vivo. While the models used may be physiological, their
properties are frequently either not reported or investigated.
Method of Approach: This study is concerned with the testing and characterisation of
previously used vessel analogue materials and the development of new materials for the
manufacture of AAA models. These materials were used in conjunction with a
previously validated injection moulding technique to manufacture AAA models of ideal
geometry. To determine the model properties (stiffness (β) and compliance) the
diameter change of each AAA model was investigated under incrementally increasing
internal pressures and compared to published in vivo studies to determine if the models
behaved physiologically. A FEA study was implemented to determine if the pressure –
diameter change behaviour of the models could be predicted numerically. ILT
analogues were also manufactured and characterised. Ideal models were manufactured
with ILT analogue internal to the aneurysm region and the effect of the ILT analogue on
the model compliance and stiffness was investigated.
Results: The wall materials had similar properties to aortic tissue at physiological
pressures (Einit 2.22MPa and 1.57MPa (aortic tissue: 1.8MPa)). ILT analogues had
similar Young’s modulus to the medial layer of ILT (0.24 and 0.33MPa (ILT:
0.28MPa)). All models had aneurysm sac compliance in the physiological range (2.62 -
8.01x10-4/mmHg (AAA in vivo: 1.8 - 9.4x10-4/mmHg)). The necks of our AAA
models had similar stiffness to healthy aortas (20.44 - 29.83 (healthy aortas in vivo:
17.5±5.5)). Good agreement was seen between the diameter changes due to
pressurisation in the experimental and FEA wall models with a maximum difference of
7.3% at 120mmHg. It was also determined that the inclusion of ILT analogue in the sac
of our models could have an effect on the compliance of the model neck.
Conclusions: Ideal AAA models with physiological properties were manufactured. The
behaviour of these models due to pressurisation was predicted using FEA, validating
this technique for the future design of realistic, physiological AAA models. Addition of
ILT analogues in the aneurysm sac was shown to affect neck behaviour. This could
have implications for endovascular AAA repair due to the importance of the neck for
stent-graft fixation.
Keywords: Abdominal aortic aneurysm, mock artery models, realistic properties.
Corbett et al.
INTRODUCTION
Abdominal aortic aneurysm (AAA) is defined as an abnormal localised dilation or bulge
in the infrarenal aorta greater than 50% of its normal diameter [1]. It affects up to 5% of
the population over the age of 55 years. Rupture of an AAA is responsible for 1.3% of
deaths in males aged 65-85 years, accounting for 10,000 deaths annually in the United
Kingdom [1-3]. Survival of ruptured abdominal aortic aneurysm (rAAA) may be as low
as 10-20%.
Many investigators have undertaken in vivo studies into specific aspects of preoperative
and post operative AAA’s using silicone rubber or latex AAA models [4-13]. There are
a number of important considerations when attempting to develop AAA models with
physiological properties. In vivo the arteries are in a state of constant longitudinal
tension [14]. As this is difficult to model in vitro it is usually not considered [4-12]. The
abdominal aorta is surrounded by visceral organs, abdominal muscle and the spine
which undoubtedly influence its deformation under pressure [15]. Because of these
parameters creating materials with a similar stress-strain response to aortic tissue may
not lead to models that behave similar to the vessel itself. Because of this, the
compliance (or stiffness) of arteries has become the attribute which researchers attempt
to mimic [10,16]. The manufacturing methods for creating both idealised and realistic
AAA models have been previously reported [10,16-21]. O’ Brien et al. [19] and Doyle
et al. [20] have verified a lost wax manufacturing technique and reported that the
variation in wall thickness for large vessel models manufactured using this technique
was between 4-11%. We have previously described a rupture study in ideal AAA
models and derived the material constants and mechanical properties of the materials
used to construct these models [22]. The ultimate tensile strength of the material and the
location of rupture of the models were the primary considerations in that study and as
such we did not examine the compliance of the models themselves. Other studies have
investigated model compliance but have not characterised the materials used to
construct the models [10,16]. Those studies were based on AAA models created by
brushing latex onto either a glass or plastic former. However, this manufacturing
method leads to an uneven wall thickness which could affect observed compliance [23].
Here we characterise our materials using strain energy functions (SEF) and perform and
validate a finite element analysis (FEA) investigation of our models so these materials
may be used in the design realistic AAA models in future in vitro studies.
Corbett et al.
Some 75% of AAAs have been shown to have Intraluminal Thrombus (ILT) present in
the aneurysm sac [24]. While many thrombus analogues ranging from gelatin to dough
have been used in in vitro studies, a repeatable manufacturing method for creating
physiological AAA models with integrated physiological thrombus analogues has not
been fully described [11,12,25]. Here we describe an injected silicone rubber based
AAA model of ideal geometry with realistic vessel stiffness with a built in silicone
rubber ILT analogue with realistic Young’s modulus. The effect of changing the
mechanical properties of the silicone rubber wall and the ILT analogue on the observed
model stiffness and compliance is also investigated.

Materials Selection and Characterisation
Silicone rubbers have been widely used as aorta analogues in in vitro studies [4-
7,9,10,19,20,22]. Wacker Elastosil RT601, (Wacker-Chemie GMBH, Munich,
Germany), was used by Doyle et al. [20] in a previous study to manufacture AAA
models. This silicone rubber was employed as the base material for this study. Different
quantities of Dow Corning 200/5CS silicone fluid (Midland, MI, USA), were added to
this material to alter its material properties.
Ten Type 2 dumb bell samples of two different wall materials, 100%RT601 (hereafter
“W1”) and 90%RT601 (hereafter “W2”) by mass, were created by injection moulding
and tested in accordance with BS ISO 37 on a uniaxial extensometer with a 1kN
loadcell (Tinius Olsen, Surrey, UK). To ensure repeatable results and to minimise the
Mullins effect the specimens were preconditioned by deforming the gauge length by
40% ten times before carrying out an actual test. This preconditioning serves to stabilise
the stress-strain response of the material [26]. A video extensometer (MESSPHYSIK,
Fürstenfeld, Austria) was used to measure the deformation of the specimen gauge length
of the material during the test.
Material characterisation was carried out using previously described methods [22].
Briefly, the force extension data from the uniaxial tensile tests were converted to
engineering stress and strain. Best fit polynomial trendlines were then fitted to the data
for each material. Using these polynomials, twenty representative engineering stress-
strain datapoints were calculated for engineering strains from 0-1 for each material and
used to define the material in the finite element solver ABAQUS 6.7-1 (Dassault
Corbett et al.
Systems, SIMULA, Providence, RI, USA). The uniaxial test was modelled in FEA with
identical geometry to the test specimens. Boundary conditions mimicking the uniaxial
test were used. Different strain energy functions were examined, and the most suitable
strain energy function selected. To ensure that the optimum strain energy function had
been selected, a comparison was made between experimentally determined true stress
and strain and the true stress and strain predicted by the finite element analysis of the
test specimen.
Silicone Rubber AAA model manufacture

Six idealised wall (AAAW) models (Fig. 1.) were manufactured for this study (three
manufactured from W1 and three from W2). The model dimensions were taken from the
EUROSTAR database [27] and are an averaged set of dimensions based on a 3413
AAAs. This idealised AAA geometry has been used extensively in previous
publications [19,22,28-30]. The lumen diameters of the ideal geometry are: neck:
24mm, aneurysm: 50mm legs: 12mm, the ideal AAA geometry has a bifurcation angle
of 60o and wall thickness of 2mm.
Fig. 1: AAA model viewed from front and side, arrows show where measurements were
taken during compliance testing.
The models were manufactured using the previously reported lost wax technique
[19,20]. Briefly, a wax lumen was cast from an aluminium mould. This was placed
within a wall mould with a 2mm cavity into which the silicone rubber wall material was
injected. The wall material was allowed to cure at 40oC for 4 hours. After curing the
wax was melted out of the wall model. The models were then cleaned to remove any
remaining wax. The full method is outlined in appendix A.
Corbett et al.
Compliance Testing
Each model was placed in the pressurised air system shown in Fig. 2. A precision
regulator (Norgren, Warwickshire, UK) was used to adjust the pressure inside the
model. A digital manometer (Druck, Leicester, UK) was used to measure the pressure in
the system. The models were pressurised from 20-160mmHg in steps of 20mmHg. A
video extensometer (MESSPHYSIK, Fürstenfeld, Austria) was used to measure the
change in diameter at each pressure increment.
Fig. 2: Pressure – diameter change measurement setup for investigating model

compliance
The models were held at each pressure for 30 seconds before a diameter measurement
was taken. The first location was in the proximal neck of the model, 30mm above the
beginning of the aneurysm section and the second location was the area of maximum
diameter of the aneurysm (Fig. 1.). The model response to pressurisation was
investigated in two planes to allow for the mean compliance to be determined for each
model, therefore minimising the effect of any variations in wall thickness on the results.
These locations were chosen to determine the compliance not just of the aneurysm sac
but also of the proximal neck of the model. The biomechanics in the proximal neck of
abdominal aortic aneurysms are of interest as endovascular stent-grafts rely on the
proximal neck for fixation within the vessel [31]. Therefore the behaviour in the
proximal neck is directly related to the success or failure of EVAR.
When the pressure diameter data was collected for all the models compliance was
computed using Eq. 1. [32].
Corbett et al.
Compliance =
(ΔA) (1)
Amax (ΔP )
Stiffness was then computed using Eq. 2. [33].
⎛P ⎞ Dmin
Stiffness β = ln⎜⎜ max ⎟⎟ (2)
⎝ Pmin ⎠ ΔD
Where D is the outer diameter of the vessel, A is the cross sectional area of the vessel
and P is the pressure within the vessel.

The FEA model was constructed in the nominal geometry of the experimental models.
As these idealised models are symmetric about two planes a quarter model was
sufficient to run the FEA study (Fig. 3.). Coefficients from the strain energy functions
determined in the material characterisation study were used to define the material
properties for the analysis. To simulate the connection of the AAA segment to the
descending aorta and the iliac bifurcation, the model was constrained in all directions on
the proximal and distal surfaces. A symmetry boundary condition was placed on
surfaces A and B in Fig. 3. with another symmetry boundary condition placed on
surface C. A uniform pressure was applied to the inside surfaces of the model and
separate analyses were run for the pressures used in the experimental study. The model
was meshed using C3D10MH ten node modified quadratic tetrahedron hybrid elements
[34]. Mesh independence was performed by increasing the number of elements in the
mesh until the difference in peak Von Mises stress was <2% of the result computed
using the previous mesh [34].
Fig 3: Quarter model used in finite element analysis
Corbett et al.
Experimental Modelling of AAAs with ILT
In order to independently examine the effect of varying wall properties and varying ILT
analogue properties on observed model compliance, two possible ILT analogues,
45%RT601 (hereafter “ILT1”) and 50%RT601 (hereafter “ILT2”) by mass, were also
created and characterised as previously described. AAA models with ILT analogue
(AAAILT) were manufactured which utilised these materials in the aneurysm sac. These
models had lumen geometry identical to AAAW models except in the aneurysm region
where the lumen diameter was 38mm; the remaining volume consisted of the ILT
analogue (Fig 4).
Fig 4 Method for manufacturing AAAILT models, Left: wax lumen, middle: ILT
analogue injected around wax lumen, right: wall material injected and allowed to cure
before wax is melted out
To manufacture the AAAILT models, 2 extra aluminium moulds were needed in addition
to the two required for manufacturing the AAAW models. The first of these was to
manufacture the wax lumen which had a smaller diameter in the sac than the AAAW
model. The second was identical in geometry to the lumen mould for the AAAW model.
The AAAILT lumen was placed in this mould leaving a cavity in the aneurysm sac region.
This mould had an inlet and vent in the aneurysm region to allow the ILT analogue to
be injected easily into the mould. When the ILT analogue was cured the lumen + ILT
analogue were placed inside the wall mould and the wall material injected as previously
described. The full method is outlined in appendix A.
Nine AAAILT models were manufactured and compliance tested for this study. Six were
manufactured from W2, three of these had ILT1 analogue and the remaining three had
ILT2 analogue in the sacs. The final three models were manufactured from W1 with
Corbett et al.
ILT1 analogue in the sacs. The model specifics of all models created are summarised in
table 1.
Table 1: List of AAA models manufactured for compliance study

Wall only With ILT (AAAILT)
Wall (AAAW) ILT1 ILT2
W1 3 3 -
(Model set: W1) (Model set: W1 + ILT1)
W2 3 3 3
(Model set: W2) (Model set: W2 + ILT1) (Model set: W2 + ILT2)
Dimensional Accuracy of AAA Models

While O’ Brien et al. [19] and Doyle et al. [20] confirmed that the injection method
produced models with acceptable variation in wall thickness the overall dimensional
accuracy of the models was not investigated in these studies. To determine if the models
were dimensionally accurate we measured the diameter of the models in the two
locations of interest (Fig. 1.) and compared them to the nominal dimensions. In both
these locations we measured the outside diameter of the model from the front and from
the side.
RESULTS
The force extension data from the uniaxial tensile tests was converted to true stress
(σ ) and true strain (ε ) using theory for large deformation of incompressible materials
(Eq. 3. & Eq. 4.).
f (1 + ε )
σ= (3)
ao
⎛ ⎛ Δl ⎞ ⎞
ε = ln⎜⎜ ⎜⎜ ⎟⎟ + 1⎟⎟ (4)
⎝ ⎝ lo ⎠ ⎠
Where f is force, ao is undeformed cross sectional area of the test length of the dumb-
bell test pieces and l is length.
The initial Young’s modulus (Einit) [35] of the wall materials was calculated by fitting a
best fit line to the linear portion of the stress-strain curves. It was determined that for the
wall materials a sixth order reduced polynomial strain energy function was most
suitable while for the ILT analogues a second order Ogden strain energy function was
most appropriate. Due to the incompressible nature of these hyperelastic materials the
reduced polynomial and Ogden strain energy functions take the form:
Corbett et al.
W = ∑ C i 0 (I 1 − 3)
N
i
Reduced Polynomial: (5)
i =1
2μ i
(λ )
def N
Ogden: W =∑ 1
αi
+ λ2αi + λ3αi − 3 (6)
i =1 α i2
Where I is a unit matrix, λ1, λ2 λ3 are the principle stretch ratios, C, μ and α are material
constants.
Figure 5. shows a comparison of the stress-strain response of each material created. The
wall analogue materials (W1 & W2) are seen to be much stiffer than the ILT analogues
(ILT1 & ILT2). Figure 5 also shows the good agreement in true stress and true strain
observed using the described numerical and experimental approaches. The initial
Young’s modulus and coefficients for the respective strain energy functions for each
material are shown in Table 2.
Fig. 5: Comparison of experimentally calculated and FEA true stress and strain in a
dumb bell sample for wall and ILT materials
Table 2: Coefficients for strain energy functions and initial Young’s modulus for each
material
Material Einit SEF Type Coefficients for strain energy functions
(MPa)
W1 2.22 R. Poly C10 = 0.2259, C20 = 0.1821, C30 = -0.3302, C40 = 0.3071,
(6th order) C50 = -0.1178, C60 = 0.01788
W2 1.57 R. Poly C10 = 0.2813 , C20 = .2263, C30 = -0.4017,
(6th order) C40 = 0.3829, C50 = -0.0148, C60 = 0.0255
ILT1 0.24 Ogden μ1 = 6.49e-4, μ2 = 9.0562, α1 = 0.0622, α2 = 2.9419
(2nd order)
ILT2 0.33 Ogden μ1 = 5.09e-4, μ2 = 7.3653, α1 = 0.0804, α2 = 1.8246
(2nd order)
Corbett et al.
Figure 6. shows a comparison of the experimental and numerical pressure-diameter
changes in the AAAW models. The maximum percentage difference between the
experimental and numerical model at the maximum pressure (160mmHg) was seen in
the aneurysm region of the W2 models (10.66%) while at a pressure of 120mmHg the
maximum percentage difference was found in the neck of the W2 models (7.3%).
Fig. 6: Comparison of mean experimental and FEA predicted pressure – diameter

change curves for wall models in the neck (left) and the aneurysm (right) regions
Figure 7 shows the experimental pressure diameter change curves for all the models in
the aneurysm and neck regions. Using these curves and Eq. 1. & Eq.2. the mean
stiffness and compliance of each set of models was computed for pressures of
120/80mmHg. These are shown in Fig. 8. and tabulated in Table 3 which also shows the
range of stiffness and compliance calculated for each model set.
Corbett et al.
Fig. 7: Mean pressure – diameter change curves for each set of models in the neck (left)
and aneurysm (right) regions (n = 3 for all cases)
Corbett et al.
Fig. 8: Mean model stiffness in the neck (left) and aneurysm (right) for each model set
(n = 3 for all cases)
To determine if there was a significant difference in the observed compliance between

different model sets, an independent samples T test was performed in the software SPSS
15.0 for windows (SPSS Inc. Chicago, IL, USA). The results of this statistical analysis
are shown in table 4.
Table 3: Mean stiffness and compliance for each model set at neck and aneurysm
regions (range shown in parentheses)
Model Set βneck βaneurysm Cneck Caneurysm
(×10-4/mmHg) (×10-4/mmHg)
W1 25.39 32.38 7.79 6.12
(23.77 - 26.95) (31.76 - 33.36) (7.31 - 8.29) (5.94 - 6.23)
W1 + ILT1 29.83 77.78 6.63 2.62
(29.12 - 30.23) (77.17 - 78.74) (6.53 - 6.79) (2.58 - 2.63)
W2 20.44 23.26 9.61 8.01
(19.64 - 21) (22.05 - 23.6) (9.38 - 9.97) (8.01 - 8.9)
W2 + ILT1 21.19 57.06 9.27 3.5
(20.78 - 21.53) (56.49 - 57.57) (9.14 - 9.45) (3.47 - 3.54)
W2 + ILT2 22.88 64.15 8.62 3.19
(22.31 - 23.21) (63.33 - 64.94) (8.52 - 8.83) (3.16 - 3.2)
Table 4: Statistical difference in stiffness and compliance between model sets

Yes = p < 0.05
W1 + ILT1 W2 W2 +ILT1 W2 + ILT2
Neck W1 Yes Yes Yes No
W1 + ILT1 - Yes Yes Yes
W2 - No Yes
W2 + ILT1 - Yes
Sac W1 Yes Yes Yes Yes
W1 + ILT1 - Yes Yes Yes
W2 - Yes Yes
W2 + ILT1 - Yes
The mean and range of diameters of each model set measured in the locations shown in
figure 3 is shown in table 5.
Corbett et al.
Table 5: Mean diameter of neck and aneurysm region measured from front and side
(range in parentheses)
All dimensions in mm
Model Set Dneck (front) Dneck (side) Daneurysm (front) Daneurysm (side)
W1 27.98 28.23 54.48 54.72
(27.9 - 28.05) (28.2 - 28.3) (54.2 - 54.8) (54.62 - 54.8)
W1 + ILT1 28.03 28.41 55.65 56.65
(28 - 28.1) (28.3 - 28.5) (55.15 - 56.5) (56.4 - 57)
W2 27.66 27.9 53.68 53.85
(27.4 - 27.9) (27.85 - 27.95) (53.3 - 54.1) (53.6 - 54.05)
W2 + ILT1 27.71 28.03 55.13 56.35
(27.65 - 27.8) (27.75 - 28.2) (54 - 56.2) (55.5 - 56.85)
W2 + ILT2 27.65 27.95 53.61 54.94
(27.6 - 27.7) (27.95 - 27.95) (53.1 - 54) (54.2 - 55.45)
DISCUSSION
Many previous reports describe in vitro studies of pre and post operative abdominal
aortic aneurysms [4-13]. The most important component in any in vitro study should be
the AAA model itself. These models should be reproducible, have consistent material
properties, consistent thickness and be physiological in behaviour. It has been proven
that the injection method creates reproducible models with relatively consistent
thickness [19,20]. However an investigation has not been undertaken into the properties
of the materials used to create these models or the properties of the models themselves.
To address this we employed a previously used silicone rubber and performed

mechanical testing. We also manufactured new wall and ILT analogue materials and
performed similar tests. Previous studies have shown that the initial Young’s Modulus
for abdominal aortic aneurysm tissue lies in the range of 0.42 – 0.56MPa and that AAA
tissue may be stiffer than healthy aorta [35,36]. Both of our wall materials have initial
Young’s modulus outside this range (2.22MPa (W1) & 1.57MPa (W2)). However
Vallabhaneni et al. [37] reported that AAA tissue has a mean Young’s modulus of
1.8MPa (0.16-4.52MPa) at stresses experienced in vivo. It is evident from the FE
analysis that at physiological pressures the walls of the AAA models are in the linear
phase of the silicone rubber stress-strain behaviour and are comparable to this published
data.
ILT has been found experimentally to be a material consisting of three separate layers
(luminal, medial and abluminal). The properties of these layers were investigated by
Wang et al. [38] They found that the Young’s moduli for the luminal and medial layers
was 0.54MPa and 0.28MPa. The abluminal layer was so degraded that they could not
perform mechanical tests on the material [38]. In this study we used single layer wall
Corbett et al.
models due to manufacturing constraints. We also considered the ILT as a single layer
due to these constraints. This approach is consistent with previously published
experimental modelling of AAA [11,12,19,20]. From the uniaxial tensile testing we
determined that our ILT analogues had Young’s modulus of 0.24MPa (ILT1) and
0.33MPa (ILT2) which is similar to values for the medial layer found by Wang et al.
[38].
The materials were characterised using previously reported techniques to obtain

material coefficients that describe the stress-strain response of each particular material.
A good comparison was seen in the pressure - diameter change behaviour between the
experimental and FEA AAAW models with a maximum difference in diameter change
of <11%. This variance was less significant at lower pressures with a maximum
difference between the FEA and experimental models of 7.3%. This validation of our
FEA model and material properties could allow for improved design of patient specific
AAA models which are usually given an arbitrary wall thickness [19,20]. Using
characterised materials, AAA models could be carefully designed utilising the finite
element method to include varying local compliance by varying local wall thickness.
This variation in stiffness could be investigated clinically by utilising improved imaging
techniques.
Mock artery models have also been used to predict rupture location and to identify the
areas of highest stress on the AAA wall [22,28]. The existing methodology for creating
patient specific or realistic AAA models from CT images does not allow for the fact that
the models are at physiological pressures during imaging. The model is usually
designed with the dimensions straight from the CT image and therefore the resulting
silicone model has pressurised dimensions when there is no pressure within the model.
Therefore when such a model is pressurised it may not be completely realistic and this
may inaccurately predict rupture location or deformation behaviour of the aneurysm.
With the assistance of a verified FEA model and manufacturing technique it is possible
to alter the design of the AAA model during the design stage so that it will have realistic
diameters when pressurised.
We examined the stiffness in two locations in our models for physiological pressures
and compared them to values from literature for in vivo studies of actual abdominal
aortas and aneurysms. Sonnesson et al. [33] reported that the stiffness of the aorta in
Corbett et al.
healthy, non-smoking individuals of age ~69 years was 17.5±5.5. While the stiffness
values for all our W1 models are outside this range, the neck region of all the W2
models and the aneurysm region of the W2 models without thrombus are within this
range. Even though the W1 models have stiffness outside the range determined by
Sonnesson et al. [33] is possible that they may be still physiological. This is because
frequently the necks of AAA are often heavily calcified and therefore the neck is stiffer
than healthy aorta [39].
Vorp et al. [32] reported that the compliance of AAAs with ILT of similar sized to our
models was 1.8 - 9.4×10-4/mmHg (mean 4×10-4/mmHg). All our models had aneurysm
compliance within this range. However in the 8 patients that Vorp et al. [32] studied
there were two distinct groups. Six of the patients had aneurysms that had compliances
between 1.8 and 3.6×10-4/mmHg while the remaining two patients had aneurysm
compliances of 6 and 9.4×10-4/mmHg. All our models are within the former range
having compliances between 2.62 – 3.5×10-4/mmHg. These values also are
comparatively close to the mean of the range outlined by Vorp et al. [32]. The models
without ILT analogue are within the latter range with compliances of 6.12 and 8.01×10-
4
/mmHg.
In our models there was a significant difference between compliance in the aneurysm
region in models with and without ILT analogue, which is to be expected. However the
presence of the ILT analogue also had an effect on the stiffness of the neck. There was a
significant difference between the compliance of the neck of W1 (AAAW) and W1/ILT1
(AAAILT) models and between the neck of W2 (AAAW) and W2/ILT2 (AAAILT) models.
There was not a significant difference between the neck compliance of W2 (AAAW) and
W2/ILT1 (AAAILT) models. It is unclear whether ILT offers support to the proximal
neck in AAAs in vivo as the authors are unaware of any study which investigates this
hypothesis. It may be an important consideration because endovascular grafts often rely
on the ability of the proximal neck to withstand excessive deformation under the radial
force of a stent to provide fixation [31]. It has already been shown that ILT present in an
AAA reduces the risk of graft migration by acting normal to the drag force of the graft
[40]. If the ILT provides structural support to the proximal neck it could also be a
positive factor in the prevention of stent-graft migration.
Corbett et al.
From the pressure - diameter change results it is clear that model geometry has a
pronounced effect on the deformation characteristics of the model. The pressure -
diameter change behaviour of the necks is almost linear while there is a nonlinear
pressure diameter change relationship in the aneurysm region, owing to the fact that the
area around the inflection points in the aneurysm region undergoing greater strain than
the area at the max diameter [28]. This has the effect of opening the radius of curvature
of the aneurysm region at low pressures, preventing a noticeable diameter change. In the
aneurysm region, the models without ILT analogue also seem to initially contract when
viewed from the front. This is not apparent when the models are viewed from the side.
A possible reason for this is the effect of the geometry of the bifurcation and the effect
of changes in geometry here during pressurisation of the model. This serves to reinforce
the point that care should be taken when attempting to ascertain mechanical properties
of tissue based on imaging techniques alone without attempting to determine the effect
of shape on the deformation behaviour of the vessel in question.
The neck diameters of all models were acceptably close to the nominal dimension of
28mm with a mean neck diameter of 27.85mm and a maximum variation in diameter of
2.14%. Likewise, the aneurysm sections of the AAAW models were acceptably close to
the nominal dimension of 54 mm with a mean diameter of 54.18mm and a maximum
variation of 1.48%. The introduction of the ILT analogue did appear to have an effect
on the dimensional accuracy of the models. The mean diameter in the aneurysm region
was 55.39mm and the maximum variation in diameter was 5.56%.
The primary limitation of this study is the idealisation of considering the ILT and the
artery wall as single layers. While this is a gross assumption, it is less significant than
using ILT analogues without investigating their mechanical properties, as previously
reported [11,12,41]. We believe that our simplified one layer ILT analogue is more
representative than materials such as dough or gelatin. The materials were evaluated in
ideal AAA geometries. It is possible that in realistic geometries there may be a
difference in observed stiffness of the models due to shape effects. However it is
probable that these differences would not be significant and therefore the silicone rubber
models would remain physiological in their behaviour. The use of our AAA wall and
ILT analogues will be examined in realistic AAA geometries in future studies in our
group. Furthermore this manufacturing approach can be extended to model other arterial
structures such as plaque laden carotid arteries [42].
Corbett et al.
CONCLUSION
Silicone rubber based models of ideal AAA geometries were manufactured with
physiological characteristics. New artery and ILT analogue materials were developed
and characterised. A methodology has been developed for creating models with
different properties from one set of moulds. Manufactured models without ILT
analogue were dimensionally accurate to 2.14%. Introduction of ILT analogue affected
the dimensional accuracy of the aneurysm region of the models. We have validated a
finite element analysis of our wall models which will allow for careful design of
realistic AAA geometries in the future.
ACKNOWLEDGEMENTS
The authors would like to thank (i) The Irish Research Council for Science Engineering
and Technology (IRCSET) Grant# RS/2005/27 (T. Corbett) and Grant # RS/2005/340
(B. Doyle) and NIH Grant # R01-HL-060670 from the United States National Heart
Lung and Blood Institute (ii) Liam Morris (Galway and Mayo Institute of Technology)
for initial model development, (iii) Stephen Broderick (CABER) for assistance with
image rendering and diagrams.
APPENDIX A – Detailed Procedure for AAA Model Manufacture

The models were manufactured as follows:
1. All mould halves were cleaned thoroughly using methanol.
2. Silicone release agent (Ambersil formula 6) was sprayed onto the surfaces of all
moulds.
3. The lumen mould halves were bolted together tightly, placed in an oven and
heated to 55oC
4. A casting wax (Castylene B581, Remet Corporation, The Netherlands) was
melted on a hotplate to a temperature of ~ 150oC
5. The Lumen Moulds were tilted at 45 degrees to horizontal and the molten wax
poured into the models. Tilting of the moulds helped to minimise splashing and
prevented air being trapped inside the casting. The mould was returned to a
vertical position to allow for complete filling of the mould.
6. When the mould was full it was tapped with a mallet for several minutes to
allow for any air bubbles in the molten wax to rise out of the mould.
7. As the wax cooled and solidified the mould was topped up with molten wax to
prevent the casting drastically losing dimensional accuracy.
8. When the wax was fully solidified (after cooling for ~2 hours) the mould halves
were unbolted and the wax model carefully removed. Any flash as a result of the
casting process was shaved off with a sharp knife.
Follow steps 9 – 17 for manufacturing AAA models with ILT: For models without
ILT proceed to step 18:
Corbett et al.
9. The ILT lumen casting was placed in the ILT mould and placed in an oven at
45oC. The ILT mould was rotated one quarter turn every ten minutes to ensure
that the wax did not sag inside the mould.
10. The components of the thrombus analogue were placed in a plastic beaker in the
following order: 1: Wacker Elastosil RT601 Part A, 2: Wacker Elastosil RT601
Part B, 3: Dow Corning 200/5CS Silicone Fluid. The components were mixed
with a glass rod for two minutes.
11. The lid was then placed on the beaker and the mixture shaken vigorously for a
further 60 seconds. The beaker was then placed in an oven at 45oC.
12. To ensure that no separation of the mixture occurred, the mixture was mixed
with a glass rod for one minute and shaken vigorously 5 times at 10 minute
intervals.
13. The liquid ILT mixture was drawn into a syringe and injected into the mould
cavity. An M8 bolt was inserted into the inlet port of the mould and an M3 bolt
was inserted into the vent of the mould to prevent leakage of the ILT mixture.
14. The mould was rotated one quarter turn every 10 minutes to prevent separation
of the ILT mixture and to prevent sagging of the ILT Lumen casting.
15. Once the mould had been at 45oC for 40 minutes it was removed from the oven,
allowed to cool and opened. The ILT lumen and ILT analogue (henceforth ‘the
lumen’) were carefully removed from the mould.
16. The wall mould was cleaned thoroughly using methanol
17. The lumen was placed inside the wall mould, the mould halves were bolted
tightly together.
18. The wall material was prepared in a beaker with the components of the wall
mixture added in the same order as before and stirred with a glass rod for two
minutes.
19. As the liquid wall material was far more viscous than the ILT material the
beaker was placed in a freezer at -20oC for 4 hours to stop the curing process and
allow bubbles due to mixing to rise out of the mixture.
20. The wall material was then drawn into a 60mm catheter tipped syringe and
injected into the wall mould. For this step the mould was injected from the
bottom up with the inlet at the aorta end of the AAA model and the vents at the
iliac legs.
21. The inlet port was again sealed with an M8 bolt.
22. The wall model was placed in a freezer for 4 hours to allow any bubbles due to
injecting to rise to the ends of the iliac legs, away from the regions of interest.
23. The mould was then transferred to an oven at 40oC and the silicone rubber was
allowed to cure for 5 hours.
24. The mould was removed from the freezer, opened and the wax lumen with
silicone wall removed.
25. The model was placed in an oven at 100oC to melt out the wax lumen leaving
the silicone rubber model behind.
REFERENCES
1. Sakalihasan, N., Limet, R., and Defawe, O. D., 2005, “Abdominal Aortic
Aneurysm,” Lancet., 365(9470), pp. 1577-1589.
2. Li. Z., and Kleinstreur. C., 2005, “Fluid-Structure Interaction Effects on Sac-
Blood Pressure and Wall Stress in a Stented Aneurysm,” Journal of
Biomechanical Engineering, 127(4), pp. 662-671.
3. Office of Population Censuses and Surveys, 1989, “Mortality Statistics, England
and Wales,” London HMSO.
Corbett et al.
4. Chong, C. K., How, T. V., Black, R. A., Shortland, A. P., and Harris, P. L.,
1998, “Development of a Simulator for Endovascular Repair of Abdominal
Aortic Aneurysms,” Annals of Biomedical Engineering, 26(5), pp. 798-802.
5. Chong, C. K., How, T. V., Gilling-Smith, G. L., and Harris, P. L., 2003
“Modelling Endoleaks and Collateral Reperfusion Following Endovascular
AAA Exclusion,” Journal of Endovascular Therapy, 10(3), pp. 424-432.
6. Chong., C. K., and How, T. V., 2004, “Flow Patterns in an Endovascular Stent
Graft for Abdominal Aortic Aneurysm Repair,” Journal of Biomechanics, 37(1),
pp. 89-97.
7. Chong, C. K., How, T. V., and Harris, P. L, 2005, “Flow Visualization in a
Model of a Bifurcated Stent-Graft,” Journal of Endovascular Therapy,” 12(4),
pp. 435-445.
8. Metha, M., Veith, F. J., Ohki, T., Lipsitz, E. C., Cayne, N. S., and Darling,
R.,C., 2003, “Significance of Endotension, Endoleak, and Aneurysm Pulsatility
After Endovascular Repair,” Journal of Vascular Surgery, 37(4), pp. 842-846.
9. Gawenda, M., Jaschke, G., Winter, S., Wassmer, G., and Brunkwall, J., 2003,
“Endotension as a Result of Pressure Transmission Through the Graft Following
Endovascular Aneurysm Repair – an In Vitro Study,” European Journal of
Vascular and Endovascular Surgery, 26(5), pp. 501-505.
10. Gawenda, M., Knez, P., Winter, S., Jaschke, G., Wassmer, G., Schmitz-Rixen,
T., and Brunkwall, J., 2004, “Endotension is Influenced by Wall Compliance in
a Latex Aneurysm Model,” European Journal of Vascular and Endovascular
Surgery, 27(1), pp. 45-50.
11. Chaudhuri, A., Ansdell, L. E., Grass, A. J., Adiseshiah, M., 2004, “Intrasac
Pressure Waveforms After Endovascular Aneurysm Repair (EVAR) are a
Reliable Marker of Type I Endoleaks, but not Type II or Combined Types: An
Experimental Study,” European Journal of Vascular and Endovascular Surgery,
28(4), pp. 373-378.
12. Chaudhuri, A., Ansdell, L. E., Grass, A. J., Adiseshiah, M., 2004, “Aneurysmal
Hypertension and its Relationship to Sac Thrombus: a Semi-Qualitative
Analysis by Experimental Fluid Mechanics,” European Journal of Vascular and
Endovascular Surgery, 27(3), pp. 305-310.
13. Corbett, T. J., Callanan, A., Morris, L. G., Doyle, B. J., Grace, P. A., Kavanagh,
E. G., and McGloughlin, T. M., 2008, “A Review of The In Vivo and In Vitro
Biomechanical Behaviour and Performance of Postoperative Abdominal Aortic
Aneurysms and Implanted Stent-Grafts,” Journal of Endovascular therapy,
15(4), pp. 498-484.
14. Nichols, W. W., and O’ Rourke, M. F., 1998, McDonald’s Blood Flow in
Arteries, Edward Arnold, London, p. 73, Chap. 4.
15. Williams, P. L., Warwick, R., Dyson, M., Bannister, L. H., 1989 Gray’s
Anatomy, Churchill Livingstone, New York, pp. 766-767, Chap. 6.
16. Walker, R.D., Smith, R. E., Sherriff, S. B., and Wood, R. F., 1999, “Latex
Vessels with Customized Compliance for use in Arterial Flow Models,”
Physiological Measurement, 20(3), pp. 277-286.
17. Chong, C. K., Rowe, C. S., Sivanesan, S., Rattray, A., Black, R. A., Shortland,
A. P., and How, T. V, 1999 “Computer Aided Design and Fabrication of Models
for In Vitro Studies of Vascular Fluid Dynamics,” Proceedings of the Institution
of Mechanical Engineers, Part H: Journal of Engineering in Medicine, 213(1),
pp. 1-4.
18. Friedman. M. H., Kuban, P., Schmalbrock, K., Smith, K., and Altan, T., 1995,
“Fabrication of Vascular Replicas from Magnetic Resonance Images,” Journal
of Biomechanical Engineering, 117(3), pp. 364-365.
Corbett et al.
19. O’ Brien, T., Morris, L., O’ Donnell, M., Walsh, M., and McGloughlin, T.,
2005, “Injection-Moulded Models of Major and Minor Arteries: the Variability
of Model Wall Thickness Owing to Casting Technique,” Proceedings of the
Institution of Mechanical Engineers, Part H: Journal of Engineering in
Medicine, 219(5), pp. 381-386.
20. Doyle, B. J., Morris, L. G., Callanan, A., Kelly, P., Vorp, D. A., and
McGloughlin, T. M., 2008, “3D Reconstruction and Manufacture of Real
Abdominal Aortic Aneurysms: from CT Scan to Silicone Model,” Journal of
Biomechanical Engineering, 130(3), pp. 034501-1 - 034501-5.
21. Chaudhuri, A., Ansdell, L. E., Richards, R., Adiseshiah, M., and Grass, A. J.,
2003, “Non-Axisymmetrical (Life-Like) Abdominal Aortic Aneurysm Models:
A Do It Yourself Approach,” Journal of Endovascular Therapy, 10(6), pp. 1097-
1100.
22. Doyle, B. J., Corbett, T. J., Callanan, A., Walsh, M.T., Vorp, D. A., and
McGloughlin, T. M, 2009 “An Experimental and Numerical Comparison of the
Rupture Locations of an Abdominal Aortic Aneurysm,” Journal of Endovascular
Therapy (In Press).
23. Morris, L. G., 2004, “Numerical and Experimental Investigation of Mechanical
Factors in the Treatment of Abdominal Aortic Aneurysms,” Ph.D. Thesis,
University of Limerick, Limerick, Ireland.
24. Harter, L. P, Gross, B. H., Callen P., W., and Barth, R. A., 1982, “Ultrasonic
Evaluation of Abdominal Aortic Thrombus,” Journal of Ultrasound in Medicine,
1(8), pp. 315-318.
25. Timaran, C. H., Ohki, T., Veith, F. J., Lipsitz, E. C., Gargiulo, N. J, Rhee, S. J.,
Malas, M. B., Suggs, W. D., and Pacanowski, J. P., 2005 “Influence of Type II
Endoleak Volume on Aneurysm Wall Pressure and Distribution in an
Experimental Model,” Journal of Vascular Surgery, 41(4), pp. 657-663.
26. Mullins, L., 1969, “Softening of Rubber by Deformation,” Rubber Chemistry
and Technology, 42(1), pp. 339-362.
27. Laheij, R., Van Marrewijk, C., and Buth, J., 2001, “Progress Report on the
Procedural and Follow up Results of 3413 Patients who Received Stent-Graft
Treatment for Infra-Renal Abdominal Aortic Aneurysms for a Period of 6
Years,” EUROSTAR Data Registry Centre.
28. Morris, L., O’Donnell, P., Delassus, P., and McGloughlin, T., 2004,
“Experimental Assessment of Stress Patterns in Abdominal Aortic Aneurysms
Using the Photoelastic Method,” Strain, 40(4), pp. 165-172.
29. O’ Brien, T., Morris, L., and McGloughlin, T., 2007, “Evidence Suggests Rigid
Aortic Grafts Increase Systolic Blood Pressure: Results of a Preliminary Study,”
Medical Engineering and Physics, 30(1), pp. 109-115.
30. O’ Brien, T. P., Walsh, M. T., Morris, L. G., Grace, P. A., Kavanagh, E. G., and
McGloughlin, T. M., 2008, “Numerical and Experimental Techniques for the
Study of Biomechanics in the Arterial System,” in Biomechanical Systems
Technology, World Scientific Publishing Co., Singapore, pp. 233-270, Chap. 7.
31. Chuter, T. A. M., 2002 “Stent-Graft Design: The Good, the Bad and the Ugly,”
Cardiovascular Surgery, 10(1), pp. 7-13.
32. Vorp, D. A., Mandarino, M., Webster, M. W., Gorcsan, J., 1996, “Potential
Influence of Intraluminal Thrombus on Abdominal Aortic Aneurysm as
Assessed by a New Non-Invasive Method,” Cardiovascular Surgery, 4(6), pp.
732-739.
33. Sonesson, B., Lanne, T., Vernersson, E., and Hansen, F., 1994, “Sex Difference
in the Mechanical Properties of the Abdominal Aorta in Human Beings,”
Journal of Vascular Surgery, 20(6), pp. 959-969.
Corbett et al.
34. Doyle, B. J., Callanan, A., McGloughlin, T. M., 2007 “A Comparison of
Modelling Techniques for Computing Wall Stress in Abdominal Aortic
Aneurysms,” Biomedical Engineering Online, 6, 38.
35. Raghavan, M. L., Webster, M. W., and Vorp, D. A., 1996, “Ex Vivo
Biomechanical Behaviour of Aabdominal Aortic Aneurysm: Assessment Using
a New Mathematical Model,” Annals of Biomedical Engineering, 24(5) pp. 573-
582.
36. He, C. M., Roach, M. R., 1994, “The Composition and Mechanical Properties of
Abdominal Aortic Aneurysms,” Journal of Vascular Surgery, 20(1), pp. 6-13
37. Vallabhaneni, S. R., Gilling-Smith, G. L., How, T.V., Carter, S. D., Brennan, J.
A., and Harris, P. L., 2004, “Heterogeneity of Tensile Strength and Matrix
Metalloproteinase Activity in the Wall of Abdominal Aortic Aneurysms,”
Journal of Endovascular Therapy, 11(4), pp. 494-502.
38. Wang, D. H. J., Makaroun, M., Webster, M. W., and Vorp, D. A., 2001
“Mechanical Properties and Microstructure of Intraluminal Thrombus from
Abdominal Aortic Aneurysm,” Journal of Biomechanical Engineering, 123(6),
pp. 536-539.
39. Fairman, R. M., Velazquez, O. C., Carpenter, J. P., Woo, E., Baum, R. A.,
Golden, M. A., Kritpracha, B. and Criado, F., 2004, “Midterm Pivotal Trial
Results of the Talent Low Profile System for Repair of Abdominal Aortic
Aneurysm: Analysis of Complicated Versus Uncomplicated Aaortic Necks,”
Journal of Vascular Surgery, 40(6), pp. 1074-1082.
40. Li, Z., and Kleinstreuer, C., 2006, “Analysis of Biomechanical Factors Affecting
Stent-Graft Migration in an Abdominal Aortic Aneurysm Model,” Journal of
Biomechanics, 39(12), pp. 2264-2273
41. Hinnen, J. W., Rixen, D. J., Koning, O. H., van Bockel, J. H., and Hamming, J.
F., 2007, “Development of Fibrinous Thrombus Analogue for In Vitro
Abdominal Aortic Aneurysm Studies,” Journal of Biomechanics. 40(2), pp. 289-
295.
42. Finn, R., McGloughlin, T., Delassus, P., and Morris, L., “An Experimental
Evaluation of the Deformation During and Post Stenting on a Realistic Coronary
Artery Phantom,” Proc. 15th Annual Conference of the Section of
Bioengineering of the Royal Academy of Medicine in Ireland, McGloughlin, T.,
and Walsh, M., eds., Limerick, Ireland, p. 10.
Corbett et al.
J ENDOVASC THER
2009;16:000–000 0
¤EXPERIMENTAL INVESTIGATION ¤
An Experimental and Numerical Comparison of the

Rupture Locations of an Abdominal Aortic Aneurysm
Barry J. Doyle, BEng1; Timothy J. Corbett, BEng1; Anthony Callanan, PhD1;
Michael T. Walsh, PhD1; David A. Vorp, PhD2; and Timothy M. McGloughlin, PhD1
1Centre for Applied Biomedical Engineering Research (CABER), Department of
Mechanical and Aeronautical Engineering, and the Materials and Surface Science
Institute, University of Limerick, Ireland. 2Departments of Surgery and Bioengineering,
Centre for Vascular Remodeling and Regeneration, McGowan Institute for Regenerative
Medicine, University of Pittsburgh, Pennsylvania, USA.
¤ ¤
Purpose: To identify the rupture locations of idealized physical models of abdominal aortic
aneurysm (AAA) using an in-vitro setup and to compare the findings to those predicted
numerically.
Methods: Five idealized AAAs were manufactured using Sylgard 184 silicone rubber, which
had been mechanically characterized from tensile tests, tear tests, and finite element
analysis. The models were then inflated to the point of rupture and recorded using a high-
speed camera. Numerical modeling attempted to confirm these rupture locations. Regional
variations in wall thickness of the silicone models was also quantified and applied to
numerical models.
Results: Four of the 5 models tested ruptured at inflection points in the proximal and distal
regions of the aneurysm sac and not at regions of maximum diameter. These findings
agree with high stress regions computed numerically. Wall stress appears to be
independent of wall thickness, with high stress occurring at regions of inflection regardless
of wall thickness variations.
Conclusion: According to these experimental and numerical findings, AAAs experience
higher stresses at regions of inflection compared to regions of maximum diameter.
Ruptures of the idealized silicone models occurred predominantly at the inflection points,
as numerically predicted. Regions of inflection can be easily identified from basic 3-
dimensional reconstruction; as ruptures appear to occur at inflection points, these findings
may provide a useful insight into the clinical significance of inflection regions. This
approach will be applied to patient-specific models in a future study.
J Endovasc Ther 2009;16:000–000
Key words: Abdominal aortic aneurysm, in-vitro model, experimental study, numerical,
rupture, inflection
¤ ¤
Aneurysms are permanent and irreversible serious ones occur in the infrarenal section of
localized dilatations1 that can form in any the abdominal aorta. If left untreated, abdom-
blood vessel, but the majority of the more inal aortic aneurysm (AAA) can eventually
This work was financially supported by the Irish Research Council for Science, Engineering and Technology (IRCSET)
Grant RS/2005/340 and also Grant #R01-HL-060670 from the US National Heart Lung and Blood Institute, Bethesda,
Maryland, USA.
The authors have no commercial, proprietary, or financial interest in any products or companies described in this article.
Address for correspondence and reprints: Prof. Timothy McGloughlin, MSG-014, MSSI Building, University of Limerick,
Ireland. E-mail: Tim.McGloughlin@ul.ie
ß 2009 by the INTERNATIONAL SOCIETY OF ENDOVASCULAR SPECIALISTS Available at www.jevt.org
Journal of Endovascular Therapy enth-16-03-18.3d 15/5/09 20:33:54 1 Cust # 09-2697

0 AAA RUPTURE LOCATIONS J ENDOVASC THER
Doyle et al. 2009;16:000–000
rupture. Approximately 500,000 AAAs are

diagnosed worldwide each year,2 resulting
in 15,000 deaths per year in the US alone.3
Typically, an AAA is surgically repaired
once it reaches/exceeds a diameter of 5 to
Figure 1 ¤ (A) Type 2 dumbbell specimen from
5.5 cm. There have been reports that this BS ISO 37. (B) Modified trouser test piece from BS
threshold may lead to inaccurate surgical- ISO 34-1; a indicates the direction of the cut. All
decision making because not only can smaller dimensions are in millimeters, and the thickness of
AAAs rupture,4–6 but also large AAAs can each specimen is 2 mm.
remain stable.6 By applying the definition of
material failure, we know that an AAA will METHODS
rupture when the locally acting wall stress
exceeds the local wall strength. Currently, Material Characterization and Validation
much research is aimed at examining the
Silicone (Sylgard 184; Dow Corning, Midland,
AAA wall stress distributions with a view to
MI, USA), an elastomeric poly(dimethylsilox-
improving rupture prediction.7–16 Researchers
ane) (PDMS), was used to construct the
have also examined methods to determine
the strength of the AAA wall, both invasively physical models for this study. This transpar-
from excised tissue17–19 and noninvasively.20 ent silicone has an ultimate tensile strength
While there is much focus on attempting to (UTS) of 7.1 MPa and a tear strength of 2.6 N/
numerically understand the mechanics of mm, according to the manufacturer’s specifi-
AAA rupture, only limited work has focused cation sheet. Supplied in 2 parts, Sylgard 184
on the development of experimental methods is mixed in a 10:1 ratio of base to curing
of determining rupture potential. agent. Although silicone rubber and arterial
Previous work by Morris et al.21 reported tissue are not identical in their stress/strain
the use of the photoelastic method in behavior, silicone remains the most suitable
determining wall stress distributions in an analogue.
idealized AAA case that used dimensions Sample preparation. Aluminum molds
taken from a comprehensive study.22 Morris were designed and manufactured to be used
et al.21 observed rupture of the photoelastic with the injection-molding method to create
model at the proximal and distal inflection the silicone tensile testing samples. The mold
points. This is the only publication in which cavity for the tensile test samples conformed
rupture of an experimental AAA model has to a type 2 tensile test specimen as outlined in
been reported. Callanan et al.23 later validat- BS ISO 37 standards, with a second set of
ed this photoelastic work numerically. Other molds designed for creating trouser test
researchers24 have also employed an ideal- samples to be used for tear testing, per BS
ized AAA in experimental studies, this time ISO 34-1 standards. The dimensions of these
using strain gauges attached to a latex samples are shown in Figure 1. The samples
model to determine the stress in the AAA prepared for tear strength testing differed
wall. The technique of creating rubber slightly in dimensions from those outlined in
models of AAAs has been reported previ- BS ISO 34-1.
ously by Doyle and colleagues,25 which was The trouser test (Fig. 2) specimens had
adapted from related techniques developed ‘‘legs’’ that were extended in opposite direc-
by O’Brien et al.26 tions to determine the tear strength of the
The primary purpose of this study is to material. The protocol for preparing samples
identify the sites of rupture in idealized AAA for testing is outlined in Appendix I. Briefly,
models by performing in-vitro rupture studies the molds were thoroughly cleaned and
and to compare these findings with those bolted together, and the necessary amounts
predicted using numerical modeling. Me- of silicone were manually mixed. Degassing
chanical characterization of the material used was required at this stage as air bubbles can
was also performed, along with the examina- get trapped in the silicone during the mixing
tion of silicone model wall thickness. process. Once all air was removed, the

J ENDOVASC THER AAA RUPTURE LOCATIONS 0
2009;16:000–000 Doyle et al.
then a 4th order polynomial curve was fitted

to the data to obtain a representative exper-
imental data curve. The polynomial curve
applied to the experimental data was then
used as the material model for the finite
element analysis (FEA), which was performed
with the ABAQUS commercial FEA solver
(version 6.7; Dassault Systems, Simulia, RI,
USA). A dumbbell model identical to that of
the tensile tests was replicated numerically,
with the 4th order polynomial curve applied to
represent the experimental data; this was
Figure 2 ¤ Illustration showing the loading direc- used to determine the most suitable strain
tions that are applied to the trouser test specimen
energy function (SEF). Various SEFs were
using the tensile test machine.
examined, such as the Neo-Hookean, Ogden,
Yeoh, and Mooney-Rivilin, to determine the
silicone was transferred from the mixing
most applicable SEF. Material constants
container to a syringe, where it was then
could then be determined for any chosen
injected into the mold cavity. Further degas-
SEF, thus allowing the material to be readily
sing in a vacuum was required after the modeled in numerical studies.
injection process. The mold was then placed
Finite element analysis. By replicating the
into an oven at 45uC for 18 hours to ensure
tensile test experiments using a numerical
adequate curing of the samples. Once cured, simulation, the accuracy of the material
the samples were removed from the mold coefficients could be examined. For this, the
and left at room temperature for 3 hours numerical model was simulated in the same
before further preparation for both tensile manner as that of the tensile tests, with a half-
and tear testing. model analyzed due to symmetry. The model
Uniaxial tensile testing (Appendix II). Using was constrained from all movement at one
a test apparatus (H25KS; Tinius Olsen, Surrey, end, with a displacement constraint identical
UK) with a 1-kN load cell, an extension rate of to that of the tensile test applied to the
500 mm/min was applied, as recommended opposite end. Mesh independence was deter-
in BS ISO 37 for type 2 specimens. Each mined by examining the stress and strain
sample was preconditioned in an initial 10- experienced in a particular node at the center
cycle, 20% extension stretch-relax program to of the gauge length over the duration of the
stabilize the stress-strain function of the loading. This was performed for mesh densi-
material and overcome the Mullin’s effect.27 ties of 260, 1752, and 5000 elements. A mesh
The sample was then stretched to failure. of 1752 elements was deemed adequate as
Nine samples were used (the BS ISO 37 there was no significant difference in results
recommends that at least 3 samples be used when the mesh density was increased to 5000
to determine material properties). elements.
Tear testing (Appendix III). The Tinius Olsen
test apparatus was used to apply an exten-
sion rate of 100 mm/min to each of 12 Rupture Modeling
samples per BS ISO 34-1. Figure 2 shows Experimental AAA model. In order to study
the load directions that were applied to the the rupture of the aneurysm in vitro, a
trouser test specimen when the legs were previously reported21,23,26 idealized AAA
pulled in opposite directions until the point of model was used. This ideal AAA was devel-
complete material failure. oped using realistic dimensions obtained
Strain energy function. To mechanically from the EUROSTAR data registry.22 In brief,
characterize the material, the experimental the ideal AAA was geometrically symmetric,
data from the tensile tests was converted to with a maximum internal diameter of 50 mm,
engineering stress and engineering strain and a total length of 260 mm, and a 2-mm

Doyle et al. 2009;16:000–000
Figure 3 ¤ Experimental setup of rupture test. (1) Retort stand, (2) plug, (3) clamp, (4) silicone
rubber AAA model, (5) plug, (6) pressure gauge, (7) mirror-walls, (8) pressure regulator, and
(9) pneumatic air-line. Mirror walls allow 360u of AAA model to be examined.
uniform wall thickness. Experimental ideal manometer, a high-speed camera and vari-
AAA silicone models were manufactured ous clamps and silicone tubing. The use of
using the lost-wax process previously report- correctly positioned and angled mirrors al-
ed by our group.25,26 Using this technique, lows 360u of the model to be viewed, which is
male and female aluminum molds were of paramount importance to the success of
designed using computer-aided design the experiments (Fig. 3). A high-speed cam-
(CAD) and machined using a computer nu- era (Olympus i-Speed; Olympus Corporation,
merical control (CNC) 3-axis milling machine. Tokyo, Japan) was used to capture the point
The female mold allowed the creation of a and location of rupture. This camera was
wax model of the AAA that upon solidification capable of recording images at rates up to
was then placed into the outer male mold. 33,000 frames per second (fps). A pixel
This male mold had a uniform wall cavity of resolution of 8003600 at 1000 fps was
2 mm that facilitated the injection of silicone deemed adequate for this application, and
around the wax model. Once the silicone images were recorded using a monochrome
rubber was cured, the complete model could image sensor. Once the AAA model was
be removed and the wax melted out from the attached to the test rig, the high speed
model. To ensure the optimum uniform wall camera was adjusted to ensure optimum
thickness of the resulting silicone rubber focus and angle. After a satisfactory setup
model, the female mold was heated to slow was established, the air pressure was in-
down the solidification process of the wax creased and the camera set to record. Air
model, which reduces the amount of shrink- pressure was incrementally increased by
age involved with the cooling process, result- 20 mmHg every 30 seconds until rupture
ing in more accurate wax models. An accu- occurred. The pressure readings were also
rate wax model allows a uniform 2-mm cavity recorded in the video image and so could be
to surround the wax upon placement in the examined post rupture.
male outer mold. Five AAA models to be used Measurement of wall thickness. In order to
in the rupture study were manufactured using determine the variations in wall thickness
this technique, with a further 4 models throughout the AAA models, 4 additional
created to determine wall thickness. silicone rubber ideal AAA models were creat-
Experimental setup. The experimental test ed. These models were sectioned at 20-mm
rig consisted of a series of mirrors, a pneu- increments along the longitudinal distance of
matic airline, a pressure regulator, a pressure each model (Fig. 4), similar to a measurement

2009;16:000–000 Doyle et al.
Figure 4 ¤ Cross-sections of ideal AAA rubber Figure 5 ¤ Representative numerical ideal AAA,
model. Cross-sections (left) correspond to slices uniform wall. (A) The finite element mesh and (B)
shown on model (right). Slices were taken at 20- loading conditions and constraints.
mm longitudinal increments. Image shows relative
uniformity of wall thickness throughout the the mesh until the peak stress is ,2% of that
rubber model. recorded with the previous mesh size. Initial-
ly, a uniform 2-mm thick model was subjected
technique previously reported.25,26 Cross sec- to an internal pressure of 364.6 mmHg, which
tions were then measured for wall thickness is an average of the 5 rupture pressures
at 4 90u equidistant locations. Results for each observed during the experiments. The model
region were averaged across the 4 models so was also tested using all 5 of the actual
that regional variations in thickness could be experimental rupture pressures. Peak von
compared. Wall thickness measurements Mises wall stress was analyzed along with
were then analyzed using SPSS 15 (SPSS peak wall tension.19 In order to gauge the
Inc., Chicago, IL, USA) to determine the 25th effect of wall thickness on the resulting wall
and 75th percentiles, maximum and minimum stress, AAA geometries were created using
wall thickness, and the mean 6 standard varying wall thickness corresponding to the
deviation. 25th and 75th percentiles, maximum and
minimum wall thickness, mean wall thick-
Numerical Modeling ness, a thick walled inflection region, and a
thin walled inflection region. This required
To correlate experimental and numerical the creation of an additional 7 numerical
results, the evaluated material constants were models. Each of the models was subjected
applied to the ideal AAA model and imple- to an internal pressure of 120 mmHg, which is
mented in the FEA software. The pressure commonly regarded as the average peak
loadings observed experimentally were then systolic pressure of the cardiac cycle. By
applied numerically. Boundary conditions applying a common internal pressure the
similar to the experimental setup were imple- stress on the AAA wall could be compared
mented in the virtual AAA model, so that the across each model. A sample numerical
model was constrained from all movement at model can be seen in Figure 5, which shows
the proximal neck and iliac legs. Due to the the finite element mesh and boundary condi-
symmetrical nature of the model, only half tions used in the study.
the model was numerically analyzed using
symmetry constraints to reduce computation-
al time. Each model was meshed using 3- RESULTS
dimensional (3D) stress elements with mesh Material Characterization and Validation
independence performed as previously re-
ported.13 This mesh independence test in- Uniaxial tensile test. Tensile testing re-
volves increasing the number of elements of vealed that the mean UTS of Sylgard 184

Doyle et al. 2009;16:000–000
¤ ¤
TABLE 1
Material Coefficients Determined for the 3rd Order
Ogden Model for Sylgard 184
m a
1 2304.235 1.2667
2 148.232 1.5962
3 157.156 0.9075
¤ ¤
where W is the strain energy density per

undeformed unit volume, (l1, l2, l3) are the
principal stretch ratios, a is a strain hardening
Figure 6 ¤ Engineering stress and engineering
exponent, and m has the interpretation of the
strain experimental data (gray lines) with 4th order
shear modulus under infinitesimal straining.
polynomial curve applied to the results (black line).
Engineering stress refers to the stress associated The resulting material coefficients are pre-
with the initial cross-sectional area of the material. sented in Table 1.
Engineering stress is defined as dE5F/A, where dE Finite element analysis. The results of the
is engineering stress, F is force, and A is cross- FEA and the uniaxial tensile tests compared
sectional area. Engineering strain is defined as favorably (Fig. 7), thus establishing confi-
eE5Dl/l, where e is engineering strain, Dl is the dence in the material characterization of
change in gauge length, and l is the original Sylgard 184.
gauge length.
was 7.736161.6597 MPa (range 4.74–9.39) Rupture Modeling

compared to 7.1 MPa on the manufacturer’s Experimental rupture tests. Of the 5 silicone
specification sheet. Figures 6 shows the models ruptured in vitro, 4 models experi-
stress/strain curves for each sample, along enced rupture at a region of inflection on the
with the 4th order polynomial curve fit. surface of the model. An inflection point is
Tear test. Tear testing showed that the defined as a point on the AAA surface at
mean tear strength of the material was which the local AAA wall shape changes from
0.41960.094 N/mm. The manufacturers re- concave outward to concave inward.29 This
ported a 2.6-N/mm tear strength on the finding is consistent with previous numerical
specification sheet, but it is known that results and experimental reports by our group13,21,23
can depend heavily on the type of tear test and others,29 who noted peak stresses occur-
sample used28; therefore, it is difficult to ring at these regions instead of at the
compare our results with those of the manu- maximum AAA diameter. A summary of the
facturer. For this study, the average tear rupture locations along with corresponding
strength of 0.419 N/mm was taken as the burst pressures is shown in Table 2. One
maximum wall tension that the material can silicone model ruptured at the iliac bifurca-
resist prior to rupture. tion. The sequence of events leading to
Strain energy function. For this particular rupture for Test 1 can be seen in Figure 8,
material, the 3rd order Ogden SEF proved to which shows the frame where the material
be the most suitable constitutive equation failure initiates, leading to rupture, and ulti-
because it provided a good curve fit to the mately complete failure of the silicone model.
data (R250.9812) and was also stable at all Wall thickness results. Wall thickness was
stresses and strains. The ABAQUS form of the assessed by slicing the 4 ideal AAA models
Ogden SEF is shown in Eqn. 1: into cross sections (Fig. 9). These 4 models
X
N were made using the same process as the
2mp
W ðl1 ,l2 ,l3 Þ ~ ðl1 ap z l2 ap z l3 ap { 3Þ, previous AAA models. Table 3 shows the
p~1
a2p average regional wall thicknesses corre-
ð1Þ sponding to the sections in Figure 9. This

2009;16:000–000 Doyle et al.
Numerical Modeling
Stress distributions on the surfaces of the
virtual AAA model reveal that high stresses
occur at the regions of inflection and not at
regions of maximum diameter. This has been
previously shown by our group for this
particular idealized AAA model both numer-
ically using FEA23 and experimentally using
the photoelastic method.21 The resulting von
Mises stress distribution on the surface of the
ideal AAA at a mean experimental pressure of
364.5 mmHg, along with the image captured
Figure 7 ¤ Comparison of results for the experi-
at the point of rupture, is shown in Figure 11.
mental tensile tests and the numerical simulation
of Sylgard 184. True stress refers to the stress Applying the same mean internal experimen-
associated with the current cross-sectional area of tal pressure to the FEA model (364.5 mmHg)
the material. True stress takes into account the returned a peak stress of 1.16 MPa located at
change in cross-sectional area. True stress is the iliac bifurcation. The figure also shows the
defined as dT5 dE*(1+eE), where dT is true stress, qualitative agreement in stress distributions
dE is engineering stress, and eE is the engineering at this pressure loading with the rupture
strain. True strain is defined as eT5ln(1+eE), where location found experimentally. Inflection
eT is true strain and eE is the engineering strain. zones experience higher stress than regions
of maximum diameter. Wall stress results can
table also shows the wall thickness used to also be expressed as wall tension results, as
create numerical models, with thick and thin the failure tension of the wall is believed to be
regions at the inflection points. The minimum a better indicator of a region’s susceptibility
wall thickness recorded in any model was to failure,19 with peak wall tension 5 peak
1.17 mm and was located at the iliac bifurca- wall stress 3 wall thickness. The numerical
tion region. The maximum wall thickness was results from the 5 ranges of pressures
2.53 mm and was also found at the iliac applied during the experimental tests result-
bifurcation region. ed in varied wall stresses and wall tensions
The effect of wall thickness on stress (Tables 4 and 5).
distributions within the AAA sac is displayed
in Figure 10. The uniform 2-mm wall model
appears to represent the overall behavior of DISCUSSION
the AAA model regardless of variations in The focus of this study was to experimentally
wall thickness, with the exception of the rupture idealized silicone AAA models and
‘‘min’’ thickness model. This study highlight- compare these locations to numerical predic-
ed that the inflection region, in particular the tions. There has been little reported on the in
proximal inflection region, experiences high- vitro rupture of AAAs, with much focus on the
er stresses than the maximum diameter computational analysis of these aneurysms.7–16
region regardless of wall thickness. Morris et al.21 observed rupture locations at
the inflection regions the using the photoelas-
¤ ¤
TABLE 2
tic method. These rupture locations correlate
Summary of Experimental Rupture Results with the sites of rupture experienced in our
study.
Test Rupture Location Rupture Pressure, mmHg
Nine models were manufactured using the
1 Proximal inflection point 254.7 injection-molding technique,25,26 of which 5
2 Proximal inflection point 278.6 were experimentally ruptured and 4 were
3 Proximal inflection point 466.2
4 Distal inflection point 278.7
used to assess wall thickness. Although
5 Iliac bifurcation 544.6 silicone rubber does not behave identically
¤ ¤ to arterial tissue when subjected to large

Doyle et al. 2009;16:000–000
Figure 8 ¤ Sequence of events leading to model rupture of Test 1. (A) The inflated model, (B) the
initial point of rupture, (C) propagation of the failure zone, and (D) complete failure of the silicone
model. A similar sequence was observed for all 4 models that ruptured at regions of inflection.
strains, it is currently the most suitable Material Characterization

analogue and has been widely used in
previous studies.25,26,30,31 Until a more suit- During tear testing, the tear in each sample
able material is available, one that possibly was smooth, possibly indicating that the
incorporates the layered structure of the bond within the PDMS chains was not strong
arterial wall and the presence of a fibrinous enough to prevent rupture under an applied
network (collagen and elastin), silicone rub- force.32 Tensile testing documented a UTS
ber will remain the most feasible arterial that correlated well with the manufacturer’s
mimic. specifications, and numerical replication of
Figure 9 ¤ (A) Key regions and (B) wall thickness measurements of the ideal AAA model.
Data points are individual measurements. Upper and lower bars represent 25th and 75th
percentiles. Middle bar is mean wall thickness.

2009;16:000–000 Doyle et al.
¤ ¤
TABLE 3
Average Regional Variations in Wall Thickness
Wall Thickness Variations by AAA Region, mm
Model 1 2 3 4 5
Uniform 2 mm 2 2 2 2 2
25th Percentile 1.94 1.89 1.63 1.7 1.55
75th Percentile 2.16 2.29 1.87 2.18 2.17
Min thickness 1.72 1.65 1.34 1.26 1.17
Max thickness 2.49 2.51 2.04 2.44 2.53
Mean thickness 2.03 2.08 1.78 1.91 1.87
Thick inflection 2 2.5 1.5 2.5 2
Thin inflection 2 1.5 2.5 1.5 2
¤ ¤
Thickness values were applied to numerical models to determine the effect of wall thickness on stress results. AAA
regions correspond to Figure 9.
the type 2 dumbbell sample used for the sures varied significantly for 2 of the 5
tensile tests agreed well with the experimen- models. Four models experienced rupture at
tal data. The 3rd order Ogden model was an inflection region. Elevated stresses at
selected as the optimum strain-energy func- regions of inflection have been previously
tion for this particular material; it allowed the reported in experimental models, using both
material to remain stable at all stresses and the photoelastic method21 and rosette strain
strains within the ranges presented. To gauge arrangements.24 Flora et al.24 observed
numerically model the actual rupture of the stresses at the inflection region nearly twice
silicone models, a different approach to the those experienced at the region of maximum
traditional continuum mechanics method diameter. The one model that ruptured at the
could be used. This could take into account iliac bifurcation was unusual; in a clinical
the supersonic rupture of rubber33 and may setting, real AAAs rarely rupture at the iliac
provide a further insight into the rupture bifurcation. Rather, rupture is typically at the
behavior of these models. For the purpose posterior wall,5 so the iliac bifurcation region
of this article, we are interested in rupture of our model may have had a localized flaw.
regions of the models, not the numerical
quantities per se.
Wall Thickness
The mean wall thickness of the models was
Rupture Modeling
comparable to the 2-mm original design of
The use of a high-speed camera to record the aluminum molds, so it was acceptable to
the point of rupture proved to be a very use a uniform 2-mm wall to numerically
powerful experimental tool. As image quality represent the ideal AAA. The statistical meth-
was of paramount concern with this work, od to determine the 25th and 75th percentiles
determining the optimum resolution and of the wall thickness allowed us to create
frame rate was crucial to the success of the realistic wall thickness variations in the
testing. The mirror-wall arrangement also numerical AAA models. The maximum and
proved to be very effective for allowing all minimum wall thickness values were also
360u of the model to be viewed. Upon used to create ‘‘worst case scenario’’ AAA
pressurization, the AAA models did not inflate models with quite large variations in wall
uniformly, resulting in asymmetry, possibly thickness. Also, two models were generated
due to localized variations in wall thickness. that had extreme changes in wall thickness
However, it was shown that variations in wall (referred to as Thick and Thin Inflection in
thickness do not significantly affect the over- Table 3). Computations of the wall stress
all stress distributions, with the exception of revealed that the inflection regions, particu-
the ‘‘min’’ thickness model. Rupture pres- larly the proximal inflection zone, experi-

Doyle et al. 2009;16:000–000
enced higher stress than the region of

maximum diameter, regardless of wall thick-
ness variations.
Numerical Modeling
The peak wall stress located immediately
prior to the iliac bifurcation was much lower
than the UTS of the material, so one would
not expect rupture of the model. Peak stress-
es at the iliac bifurcation are common in
numerical computations of AAA wall stress. It
Figure 10 ¤ Effect of varying wall thickness on is known that stress artifacts arise due to the
wall stress results. X-axis refers to the longitudinal inclusion of this bifurcation, and a method of
distance from above the AAA sac to immediately evaluating and refining this localized stress
above the iliac bifurcation. The line on the AAA region has yet to be developed.34 The peak
model (between the brackets) indicates the path of stress obtained numerically was on average
stress results. The ‘‘25th’’ and ‘‘75th’’ refer to only 13.8% of the UTS, with the stress at the
models created using the 25th and 75th percentiles
inflection regions even less (8.4%). The
of the wall thickness results; min, max, and mean
refer to models created using the corresponding
manufacturers state that UTS values can have
wall thickness values recorded. a very large range, so the stated 7.1-MPa UTS
can actually be as low as 3.5 MPa.
Wall tension in these models exceeded the
tear strength of the material, which may
suggest that the model would experience
rupture in all cases examined, if indeed tear
strength is to be considered a possible
indicator of material failure strength. At
elevated pressures, silicone AAA models
may rupture due to the pressure applied to
the inner surface, causing an increase in wall
tension and propagation of a nick or tear,
ultimately leading to complete failure of the
silicone wall. In realistic situations, calcified
deposits embedded in the AAA wall may raise
stress and could act as rupture initiators.
Ruptures may occur at these areas, in partic-
ular if calcifications occur at regions of
inflection. Analogue models to replicate cal-
cified deposits can be included in future
rubber models using available techniques at
our laboratory.
Figure 11 ¤ In vitro rupture locations and FEA Good agreement was observed when com-
stress patterns. High stress occurs at inflection paring the experimental rupture locations
points that correlate with rupture locations in with the high stress regions found using
experimental models. The undeformed image
FEA. It is believed that the geometry of the
shows stress contours displayed on the AAA
idealized AAA model may play a role in
model prior to loading, with the deformed illustra-
tion showing the stress contours displayed on the rupture behavior. The 3D nature of the model,
model when an internal loading of 364.5 mmHg is along with slight deviations in wall thick-
applied. There is a noticeable change in geometry ness,25,26 may increase the likelihood of
with increased internal pressure, particularly at the rupture. Another interesting point to note is
inflection regions. that when the experimental model is inflated,

2009;16:000–000 Doyle et al.
¤ ¤
TABLE 4
Summary of FEA Wall Stress Results
Rupture Test Pressure, mmHg dmax, MPa dinflection, MPa dmax, % of UTS dinflection, % of UTS
1 254.7 0.6808 0.4156 9 5
2 278.6 0.7375 0.4586 10 6
3 466.2 1.337 0.8668 17 11
4 278.7 0.7375 0.4586 10 6
5 544.6 1.733 1.1033 23 14
¤ ¤
d: stress.
UTS: ultimate tensile strength (7.7 MPa) of the material as determined from tensile testing.
the radius of curvature of the inflection region representative calcified deposits. Numerical
changes, while the numerical model does not modeling has shown that thrombus reduces
experience such a significant change. This wall stress,13,15,36 but it has a negligible effect
may also contribute to the difference in the on wall pressures in vivo,37 so there is also a
wall stress results. need to examine the role of ILT on in-vitro
Also, it is known that no silicone model is rupture tests. Calcifications can affect numer-
absolutely free from defects and imperfec- ically computed wall stress and also decrease
tions. The manufacturing process results in a the biomechanical stability of the AAA,38,39 so
silicone model that may contain microscopic they should be included in future work.
flaws. Tensile testing determined the UTS of
the material using planar dog-bone shapes,
Limitations
whereas the AAA model has regions of
inflection, along with a bifurcation branching A more suitable AAA wall analogue could
into the iliac arteries. These factors may all be employed. Use of silicones with lower
contribute to the propensity of the silicone tensile strengths would allow lower, more
model to rupture, and hence the FEA results realistic pressures to cause rupture. Work has
would differ somewhat. It has also been begun in our group on the use of more
reported31 using the Law of Laplace that suitable arterial vessel analogues.40 Also, the
actual AAAs can withstand up to 982 mmHg use of a realistic cardiac pressure wave
before rupture, with much variability between instead of static air pressure may alter the
intrasac rupture pressures.35 results. These factors are to be examined in
Our group has begun work on the rupture future work.
of realistic AAA models and the correlation
with both FEA-predicted and photoelastic-
Conclusion
predicted high stress regions. Also, our group
has developed a successful method of creat- According to the experimental and numer-
ing silicone AAA models complete with a ical findings presented here, AAAs experience
region of intraluminal thrombus (ILT) and higher stresses at regions of inflection com-
¤ ¤
TABLE 5
Summary of FEA Wall Tension Results
Rupture Test Pressure, mmHg Vmax, N/mm Vinflection, N/mm
1 254.7 1.3616 0.8312

2 278.6 1.475 0.9172
3 466.2 2.674 1.7336
4 278.7 1.475 0.9172
5 544.6 3.466 2.2066
¤ ¤
V: wall tension.
Average tear strength 5 0.419 N/mm.

Doyle et al. 2009;16:000–000
pared to regions of maximum diameter. 6. Vorp DA. Biomechanics of abdominal aortic

Ruptures of the idealized silicone models aneurysm. J Biomech. 2007;40:1887–1902.
occurred predominantly at the inflection 7. Raghavan ML, Vorp DA, Federle MP, et al. Wall
points, as numerically predicted. The use of stress distribution on three-dimensionally re-
constructed models of human abdominal aor-
a high speed camera is a useful experimental
tic aneurysm. J Vasc Surg. 2000;31:760–769.
tool in observing rubber AAA rupture loca-
8. Venkatasubramaniam AK, Fagan MJ, Mehta T,
tions. To improve the method described, et al. A comparative study of aortic wall stress
more suitable arterial analogues that mimic using finite element analysis for ruptured and
arterial properties more closely are required, non-ruptured abdominal aortic aneurysms.
thus possibly leading to an improved under- Eur J Vasc Endovasc Surg. 2004;28:168–176.
standing of AAA rupture. Regions of inflec- 9. Fillinger MF, Marra SP, Raghavan ML, et al.
tion can be easily identified from basic 3D Prediction of rupture risk in abdominal aortic
reconstruction, and as ruptures appear to aneurysm during observation: wall stress ver-
occur at inflection points, these findings, sus diameter. J Vasc Surg. 2003;37:724–732.
along with future work, may provide a useful 10. Fillinger MF, Raghavan ML, Marra SP, et al. In
vivo analysis of mechanical wall stress and
insight into the clinical significance of inflec-
abdominal aortic aneurysm rupture risk. J Vasc
tion regions. Surg. 2002;36:589–597.
11. Truijers M, Pol JA, SchultzeKool LJ, et al. Wall
Acknowledgments: The authors would like to thank (1) stress analysis in small asymptomatic, symp-
the Irish Research Council for Science, Engineering and tomatic and ruptured abdominal aortic aneu-
Technology (IRCSET) for grant RS/2005/340; (2) the US rysms. Eur J Vasc Endovasc Surg. 2007;33:
National Heart Lung and Blood Institute for grant #R01-
401–407.
HL-060670; (3) Dr. Liam Morris from the Galway Medical
Technology Centre, Galway, Ireland; (4) Kevin O’Flanagan
12. Mower WR, Quinones WJ, Gambhir SS. Effect
from the Department of Manufacturing and Operations of intraluminal thrombus on abdominal aortic
Engineering, University of Limerick, Ireland, for his aneurysm wall stress. J Vasc Surg. 1997;26:
assistance with the high-speed imaging; (5) Maria Ryan 602–608.
for her help with the experimental rupture modeling; (6) 13. Doyle BJ, Callanan A, McGloughlin TM. A
the Department of Vascular Surgery in the Midwestern comparison of modelling techniques for com-
Regional Hospital, Ireland, in particular, Mr. Eamon puting wall stress in abdominal aortic aneu-
Kavanagh, Mr. Paul Burke, and Prof. Pierce Grace; and rysms. Biomed Eng Online. 2007;6:38.
(7) Michel S. Makaroun, MD, Department of Surgery, 14. Rodriguez JF, Ruiz C, Doblare M, et al. Me-
University of Pittsburgh, Pennsylvania, USA.
chanical stresses in abdominal aortic aneu-
rysms: influence of diameter, asymmetry, and
REFERENCES material anisotropy. J Biomech Eng. 2008;130:
021023.
1. Sakalihasan N, Limet R, Defawe OD. Abdomi- 15. Doyle BJ, Callanan A, Walsh MT, et al. A finite
nal aortic aneurysm. Lancet. 2005;365:1577– element analysis rupture index (FEARI) as an
1589. additional tool for abdominal aortic aneurysm
2. Vande Geest JP, Sacks MS, Vorp DA. A planar burst prediction. Vasc Dis Prev. 2009;6:114–121.
biaxial constitutive relation for the luminal 16. Doyle BJ, Callanan A, Burke PE, et al. Vessel
layer of intra-luminal thrombus in abdominal asymmetry as an additional diagnostic tool in
aortic aneurysms. J Biomech. 2006;39:2347– the assessment of abdominal aortic aneu-
2354. rysms. J Vasc Surg. 2009;49:443–454.
3. Kleinstreuer C, Li Z. Analysis and computer 17. Thubrikar MJ, Labrosse M, Robicsek F, et al.
program for rupture-risk prediction of abdom- Mechanical properties of abdominal aortic an-
inal aortic aneurysms. Biomed Eng Online. eurysm wall. J Med Eng Tech. 2001;25:133–142.
2006;5:19. 18. Raghavan ML, Webster MW, Vorp DA. Ex vivo
4. Nicholls SC, Gardner JB, Meissner MH, et al. biomechanical behaviour of abdominal aortic
Rupture in small abdominal aortic aneurysms. aneurysm: assessment using a new mathemat-
J Vasc Surg. 1998;28:884–888. ical model. Ann Biomed Eng. 1996;24:573–582.
5. Darling RC, Messina CR, Brewster DC, et al. 19. Raghavan ML, Kratzberg J, de Tolosa EM, et al.
Autopsy study of unoperated abdominal aortic Regional distribution of wall thickness and
aneurysms. The case for early resection. Cir- failure properties of human abdominal aortic
culation. 1977;56:161–164. aneurysm. J Biomech. 2006;39:3010–3016.

2009;16:000–000 Doyle et al.
20. Vande Geest JP, Wang DH, Wisniewski SR, 33. Marder M. Supersonic rupture of rubber.
et al. Towards a non-invasive method for J Mech Phys Solids. 2006;54:491–532.
determination of patient-specific wall strength 34. Fillinger M. The long-term relationship of wall
distribution in abdominal aortic aneurysms. stress to the natural history of abdominal
Ann Biomed Eng. 2006;34:1098–1106. aortic aneurysms (finite element analysis and
21. Morris L, O’Donnell P, Delassus P, et al. other methods). Ann NY Acad Sci. 2006;1085:
Experimental assessment of stress patterns in 22–28.
abdominal aortic aneurysms using the photo- 35. Vallabhaneni SR, Gilling-Smith GL, How TV,
elastic method. Strain. 2005;40:165–172. et al. Heterogeneity of tensile strength and
22. Laheij R, van Marrewijk C, Buth J. Progress matrix metalloproteinase activity in the wall of
report on the procedural and follow up results abdominal aortic aneurysms. J Endovasc Ther.
of 3413 patients who received stent graft 2004;11:494–502.
treatment for infrarenal aortic aneurysms for 36. Wang DH, Makaroun MS, Webster MW, et al.
a period of 6 years. EUROSTAR Data Registry Effect of intraluminal thrombus on wall stress
Centre, 2001. in patient-specific models of abdominal aortic
23. Callanan A, Morris LG, McGloughlin TM. aneurysm. J Vasc Surg. 2002;36:598–604.
Numerical and experimental analysis of an 37. Schurink GW, van Baalen JM, Visser MJ, et al.
idealised abdominal aortic aneurysm. Present- Thrombus within an aortic aneurysm does not
ed at: European Society of Biomechanics. S- reduce pressure on the aneurysmal wall. J Vasc
Hertogenbosch, Netherlands; July 4–7, 2004. Surg. 2000;31:501–506.
24. Flora HS, Talie-Faz B, Ansdell L, et al. Aneurysm 38. Speelman L, Bohra A, Bosboom EM, et al.
wall stress and tendency to rupture are features Effects of wall calcifications in patient-specific
of physical wall properties: an experimental wall stress analyses of abdominal aortic aneu-
study. J Endovasc Ther. 2002;9:665–675. rysms. J Biomech Eng. 2007;129:105–109.
25. Doyle BJ, Morris LG, Callanan A, et al. 3D 39. Li Z-Y, U-King-Im J, Tang TY, et al. Impact of
reconstruction and manufacture of real ab- calcification and intraluminal thrombus on the
dominal aortic aneurysms: from CT scan to computed wall stresses of abdominal aortic
silicone model. J Biomech Eng. 2008;130: aneurysm. J Vasc Surg. 2008;47:928–935.
034501. 40. Doyle BJ, Corbett TJ, O’Donnell MR, et al.
26. O’Brien T, Morris L, O’Donnell M, et al. Injec- Design and development of a range of silicone
tion-moulded models of major and minor elastomers for use in experimental studies.
arteries: the variability of model wall thickness Proceedings of the Fifteenth Annual Meeting of
owing to casting technique. Proceedings of the the Section of Bioengineering of the Royal
International Mechanical Engineers, Part H: Academy of Medicine in Ireland. 2009;112.
Journal of Engineering in Medicine. 2005;219.
27. Mullins L. Softening of rubber by deformation.
Rubber Chem Tech. 1969;42:339–362. APPENDIX I
28. Brown R. Physical Testing of Rubber, 4th ed. Sample Preparation
New York: Springer Science+Business+Media,
Inc.; 2006:159–167.
1. Clean the molds thoroughly using ace-
29. Vorp DA, Raghavan ML, Webster MW. Me-
chanical wall stress in abdominal aortic aneu- tone.
rysm: influence of diameter and asymmetry. 2. Bolt the molds tightly together and seal
J Vasc Surg. 1998;27:632–639. the edges to prevent leaks.
30. Marins PA, Natal Jorge RM, Ferreira AJ. A 3. Weigh appropriate amounts of compo-
Comparative study of several material models nents, i.e., silicone and catalyst (10:1
for prediction of hyperelastic properties: appli- ratio), depending on the quantity of
cation to silicone-rubber and soft tissues. silicone material desired.
Strain. 2006;42:135–147. 4. Combine the components in a mixing
31. Shergold OA, Fleck NA, Radford D. The
container and thoroughly mix for 3 min-
uniaxial stress versus strain response of pig
utes.
skin and silicone rubber at low and high strain
rates. Int J Imp Eng. 2006;32:1384–1402. 5. Degas the mixture at a pressure of 100
32. Aziz T, Waters M, Jagger R. Analysis of the mBar for 15 minutes.
properties of silicone rubber maxillofacial 6. Remove from vacuum and suck mixture
prosthetic materials. J Dent. 2003;31:67–74. into a 60-mL syringe.

Doyle et al. 2009;16:000–000
7. Slowly inject the mixture into the mold, 5. On failure of sample, the machine resets
ensuring no leaks. to pre-determined setting.
8. Place the mold into vacuum to further 6. Software automatically generates UTS
degas for 15 minutes at a pressure of (MPa) and other user-defined variables.
350 mBar. 7. Record data such as UTS and repeat test
9. Remove and cure in oven for 18 hours at for the next sample.
45uC.
10. Upon curing, remove the samples from
the mold and leave at room temperature APPENDIX III
for 3 hours (BS ISO 37). Tear Testing
1. Set up the tensile machine apparatus as

APPENDIX II recommended by the manufacturer.
Tensile Testing 2. Measure the thickness of each sample
before testing.
1. Set up the tensile test machine apparatus 3. Place the specimen in clamps and en-
as recommended by the manufacturer. sure proper location, i.e., the specimen
2. Measure the thickness and width of each is clamped in the same position each
sample before testing. time.
3. Place the specimen in clamps and en- 4. Set up the software with the pre-deter-
sure proper location, i.e., the specimen mined loading and failure criteria.
is clamped in the same position each
time. a. Dimensions 5 Input thickness of
4. Set up the software with the pre-deter- sample (mm)
mined loading and failure criteria. b. Extension rate 5 100 mm/min (BS
ISO 34-1)
a. Dimensions 5 Input thickness and
width of sample (mm) 5. On failure of sample, the machine resets
b. Extension rate 5 500 mm/min (BS to pre-determined setting.
ISO 37) 6. Software automatically generates tear
c. Preconditioning 5 10 cycles at 20% strength (N/mm) and other user-defined
extension variables.
d. End criteria 5 Stretch sample to 7. Record data and repeat the test for the
failure next sample.

Ir J Med Sci
DOI 10.1007/s11845-009-0318-4
ORIGINAL ARTICLE
Improved assessment and treatment of abdominal aortic

aneurysms: the use of 3D reconstructions as a surgical guidance
tool in endovascular repair
B. J. Doyle Æ P. A. Grace Æ E. G. Kavanagh Æ
P. E. Burke Æ F. Wallis Æ M. T. Walsh Æ
T. M. McGloughlin
Received: 21 May 2008 / Accepted: 3 March 2009

Ó Royal Academy of Medicine in Ireland 2009
Abstract Wall stress was determined on one particular case using

Background Endovascular repair is fast becoming the finite element analysis.
treatment of choice for abdominal aortic aneurysms in Results 3D reconstruction allows measurements to be
anatomically suitable patients. 3D reconstructions not only obtained that can be difficult to determine using 2D ima-
aid conventional 2D measurements but also allow further ges. This method complements traditional 2D approaches.
analyses of the vessel anatomy. Reconstructions also provided imaging of potential ana-
Methods Computed tomography scan data for four male tomical problems. Wall stress results showed key regions
patients awaiting endovascular repair were obtained. 3D that may be possible rupture sites.
reconstructions were performed to determine measurements. Conclusion 3D reconstructions greatly aid surgical plan-
ning. As stent-graft devices evolve, anatomical difficulties
previously considered contraindications to endovascular
B. J. Doyle P. A. Grace E. G. Kavanagh repair can now be overcome with careful planning. 3D
P. E. Burke F. Wallis M. T. Walsh T. M. McGloughlin (&)
reconstruction is a useful adjunct to assessment and plan-
Centre for Applied Biomedical Engineering Research (CABER),
Department of Mechanical and Aeronautical Engineering, ning of endovascular repair.
Materials and Surface Science Institute, University of Limerick,
Limerick, Ireland Keywords Abdominal aortic aneurysm
e-mail: tim.mcgloughlin@ul.ie
3D reconstruction Endovascular repair Surgical planning
B. J. Doyle
e-mail: barry.doyle@ul.ie
P. A. Grace Introduction
e-mail: pagrace@eircom.net
E. G. Kavanagh An abdominal aortic aneurysm (AAA) can be defined as a
e-mail: eamon.kavanagh@hse.ie
permanent and irreversible localised dilatation of the aorta
P. E. Burke [1]. Aneurysms form due to alterations of the connective
e-mail: peburke1@gmail.com
tissue in the aortic wall. This degradation of the aortic
F. Wallis wall is attributed to risk factors such as tobacco smoking,
e-mail: fintan.wallis@hse.ie
sex, age, hypertension, chronic obstructive pulmonary
M. T. Walsh disease, hyperlipidaemia and family history of the disor-
e-mail: michael.walsh@ul.ie
der [1]. With recent advancements in non-invasive
P. A. Grace E. G. Kavanagh P. E. Burke diagnostic imaging more AAAs are being detected.
Department of Vascular Surgery, Midwestern Regional Hospital, Approximately 150,000 new cases are diagnosed each
Limerick, Ireland year in the US [2–4], and 15,000 deaths per year are
attributed to AAA rupture [5]. The normal diameter of the
F. Wallis
Department of Radiology, Midwestern Regional Hospital, abdominal aorta can vary with increasing age, male sex
Limerick, Ireland and bodyweight [1, 2], and decreases progressively from
123
its entry into the abdominal cavity to the iliac bifurcation. aneurysmal disease. Also, included in this study is the
The infrarenal abdominal aortic diameter ranges from 15 methodology and results of a finite element analysis (FEA)
to 24 mm for most elderly men [1] and it was proposed of a particularly extreme case of aneurysmal disease. FEA
[6] that an AAA is an aorta with an infrarenal diameter is a commonly used engineering technique to determine
greater than 30 mm. As the diameter of the infrarenal stress distributions experienced in complex structures and
aorta is dependent on age and sex [7], a size of at least is used widely in biomedical engineering applications. The
1.5 times greater than the expected normal diameter of use of FEA in assessing the propensity of an aneurysm to
the infrarenal aorta [8] has been suggested as an appro- rupture is becoming more widespread amongst researchers
priate definition of an aneurysm. [11, 12, 19–22], and helps provide a useful insight into the
Currently, the indication for surgical intervention is biomechanical behaviour of the aorta. Ultimately, these
based on the maximum diameter of the AAA in the reconstructions provide the clinician with valuable data
context of the patient’s general health status with most regarding dimensions, geometry and possible problem
repairs performed when the aneurysm exceeds 50– areas that may arise during endovascular surgical repair.
60 mm [9–12]. The aim of AAA treatment is to totally
exclude the aneurysm, isolating the aneurysm sac from
systemic blood pressure and flow so as to minimise the Methods
risk of rupture. As a result of AAA treatment, most
aneurysms should stabilise or shrink [13], although, it 3D reconstructions
has been reported that aneurysms can often enlarge after
AAA treatment due to endotension [14]. Surgical repair Computed tomography (CT) scan data were obtained from
can be either traditional open repair or endovascular the Midwestern Regional Hospital, Limerick, and St.
repair (EVAR). With traditional surgical repair, the James’s Hospital, Dublin, for four patients either awaiting
abdomen is entered via a long midline or a wide trans- or had undergone surgical repair of an AAA. All patients
verse incision. Clamping of the infrarenal, or rarely the were male with a mean age of 77.5 years (range 59–
suprarenal, aorta is necessary whilst suturing either a 87 years). CT scans were obtained using a Somotom Plus 4
Dacron or expanded polytetrafluoroethylene (ePTFE) (Siemens AG, D-91052 Erlangen, Germany). The mean
graft to the neck of the aorta proximal to the aneurysm pixel size of the CT scans was 0.675 mm with all scans
and either the distal aorta (tube graft) or to both iliac taken using a 3-mm slice increment. The CT scans were
arteries (bifurcated graft) distal to the aneurysm which is imported into the commercially available software Mimics
effectively removed from the circulation. Endovascular v12 (Materialise, Belguim) for reconstruction. The full
repair consists of the placement of a stent-graft via an reconstruction technique has been reported previously by
intraluminal introducer across the aneurysm and its fix- our group [23]. Briefly, a thresholding technique is applied
ation to the normal aortic and iliac walls with self- to each scan in the series. This process assigns a pixel
expanding stents at both ends [1]. This treatment intensity value measured in Hounsfield units (HU) to each
involves the surgical exposure of the common femoral pixel in the image. From this, the HU value can be con-
arteries where the endovascular graft can be inserted by trolled so that only the regions of interest are thresholded.
an over-the-wire technique. The early benefit of the Following this thresholding, segmentation of the image is
minimally invasive EVAR approach has been proven in possible. This assigns a certain colour to a certain region of
several randomised trials in reducing the perioperative interest, in this case the diseased aorta. By applying this
mortality compared with open repair [14–17]. It is algorithm to the series of CT scans, a complete 3D
thought that this benefit is a result of the avoidance of reconstruction can be generated. This process can be seen
the large abdominal incision and consequent systemic in Fig. 1.
inflammatory response, respiratory and wound compli- Assigning regions into different layers allows certain
cations as well as the avoidance of aortic clamping with areas to be highlighted or removed depending on the
resultant cardiovascular complications. desired region of interest. An example of this layered
The conversion of computed tomography (CT) scans approach to reconstruction can be seen in Fig. 2. In this
into 3D reconstructions of patient-specific regions is of particular reconstruction, all major internal structures were
great use to surgeons and is becoming widely used [18]. generated for illustration purposes.
For many years, regions of the human body have been For these particular reconstructions, only the lumen
reconstructed in order to give clinicians a further insight regions were of clinical importance as this is the region to
into the particular problem region or condition. The focus which the stent-graft is deployed and fixated. For the pur-
of this paper is on the use of 3D reconstruction of poses of FEA, the full aneurysmal sac of the extremely
the human abdominal aorta in patients suffering from diseased aorta was reconstructed.
123
Fig. 1 a Typical CT scan
showing aneurysmal aorta in
centre of image, b algorithm
defines regions of interest and c
software generates 3D
reconstruction of desired
regions
Fig. 2 Full 3D reconstruction

of CT scan data set. Image
shows the use of layers to
highlight/remove particular
regions of interest depending on
desired areas
Finite element analysis (FEA) constrained from movement at the entrance to the aortic
arch and the iliac arteries to represent tethering to the
The process of determining wall stresses experienced remainder of the ascending aorta and the femoral arteries.
within the aortic wall from CT scan data sets has been The aortic wall was assumed to have a uniform thickness of
previously reported by this group [21, 22, 24]. The FEA 1.9 mm [10, 20, 25], and was modelled as a non-linear
technique works by subdividing the structure into smaller hyperelastic material [10]. These boundary conditions have
regions called elements with each element connected via been implemented in many previous studies [10, 19, 21, 22,
nodes. The resulting subdivision is known as a mesh. The 24, 26–28].
stress at each node is then mathematically computed using
the appropriate software, in this case ABAQUS v6.7
(Dassault Systemes, SIMULIA, RI, USA). Realistic Results
boundary conditions were then applied to the aorta. In this
study, FEA was used to determine the resulting wall stress 3D reconstructions
distributions for the extremely diseased case shown in
Fig. 3. The model was meshed using shell elements. The Four patients awaiting surgical repair were reconstructed
intraluminal thrombus (ILT) was omitted from the analysis. into virtual 3D models for examination. Reconstruction
The ILT has been shown to reduce wall stress by acting times can range from as little as 2 min, for a basic model
like a mechanical buffer between the internal blood pres- where minor details are ignored, to 1 h where all details are
sure and the aortic wall [24]. An internal peak systolic included such as ILT and surface indentations. The critical
blood pressure of 120 mmHg (16 KPa) was applied to the dimensions when concerned with EVAR stent-grafts are
internal region of the aorta, with the reconstruction the neck diameter of the stent-graft, the iliac diameter and
123
Fig. 3 Meshed reconstruction
for use in FEA. Case shown is
that of an extremely diseased
aorta with TAA, AAA and also
IAA. Case shown here is from
the posterior viewpoint
Fig. 4 Example reconstruction showing measurements used to size

stent-graft. All measurements are in millimetres (mm). Total length of
stent-graft was determined to be 159 mm
also the length of the device. Clinicians can usually

determine the neck diameter without difficulty using 2D
CT scans, but difficulties can arise when determining
length. 3D reconstruction allows the overall length to be
easily measured, thus allowing the exact sizing of the
medical device. An example of this AAA length mea-
surement can be seen in Fig. 4. The measurements are
taken from below the lowest renal artery to a point in the
iliac arteries where the clinician feels comfortable will
maintain an adequate fixation of the device.
Concerns arose about one particular patient when CT
scans revealed an extremely tortuous proximal neck that
may cause difficulties during the surgical procedure. 3D
reconstruction highlighted the degree of curvature of the
neck and allowed the clinician to decide on the most
suitable approach before EVAR. As shown in Fig. 5, the
proximal neck of the AAA required special attention
during the operation. As the upper portion of the stent-
graft is embedded into the proximal neck, the tortuous
nature of this could complicate placement and fixation of
the device.
The patient-specific reconstruction shown in Fig. 6 was
a particularly extreme case. In this reconstruction, the full Fig. 5 3D reconstruction revealed extremely tortuous proximal neck
which can be identified using 2D CT scans. The 3D reconstruction
extent of the aneurysmal aorta was revealed. This patient
however, showed exactly the degree of tortuosity, thus allowing the
had a descending thoracic abdominal aneurysm (TAA), clinician to decide on the most suitable means of approach during
AAA and also an iliac aortic aneurysm (IAA). EVAR
123
Fig. 6 3D reconstruction
showing thoracic aortic
aneurysm (TAA), abdominal
aortic aneurysm (AAA) and iliac
aortic aneurysm (IAA) from
various viewpoints
Finite element analysis (FEA)
The resulting wall stress distributions revealed that the

TAA of this case experienced higher stresses than that of
both the AAA and IAA. In this case, the peak wall stress
was 1.336 MPa and was located at an inflection point at the
proximal region of the TAA, as shown in Fig. 7. Inflection
points have been previously reported to act as stress raisers
on the surface of aneurysms [19, 21, 22, 24], with inflection
points defined as regions where the curvature changes from
concave to convex. The peak wall stress was computed to
be 1.336 MPa, and the failure strength of thoracic aortic
aneurysmal tissue has been reported to be approximately
1.19 MPa [25]. Since it is known that rupture will occur
when local wall stress exceeds local wall strength, these
findings provided further support for the clinical
intervention.
Discussion
Fig. 7 Resulting wall stress distribution for case examined using
The focus of this paper was to highlight the use of 3D FEA. Peak stress was located at the proximal inflection point of the
reconstruction as a surgical guidance tool for vascular TAA. High stress is shown in red in the figure. Lower stresses were
surgeons. Currently, many clinicians use 2D CT scans to observed in both the AAA and the IAA
determine the correct sizing of stent-grafts and also to
visualise the aneurysm prior to surgery. 3D reconstructions outcome of EVAR. This increases surgical confidence in
provide a useful addition to 2D measurements for stent- the operation, as the clinician will not encounter any
graft sizing, and therefore can possibly help to improve the unforeseen geometrical obstacles to the already difficult
123
procedure. Measurements obtained from 2D CT scans are and curvature, like the cases presented in Figs. 5 and 6.
often accurate, in particular when measuring diameter, as Enabling the clinician to visualise the AAA prior to surgery
the measurement must be taken in a plane perpendicular to increases confidence in the procedure as the degree of
the true lumen of the vessel. Many clinicians measure curvature of the neck or iliac arteries can readily be
stent-graft length from 2D images as the tortuous nature of determined. The case presented in Fig. 5 was repaired
the anerysmatic aorta can often be reduced with the using the AorifxTM stent-graft by Lombard due to the
introduction of the stiff guide wire. A combination of the tortuous nature of the anatomy. This stent-graft showed no
methods may provide a better approximation of the exact complications 1 year post-operation. This patient later died
length, thus allowing more precise stent-graft lengths to be due to unrelated causes. Extreme cases, such as that of
obtained. Measurements from the virtual 3D model, how- Fig. 6, highlight the advantages of 3D reconstruction as a
ever, offer further insight into the morphology of the surgical tool. From the reconstruction, precise measure-
diseased aorta. Iliac bifurcation angles can now be accu- ments can be determined, such as length, degree of
rately measured in order to determine the correct stent-graft curvature, angles and exact distances over surfaces, there-
for the particular application. Also, when fenestrated, stent- fore aiding in the choice of a patient-specific stent-graft
grafts are to be utilised, the distances between the mesen- from the range currently available. Cases such as that
teric arteries and the renal arteries can be accurately presented in Fig. 6 may often require tailor-made stent-
obtained. These exact measurements can be difficult to grafts, and the use of reconstructions aid this task. Tailor-
determine from 2D images. It is also common practice to made stent-grafts can be designed using the centrelines
use a graduated measuring catheter placed over the stiff generated from the reconstructions, and the resulting flow
guide wire prior to deployment of the device, in order to patterns can be identified [30]. Clinicians will usually opt
ensure the correct length stent-graft is to be introduced. to have a range of device sizes at hand during the operation
The case presented in Fig. 4 was repaired using the Tal- so that the correct device can be obtained in the event of
entTM stent-graft by Medtronic with no complications unforeseen circumstances. Full examinations using 3D
2 years post-operation. Reconstruction times can vary reconstructions can help to alleviate this worry.
depending on the complexity of the case. Most recon- 3D reconstructions also allow the clinical problem to be
structions can be performed under 1 h, which should be examined using various software applications. Diseased
adequate as the majority of CT scans are taken a consid- aortas can now be modelled using complex mathematical
erable time prior to the actual operation. Therefore, the software to determine stresses and strains experienced in
clinician can often afford to examine a 3D reconstruction the wall, and also the complex flow patterns of the blood
without sacrificing the health status of the patient. through the diseased artery allowing for stent-graft design
The reconstruction shown in Fig. 6 revealed a tortuous [30]. The results of the FEA studied here show that the
proximal neck and iliac aneurysm, both of which increase aorta is experiencing a peak stress of 1.356 MPa at the
the difficulty of EVAR. Fixation of the distal end of the proximal inflection point of the TAA, which exceeds the
device in the iliac arteries is hampered by the aneurysmal average failure strength of tissue in this area as reported
or tortuous artery, and therefore the device may dislodge previously [25]. FEA also identifies hotspots on the dis-
and ultimately fail. Calcified or tortuous iliac arteries may eased aorta that may indicate possible rupture sites in vivo.
also rupture upon insertion or withdrawal of the introducer. The reconstruction of the stent-graft after placement within
The problems associated with irregular iliac arteries were the TAA can be seen in Fig. 8. The device now excludes
highlighted in 2003 when Guidant were forced to withdraw the region of high stress and thus aids to return the aorta to
the AncureTM stent-graft from the market [29]. The a lower stress state that is relatively safe from rupture. This
diameter of the iliac arteries is of obvious importance when reconstruction was generated from post-operative CT scans
sizing stent-grafts, but so too is the tortuosity, as the that help to monitor the diseased aorta, and help to high-
introducer can place large stresses on the artery wall whilst light any additional aneurysm growth from endotension
being navigated through the vessel. Nowadays, clinicians [14], or further progression of the AAA. Post-operative 3D
are tackling more challenging anatomy with EVAR com- reconstructions can also aid the clinician to observe the
pared to recent years, and therefore the ability to view the outcome of the EVAR procedure at regular intervals. Stent-
exact morphology of the iliac arteries has clinical rele- graft limbs can often become kinked, twisted or occluded
vance, and may not always be possible with 2D CT scans. by thrombosis [17], which can complicate the flow of blood
Reconstruction also gives the clinician valuable infor- and may require further intervention by the clinician.
mation about the aneurysm before surgery. This ensures Although these complications can be identified using 2D
that the clinician does not encounter any unforeseen CT scans, 3D reconstructions allow exact visualisation of
problems upon surgical repair. This was particularly useful the problem and aid towards a better understanding of the
for cases where the aneurysm displays extreme tortuosity overall situation of the patient. Currently, there is no
123
Fig. 8 Post-operative EVAR is quick to perform and could not only help towards stan-
illustration showing deployed dardising the measurement and sizing of stent-grafts, but also
TAA stent-graft excluding TAA
sac. Stent-graft now returns aid surgeons in improving the treatment of AAAs.
blood flow in the descending
aorta to a relatively normal Acknowledgments The authors would like to thank (1) the Irish
state. Transparency of the aorta Research Council for Science, Engineering and Technology (IRC-
in the image allows easy SET) Grant RS/2005/340 (2) Grant #R01-HL-060670 from the US
visualisation of the device. National Heart Lung and Blood Institute (3) the Department of
Model shown from the posterior Vascular Surgery in the HSE Midwestern Regional Hospital, Ireland
viewpoint (4) the Department of Radiology in the HSE Midwestern Regional
Hospital, Ireland (5) Prof. David A. Vorp from the Departments of
Surgery and Bioengineering, Centre for Vascular Remodelling and
Regeneration, McGowan Institute of Regenerative Medicine, Uni-
versity of Pittsburgh, USA and (6) Mr. Prakash Madhavan, Consultant
Vascular Surgeon, St. James Hospital, Dublin, Ireland.
References
1. Sakalihasan N, Limet R, Defawe OD (2005) Abdominal aortic

aneurysm. Lancet 365:1577–1589. doi:10.1016/S0140-6736(05)
66459-8
2. Bengtsson H, Nilsson P, Bergqvist D (1998) Natural history of
abdominal aortic aneurysm detected by screening. Br J Surg
80:718–720. doi:10.1002/bjs.1800800613
3. Vorp DA (2007) Biomechanics of abdominal aortic aneurysm. J
Biomech 40:1887–1902. doi:10.1016/j.jbiomech.2006.09.003
4. Ouriel K, Green RM, Donayre C et al (1992) An evaluation of
new methods of expressing aortic aneurysm size: relationship to
rupture. J Vasc Surg 15:12–20. doi:10.1067/mva.1992.32982
5. Kleinstreuer C, Li Z (2006) Analysis and computer program for
rupture-risk prediction of abdominal aortic aneurysms. Biomed
institution-wide standard regarding stent-graft sizing with Eng Online 5:19. doi:10.1186/1475-925X-5-19
many clinicians opting for their own methods of mea- 6. McGregor JC, Pollock JG, Anton HC (1975) The value of
surements, particularly lengths. There is still a degree of ultrasonography in the diagnosis of abdominal aortic aneurysm.
Scott Med J 20:133–137
learning amongst the EVAR community when concerned 7. Grimshaw GM, Thompson JM (1997) Changes in diameter of the
with AAA morphology and stent-graft behaviour. The abdominal aorta with age: an epidemiological study. J Clin
latest generation of the Medtronic EndurantTM graft Ultrasound 25:7–13. doi:10.1002/(SICI)1097-0096(199701)25:
appears to shorten more in length on deployment than its 1\7::AID-JCU2[3.0.CO;2-M
8. Johnston KW, Rutherford RB, Tilson MD et al (1991) Suggested
predecessor, TalentTM, due to design modifications. This standards for reporting on arterial aneurysms. Subcommittee on
consequently affects the recommended measurements and reporting standards for arterial aneurysms, Ad hoc committee on
sizing by the manufacturer. Ultimately, better imaging reporting standards, Society for Vascular Surgery and North
and 3D reconstructions can only add to the quality of the American Chapter, International Society for Cardiovascular
Surgery. J Vasc Surg 13:452–458. doi:10.1067/mva.1991.26737
endovascular repair. 9. Sayers RD (2002) Aortic aneurysms, inflammatory pathways and
nitric oxide. Ann R Coll Surg Engl 84(4):239–246. doi:
10.1308/003588402320439667
Conclusion 10. Raghavan ML, Vorp DA (2000) Toward a biomechanical tool to
evaluate rupture potential of abdominal aortic aneurysm: identi-
fication of a finite strain constitutive model and evaluation of its
3D reconstruction of AAAs is a powerful and useful surgical applicability. J Vasc Surg 33:475–482
guidance tool. Reconstruction allows the clinician to obtain 11. Fillinger MF, Marra SP, Raghavan ML et al (2003) Prediction of
measurements useful for the sizing of stent-grafts, and also rupture risk in abdominal aortic aneurysm during observation:
wall stress versus diameter. J Vasc Surg 37:724–732. doi:
allows the clinician to visualise the aneurysm prior to sur- 10.1067/mva.2003.213
gery. Reconstructions are also necessary for further use with 12. Fillinger MF, Raghavan ML, Marra SP et al (2002) In vivo
FEA which has been shown to be a good method of deter- analysis of mechanical wall stress and abdominal aortic aneurysm
mining wall stress in the diseased aorta. FEA provides an rupture risk. J Vasc Surg 36:589–597. doi:10.1067/mva.2002.
125478
additional source of information to the clinician and helps 13. Morris LG (2004) Numerical and experimental investigation of
towards a greater understanding of the biomechanical mechanical factors in the treatment of abdominal aortic aneu-
behaviour of the anatomy prior to surgery. 3D reconstruction rysms. PhD Thesis 2004. University of Limerick
123
14. EVAR Trial Participants (2004) Comparison of endovascular 23. Doyle BJ, Morris LG, Callanan A et al (2008) 3D reconstruction
aneurysm repair with open repair in patients with abdominal and manufacture of real abdominal aortic aneurysms: from CT
aortic aneurysm (EVAR trial 1), 30-day operative mortality scan to silicone model. J Biomech Eng 130:034501–034505. doi:
results: randomised controlled trial. Lancet 364:843–848. doi: 10.1115/1.2907765
10.1016/S0140-6736(04)16979-1 24. Doyle BJ, Callanan A, McGloughlin TM (2007) A comparison of
15. EVAR trial participants (2005) Endovascular aneurysm repair modelling techniques for computing wall stress in abdominal
versus open repair in patients with abdominal aortic aneurysm aortic aneurysms. Biomed Eng Online 6:38. doi:10.1186/
(EVAR trial 1): randomised control trial. Lancet 365:2179–2186 1475-925X-6-38
16. Schermerhorn M, O’Malley J, Jhaveri A et al (2008) Endovas- 25. Vorp DA, Shiro BJ, Ehrlich MP et al (2003) Effect of aneurysm
cular vs. open repair of abdominal aortic aneurysms in the on the tensile strength and biomechanical behaviour of the
Medicare population. N Engl J Med 358:464–474. doi:10.1056/ ascending thoracic aorta. Ann Thorac Surg 75:1210–1214. doi:
NEJMoa0707348 10.1016/S0003-4975(02)04711-2
17. Corbet TJ, Callanan A, Morris LG et al (2008) In vivo and in 26. Thubrikar MJ, Al-Soudi J, Robicsek F (2001) Wall stress studies
vitro biomechanical behaviour and performance of post-operative of abdominal aortic aneurysm in a clinical model. Ann Vasc Surg
abdominal aortic aneurysms and implanted stent-grafts: a review. 15:355–366. doi:10.1007/s100160010080
J Endovasc Ther 15(4):468–484. doi:10.1583/08-2370.1 27. Truijers M, Pol JA, SchultzeKool LJ et al (2007) Wall stress
18. Resch TA. TeraRecon clinical case studies: endograft planning analysis in small asymptomatic, symptomatic and ruptured
for complex aortic aneurysms. Available via TeraRecon. abdominal aortic aneurysms. Eur J Vasc Endovasc Surg 33:401–
http://www.terarecon.com/downloads/news/clinical_case_studies_ 407. doi:10.1016/j.ejvs.2006.10.009
v2/casestudy_Resch-EndograftPlanning.pdf. Accessed 10 March 28. Speelman L, Bohra A, Bosboom EMH et al (2007) Effects of wall
2008 calcifications in patient-specific wall stress analyses of abdominal
19. Vorp DA, Raghavan ML, Webster MW (1998) Mechanical wall aortic aneurysms. J Biomech Eng 129:1–5. doi:10.1115/1.
stress in abdominal aortic aneurysm: influence of diameter and 2401189
asymmetry. J Vasc Surg 27(4):632–639. doi:10.1016/S0741- 29. Faries PL, Dayal R, Rhee J et al (2004) Stent graft treatment for
5214(98)70227-7 abdominal aortic aneurysm repair: recent developments in
20. Raghavan ML, Vorp DA, Federle MP et al (2000) Wall stress therapy. Diseases of the aorta, pulmonary, and peripheral vessels.
distribution on three-dimensionally reconstructed models of Curr Opin Cardiol 19(6):551–557. doi:10.1097/01.hco.
human abdominal aortic aneurysm. J Vasc Surg 31:760–769. doi: 0000141267.26503.ed
10.1067/mva.2000.103971 30. Molony DS, Callanan A, Morris LG et al (2008) Geometrical
21. Doyle BJ, Callanan A, Walsh MT et al (2009) A finite element enhancements for abdominal aortic stent grafts. J Endovasc Ther
analysis rupture index (FEARI) as an additional tool for abdom- 15:518–529
inal aortic aneurysm burst prediction. Vasc Dis Prev 6:75–82
22. Doyle BJ, Callanan A, Burke PE et al (2009) Vessel asymmetry
as an additional diagnostic tool for the assessment of abdominal
aortic aneurysms. J Vasc Surg 43:443–454
123
114 Vascular Disease Prevention, 2009, 6, 114-121
Open Access
A Finite Element Analysis Rupture Index (FEARI) as an Additional Tool
for Abdominal Aortic Aneurysm Rupture Prediction
Barry J. Doyle1, Anthony Callanan1, Michael T. Walsh1, Pierce A. Grace1,2 and

Timothy M. McGloughlin1,*
1
Centre for Applied Biomedical Engineering Research (CABER), Department of Mechanical and Aeronautical
Engineering, and Materials and Surface Science Institute, University of Limerick, Ireland; 2Department of Vascular
Surgery, HSE Midwestern Regional Hospital, Limerick, Ireland
Abstract: Currently, abdominal aortic aneurysms (AAAs), which are a permanent dilation of the aorta, are treated surgi-
cally when the maximum transverse diameter surpasses 5.5cm. AAA rupture occurs when the locally acting wall stress
exceeds the locally acting wall strength. There is a need to review the current diameter-based criterion, and so it may be
clinically useful to develop an additional tool to aid the surgical decision-making process. A Finite Element Analysis Rup-
ture Index (FEARI) was developed.
Ten patient-specific AAAs were reconstructed, and the corresponding wall stress computed. Previous experimental work
on determination of ultimate tensile strengths (UTS) from AAA tissue samples was implemented in this study. By com-
bining peak wall stress along with average regional UTS, a new approach to the estimation of patient-specific rupture risk
has been developed.
Ten cases were studied, all of which were awaiting or had previously undergone surgical AAA repair. A detailed exami-
nation of these ten cases utilising the FEARI analysis suggested that there was a possibility that some of the AAAs may
have been less prone to rupture than previously considered.
It is proposed that FEARI, used alongside other rupture risk factors, may improve the current surgical decision-making
process. The use of FEARI as an additional tool for rupture prediction may provide a useful adjunct to the diameter-based
approach in surgical decision-making.
Keywords: Aneurysm, rupture, prediction, stress, strength.
INTRODUCTION Currently, the rupture risk of AAAs is regarded as a con-

tinuous function of aneurysm size. Traditionally, surgeons
Cardiovascular disease is a leading cause of death in the
operate once the AAA exceeds a transverse diameter of
Western world. Aneurysms, which are dilations or widening
50mm [7-10], but previous work has shown that AAAs
of blood vessels, form a significant portion of these deaths.
smaller than 50mm can also rupture [11, 12]. It is believed
The healthy infrarenal abdominal aorta diameter ranges from by many researchers that there is a need to review the deter-
15mm to 24mm for most elderly men [1], with abdominal
mination of the timing of surgical intervention based solely
aortic aneurysms (AAAs) defined as an aorta with an in-
on aneurysm diameter, and consider the inclusion of other
frarenal diameter 1.5 times that of the normal aortic diameter
relevant risk factors.
[2]. The actual mechanisms resulting in AAA formation are
still not clearly understood. It is believed that these aneu- It is known that AAA rupture occurs when the locally
rysms form due to alterations of the connective tissue in the acting wall stress exceeds the locally acting wall strength.
aortic wall. This degradation of the aortic wall can be attrib- Therefore, the AAA tissue strength must play an equal role
uted to risk factors such as tobacco smoking, sex, age, hyper- to AAA wall stress in determining failure. A region of AAA
tension, chronic obstructive pulmonary disease, hyperlipi- wall that is under elevated wall stress may also have high
daemia, and family history of the disorder [1]. Increased wall strength, thus equalising its rupture potential. Many
screening of subjects and improved imaging technologies, researchers have proposed alternative methods of determin-
have given rise to increased detection of AAAs. Approxi- ing AAA rupture potential. These additional risk factors
mately 150,000 new cases are diagnosed each year in the could, for example, include: AAA wall stress [9, 10]; AAA
USA [3-5], resulting in 15,000 deaths per year due to AAA expansion rate [6, 13, 14]; degree of asymmetry [15-17];
rupture [6]. presence of intraluminal thrombus (ILT) [18] which is a fi-
brin structure incorporated with blood cells, platelets, blood
proteins, and cellular debris, and are found in most AAAs; a
*Address correspondence to this author at the MSG-014, MSSI Building,
University of Limerick, Limerick, Ireland; Tel: +353 61 202369; Fax: +353
rupture potential index (RPI) [19, 20]; biomechanical factors
61 202944; E-mail: tim.mcgloughlin@ul.ie and computer analysis [6]; growth of thrombus [21]; geomet-
1567-2700/09 2009 Bentham Open

A Finite Element Analysis Rupture Index (FEARI) Vascular Disease Prevention, 2009, Volume 6 115
rical parameters [22]; and also a method of determining basis for the FEM stress analysis. The pre-operative details
AAA growth and rupture based on a mathematical model of the ten cases examined can be seen in Table 1.
[23]. 0.2892
Diamax Eqn.1
The purpose of this study is to examine the use of a new t = 3.9
additional tool to assist in the assessment of AAA rupture 2
risk. This new approach focuses on a combination of the Wall Stress Analysis
finite element method (FEM) coupled with published ulti-
mate tensile strength (UTS) data from AAA tissue reported The ten 3D reconstructions were imported into the com-
by previous researchers [13, 24, 25]. The FEM is a numerical mercial finite element solver ABAQUS v6.7 (Dassault Sys-
technique widely used in engineering that divides complex temes, SIMULIA, R.I., USA) for stress analysis. In order to
3D structures into small areas called elements, for which the simulate in vivo wall stress in the AAA wall, realistic bound-
stress distribution can be more easily studied. This new ap- ary conditions were applied to each model. The AAA wall
proach which we have described as the Finite Element was modelled as a homogenous isotropic hyperelastic mate-
Analysis Rupture Index (FEARI), may be clinically useful in rial using the finite strain constitutive model proposed by
aiding surgeons as to the most appropriate time to surgically Raghavan and Vorp [8]. These material properties have been
intervene, and may serve as a useful adjunct to maximum utilised in many previous stress analysis studies [9, 10, 17,
diameter. 18, 29, 30-35]. The aorta is also known to be nearly incom-
pressible with a Poisson’s ratio of 0.49. The ILT was mod-
MATERIALS AND METHODS elled as a hyperelastic material using the material characteri-
Study Subjects and Reconstructions sation derived from 50 ILT specimens from 14 patients per-
formed by Wang et al. [36]. Each AAA was constrained in
Computed tomography (CT) scan data was obtained for the proximal and distal regions to simulate tethering to the
10 patients (male, n=6; female, n=4). These patient scans aorta at the renal and iliac bifurcations. The blood pressure
were obtained from the Midwestern Regional Hospital, Lim- within the AAA acts on the luminal contour of the ILT and
erick, Ireland, and the University of Pittsburgh Medical Cen- therefore, pressure was applied to the inner surface of the
tre, Pittsburgh, PA, USA. All ten patients were awaiting or computational AAA model. A static peak systolic pressure of
had received AAA repair at the time of CT scan, as AAA 120mmHg (16KPa) was used, as employed in most AAA
diameters had reached or exceeded the current 5.0cm thresh- stress analyses [15, 22, 33, 37, 38]. It is known that patient-
old for repair. 3D reconstructions were performed using the specific blood pressures may be higher than 120mmHg, but
commercial software Mimics v10.0 (Materialise, Belgium). for the purpose of this study a standard value was more ap-
This reconstruction technique was validated with previous propriate so as to eliminate some of the unknown variables
work performed by our group [26, 27]. The iliac arteries in the analysis. The shear stress induced by blood flow was
were omitted from the reconstruction as they have been neglected in this study, although the effects of blood flow
shown not to significantly alter the resulting wall stress dis- have been shown to reduce wall stress in idealised AAA
tributions [10]. Patient-specific wall thickness was obtained models [16]. Residual stresses that may exist within the aor-
using the equation proposed by Li and Kleinstreuer [28] tic wall in vivo and tethering forces on the posterior surface
shown in Eqn. 1, where t is wall thickness and Diamax is the caused by the lumbar arteries were also neglected.
maximum diameter. The ILT was also included in the recon-
structions as this structure has been shown to reduce wall Once adequate boundary conditions were applied to each
stress [18]. The models developed by this method formed the model, a mesh was generated on each AAA. Mesh independ-
Table 1. Pre-Operative Patient Details for Each Study Subject. Note that Ø is Diameter
Max Ø Wall Thickness AAA Length AAA Volume ILT Volume

Patient Sex
(cm) (mm) (cm) (cm3) (cm3)
1 Male 5.1 1.53 13.2 176.7 54.1
2 Male 5.75 1.48 11.2 192.4 122.8
3 Male 5.85 1.47 11.6 194.9 96.8
4 Female 5 1.54 9.3 136.9 38.4
5 Male 5.9 1.47 12.8 220.9 116.8
6 Male 7.4 1.37 14.3 311.6 147.6
7 Female 5.3 1.51 10.5 137.9 82.5
8 Female 6.2 1.48 9.7 61.6 29.1
9 Female 6.5 1.43 10.5 217.3 150.2
10 Male 9 1.29 11.7 445.5 124.5

116 Vascular Disease Prevention, 2009, Volume 6 Doyle et al.
ence was performed for all AAA models in order to deter- Table 2. Regional UTS Values Obtained by Combining and
mine the optimum number of elements, and therefore, the Averaging Previous Experimental Data [13, 24, 25]
optimum mesh. Mesh independence was performed by in-
creasing the number of elements in the mesh until the differ-
AAA Region UTS (MPa)
ence in peak stress was less than 2% of the previous mesh
[17, 18, 34, 38]. This method of determining wall stress dis-
Anterior 0.7744
tributions has been previously shown to be the most effective
in computing accurate results compared with other ap- Posterior 0.8658
proaches [34].
Left 0.9221
FEARI Right 0.9187
The Finite Element Analysis Rupture Index (FEARI) is Anterior/Left 0.8482
defined by Eqn. 2. In this equation, the peak wall stress is
computed using the FEM, whereas, the wall strength values Anterior/Right 0.8465
are obtained from previous research on experimental testing Posterior/Left 0.8939
of AAA wall specimens [13, 24, 25].
Posterior/Right 0.8922
FEA Wall Stress Eqn. 2
FEARI =
RESULTS
This equation is based on the simple engineering defini-
tion of material failure, that is, failure will occur when the Wall Stress Results
stress acting on the material exceeds the strength of the ma- The finite element analyses using ABAQUS v6.7 pro-
terial. This index then returns a value ranging from 0 to 1, duced detailed stress distributions on each of the models
where 0 indicates a very low rupture potential, and a value under the pressure loading [19]. The von Mises stress is a
close to 1 indicates a very high rupture potential. stress index especially suited for failure analysis, as stress is
In order to determine strength values for the AAA wall, a tensor quantity with nine components, with the von Mises
the previous research of both Raghavan et al. [24, 25] and stress being a combination of these components. By observ-
Thubrikar et al. [13] were analysed. These publications are ing the stress distributions, it was noted that the regions of
the most detailed reports of experimental uniaxial testing of elevated and peak wall stresses occurred at inflection points
AAA tissue. Raghavan et al. [24] tensile tested 52 specimens on the AAA surface, and not at regions of maximum diame-
of AAA tissue and found that the average UTS of this aneu- ter. This observation was also observed by previous re-
rysmal tissue was 0.942 MPa. Thubrikar et al. [13] later searchers in idealised models, both experimentally [39] and
segmented AAAs into posterior, anterior and lateral regions numerically [15, 40] and also in realistic models [17, 34].
and tensile tested 49 tissue specimens. Average regional Inflection points are defined as points on the AAA surface at
UTS values were shown to be 0.46 MPa, 0.45 MPa and 0.62 which the local AAA wall shape changes from concave out-
MPa, respectively. Raghavan et al. [25] subsequently fur- ward to concave inward [15]. The peak wall stresses found in
thered their work, and also divided each AAA into regional this study ranged from 0.3167 – 1.282 MPa, with a mean ±
sections and tested 48 samples. They showed that the UTS of standard deviation of 0.6201 ± 0.2836 MPa. The results of
AAA tissue can range from 0.336 – 2.35 MPa. They also the computational stress analysis, along with the maximum
reported that the regional variations in UTS for the anterior, diameter and UTS for each case can be seen in Table 3.
posterior, left and right regions, were 1.099 MPa, 1.272 There was no statistical significance between peak wall
MPa, 1.217 MPa and 1.224 MPa, respectively. stress and any geometrical parameters analysed here. There
was however a significant relationship between both FEARI
By combining all the previously published experimental and maximum diameter (P=0.043), and FEARI and AAA
data [13, 24, 25], average regional UTS values for the four volume (P=0.036).
main regions of the AAA could be obtained. By subdividing
the AAA into a further four sections, eight in total, more FEARI Results
regionally accurate strength estimates were obtained, thus Once all peak wall stresses, locations of peak wall stress,
allowing FEARI values to be calculated. The method of di- and corresponding regional UTS values were obtained and
viding each AAA into regions is shown in Fig. (1), with the grouped, FEARI results using Eqn. 2 could be calculated.
resulting UTS values for the varying regions shown in Table Table 4 and Fig. (2) show the resulting FEARI values with
2. the corresponding maximum diameters of each patient. In
Once the combined UTS values of AAA tissue were cal- this figure, an indication of the possible rupture risk using
culated, these values could be coupled with the computed the FEARI analysis of each patient is shown. An example
wall stress results from the FEM. By recording the location calculation of this index for Patient 9 can be seen below.
of peak stress in each AAA model, the area can also be as- FEA Wall Stress Eqn. 2
signed a regional UTS value. FEARI was then computed for FEARI =
each of the ten cases. Statistical analysis was performed on Experimental Wall Strength
all results using a Pearson’s correlation, with P<0.05 ac- Where, peak FEA wall stress is 0.5263 MPa, and this
cepted as significant. peak stress occurred in the anterior region of the AAA.
0.526 MPa
A
0.004 MPa
A/R A/L
A: Anterior
P: Posterior
L: Left
R: Right
R L
A/L: Anterior/Left
A/R: Anterior/Right
P/L: Posterior/Left
P/R: Posterior/Right
P/R P/L
Fig. (1). Illustration showing a representative AAA stress distribution for Patient 9 (ellipse passes through region of peak stress, indicated by
black circle) and how each AAA model was segmented in order to determine the corresponding UTS value for the particular region of peak
stress. In this case, peak stress occurs on the anterior wall of the AAA. AAA model shown in the anterior view.
Table 3. Maximum Diameter, FEA Computed Peak Wall Stress, Location of Peak Wall Stress, and UTS of Peak Stress Region in
All Ten Cases Examined. Wall Thickness was Incorporated in Peak Wall Stress Calculations
Patient Max Diameter (cm) Peak Wall Stress (MPa) Location UTS (MPa)
1 5.1 0.4291 Anterior 0.7744
2 5.75 0.3167 Left 0.9221
3 5.85 0.4346 Anterior/Right 0.8465
4 5.0 0.6641 Anterior/Right 0.8465
5 5.9 0.5866 Anterior/Right 0.8465
6 7.4 0.707 Right 0.9187
7 5.3 0.4 Anterior/Right 0.8465
8 6.2 1.282 Anterior 0.7744
9 6.5 0.5263 Anterior 0.7744
10 9.0 0.855 Posterior 0.8658
Therefore, the UTS of the peak stress region is 0.7744 MPa may be safe from rupture, and should be rigorously moni-
using Table 2. Eqn. 2 now becomes Eqn. 3. tored [1]. The results presented here suggest that the inclu-
sion of the FEARI monitoring approach to AAA manage-
0.5263 Eqn. 3
FEARI = = 0.6796 ment may be useful in the assessment of all size AAAs,
0.7744 rather than repairing the aneurysm based on maximum di-
The resulting FEARI for Patient 9 is 0.6796, representing ameter alone.
a possible 68% chance of rupture. It is known that AAA rupture occurs when the locally
DISCUSSION acting wall stress exceeds the locally acting wall strength,
and thus the inclusion of wall strength as a rupture parameter
Considering the high level of mortality associated with may also be useful. The FEARI approach presented here
AAA rupture, improved surgical decision tools may lead to utilises previously published experimental data from tissue
better clinical outcomes. At present, the timing of interven- specimens [13, 24, 25] which were applied to establish
tion is determined on the basis of maximum diameter alone. FEARI results. The wall stress was computed using the
However, it was reported that AAAs with diameters <5.0cm commercial FEA solver, ABAQUS v6.7, from which, peak
Table 4. Resulting FEARI Values Compared to the Corre- hand, would indicate that the risk of rupture is low, and
sponding Maximum Diameters for each AAA should be monitored closely through regular ultrasound or
Studied CT scanning. All ten cases examined had reached or ex-
ceeded the current 5.0cm threshold for surgical repair. At the
Maximum Diameter time of this study, all cases were awaiting or had undergone
Patient FEARI traditional open repair or endovascular aneurysm repair
(cm)
(EVAR). The peak stresses show that smaller AAAs can
1 5.1 0.5541 have higher peak wall stresses than larger AAAs (see Table
3). Wall stress is closely related to geometrical parameters,
2 5.75 0.3435 with highly stressed areas occurring at regions of inflection
3 5.85 0.5133
on the surface of the AAA. Tortuous and irregularly shaped,
or asymmetric, AAAs can experience many regions of in-
4 5.0 0.7845 flection, and thus, many regions acting as stress raisers on
the surface.
5 5.9 0.6929
Although FEARI results presented here are preliminary,
6 7.4 0.7696 in that they currently use experimental data from other re-
7 5.3 0.4725 searchers [13, 24, 25], the approach may be clinically useful.
In order to gauge the effect of the UTS on the resulting
8 6.2 1.6554 FEARI, the “worst case scenario” in terms of UTS was ex-
amined. Raghavan et al. [25] reported that the UTS of AAA
9 6.5 0.6796
tissue can range from 0.336 – 2.35 MPa, and therefore, this
10 9.0 0.9876 minimum strength of 0.336 MPa was applied to our FEARI
equation. The results can be seen in Fig. (3). This resulted in
wall stress values and the location of peak stress could be much higher rupture potentials for all ten cases, and sug-
recorded for each case. Convergence studies were also per- gested that all patients were suitable for immediate repair.
formed on each model to establish confidence in the finite Although this minimum UTS value was recorded, only 8 of
element mesh size and accuracy of the results [17, 18, 34, the 48 specimens tested by Raghavan et al. [25] resulted in
38]. Patient-specific wall thickness values were applied to UTS values below 0.5 MPa, with 0.336 MPa being the abso-
each case using the equation proposed by Li and Kleinstreuer lute minimum displayed by any specimen. Therefore, to as-
[28] to provide more patient-specific results. Peak stresses sume that all AAAs, in all regions experience this low failure
varied throughout the ten cases, and were independent of the stress would appear to be overly conservative.
maximum diameter of the AAA. There was no statistical It is proposed that FEARI could serve as a useful adjunct
significance between peak wall stress and any of the meas- to diameter-based surgical decision making. Diameter, and
ured parameters (maximum diameter, P=0.227; sex, ultimately size, of the AAA is an obvious concern for the
P=0.404; AAA volume, P=0.936; AAA length, P=0.501; clinician, and must remain a consideration. The overall ge-
ILT volume, P=0.247; wall thickness, P=0.375; location of ometry of the aneurysm should also play a role. Asymmetry
peak stress, P=0.54). FEARI results were also statistically has been shown to affect wall stress in idealised AAA mod-
compared with the same parameters. Of these comparisons, els, both numerically [15, 16, 40] and experimentally[39],
FEARI was statistically significant with maximum diameter and may also affect realistic cases [17]. Other researchers
(P=0.043) and AAA volume (P=0.036), whereas, there was have proposed the Rupture Potential Index (RPI) [20] which
no significance with other parameters (sex, P=0.961; AAA uses a statistical modelling technique with the inclusion of
length, P=0.804; ILT volume, P=0.881; wall thickness, factors such as age, gender, family history, smoking status,
P=0.056; location of peak stress, P=0.755). Significance among others, to deduce patient-specific wall strength [41].
was obtained when P<0.05. These results alone suggest that This RPI approach uses a more theoretical method of calcu-
other rupture indicating parameters should be included in the lating wall strength, compared to the experimental approach
surgical decision-making process. From the ten cases exam- of FEARI, and combining the two approaches may lead to
ined, all AAAs experienced peak wall stress at areas where improved predictions. Ultimately, the decision to surgically
the ILT was regionally thinner, therefore supporting the hy- intervene may include a combination of factors including
pothesis that ILT reduces wall stress by acting like a “me- diameter, asymmetry, RPI and FEARI, along with clinical
chanical cushion” [18, 19, 38] for the AAA wall. Another experience, and may determine the most suitable approach to
interesting observation relates to Patient 8. In this case, the a particular AAA.
AAA geometry had a rapid change in diameter from 33.4mm The use of fluid-structure interaction (FSI) software, as
to 61.4mm over 12.5mm. This sudden change in geometry opposed to the current static solid-mechanics approach may
resulted in a large peak stress of 1.282 MPa, which is ap- reduce peak stresses [16]. The effect of varying blood pres-
proximately 60% higher than the similarly sized-diameter sure may also alter results. Blood pressure can change over
AAA of Patient 9. time, often as a result of stress or exercise. Increasing or de-
A FEARI value was determined for each case from Eqn. creasing the applied pressure to the stress analyses would
2. A FEARI value close to or above 1, suggests that the increase or decrease the FEARI results respectively. Moni-
AAA may be a high rupture-risk AAA, and should be re- toring a patients’ blood pressure over the course of a day
paired immediately. FEARI values closer to 0, on the other may allow a more accurate loading condition to be devel-
oped, and thus provide more accurate patient-specific FEARI
1.8
6.2cm
1.6
1.4
1.2
9.0cm
FEARI
1.0
5.0cm 7.4cm
0.8 5.9cm 6.5cm
0.6 5.1cm 5.9cm 5.3cm
0.4 5.8cm
0.2
0
1 2 3 4 5 6 7 8 9 10
Patient
Fig. (2). Graph displaying FEARI results for all ten cases. Horizontal line indicates possible AAA rupture based on the FEARI model. Di-
ameters of each case are presented also, indicating how diameter is not related to FEARI.
4.5
4.0
3.5
3.0
FEARI
2.5
2.0
1.5
1.0
0.5
0
1 2 3 4 5 6 7 8 9 10
Patient
Fig. (3). FEARI results for the “worst case scenario” of extremely low failure strengths of the AAA tissue. Horizontal line indicates possible
AAA rupture using the FEARI model.
results. It is known that the realistic AAA has a non-uniform strength, which could then be averaged over the population.
wall thickness [25] and also non-uniform material properties Biaxial tensile testing of tissue is also necessary. It has been
due to regions of calcifications [15] which can lead to altera- reported that AAA tissue is anisotropic, and that 3D multi-
tions in stress distributions [33]. The refinement of the tech- axial mechanical evaluation would allow for more appropri-
nique used to compute wall stress may therefore produce ate modelling of aneurysmal tissue [42]. There have been
more realistic FEARI results even when considering the recent reports on the use of anisotropic material data in wall
“worst case scenario” regarding tissue strength. Along with stress analyses which may also yield better results [43]. By
these limitations in computing the peak wall stress, further actually measuring AAA tissue harvested from unruptured
experimental testing of AAA tissue is necessary. Both AAAs during surgery, tensile strength data can be gathered.
Raghavan et al. [24, 25] and Thubrikar et al. [13] recorded By expanding the database to many institutions, larger popu-
varying UTS values for the differing regions of the AAA, lations can be covered, thus increasing the effectiveness of
and therefore, a more widespread study on the UTS of AAA the AAA strength database. This may lead to a more accurate
tissue is required. Collation of data from other centres FEARI measure in order to improve the assessment of pa-
worldwide, would lead to a much larger database of tissue tients with AAAs.
CONCLUSIONS [12] Darling RC, Messina CR, Brewster DC, et al. Autopsy study of
unoperated abdominal aortic aneurysms. The case for early resec-
FEARI may be clinically useful due to the simplicity of tion. Circulation 1977; 56(3 Suppl): 11161-4.
the approach. Rupture occurs when the AAA tissue cannot [13] Thubrikar MJ, Labrosse M, Robicsek F, et al. Mechanical proper-
ties of abdominal aortic aneurysm wall. J Med Eng Tech 2001; 25:
withstand the locally acting wall stress exerted, and there- 133-42.
fore, tissue strength must be considered when assessing [14] Hirose Y, Takamiya M. Growth curve of ruptured aortic aneurysm.
AAA rupture potential. Peak wall stresses were computed, J Cardiovasc Surg 1998; 39: 9-13.
along with the location, and therefore UTS, of peak wall [15] Vorp DA, Raghavan ML, Webster MW. Mechanical wall stress in
stress region. FEARI results indicate that surgery may not be abdominal aortic aneurysm: influence of diameter and asymmetry.
J Vasc Surg 1998; 27: 632-9.
necessary for all cases, but rather continued monitoring may [16] Scotti CM, Shkolnik AD, Muluk SC, et al. Fluid-structure interac-
suit particular patients. It is proposed to couple FEARI to- tion in abdominal aortic aneurysms: effect of asymmetry and wall
gether with diameter and other important factors in AAA thickness. Biomed Eng Online 2005; 4: 64.
assessment to allow the clinician a greater understanding of [17] Doyle BJ, Callanan A, Burke PE, et al. Vessel asymmetry as an
additional diagnostic tool in the assessment of abdominal aortic an-
the severity of an individual AAA before deciding on surgi- eurysms. J Vasc Surg 2009; 49: 443-54.
cal intervention. This preliminary study of a FEARI suggests [18] Wang DHJ, Makaroun MS, Webster MW, et al. Effect of intralu-
that further work into this approach may yield more accurate minal thrombus on wall stress in patient-specific models of ab-
results, and may provide a useful adjunct to the diameter- dominal aortic aneurysm. J Vasc Surg 2002; 36: 598-604.
based approach in surgical decision-making. [19] Vorp DA, Vande Geest J. Biomechanical determinants of abdomi-
nal aortic aneurysm rupture. Artherioscler Thromb Vasc Biol 2005;
25: 1558-66.
ACKNOWLEDGEMENTS [20] Vande Geest JP, Di Martino ES, Bohra A, et al. A biomechanics-
based rupture potential index for abdominal aortic aneurysm risk
The authors would like to thank (i) the Irish Research assessment. Ann N Y Acad Sci 2006; 1085: 11-21.
Council for Science, Engineering and Technology (IRCSET) [21] Stenbaek J, Kalin B, Swedenborg J. Growth of thrombus may be a
Grant RS/2005/340 (ii) Grant #R01-HL-060670 from the US better predictor of rupture than diameter in patients with abdominal
aortic aneurysms. Eur J Vasc Endovasc Surg 2000; 20: 466-9.
National Heart Lung and Blood Institute (iii) the Department [22] Giannoglu G, Giannakoulas G, Soulis J, et al. Predicting the risk of
of Vascular Surgery in the Midwestern Regional Hospital, rupture of abdominal aortic aneurysms by utilizing various geomet-
Ireland, in particular, Mr. Eamon Kavanagh, Mr. Paul Burke rical parameters: revisiting the diameter criterion. Angiology 2006;
and Prof. Pierce Grace (iv) Aidan Cloonan for his assistance 57: 487-94.
with the numerical analyses (iv) Prof. David A. Vorp from [23] Volokh KY, Vorp DA. A model of growth and rupture of abdomi-
nal aortic aneurysm. J Biomech 2008; 41: 1015-21.
the Centre for Vascular Remodelling and Regeneration, Uni- [24] Raghavan ML, Webster MW, Vorp DA. Ex vivo biomechanical
versity of Pittsburgh and (v) Michel S. Makaroun, MD, De- behaviour of abdominal aortic aneurysm: assessment using a new
partment of Surgery, University of Pittsburgh. mathematical model. Ann Biomed Eng 1996; 24: 573-82.
[25] Raghavan ML, Kratzberg J, de Tolosa EMC, et al. Regional distri-
bution of wall thickness and failure properties of human abdominal
REFERENCES aortic aneurysm. J Biomech 2006; 39: 3010-6.
[1] Sakalihasan N, Limet R, Defawe OD. Abdominal aortic aneurysm. [26] Doyle BJ, Morris LG, Callanan A, et al. 3D reconstruction and
Lancet 2005; 365: 1577-89. manufacture of real abdominal aortic aneurysms: from CT scan to
[2] Johnston KW, Rutherford RB, Tilson MD, Shah DM, Hollier L, silicone model. J Biomech Eng 2008; 130: 034501-5.
Stanley JC. Suggested standards for reporting on arterial aneu- [27] Morris L, Delassus P, Callanan A, et al. 3D numerical simulation
rysms. Subcommittee on reporting standards for arterial aneurysms, of blood flow through models of the human aorta. J Biomech Eng
Ad hoc committee on reporting standards, Society for Vascular 2005; 127: 767-75.
Surgery and North American Chapter, International Society for [28] Li Z, Kleinstreur C. A new wall stress equation for aneurysm-
Cardiovascular Surgery. J Vasc Surg 1991; 13: 452-8. rupture. Ann Biomed Eng 2005; 33: 209-13.
[3] Vorp DA. Biomechanics of abdominal aortic aneurysm. J Biomech [29] Raghavan ML, Vorp DA, Federle MP, et al. Wall stress distribu-
2007; 40: 1887-902. tion on three-dimensionally reconstructed models of human ab-
[4] Ouriel K, Green RM, Donayre C, Shortell CK, Elliott J, DeWesse dominal aortic aneurysm. J Vasc Surg 2000; 31: 760-9.
JA. An evaluation of new methods of expressing aortic aneurysm [30] Venkatasubramaniam AK, Fagan MJ, Mehta T, et al. A compara-
size: relationship to rupture. J Vasc Surg 1992; 15: 12-20. tive study of aortic wall stress using finite element analysis for rup-
[5] Bengtsson H, Sonesson B, Bergqvist D. Incidence and prevalence tured and non-ruptured abdominal aortic aneurysms. Eur J Vasc
of abdominal aortic aneurysms, estimated by necroscopy studies Endovasc Surg 2004; 28: 168-76.
and population screening by ultrasound. Ann NY Acad Sci 1996; [31] Leung JH, Wright AR, Cheshire N, et al. Fluid structure interaction
800: 1-24. of patient specific abdominal aortic aneurysms: a comparison with
[6] Kleinstreuer C, Li Z. Analysis and computer program for rupture- solid stress models. Biomed Eng Online 2006; 5: 33.
risk prediction of abdominal aortic aneurysms. Biomed Eng Online [32] Papaharilaou Y, Ekaterinaris JA, Manousaki E, et al. A decoupled
2006; 5: 19. fluid structure approach for estimating wall stress in abdominal
[7] Sayers RD. Aortic aneurysms, inflammatory pathways and nitric aortic aneurysms. J Biomech 2007; 40: 367-77.
oxide. Ann R Coll Surg Eng 2002; 84: 239-46. [33] Speelman L, Bohra A, Bosboom EMH, et al. Effects of wall calci-
[8] Raghavan ML, Vorp DA. Toward a biomechanical tool to evaluate fications in patient-specific wall stress analyses of abdominal aortic
rupture potential of abdominal aortic aneurysm: identification of a aneurysms. J Biomech Eng 2007; 129: 1-5.
finite strain constitutive model and evaluation of its applicability. J [34] Doyle BJ, Callanan A, McGloughlin TM. A comparison of model-
Vasc Surg 2000; 33: 475-82. ling techniques for computing wall stress in abdominal aortic aneu-
[9] Fillinger MF, Raghavan ML, Marra SP, et al. In vivo analysis of rysms. Biomed Eng Online 2007; 6: 38.
mechanical wall stress and abdominal aortic aneurysm rupture risk. [35] Raghavan ML, Fillinger MF, Marra SP, et al. Automated method-
J Vasc Surg 2002; 36: 589-97. ology for determination of stress distribution in human abdominal
[10] Fillinger MF, Marra SP, Raghavan ML, et al. Prediction of rupture aortic aneurysm. J Biomech Eng 2005; 127: 868-71.
risk in abdominal aortic aneurysm during observation: wall stress [36] Wang DHG, Makaroun MS, Webster MW, et al. Mechanical prop-
vs. diameter. J Vasc Surg 2003; 37: 724-32. erties and microstructure of intraluminal thrombus from abdominal
[11] Nicholls SC, Gardner JB, Meissner MH, et al. Rupture in small aortic aneurysm. J Biomech Eng 2001; 123: 536-9.
abdominal aortic aneurysms. J Vasc Surg 1998; 28: 884-8.
[37] Inzoli F, Boschetti F, Zappa M, et al. Biomechanical factors in pean Society of Biomechanics S-Hertogenbosch, Netherlands
abdominal aortic aneurysm rupture. Eur J Vasc Surg 1993; 7: 667- 2004.
74. [41] Vande Geest JP, Wang DHJ, Wisniewski SR, et al. Towards a non-
[38] Truijers M, Pol JA, SchultzeKool LJ, et al. Wall stress analysis in invasive method for determination of patient-specific wall strength
small asymptomatic, symptomatic and ruptured abdominal aortic distribution in abdominal aortic aneurysms. Ann Biomed Eng
aneurysms. Eur J Vasc Endovasc Surg 2006; 33: 401-07. 2006; 34: 1098-106.
[39] Morris L, O’Donnell P, Delassus P, et al. Experimental assessment [42] Vande Geest JP, Sacks MS, Vorp DA. The effects of aneurysm on
of stress patterns in abdominal aortic aneurysms using the photoe- the biaxial mechanical behaviour of human abdominal aorta. J
lastic method. Strain 2005; 40: 165-72. Biomech 2006; 39: 1324-34.
[40] Callanan A, Morris LG, McGloughlin TM. Numerical and experi- [43] Rodriguez JF, Ruiz C, Doblare M, et al. Mechanical stresses in
mental analysis of an idealised abdominal aortic aneurysm. Euro- abdominal aortic aneurysms: influence of diameter, asymmetry,
and material anisotropy. J Biomech Eng 2008; 130: 0210231-10.
Received: June 16, 2008 Revised: August 05, 2008 Accepted: August 17, 2008
© Doyle et al.; Licensee Bentham Open.

This is an open access article licensed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-
nc/3.0/) which permits unrestricted, non-commercial use, distribution and reproduction in any medium, provided the work is properly cited.
BASIC RESEARCH STUDIES
Vessel asymmetry as an additional diagnostic tool

in the assessment of abdominal aortic aneurysms
Barry J. Doyle, B.Eng,a Anthony Callanan, B.Eng,a Paul E. Burke, FRCS,a,b Pierce A. Grace, FRCSI,a,b
Michael T. Walsh, PhD,a David A. Vorp, PhD,c Timothy M. McGloughlin, PhD,a Limerick, Ireland; and
Pittsburgh, Pa
Objective: Abdominal aortic aneurysm (AAA) rupture is believed to occur when the local mechanical stress exceeds the
local mechanical strength of the wall tissue. On the basis of this hypothesis, the knowledge of the stress acting on the wall
of an unruptured aneurysm could be useful in determining the risk of rupture. The role of asymmetry has previously been
identified in idealized AAA models and is now studied using realistic AAAs in the current work.
Methods: Fifteen patient-specific AAAs were studied to estimate the relationship between wall stress and geometrical
parameters. Three-dimensional AAA models were reconstructed from computed tomography scan data. The stress
distribution on the AAA wall was evaluated by the finite element method, and peak wall stress was compared with both
diameter and centerline asymmetry. A simple method of determining asymmetry was adapted and developed. Statistical
analyses were performed to determine potential significance of results.
Results: Mean von Mises peak wall stress ⴞ standard deviation was 0.4505 ⴞ 0.14 MPa (range, 0.3157-0.9048 MPa).
Posterior wall stress increases with anterior centerline asymmetry. Peak stress increased by 48% and posterior wall stress
by 38% when asymmetry was introduced into a realistic AAA model.
Conclusion: The relationship between posterior wall stress and AAA asymmetry showed that excessive bulging of one
surface results in elevated wall stress on the opposite surface. Assessing the degree of bulging and asymmetry that is
experienced in an individual AAA may be of benefit to surgeons in the decision-making process and may provide a useful
adjunct to diameter as a surgical intervention guide. ( J Vasc Surg 2009;49:443-54.)
Clinical Relevance. There is much debate about the most appropriate time to intervene with surgical treatment of
abdominal aortic aneurysms. Currently, maximum diameter is deemed the most accurate indicator of rupture potential
because size is not only an obvious factor in the decision-making process but is also easy for the clinician to determine
from computed tomography scans. The method of determining vessel asymmetry proposed here is easy to interpret and
was shown to be as significant as diameter in the cases examined. Therefore, asymmetry could become a useful adjunct to
diameter in the decision-making process of the clinician.
There is currently much debate about the most appro- diameter. Previous research8,9 has shown how AAAs ⬍5.5
priate time to surgically intervene and repair an abdominal cm in maximum diameter can also rupture. The reliability of
aortic aneurysm (AAA).1-7 Surgery is often performed the maximum diameter as the main criterion for rupture has
when the detected AAA is ⬎5.0 to 5.5 cm in maximum been questioned recently, and a need for a more reliable
clinical predictor of AAA rupture has been identified.1-7,10-12
From the Centre for Applied Biomedical Engineering Research (CABER),
Department of Mechanical and Aeronautical Engineering, and the Mate-
Previous work12,13 has identified the importance of asym-
rials and Surface Science Institute, University of Limerick,a and the metry in idealized AAA models and also indicated the need
Department of Vascular Surgery, HSE Midwestern Regional Hospital, to investigate this aspect in realistic models. In this study,
Limerickb; and the Centre for Vascular Remodelling and Regeneration, we have examined the role of asymmetry and resulting wall
McGowan Institute for Regenerative Medicine, Departments of Surgery
stress in realistic patient-specific AAA cases.
and Bioengineering, University of Pittsburgh, Pittsburgh.c
Supported by the Irish Research Council for Science, Engineering and
Technology (IRCSET) Grant RS/2005/340, and Grant No. R01-HL- METHODS
060670 from the United States National Heart Lung and Blood Institute.
Computed tomography (CT) scan data was obtained
Competition of interest: none.
Reprint requests: Prof Timothy M. McGloughlin, Centre for Applied Bio- for 22 patients. For this study, AAAs that were asymmetric
medical Engineering Research, University of Limerick, National Techno- in the anterior–posterior plane were deemed applicable.
logical Park, Castletroy, Limerick, Ireland (e-mail: tim.mcgloughlin@ This criterion was used to exclude 7 of the 22 patients from
ul.ie). the analysis because their AAAs were asymmetric in other
0741-5214/$36.00
Copyright © 2009 Published by Elsevier Inc. on behalf of The Society for
directions. The resulting cohort of 15 patients (10 men, 5
Vascular Surgery. women) and a mean age ⫾ standard deviation (SD) of 73.2
doi:10.1016/j.jvs.2008.08.064 ⫾ 6.7 years. Scans for these patients were obtained from the
443
JOURNAL OF VASCULAR SURGERY
444 Doyle et al February 2009
tions allowed the computation of stress distributions within

the geometries. The patient-specific details for the cases
studied were obtained using the schematic of Fig 1 and are
compiled in Table I.
Three-dimensional reconstruction procedure. Spiral
CT data were used to reconstruct the infrarenal section of
the aorta. Because CT scanning is routine for AAA patients
scheduled for repair, collection of this information in-
volved no extra participation by the study subjects. Dig-
ital files in the digital imaging and communications in
medicine (DICOM) file format, containing cross-sectional
information, were imported to the Mimics 10.0 software
for reconstruction. All reconstructions were developed
from scan positions immediately distal to the lowest renal
artery to immediately proximal to the iliac bifurcation. The
intraluminal thrombus (ILT) was neglected in this study, as
with previous approaches.6,7,11-13,15 The thickness of the
aorta wall is not easily identifiable from CT scans; therefore,
the wall was assumed to be uniform throughout the model
and set to 2 mm.16
Once regions of interest were identified, the three-
dimensional (3D) reconstructions were generated. The
reconstruction method used here was validated and re-
ported in previous work performed by our group17-19
along with the effect of geometry smoothing on resulting
wall stress.20 All AAAs underwent the same degree of
smoothing. The iliac bifurcation was omitted from this
study, as in previous stress analysis work, because it does not
significantly affect the wall stress results of the AAA.7 The
influence of asymmetry compared with a symmetric AAA
was also examined. The reconstruction of patient 2 was
modified using ProEngineer Wildfire 3.0 software (Para-
metric Technology Corp, Needham, Mass) so that the AAA
now formed along a straight central axis, becoming an
Fig 1. Schematic of representative abdominal aortic aneurysm axisymmetric fusiform aneurysm. This symmetric AAA was
(AAA) shows how the dimensions are obtained from each AAA created using the same diameter information as the original
model. Surface area and volume both encompass the total surface case. The two forms of this AAA can be seen in Fig 2.
area or volume of the AAA from immediately below the renal
Biomechanical material properties. The AAA mate-
arteries to immediately before the iliac bifurcation. Dianorm is the
rial was assumed to be homogenous and isotropic, with
infrarenal aortic diameter of the particular case.
nonlinear realistic material properties5 that have been im-
plemented in many previous publications.3,6,7,10,11,20-22
Midwestern Regional Hospital, Limerick, Ireland, and the The aorta is also known to be nearly incompressible with a
University of Pittsburgh Medical Center, Pittsburgh, Pa. Poisson’s ratio of 0.49.
All 15 patients were either awaiting or had received AAA Finite element mesh generation. Once the AAA sur-
repair because their AAA diameters had reached or ex- faces were imported into ABAQUS 6.6-2 software (Das-
ceeded the current 5-cm threshold for repair. sault Systemes, SIMULIA, Providence, RI) for stress anal-
The CT scans were acquired using Somatom Plus 4 ysis, a finite element mesh was generated on the AAA
(Siemens AG, D-91052 Erlangen, Germany) and Light- model. Because wall thickness cannot be fully determined
Speed Plus (General Electric Medical Systems, Wauke- from the AAA scan data, each shell element was assigned a
sha, Wisc) imaging equipment. All scans were single uniform thickness of 2 mm.16 Mesh independence was
slices with a standard width ⫻ height of 512 ⫻ 512 performed by increasing the number of elements in the
pixels. Mean ⫾ SD pixel size of scans was 0.742 ⫾ 0.072 mesh until the difference in peak stress was ⬍2% of the
mm. The bodily structures of each subject were made previous mesh.10,20,23
visible using Optiray nonionic contrast dye (Mallinck- Forces and boundary conditions. The blood pres-
rodt Inc, Hazelwood, Mo). sure within the AAA acts on the AAA inner wall, and
These CT data sets were then reconstructed using the therefore pressure was applied to the inner surface of the
commercially available software, Mimics 10.0 (Materialise computational AAA model. A static peak systolic pressure
Technologies, Leuven, Belgium), and these reconstruc- of 120 mm Hg (16 KPa) was used. To simulate the teth-
Volume 49, Number 2 Doyle et al 445
Table I. Details for study patients
Patient Sex Age Max diameter, cm Total length, cm Total volume, cm3 Total SA,a cm2 Diam/length ROD
1 M 66 5.6 13.2 176.7 19.8 0.425 1.533

2 M 78 6.1 11.2 192.4 18.5 0.544 2.071
3 M 70 5.7 13 194.9 19.1 0.439 1.752
4 F 65 5.6 9.3 136.9 14.8 0.606 2.474
5 M 81 5.9 12.8 220.9 21.8 0.463 1.772
6 F 68 5.7 10 148.2 16.3 0.570 2.953
7 F 67 5.3 10.5 137.9 15.5 0.505 2.000
8 M 70 6.0 11.1 216.7 20.4 0.541 1.508
9 F 77 5.8 8.8 94.2 12.6 0.661 2.535
10 M 87 9.0 11.7 445.5 32.4 0.769 1.768
11 M 66 6.5 10.5 207.6 19.2 0.617 1.952
12 M 81 6.8 14 267.6 25.2 0.484 1.994
13 M 77 6.2 16.8 320.8 27 0.371 1.757
14 F 72 5.7 11.4 143.8 16.7 0.497 2.675
15 M 73 7.9 11.8 286.5 24.3 0.671 2.865
F, Female; M, male.
a
SA is the total AAA surface area, Diam/length is the ratio of maximum diameter to total abdominal aortic aneurysm (AAA) length, and ROD is the ratio of
maximum AAA diameter to infrarenal diameter of that patient.14
point on the centerline. Figs 3-5 show how these asymme-

try measures are obtained.
This method of determining asymmetry was adapted
from previous work by Young Suh et al.24 Starting with the
3D AAA model in Fig 3, A, a centerline is automatically
created through the polyline centroids of Fig 3, B, thus
creating Fig 3, C. Then these polylines are exported from
Mimics 10.0 to ProEngineer Wildfire 3.0. Next, the soft-
ware is used to connect the endpoints of the centerline with
a straight line (Fig 3, D), and a perpendicular line is
extended from this connecting line to predetermined
points along the centerline (Fig 3, E and F).
The asymmetry at a specified distance along the AAA
model is regarded as this perpendicular distance and is
measured in millimeters. This method of determining AAA
asymmetry is shown for patient 3 in Figs 4 and 5. Fig 4
shows the creation of the polylines on the CT scan after the
thresholding and segmentation process in Mimics 10.0 and
also the resulting model of polylines and AAA centerline.
Fig 2. Left, Original asymmetric abdominal aortic aneurysm Fig 5 shows the measurement process of asymmetry for
(AAA) of patient 2 and (right) the modified axisymmetric AAA. patient 3 and the resulting asymmetry plot. Maximum
asymmetry for this case was 24.5 mm.
Statistical analysis. The statistical significance of the
wall stress results was evaluated with SPSS 14.0 software
ering of the AAA to the aorta at the renal junction and iliac
(SPSS Inc, Chicago, Ill). This allowed any significant cor-
bifurcation, the AAA model was fully constrained in the relations within the results to be identified. Correlations
proximal and distal regions. between various geometric parameters with both peak wall
Asymmetry definition. Most AAAs are constrained stress and posterior wall stress were assessed for signifi-
from radial expansion in the posterior direction due to the cance. Mean values are presented with the SD.
spinal column; therefore, AAAs predominantly dilate in the
anterior plane. All AAAs studied in this analysis were natu-
rally asymmetric in the anterior–posterior direction. To RESULTS
examine the effect asymmetry has on wall stress, the cen- The results from the geometrical examination of each case
terline of each AAA was automatically found using the are summarized in Table I. The finite element analysis using
Mimics 10.0 software. The centerline passes through the ABAQUS 6.6-2 software produced a detailed stress pattern
centroid of each polyline slice in the series. Asymmetry is on each of the aneurysmal models under the pressure load-
defined, in this case, as the perpendicular distance from the ing.25 These stress results could be used to examine factors
proximal and distal points of the centerline to a defined affecting wall stress, in particular, the role of asymmetry.
Fig 3. Simple illustration showing the method of obtaining asymmetry measurements. A, Three-dimensional abdom-
inal aortic aneurysm (AAA) model and (B) line drawing. C, A centerline is automatically created through the polyline
centroids. D, The endpoints of the centerline are connected with a straight line. E, and F, A perpendicular line is
extended from this connecting line to predetermined points along the centerline.
Fig 4. Computed tomography scans for patient 3 show the creation of polylines on each slice, together with the center
point of each polyline. Once the polylines are created on each scan in the series, the slices are stacked to form the model
on the right.
Fig 5. Diagrams depict how measurements of asymmetry are determined once the abdominal aortic aneurysm (AAA)
centerline is created. The example shown is for patient 3. Left, Endpoints of the centerline are connected. Middle,
Perpendicular distances from the centerline are measured. Right, Asymmetry measurements along the length of the
AAA are plotted.
Table II. Aneurysm diameters and wall stress variables 0.14 MPa (range, 0.3157-0.9048 MPa). The mean cir-
for study patients cumferential stress was also recorded for each case and was
0.1176 ⫾ 0.061 MPa (range, 0.07-0.3271 MPa).
Max Peak wall Diameter at Wall stress–asymmetry relationship. Figs 6-8 show
diameter, stress, Location of peak wall
Patient cm MPa peak wall stress stress, cm how the von Mises wall stress varies with respect to the
asymmetry of the AAA centerline. Regions of elevated
1 5.6 0.4018 Anterior-right 3.8 centerline asymmetry experienced a region of elevated pos-
2 6.1 0.4213 Anterior-right 4.1 terior wall stress.
3 5.7 0.5524 Posterior 5.3
Effect of asymmetry on wall stress. To gauge the
4 5.6 0.3157 Posterior 4.9
5 5.9 0.3822 Posterior 5.3 effect of asymmetry on resulting wall stress, the AAA of
6 5.7 0.3823 Left 5.8 patient 2 was modified into a symmetric aneurysm, as
7 5.3 0.3621 Left 4.8 described earlier. The symmetric wall stress can be seen
8 6.0 0.3872 Posterior 4.6
compared with that of the posterior wall stress in the
9 5.8 0.4093 Posterior-right 5.5
10 9.0 0.9048 Posterior 6.2 asymmetric case in Fig 9. Peak stress increased from 0.2186
11 6.5 0.4608 Left 5.8 MPa to 0.4213 MPa when asymmetry was introduced into
12 6.8 0.4991 Left 5.6 the AAA. This resulted in a 48% increase in peak wall stress
13 6.2 0.4523 Left 6 in the asymmetric model. There was also a noticeable
14 5.7 0.3703 Posterior 5.4
15 7.9 0.4747 Left 7.6 increase of 38% from 0.2186 to 0.3527 MPa in posterior
wall stress between the two models.
Statistical analysis. A Spearman ␳ correlation test was
Peak wall stress. The computed stress results showed considered to assess any relationships evident between both
that the regions of peak wall stress occurred at regions of peak and posterior wall stress and various patient-specific
inflection on the surface of the AAA models. Inflection measurable parameters. Correlations are deemed signifi-
points are defined as points on the AAA surface at which the cant when P ⬍ .05. The P values of this study are compiled
local AAA wall shape changes from concave outward to in Tables III and IV. No significant correlation was noted
concave inward.12 The peak stress occurring at regions of between peak circumferential stress and either maximum
inflection was also observed by previous researchers in asymmetry (P ⫽ .0708) or maximum diameter (P ⫽
idealized models.12,13,26,27 The von Mises peak wall stress, .5197). The relationship between posterior wall stress with
diameter at peak stress, and location were recorded for each both asymmetry and diameter was also examined using a
AAA and were compared with the maximum diameter in Spearman ␳ correlation test. Coefficients were found using
Table II. All peak stress values are recorded at the peak a bivariate correlation test to compare posterior wall stress
systolic pressure of 120 mm Hg (16 KPa). Shown also is the with both asymmetry and diameter at 10-mm intervals
diameter of the region through which the peak wall stress along the longitudinal distance of each patient. These
occurred. The mean von Mises peak wall stress was 0.4505 ⫾ results can be seen in Table V. The significance of the
Fig 6. Relationships between (left column) posterior wall stress and anterior asymmetry and (right column)
posterior wall stress and diameter for patients 1 through 5.
Table IV. Statistical analysis of patient-specific

parameters and posterior wall stress
Variable P
AAA volume .0002

AAA surface area .0008
Lumen volume .0012
Maximum CSA .0013
Peak stress .0021
Maximum diameter .0028
Sex .0081
Asymmetry at peak circumferential stress .0136
AAA length .0144
Peak circumferential stress .0378
AAA volume/ILT volume .0983
Fig 9. Comparison of the posterior wall stress for the symmetric ILT volume .1728
and asymmetric (circles) abdominal aortic aneurysm (AAA) for ROD .2597
patient 2. Peak stress for this asymmetric AAA was located on the Peak stress location .5399
anterior-right wall. Peak asymmetry .6025
Asymmetry at peak von Mises stress .7466
AAA diameter/AAA length .9295
Table III. Statistical analysis of patient-specific Average asymmetry .9496
parameters and peak wall stress AAA, Abdominal aortic aneurysm; CSA, cross-sectional area at the region of
maximum diameter; ILT, intraluminal thrombus; ROD, ratio of the maxi-
Variable P mum diameter to the infrarenal diameter.
Maximum diameter .0003

Peak posterior stress .0021 Table V. Correlation coefficients for posterior wall stress
Asymmetry at peak circumferential stress .0036 and asymmetry and diameter
AAA volume .0043
Sex .0061 Patient Asymmetry P Diameter P
Maximum CSA .0126
AAA surface area .0130 1 0.574 .032 0.420 .135
ILT volume .0232 2 0.781 .003 0.895 .000
AAA length .0961 3 0.175 .587 0.755 .005
Peak circumferential stress .1580 4 0.37 .293 0.733 .016
Lumen volume .1658 5 0.862 .000 0.820 .000
ROD .3307 6 0.82 .002 0.609 .047
Asymmetry at peak von Mises stress .4384 7 0.464 .151 0.573 .066
AAA diameter/AAA length .5409 8 0.834 .001 0.573 .051
Peak stress location .5814 9 0.474 .166 0.529 .116
Peak asymmetry .6384 10 0.443 .130 0.505 .078
AAA volume/ILT volume .8994 11 0.683 .014 0.811 .001
Average asymmetry .9345 12 0.411 .128 0.593 .020
AAA, Abdominal aortic aneurysm; CSA, cross-sectional area at the region of 13 0.411 .128 0.593 .020
maximum diameter; ILT, intraluminal thrombus; ROD, ratio of maximum 14 0.834 .000 0.709 .007
AAA diameter to infrarenal diameter. 15 0.667 .013 ⫺0.132 .668
relationship between asymmetry and diameter was also

assessed using a nonparametric correlation test (Table VI). using the finite element method. From the von Mises wall
In 11 of the 15 AAAs, a significant correlation was revealed stress distributions, the peak stress occurred at regions of
between asymmetry and diameter. Patient age also corre- inflection. This finding is consistent with previous research,
lated well with both maximum diameter (P ⫽ .009) and both numeric12,26 and experimental.27 The mean values of
peak posterior wall stress (P ⫽ .028). peak wall stresses found in this study were 0.482 ⫾ 0.197
The statistical significance between peak stress and MPa (range, 0.3157-0.9584 MPa). The AAA in patient 4
other relevant parameters was also assessed. The rate of had the lowest peak wall stress (0.3157 MPa) and also had
change of both asymmetry and diameter along the length the second smallest maximum AAA diameter (5.6 cm);
of the AAA were not statistically significant (P ⫽ .089 and whereas, the AAA in patient 10 had the highest peak stress
P ⫽ .501, respectively). The diameter at which peak stress (0.9048 MPa) and had the largest AAA diameter (9 cm).
occurred was significant (P ⫽ .039). These findings may suggest that the maximum diame-
ter criterion may be a good predictor of AAA rupture. From
DISCUSSION the previous research into this hypothesis,9 it is known that
This study reconstructed 15 patient-specific AAAs, and this may not always be the case, because small AAAs can
wall stress distributions in each aneurysm were estimated also rupture. Fillinger et al7 performed stress analysis on an
Table VI. Correlation between abdominal aortic also reported that failure tension might be a better indicator
aneurysm asymmetry and diameter for each patient of rupture rather than failure stress, with failure tension
described as: peak wall tension ⫽ peak wall stress ⫻ wall
Patient Coefficient P thickness. Applying this failure tension to this study results
in failure tensions ranging from 0.6314 to 1.8096 N/mm,
1 0.801 .001
2 0.823 .001 compared with the range of 0.42 to 1.48 N/mm observed
3 0.755 .005 by Fillinger et al.6 Actual failure stress of AAA tissue has
4 0.612 .060 been shown to be a median of 1.266 MPa (range, 0.336-
5 0.935 .000 2.351 MPa).29
6 0.752 .008
7 0.9 .000
Giannoglou et al30 also determined that mean AAA
8 0.806 .002 curvature may be a better predictor of AAA rupture risk,
9 0.799 .006 although this previous study implemented linearly elastic
10 0.627 .022 material properties. AAA centerline curvature was also an-
11 0.746 .005 alyzed as part of this present study using the curvature
12 0.391 .150
13 0.391 .150 analysis function in ProEngineer Wildfire 3.0. The resulting
14 0.889 .000 centerline curvature readings are difficult to interpret be-
15 0.066 .830 cause minor changes in the centerline result in large spikes
of curvature. Gaussian surface curvature was also examined
using ProEngineer Wildfire 3.0, as previously reported.22
AAA with a 4.8-cm diameter, which was smaller than the Rapid changes in surface curvature may indicate regions of
current 5.5-cm threshold. This particular AAA experienced high wall stress.22 As with centerline curvature, the model
a peak wall stress of 0.335 MPa, which is within 10% of the geometries are too complex to achieve surface curvature
peak stress found in patients 4, 7 and 14 of this study. results with any quantifiable meaning. The definition of
It is important to note that ruptures do not necessarily asymmetry in this study is a measurement that is easy to
occur at the region of peak wall stress, but in fact occur interpret and calculate and shows good agreement with
where the locally acting wall stress exceeds the locally acting wall stress results.
wall strength. This study examined the role of realistic As the anterior region of the AAA bulges outwards, the
asymmetry on posterior wall stress in patient-specific cases. posterior region is often constrained from radial expansion
To determine the effect the asymmetry of the AAA has on by the spinal column and results in elevated posterior wall
posterior wall stress, a simple method of calculating asym- stress. AAAs may also rupture at regions experiencing a wall
metry was established. Figs 3-5 show the approach used to stress that is less than that of the peak wall stress because
plot the degree of asymmetry in each AAA, and these results AAAs are known to rupture when the local stress exceeds
were then coupled with the posterior wall stress results. the local wall strength, with AAAs experiencing regional
These stress-asymmetry relationships are shown in Figs 6-8, variations in wall strength.29
which also show the relationship between diameter and Two AAAs in this study experienced peak wall stress on
posterior wall stress. the anterior-right wall. Peak stress can occur at any region
Raghavan et al28 reported that the posterior wall tends along the AAA surface but is predominantly found at
to be the higher stressed region and also the rupture site, regions where there is high local surface curvature or asym-
even though the bulge is predominantly anterior. Vorp et metry. Therefore, when patient-specific AAAs are analyzed,
al12 identified the link between asymmetry and wall stress in it is difficult to predetermine the location of peak stress.
idealized AAA models using finite element mesh, with Results in six cases showed peak stress on the left wall of the
Scotti et al13 using a fluid-structure interaction approach to AAA. Again, these locations of elevated stress are due to the
highlight the relationship in idealized models. Scotti et al13 local topology of the surface. Even though peak stress
concluded that AAAs experiencing asymmetry may be ex- does not necessarily occur along the posterior wall, in all
posed to higher mechanical stresses and increased risk of AAAs examined there was an increase in posterior stress
rupture than more fusiform AAAs. This current work along the length of the particular AAA in relation to
agrees with this hypothesis and has furthered the work of anterior asymmetry.
idealized AAAs to that of realistic AAA geometries, with From the results of the statistical analysis summarized
results suggesting that an intrinsic relationship may exist in Tables III and IV, one can determine that the significant
between asymmetry and posterior wall stress. parameters that relate to peak wall stress are maximum
Darling et al9 determined from 118 AAA autopsies that diameter (P ⫽ .0003), peak posterior wall stress (P ⫽
82% of ruptures occur on the posterior wall, indicating that .0021), asymmetry at the region of peak circumferential
these ruptures may have resulted from elevated posterior stress (P ⫽ .0036), AAA volume (P ⫽ .0043), sex (P ⫽
wall stress. Of the 22 patients examined for this study, 15 .061), maximum cross-sectional area (P ⫽ .0126), AAA
(68%) experienced posterior–anterior bowing, resulting in surface area (P ⫽ .013), and also the ILT volume (P ⫽
elevated posterior wall stress. Raghavan et al29 recently .0232). In comparison, the significant relationships with
reported that the posterior and right regions of AAAs are posterior wall stress are AAA volume (P ⫽ .0002), AAA
regionally thinner than the anterior and left regions. They surface area (P ⫽ .0008), lumen volume (P ⫽ .0012),
maximum cross-sectional area (P ⫽ .0013), peak wall stress the maximum diameter criterion, and may ultimately lead
(P ⫽ .0021), maximum diameter (P ⫽ .0028), sex (P ⫽ to improved surgical decision making.
.0081), asymmetry at the region of peak circumferential Our study has some limitations. Similar to previous
stress (P ⫽ .0136), AAA length (P ⫽ .0144), and also peak work,6,7,11-13,15 this study did not include ILT in the AAA
circumferential stress (P ⫽ .0378). From these results, max- 3D reconstructions. The ILT has been shown to reduce
imum diameter appears to be significantly related to peak wall stress by up to 30%10,30 and can act as a “mechanical
stress, but if the sample size of 15 used in this study were cushion”31 for the AAA wall. The realistic AAA has a
increased to much larger numbers, this relationship might not nonuniform median wall thickness29 of 1.48 mm, varying
be as strong. Most of these parameters are also based on regionally from 0.23 mm at a rupture site to 4.26 mm at a
diameter; therefore, if diameter returns a strong correlation calcified site, and also has nonuniform material properties
with peak stress, it may be obvious that similar parameters will due to regions of calcifications,32 which can lead to alter-
also score highly. No statistical significance was noted be- ations in stress distributions.32,33
tween peak wall stress and the rate of change of diameter (P ⫽ This study examined AAA wall stress using a static
.501) or rate of change of asymmetry (P ⫽ .089). analysis. It is possible that a dynamic loading, such as a
Closer examination of the relationship between both realistic infrarenal aortic pulse, may influence the stress
asymmetry and diameter with posterior wall stress involved distributions. Researchers have shown that the use of fluid-
analyses using a nonparametric Spearman ␳ correlation. structure interaction methods to determine wall stress can
These results are presented in Table V and show how give more accurate results.3,13,34 Computational time is
asymmetry and diameter are both comparable in their increased by as much as 2500-fold13 from that of a static
significance towards posterior wall stress. From the result- pressure finite element analysis, with maximum stress loca-
ing correlations in the 15 AAAs, eight show asymmetry is tions found to be the same using both methods. Variations
significant and nine show that diameter is significant. These in maximum wall stress in realistic models have been re-
results suggest that if posterior wall stress is to gain clinical ported in prior studies to range from 1% to 25%.3,21,33
acceptance as a possible high-risk rupture indicator, asym- To establish the suitability of this method for clinical
metry and diameter may both be as important in determin- applications, a larger cohort of patient data is required. We
ing the posterior wall stress and therefore may both equally are investigating the possibility of applying this method to a
contribute to AAA rupture. Both Vorp et al25 and Fillinger database of previously screened patients, with a view to
et al6,7 have previously postulated that the biomechanics enhancing confidence in the asymmetry approach. Apply-
of the AAA may provide useful clinical guidance over ing this study to a larger cohort may also significantly alter
the maximum diameter criteria. This work supports this the statistical results because only 15 patients were studied
biomechanics-based approach and in particular suggests here.
that posterior wall stress may be clinically important. The
results presented also suggest that if peak wall stress is to CONCLUSIONS
remain the primary purpose of AAA stress analyses, then Most AAA ruptures occur on the posterior wall. We
diameter remains a significant factor. have showed here how posterior wall stress can be related to
Although ideally, stress analysis should be done on anterior asymmetry in patient-specific cases. Results sug-
every AAA detected, the reality is that the decision to repair gest that an increase in asymmetry may cause increases in
lies with the surgeon. The use of the maximum diameter posterior wall stress. Statistical analyses revealed that the
criterion is very easy to implement for the surgeon, in that maximum diameter still significantly influences wall stress,
he or she must simply measure the maximum diameter particularly peak wall stress, but that asymmetry may also
from CT scans. The asymmetry condition described in this have a significant role in posterior wall stress. This study
study could also be readily incorporated into the surgeon’s suggests that AAA asymmetry may be an important
decision making. The clinician can identify this dilation criterion in AAA assessment and could possibly be in-
and, ultimately, asymmetry from a basic 3D reconstruction, cluded as a factor in the clinicians’ decision to surgically
which could greatly aid in the decision to surgically inter- intervene. Further evaluation is needed to determine
vene. A method of determining asymmetry from 2D CT clinical applicability.
scans is also currently under development within our
group. We would like to thank the Department of Vascular
Our group is also developing an approach that accounts Surgery in the Midwestern Regional Hospital, Ireland; Dr
for asymmetry in all directions, and therefore, the relation- Liam Morris in the Galway Medical Technology Center,
ship between asymmetry and wall stress can be assessed in Galway, Mayo institute of Technology, Ireland; Samarth
all AAAs regardless of orientation. Once detected, the Shah from the Centre for Vascular Remodelling and Re-
degree of bulging could be incorporated into the surgeon’s generation, and Michel S. Makaroun, MD, Department of
decision-making process, and may refine and improve the Surgery, University of Pittsburgh.
current system of deciding on surgical intervention solely
on the basis of maximum diameter. It is suggested that to AUTHOR CONTRIBUTIONS
include AAA asymmetry as another means of assessing the Conception and design: BD
rupture potential of AAAs could serve as a useful adjunct to Analysis and interpretation: BD, AC, TM
Data collection: PG, DV, PB 17. Morris L, Delassus P, Callanan A, Walsh M, Wallis F, Grace P, et al. 3D
Writing the article: BD numerical simulation of blood flow through models of the human aorta.
J Biomech Eng 2005;127:767-75.
Critical revision of the article: BD, TM, AC, MW, PB, PG, DV 18. Morris LG. Numerical and experimental investigation of mechanical
Final approval of the article: TM factors in the treatment of abdominal aortic aneurysms [PhD thesis].
Statistical analysis: BD Limerick, Ireland: University of Limerick; 2004.
Obtained funding: TM, DV 19. Doyle BJ, Morris LG, Callanan A, Kelly P, Vorp DA, McGloughlin TM.
3D reconstruction and manufacture of real abdominal aortic aneurysms:
Overall responsibility: TM
from CT scan to silicone model. J Biomech Eng 2008;130:034501-5.
REFERENCES techniques for computing wall stress in abdominal aortic aneurysms.
1. Vande Geest JP, Wang DHJ, Wisniewski SR, Makaroun MS, Vorp DA. Biomed Eng Online 2007;6:38.
Towards a non-invasive method for determination of patient-specific 21. Papaharilaou Y, Ekaterinaris JA, Manousaki E, Katsamouris AN. A
wall strength distribution in abdominal aortic aneurysms. Ann Biomed decoupled fluid structure approach for estimating wall stress in abdom-
Eng 2006;34:1098-106. inal aortic aneurysms. J Biomech 2007;40:367-77.
2. Kleinstreuer C, Li Z. Analysis and computer program for rupture-risk 22. Sacks MS, Vorp DA, Raghavan ML, Federle MP, Webster MW. In vivo
prediction of abdominal aortic aneurysms. Biomed Eng Online 2006; three-dimensional surface geometry of abdominal aortic aneurysms.
5:19. Ann Biomed Eng 1999;27:469-79.
3. Leung JH, Wright AR, Cheshire N, Crane J, Thom SA, Hughes AD, 23. Truijers M, Pol JA, SchultzeKool LJ, van Sterkenburg SM, Fillinger
et al. Fluid structure interaction of patient specific abdominal aortic MF, et al. Wall stress analysis in small asymptomatic, symptomatic and
aneurysms: a comparison with solid stress models. Biomed Eng Online ruptured abdominal aortic aneurysms. Eur J Vasc Endovasc Surg 2007;
2006;5:33. 33:401-7.
4. Sayers RD. Aortic aneurysms, inflammatory pathways and nitric oxide. 24. Young Suh G, Choi G, Draney Blomme M, Taylor C. Quantification of
Ann R Coll Surg Engl 2002;84:239-46. three-dimensional motion of the renal arteries using image-based mod-
5. Raghavan ML, Vorp DA. Toward a biomechanical tool to evaluate elling techniques. Proceedings of the ASME 2007 Summer Bioengi-
rupture potential of abdominal aortic aneurysm: identification of a finite neering Conference. New York, NY: ASME International, 2007.
strain constitutive model and evaluation of its applicability. J Vasc Surg 25. Vorp DA. Biomechanics of abdominal aortic aneurysm. J Biomech
2000;33:475-82. 2007;40:1887-902.
6. Fillinger MF, Marra SP, Raghavan ML, Kennedy FE. Prediction of 26. Callanan A, Morris LG, McGloughlin TM. Numerical and experimental
rupture risk in abdominal aortic aneurysm during observation: wall analysis of an idealised abdominal aortic aneurysm. Presented at: Euro-
stress versus diameter. J Vasc Surg 2003;37:724-32. pean Society of Biomechanics, Hertogenbosch, Netherlands, Jul 4-7,
7. Fillinger MF, Raghavan ML, Marra SP, Cronenwett JL, Kennedy FE. In 2004.
vivo analysis of mechanical wall stress and abdominal aortic aneurysm 27. Morris L, O’Donnell P, Delassus P, McGloughlin T. Experimental
rupture risk. J Vasc Surg 2002;36:589-97. assessment of stress patterns in abdominal aortic aneurysms using the
8. Nicholls SC, Gardner JB, Meissner MH, H Johansen K. Rupture in photoelastic method. Strain 2005;40:165-72.
small abdominal aortic aneurysms. J Vasc Surg 1998;28:884-8. 28. Raghavan ML, Kratzberg JA, Golzarian J. Introduction to biomechan-
9. Darling RC, Messina CR, Brewster DC, Ottinger LW. Autopsy study of ics related to endovascular repair of abdominal aortic aneurysm. Tech
unoperated abdominal aortic aneurysms. The case for early resection. Vasc Interv Radiol 2005;8:50-5.
Circulation 1977;56:161-4. 29. Raghavan ML, Kratzberg J, de Tolosa EMC, Hanaoka MM, Walker P,
10. Wang DHJ, Makaroun MS, Webster MW, Vorp DA. Effect of intralu- da Silva ES. Regional distribution of wall thickness and failure properties
minal thrombus on wall stress in patient-specific models of abdominal of human abdominal aortic aneurysm. J Biomech 2006;39:3010-6.
aortic aneurysm. J Vasc Surg 2002;36:598-604. 30. Giannoglu G, Giannakoulas G, Soulis J, Chatzizisis Y, Perdikides T,
11. Raghavan ML, Vorp DA, Federle MP, Makaroun MS, Webster MW. Melas N, et al. Predicting the risk of rupture of abdominal aortic
Wall stress distribution on three-dimensionally reconstructed models of aneurysms by utilizing various geometrical parameters: Revisiting the
human abdominal aortic aneurysm. J Vasc Surg 2000;31:760-9. diameter criterion. Angiology 2006;57:487-94.
12. Vorp DA, Raghavan ML, Webster MW. Mechanical wall stress in 31. Vorp DA, Vande Geest J. Biomechanical determinants of abdominal
abdominal aortic aneurysm: influence of diameter and asymmetry. J aortic aneurysm rupture. Artherioscler Thromb Vasc Bio 2005;25:
Vasc Surg 1998;27:632-9. 1558-66.
13. Scotti CM, Shkolnik AD, Muluk SC, Finol E. Fluid-structure interac- 32. Vorp DA, Mandarino WA, Webster MW, Gorcsan J 3rd. Potential
tion in abdominal aortic aneurysms: effect of asymmetry and wall influence of intraluminal thrombus on abdominal aortic aneurysm as
thickness. Biomed Eng Online 2005;4:64. assessed by a new non-invasive method. Cardiovasc Surg 1996;4:732-9.
14. Hatakeyama T, Shigematsu H, Muto T. Risk factors for rupture of 33. Speelman L, Bohra A, Bosboom EMH, Schurink GWH, van de Vosse
abdominal aortic aneurysm based on a three-dimensional study. J Vasc FN, Makaroun MS, et al. Effects of wall calcifications in patient-specific
Surg 2001;33:453-61. wall stress analyses of abdominal aortic aneurysms. J Biomech Eng
15. Thubrikar MJ, Al-Soudi J, Robicsek F. Wall stress studies of abdominal 2007;129:1-5.
aortic aneurysm in a clinical model. Ann Vasc Surg 2001;15:355-66. 34. Scotti CM, Finol EA. Compliant biomechanics of abdominal aortic
16. Venkatasubramaniam AK, Fagan MJ, Mehta T, Mylankal KJ, Ray B, aneurysms: a fluid-structure interaction study. Comp Struc 2007;85:
Kuhan G, et al. A comparative study of aortic wall stress using finite 1097-113.
element analysis for ruptured and non-ruptured abdominal aortic aneu-
rysms. Eur J Vasc Endovasc Surg 2004;28:168-76. Submitted Apr 17, 2008; accepted Aug 23, 2008.
518 J ENDOVASC THER
2008;15:518–529
¤EXPERIMENTAL INVESTIGATION ¤
Geometrical Enhancements for Abdominal Aortic

Stent-Grafts
David S. Molony, BEng1; Anthony Callanan, BEng1; Liam G. Morris, PhD2; Barry J. Doyle,
BEng1; Michael T. Walsh, PhD1; and Tim M. McGloughlin, PhD1
1Centre for Applied Biomedical Engineering Research (CABER) and the Materials and
Surface Science Institute, University of Limerick, Ireland. 2Galway Medical Technology
Centre, Galway Mayo Institute of Technology, Galway, Ireland.
¤ ¤
Purpose: To compare the function of 2 stent-graft designs for endovascular abdominal
aortic aneurysm repair.
Methods: Computational fluid dynamics was used to investigate the performance of a
conventional stent-graft versus one with a novel tapered configuration (equal area ratios at
the inlet and bifurcation). Idealized geometries (uniplanar) were formed first for both
devices. To mimic the clinical setting with pulsatile blood flow, a realistic model
(multiplanar) was created for the conventional stent-graft based on computed tomography
scans from 3 patients with different aortic geometries. A similar model was created for the
tapered stent-graft by mimicking the deployment of the conventional stent-graft through
its centerline.
Results: The tapered stent-graft model demonstrated reduced secondary flow vortices and
wall shear stresses in the iliac limbs compared to the conventional graft in the idealized
scenario. The drag forces in the idealized models were similar for both designs, though the
tapered stent-graft showed a 4% reduction. Flow was split more evenly between the
tapered stent-graft limbs in the realistic scenario.
Conclusion: The novel tapered design reduced flow velocities and secondary flows due to
its smooth trunk-to-limb transition, while also splitting the flow between the iliac limbs
more evenly. In multiplanar models, the out-of-plane curvature was the greatest cause of
skewed flow, which reduced the benefits of the tapered stent-graft.
Key words: endovascular aneurysm repair, stent-graft, tapered graft, computational fluid
dynamics, bifurcation, secondary flow, wall shear stress, out-of-plane curvature
¤ ¤
Abdominal aortic aneurysm (AAA) constitutes traditional open repair and the minimally
a serious health problem in both the US and invasive procedure known as endovascular
Western Europe. If left untreated, these aneurysm repair (EVAR).1,2 The endovascular
typically asymptomatic aneurysms may rup- technique has been widely applied in clinical
ture, possibly leading to death. There are practice, but several problems remain.2–4
currently two surgical treatments for AAA: Chief among these are device migration,
Funding support was provided by an Enterprise Ireland Technology Development Grant (no. CFTD 05 121).
Anthony Callanan, Liam G. Morris, Michael T. Walsh, and Tim M. McGloughlin are named as co-inventors on a patent
application for a novel bifurcated graft for the treatment of abdominal aortic aneurysm (International Publication Number
WO 2006/103641 A1). The other authors have no commercial, proprietary, or financial interest in any products or
companies described in this article.
Address for correspondence and reprints: Tim M. McGloughlin, Professor, Centre for Applied Biomedical Engineering
Research, Materials and Surface Science Institute, University of Limerick, Ireland. E-mail: tim.mcgloughlin@ul.ie

J ENDOVASC THER GEOMETRICAL ENHANCEMENTS FOR AAA GRAFTS 519
2008;15:518–529 Molony et al.
Figure 1 ¤ Conventional (A) and tapered (B) stent-grafts.
endoleaks, and thrombotic occlusion. It has Our group previously investigated the po-
been suggested that re-designing current tential benefits of a newly designed tapered
stent-graft devices may provide a remedy to stent-graft that maintains an area ratio at its
these problems.5,6 inlet equal to that at the bifurcation of the
The design of conventional grafts and device.13 In an in vitro study, the tapered
stent-grafts has normally featured a constant design was associated with a reduction in
proximal diameter with a sudden diameter blood pressure compared to a conventional
reduction at the bifurcation. In the healthy graft.14 Further to this line of research, we
aorta, the geometry tapers smoothly into the now compare this novel tapered stent-graft to
iliac arteries.7 The rapid cross-section area a conventional bifurcated stent-graft using
change with conventional bifurcated devices both ideal uniplanar and realistic out-of-plane
causes the flow to converge, promoting flow model configurations.
separation and recirculation in the iliac limbs
that may be associated with graft limb
thrombosis.8,9 Occlusions are believed to be METHODS
caused by thrombosis as well as kinks, even Idealized Models
though this is not a very common problem.10
Computational fluid dynamics (CFD) has Separate models were created of a conven-
been used to examine stent-graft perfor- tional Dacron stent-graft and a tapered stent-
mance in the past, providing useful informa- graft (Fig. 1). The conventional stent-graft,
tion on flow patterns and forces affecting the whose dimensions were based on previous
device.11,12 Drag forces, which act on stent- work,15 has a sudden change in area at the
grafts as a result of shear and pressure forces bifurcation point (an area ratio of 2:1), while
over the pulse cycle, have been implicated in the tapered stent-graft is characterized by a
graft migration. Drag is dominated by arterial tapering section from the proximal neck into
pressure and is influenced by several factors, the iliac limbs. The novelty of this tapered
such as iliac angle, neck angle, and neck design lies in the blended transition outward
diameter.12 CFD can be a cost-effective meth- as the iliac limbs separate from the trunk,
od for investigating alternative stent-graft which achieves a 1:1 area ratio at the
designs. bifurcation point, i.e., the same area at the
520 GEOMETRICAL ENHANCEMENTS FOR AAA GRAFTS J ENDOVASC THER
Molony et al. 2008;15:518–529
entrance of the stent-graft and at the bifurca-

tion. Using Pro/Engineer software (Parametric
Technology Corporation, Needham, MA,
USA), the tapered stent-graft model was
constructed to match the dimensions of the
conventional stent-graft, such that both had a
24-mm proximal inlet diameter and a 12-mm
distal outlet diameter. Each stent-graft had
the same 30u iliac bifurcation angle.
Realistic Models
Realistic stent-graft models were recon-
structed based on postoperative computed
tomography (CT) scans taken from 3 patients
with different stent-graft geometries. To pre-
pare the CT data for modeling, the digital files Figure 2 ¤ Geometries of the conventional (left)
(in Digital Imaging and Communications in and tapered (right) stent-grafts, respectively, of
Medicine format) containing the cross-section patients A, B, and C.
information were imported into Mimics soft-
ware (Materialise, Belgium) for reconstruc- Computational Fluid Dynamics
tion. The region of interest (stent-graft lumen)
was identified using a thresholding tech- Model construction. All reconstructed mod-
nique. Wall thickness was neglected and els from Pro/Engineer were exported in IGES
assumed constant. Next, 3-dimensional (3D) format to a meshing software (Gambit; Fluent
models were constructed from the cross Europe, Germany). Meshing was achieved
sections. To achieve a more precise repre- using the Cooper algorithm where the origi-
sentation, a smoothing factor based on nal pave mesh at the inlet was swept through
previous studies was applied to provide the volume. Finally, the models were export-
optimum smoothing without an unnecessary ed to Fluent 6.2.16, where the Navier-Stokes
reduction in volume and surface area.16 The equations for fluid flow were solved. The fluid
stent-graft was fitted with a centerline, which was assumed as laminar, incompressible,
was exported along with the cross sections in and Newtonian,15,17 which is an acceptable
Initial Graphics Exchange Specification (IGES) assumption for large diameter arteries. Blood
format for further use. was modeled with a density of 1050 kg/m3
Based on the CT data, 2 realistic models and viscosity of 0.0035 Pa s.18 The graft walls
were created in Pro/Engineer from each were considered rigid, with no slip.19 Velocity
patient dataset. For the conventional stent- and pressure pulses at the inlet and outlets,
graft, a constant cross-section volume was respectively (Fig. 3), resembled those record-
swept along the length of the model until the ed from a typical abdominal aorta over a
bifurcation point, where it split into the iliac cardiac cycle by previous authors who had
limbs. A constant volume was then swept assumed laminar flow for the velocity pro-
through each iliac limb centerline. The cross- file.15,17
section diameter was based on the stent-graft The maximum Reynolds number is given
lumens from the scans. For the tapered stent- by
graft, the same outlet and inlet diameters d
ru
were used with the stent-graft tapered in Re ~ ð1Þ
m
between. The difference in the configurations
can be appreciated in Figure 2; the conven- where r is the density of the fluid, u is the
tional stent-graft splits suddenly into the iliac maximum inlet velocity, d is the inlet diam-
limbs, whereas the tapered graft has a eter, and m the viscosity of the fluid. The value
blended tapering into the bifurcation. for these models ranged from 3600 to 4500
2008;15:518–529 Molony et al.
Figure 3 ¤ Velocity (A) and pressure (B) profiles.
depending on the inlet diameter, which RESULTS

would normally indicate turbulent flow, but
for pulsating flow, turbulence occurs for a Idealized Models
much greater Reynolds number.15 At a steady flow input velocity of 0.4 m/s
The critical Reynolds number for unsteady (approximate to the peak velocity for the
flow is given by Rec 5 constant*a, where the pulse wave), secondary flows were greatly
constant ranges from 250 to 1000 and a is the reduced in the tapered model, with re-circu-
Womersley number. The Womersley number lations downstream eliminated in the axial
is defined as direction of flow (Fig. 5A). Where the conven-
rffiffiffiffiffiffiffi tional stent-graft began to narrow (Fig. 5B,
d vr
a~ ð2Þ slice 1), the rapid area change led to greater
2 m
vortices. Flows in both models showed
where v is the frequency equal to 2p/T, in similar skewing at the bifurcation and
which T is the period of the wave. This gives a 20 mm downstream in the iliac limbs. Using
Womersley number range of 15 to 18.8, which the average across the velocity line plots
is within the normal range of the human (Fig. 6) to calculate the percentage difference
aorta.20 This Womersley number range gives for each model, the conventional stent-graft
a critical Reynolds number between 3750 and showed a greater increase in velocity after the
4700 for the lower constant. The maximum bifurcation than the tapered stent-graft (16%
Reynolds number is within this range, hence versus 6%). Also the conventional model had
laminar flow can be assumed. 18% and 31% greater velocities than the
Grid Independence. Grid size ranged from
535,000 to 850,000 cells for these models, and
wall shear stress (WSS) was used to deter-
mine grid independence. WSS was measured
across the circumference of each model
(Fig. 4). Values were found to be within 5%
of each other. A time step of 0.0025 seconds
was used, which gave 480 time steps per
pulse cycle. Four pulse cycles were required
for an accurate solution. Convergence criteria
were set to 131024 for mass and 131025 for
momentum. Independence was achieved for
both idealized models with grids of ,180,000
hexahedral cells. Time step and pulse cycle Figure 4 ¤ Grid independence for patient B show-
independence were also reached for the ing wall shear stress (WSS) across the circumfer-
realistic models. ence of the model.
Molony et al. 2008;15:518–529
Figure 6 ¤ Axial velocity plots at the bifurcation

and the right iliac limb for the idealized geometry.
Only one iliac limb is shown because the model
was symmetrical.
model. Because blood pressure, inlet diame-

ter, and stent-graft angulation were constant
in these models, the drag force (Fig. 7A)
remained similar throughout for both, but
there was a slightly higher peak force for the
conventional model. Pressure, which was
measured ,10 mm before and after the
bifurcation point, dropped more for the
conventional graft than the tapered design
over the same distance (Fig. 7B).
Realistic Models
In the 3 patient-specific, multiplanar, pulsa-
tile flow models, cross-section axial velocities
Figure 5 ¤ The (A) axial velocity vector profiles were measured at the iliac bifurcation and
(m/s), (B) cross sections showing secondary flows, ,10 mm downstream at 2 time points: sys-
and (C) wall shear stress (WSS) profiles (Pa) for the
tolic peak (t50.305) and maximum decelera-
idealized models of the conventional and
tapered stent-grafts.
tion (t50.47) shown in Figure 3A. In patient A
(Fig. 8), a higher velocity (,10%) was seen in
tapered model above and below the bifurca- the tapered stent-graft than in the conven-
tion, respectively. tional model at the bifurcation (Fig. 8A,D).
Wall shear stress (Fig. 5C) was increased in Downstream, this situation was reversed; in
the iliac limbs for both stent-graft models, but the iliac limbs, the velocity was lower in the
particularly so for the conventional stent- tapered stent-graft (Fig. 8B,C,E,F), with differ-
graft, which also had a larger spatial distribu- ences here varying from 16% to 61%. Notice-
tion of high WSS compared to the tapered ably, there was a very large flow disturbance
2008;15:518–529 Molony et al.
tapered stent-graft than the conventional

model. Skewing of the flow was similar for
both grafts, except for the conventional
model’s right iliac limb in the deceleration
phase, where a highly disturbed flow pattern
was observed. Percentage velocity increases
for the conventional stent-graft over the
tapered design ranged from 18% to 33% at
the bifurcation and between 76% and 102% in
the iliac limbs.
WSS profiles (Fig. 11) taken at maximum
deceleration (t50.47) in patients A, B, and C
demonstrated greater WSS in the iliac limbs
owing to flow convergence and resultant
increased velocity. In patients A and B, the
spatial distribution and magnitude of WSS
were quite similar for both models, but in
patient C, the WSS appeared to be much less
in the tapered stent-graft (the conventional
model had a greater distribution of green
bands).
Both the tapered and conventional stent-
grafts had almost identical drag forces in the
realistic scenario (Fig. 12). For patient C, both
stent-grafts had a peak drag force of 7.9 N at
the pressure pulse peak. Drops in the pres-
sure ,10 mm either side of the bifurcation
were observed (Fig. 13) at the peak of the
Figure 7 ¤ (A) Drag force acting on conventional pressure wave pulse. The tapered stent-graft
and tapered stent-grafts. (B) Pressure drop at the had a more gentle pressure drop in all
bifurcation for conventional and tapered models. patients owing to the equal area ratio, while
the sudden area change at the bifurcation in
in the left iliac limb in the conventional stent- the conventional stent-graft caused the steep-
graft during flow deceleration (Fig. 8E). er drop in pressure.
Patient B (Fig. 9) showed a similar pattern. The volume flow rates (Table) through both
Once again, at both the systolic peak and iliac limbs measured at the outlet of the stent-
deceleration time points, the velocity at the grafts in 3 phases of the pulse (peak systolic,
bifurcation was greater in the tapered graft, maximum systolic deceleration, and begin-
ranging from 18% to 38% higher. In the iliac ning of diastole) were not equally distributed
limbs, flow patterns were quite similar. The due to the out-of-plane curvature of these
flow was similarly skewed, and magnitudes stent-grafts. However, the tapered design
were almost equal, particularly in the right appeared to maintain a more uniform flow
iliac limb. The one exception was the left iliac of blood in most cases, particularly in patients
limb at maximum velocity for the tapered B and C. The greatest disparity occurred
stent-graft, which had a more disturbed during diastole. In patient B, the flow rate at
profile but also a lower velocity compared to the beginning of diastole in the conventional
the conventional stent-graft. For the conven- stent-graft was 3.4 times greater entering the
tional design, the velocity increase over the right limb than the left. In the tapered design,
tapered stent-graft was 36%. the corresponding value was 2.4. Similar
The flow patterns in patient C (Fig. 10) results were seen for the systolic deceleration
differed from the others. At all locations and phase. Flow splitting toward one limb was
time points, the velocity was lower in the reduced during phases of quicker flow; how-
Molony et al. 2008;15:518–529
Figure 8 ¤ Axial velocity plots (patient A) for the (A) bifurcation at peak systole, (B) left iliac
limb at peak systole, (C) right iliac limb at peak systole, (D) bifurcation at maximum
deceleration, (E) left iliac limb at maximum deceleration, and (F) right iliac limb at
maximum deceleration.
ever, there was a more even flow split in the tional stent-graft. There was a large velocity
conventional stent-graft in patients A and C at difference in the iliac limbs for patient C for
the systolic peak. the conventional stent-graft over the tapered
design.
The tapered stent-graft does not appear to
DISCUSSION
significantly reduce skewing of the flow in
Our first impression from the uniplanar realistic models. Because both stent-grafts
models was that the tapered stent-graft could follow the same deployment path, the out-of-
demonstrate reduced velocities and vortices plane curvature skews the flow against the
due to the smooth trunk-limb transition. graft wall, thus the natural curvature of the
However, the velocity plots in the patient- stent-graft tends to dictate the flow patterns. In
specific (multiplanar) models were more the idealized scenario, secondary flow and
mixed. In patients A and B, the velocity was skewing was created by the sudden cross-
actually greater in the tapered stent-graft than section change in the conventional stent-graft,
the conventional design just prior to the which the tapered stent-graft eliminated. It
bifurcation. This situation was reversed in appears that the multiplanar geometry or
the iliac limbs, as the large area change in the curvature of the graft has a greater influence
conventional stent-graft caused an increase in on flow skewness and the creation of vortices
velocity. In the third patient, the velocity was than graft design. Chong and How8 saw these
reduced at all locations and times in the types of flow disturbances in the iliac limbs of
tapered stent-graft compared to the conven- their in vitro model, which they surmised could
2008;15:518–529 Molony et al.
Figure 9 ¤ Axial velocity plots (patient B) for the (A) bifurcation at peak systole, (B) left iliac
explain why certain grafts fail at this location It was expected that the drag force acting
due to thrombosis. Overall, the profiles are on both types of stent-grafts would be similar
much more disturbed in the deceleration given their identical inlet diameter and flow
phase than at maximum velocity, which agrees path, which affect the drag on the graft.12
with previous work on aortic pulse profiles.21 However, the tapered stent-graft showed a
While there were pockets of similar high 4% reduction in drag over the conventional
wall shear stress in both types of stent-grafts stent-graft in the uniplanar models. This small
(patients A and B), the overall distribution of difference could aid in preventing stent-graft
high WSS appeared to be greater in the migration. Part of this decrease may be
conventional stent-graft (patient C). This high explained by the reduction in velocity in the
WSS was concentrated almost completely in tapered stent-graft, which aids in reducing
the iliac limbs for the conventional stent-graft, viscous forces acting on the graft. In the
while there was a more even spatial distribu- multiplanar situation, drag force was un-
tion in the tapered design. High shear condi- changed between both stent-grafts.
tions may overstimulate platelet-induced The tapered stent-graft had a much gentler
thrombosis,22 so the more widespread high pressure transition from inlet to outlet, which
WSS in the conventional stent-graft, suggest- was most significant at the bifurcation in the
ed by its higher velocities, may increase the conventional stent-graft, where the sudden
risk of occlusion. Once again, the out-of-plane area change caused a large pressure drop.
curvature reduced the benefits of the tapered This also created higher overall pressures in
stent-graft in reducing WSS. the conventional stent-graft.
Molony et al. 2008;15:518–529
Figure 10 ¤ Axial velocity plots (patient C) for the (A) bifurcation at peak systole, (B) left iliac
The lower flow rate in the conventional would have specified the flow rate through
stent-graft during diastole may make it more each limb or used impedance at these
prone to distal disease formation. Patient A outlets.7 Even with impedance outlets, differ-
had almost the same flow in both iliac limbs ing limb flow rates have been seen due to
for both stent-graft models, likely due to the patient geometry.24
fact that both limbs are quite straight. Con- Other limitations of this work are the
versely, patient B had differing flow rates in assumption that the stent-graft is a rigid
the iliac limbs, but the limbs in this patient material. This and the presence of an aneu-
were severely twisted. The tapered stent-graft rysm were neglected even though they may
reduced this flow disparity here and in the influence results.11 Blood was assumed to be
other patients, most likely due to its blended a Newtonian fluid, which is a common
bifurcation. Previous researchers have also assumption, but blood tends to exhibit non-
noted differing flow rates in the iliac Newtonian behavior at shear rates below
limbs.19,23 A word of caution should be given 100 s21. However, Shipkowitz et al.7 did not
since these observations may be due to the find the non-Newtonian characteristics of
boundary condition; in physiological flow, the blood to be an important factor in general
split at the iliac limbs may be weighed more flow field parameters. Leuprecht and Perk-
evenly. Pressure boundary conditions mean told25 suggested that it may be necessary to
that the flow split is dictated solely by the model non-Newtonian flow only in patholog-
resistance to flow due to the geometry of the ically altered configurations such as aneu-
model, neglecting the demands of down- rysms. In our study, the stent-graft resembled
stream arteries.24 A more accurate simulation the contours of the natural aorta and was
2008;15:518–529 Molony et al.
Figure 11 ¤ Wall shear stress (Pa) profile of

patients A, B, and C with conventional (left) and
tapered (right) stent-grafts.
Figure 13 ¤ Pressure results for patients A, B,
large enough for a Newtonian model to be and C.
used.
Gawenda et al.26 have shown that endo- compliancy study of aortic stent-grafts,13 our
vascular AAA repair may increase blood group found that a compliant tapered graft
pressure; hence, any stent-graft that reduces will reduce pressure by 11 mmHg over a
blood pressure would be of great benefit. In a compliant conventional design. Both grafts
in this study were made of the same material,
which eliminated any pressure differences
caused by compliance differences; thus, the
pressure differences can be explained solely
by the geometrical differences in devices. The
reduction in pressure produced by the ta-
pered graft created a potential pressure
benefit of 5% to 10%. The incorporation of a
taper more closely mimics the real aorta-to-
iliac area transition, whereas the conventional
graft area ratio creates a throttle effect,
resulting in increased pressure.
Conclusion
Based on CFD analysis, the novel tapered
Figure 12 ¤ Drag force acting on conventional stent-graft improved graft hemodynamics
and tapered stent-grafts for patient C. compared to the conventional design. In
Molony et al. 2008;15:518–529
¤ ¤
TABLE
Volume Flow Rates in Iliac Limbs
Flow Rates, L/s
Patient A Patient B Patient C
C T C T C T
Systolic Peak
Left 0.182 0.179 0.105 0.105 0.102 0.104
Right 0.187 0.188 0.123 0.123 0.096 0.092
Diastole
Left 0.043 0.046 0.014 0.018 0.028 0.028
Right 0.055 0.051 0.048 0.044 0.019 0.019
Systolic Deceleration
Left 0.029 0.029 0.031 0.027 0.021 0.018
Right 0.038 0.038 0.011 0.015 0.015 0.018
¤ ¤
C: conventional, T: tapered.
idealized models, the tapered stent-graft re- 3. Parodi JC, Palmaz JC, Barone HD. Transfemo-
duced velocity and shear stress impinging on ral intraluminal graft implantation for treat-
the graft wall. Drag force was also slightly ment of abdominal aortic aneurysms. Ann
reduced in the ideal case. In the realistic Vasc Surg. 1991;5:491–499.
4. Egelhoff CJ, Buduig RS, Elger DF, et al. Model
scenario, the out-of-plane curvature of the
studies of the flow in abdominal aortic aneu-
aneurysms caused most flow disturbances,
rysms during resting and exercise conditions.
which reduced the positive impact of the J Biomech. 1999;32:1319–1329.
tapered stent-graft. That is to say, in patients 5. Torella F. Effect of improved endograft design
with more irregular geometry, some of the on outcome of endovascular aneurysm repair.
benefits from the tapered design are dimin- J Vasc Surg. 2004;40:216–221.
ished or obviated. A study of more patient- 6. Chuter TA. Stent-graft design: the good, the
specific models is necessary to fully ascertain bad and the ugly. Cardiovasc Surg. 2002;10:
the benefits of the tapered stent-graft in vivo. 7–13.
In vitro, a tapered graft can reduce blood 7. Shipkowitz T, Rodgers V, Frazin L, et al.
pressure over a conventional graft with a 2:1 Numerical study on the effect of steady axial
bifurcation area ratio. Based on these obser- flow development in the human aorta on local
shear stresses in abdominal aortic branches.
vations, the tapered graft may have a clinical
J Biomech. 1998;31:995–1007.
application in both open and endovascular 8. Chong CK, How TV. Flow patterns in an
aortic repairs, particularly when the graft will endovascular stent-graft for abdominal aortic
undergo very little bending or twisting. aneurysm repair. J Biomech. 2004;37:89–97.
9. Jacobowitz GR, Lee AM, Riles TS. Immediate
Acknowledgements: The authors extend their thanks to and late explantation of endovascular aortic
David A. Vorp, PhD, the Center for Vascular Remodeling grafts: the EndoVascular Technologies experi-
and Regeneration, and to Michel S. Makaroun, MD, ence. J Vasc Surg. 1999;29:309–316.
Department of Surgery, University of Pittsburgh, Penn- 10. Corbett TJ, Callanan A, Morris LG, et al. A
sylvania, USA, for providing the CT data. review of the in vivo and in vitro biomechanical
behavior and performance of postoperative
REFERENCES abdominal aortic aneurysms and implanted
stent-grafts. J Endovasc Ther. 2008;15:468–
1. Sakalihasen N, Limet R, Defawe OD. Abdominal 484.
aortic aneurysm. Lancet. 2005;365:1577–1589. 11. Li Z, Kleinstreuer C. Blood flow and structure
2. Kamineni R, Heuser R. Abdominal aortic aneu- interactions in stented abdominal aortic aneu-
rysm: a review of endoluminal treatment. rysm model. Med Eng Phys. 2005;27:369–
J Interv Cardiol. 2004;17:437–445. 382.
2008;15:518–529 Molony et al.
12. Howell BA, Kim T, Cheer A, et al. Computa- 19. Morris L, Delassus P, Grace P, et al. Effects of
tional fluid dynamics within bifurcated abdom- flat, parabolic and realistic steady flow inlet
inal aortic stent-grafts. J Endovasc Ther. profiles on idealised and realistic stent graft fits
2007;14:138–143. through abdominal aortic aneurysm (AAA).
13. Morris L, McGloughlin T, Dellasus P, et al. A Med Eng Phys. 2006;28:19–26.
vascular graft. International Patent WO 2006/ 20. Artoli AM, Hoekstra AG, Sloot PM. Mesoscopic
103641 A1. November 5, 2006. simulations of systolic blood flow in human
14. O’Brien T, Morris L, McGloughlin T. Evidence abdominal aorta. J Biomech. 2006;39:873–
suggests rigid aortic graft increase systolic 884.
blood pressure: results of a preliminary study. 21. Morris L, Delassus P, Callanan A, et al. 3-D
Med Eng Phys. 2008;30:109–115. numerical simulation of blood flow through
15. Morris L, Delassus P, Walsh M, et al. A models of the human aorta. J Biomech Eng.
mathematical model to predict in vivo pulsatile 2005;127:767–775.
drag forces acting on bifurcated stent grafts 22. Ku DN. Blood flow in arteries. Annu Rev Fluid
used in endovascular treatment of abdominal Mech. 1997;29:399–434.
aortic aneurysms (AAA). J Biomech. 2004;37: 23. Frauenfelder T, Lotfey M, Boehm T, et al.
1087–1095. Computational fluid dynamics: hemodynamic
16. Doyle B, Morris L, Callanan A, et al. 3D changes in abdominal aortic aneurysm after
reconstruction of patient-specific abdominal stent-graft implantation. Cardiovasc Intervent
aortic aneurysms: from CT scan to silicone Radiol. 2006;29:613–623.
model. Presented at the American Society of 24. Vignon-Clementel IE, Figueroa CA, Jansen KE,
Mechanical Engineers (ASME) Summer Bioen- et al. Outflow boundary conditions for three-
gineering Conference; Keystone, Colorado, dimensional finite element modeling of blood
USA; June 20–24, 2007. flow and pressure in arteries. Comput Methods
17. Di Martino ES, Guadagni G, Fumero A, et al. Appl Mech Eng. 2006;195:3776–3796.
Fluid-structure interaction within realistic 25. Leuprecht A, Perktold K. Non-Newtonian ef-
three-dimensional models of the aneurysmatic fects on blood flow in large arteries. Comp
aorta as a guidance to assess the risk of Methods Biomech Biomed Eng. 2000;4:149–
rupture of the aneurysm. Med Eng Phys. 163.
2001;23:647–655. 26. Gawenda M, Jaschke G, Winter S, et al.
18. Scotti CM, Finol EA. Compliant biomechanics Endotension as a result of pressure transmis-
of abdominal aortic aneurysms: a fluid-struc- sion over the graft following endovascular
ture interaction study. Comput Struct. 2007;85: aneurysm repair - an in vitro study. Eur J Vasc
1097–1113. Endovasc Surg. 2003;26:501–505.
J ENDOVASC THER
2008;15:000–000 0
¤ENDOVASCULAR THERAPY REVIEWS ¤
A Review of the In Vivo and In Vitro Biomechanical

Behavior and Performance of Postoperative Abdominal
Aortic Aneurysms and Implanted Stent-Grafts
Timothy J. Corbett, BEng1; Anthony Callanan, BEng1; Liam G. Morris, PhD2;
Barry J. Doyle, BEng1; Pierce A. Grace, MCh, FRCSI3; Eamon G. Kavanagh, MD, FRCSI3;
and Tim M. McGloughlin, PhD1
1Centre for Applied Biomedical Engineering Research, MSSI, Department of Mechanical
and Aeronautical Engineering, University of Limerick, Ireland. 2Department of Mechan-
ical and Industrial Engineering, Galway and Mayo Institute of Technology, Galway,
Ireland. 3Department of Vascular Surgery, Mid-Western Regional Hospital, Dooradoyle,
Limerick, Ireland.
¤ ¤
Endovascular repair of abdominal aortic aneurysms has generated widespread interest
since the procedure was first introduced two decades ago. It is frequently performed in
patients who suffer from substantial comorbidities that may render them unsuitable for
traditional open surgical repair. Although this minimally invasive technique substantially
reduces operative risk, recovery time, and anesthesia usage in these patients, the
endovascular method has been prone to a number of failure mechanisms not encountered
with the open surgical method. Based on long-term results of second- and third-generation
devices that are currently becoming available, this study sought to identify the most
serious failure mechanisms, which may have a starting point in the morphological changes
in the aneurysm and stent-graft. To investigate the ‘‘behavior’’ of the aneurysm after stent-
graft repair, i.e., how its length, angulation, and diameter change, we utilized state-of-the-
art ex vivo methods, which researchers worldwide are now using to recreate these failure
modes.
Key words: abdominal aortic aneurysm, endovascular aneurysm repair, stent-graft
postoperative performance
¤ ¤
Abdominal aortic aneurysm (AAA) is defined as accounting for 10,000 deaths annually in the
an abnormal localized dilation or bulge in the United Kingdom.1–3 Survival after ruptured
aorta greater than 50% of its normal diameter.1 AAA is in the region of 10% to 20%. In the case
It affects up to 5% of the population over the age of small aneurysms (diameter ,5.5 cm), the
of 55 years. Rupture of an AAA is responsible usual treatment is careful and frequent obser-
for 1.3% of deaths in men aged 65 to 85 years, vation. However, if the aneurysm diameter
This work is funded by a grant from the Irish Research Council for Science, Engineering and Technology (Grant #RS/2005/
27/Timothy Corbett and Grant #RS/2005/340/Barry Doyle) and by a Technology Development Grant from Enterprise
Ireland.
Tim McGloughlin, Anthony Callanan, and Liam Morris are named as co-inventors on a patent application for a novel
bifurcated graft for the treatment of AAA (International Publication Number WO 2006/103641 A1). The other authors have
no commercial, proprietary, or financial interest in any products or companies described in this article.
Address for correspondence and reprints: Professor Tim McGloughlin, MSG-014, University of Limerick, Ireland. E-mail:
tim.mcgloughlin@ul.ie

0 POSTOPERATIVE PERFORMANCE OF STENT-GRAFTS J ENDOVASC THER
Corbett et al. 2008;15:000–000
exceeds 5.5 cm or is rapidly growing, aneurysm vascular, endovascular, biomedical engineer-

repair is usually recommended.1 ing, and science material on the topic. Articles
The gold standard in AAA repair remains published before January 2000 were excluded
the open surgical method. The long-term unless they were heavily cited (Science
durability of open repair (OR) is excellent, Direct) or unless the work contained in an
with little need for aneurysm-related reinter- article was not repeated or further developed
vention.4,5 However, open repair is a major after January 2000. The search terms used
surgical procedure and involves prolonged were abdominal aortic aneurysm, endovas-
cross clamping of the aorta, which poses the cular repair, EVAR, stent-graft, and aneurysm
risk of serious operative complications.1,6 rupture. Using these keywords, 4472 articles
Therefore, it is considered high risk if per- were identified in the Elsevier and Science
formed in patients with excessive comorbid- Direct databases. The abstract of each surgi-
ities, such as symptomatic congestive heart cal article was read by the authors, who
failure, valvular heart disease, cardiac ar- selected articles that contained information
rhythmia, obstructive pulmonary disease, on one or more of the following headings:
chronic renal failure, or high anesthetic risk.6 device failures, reintervention procedures,
Endovascular aneurysm repair (EVAR) was aneurysm/stent-graft morphology, and stent-
introduced in the late 1980s.7 It has evolved graft type. Engineering and basic science
since then to become the preferred treatment articles were classified as experimental, com-
in older and unfit patients. This technique is putational, or mathematical modeling. All
minimally invasive and involves inserting a relevant articles were downloaded in full.
collapsed graft through a femoral artery Key citations pertinent to the topic in these
access and over a guidewire to the aneurysm selected articles were investigated, and a
location. Here, the stent-graft is allowed to further 12 online databases were used to
expand to form a new conduit for blood flow identify a total of 148 full-text papers for
excluding the aneurysm sac. Due to the critical review.
minimally invasive nature of this procedure,
operation time and blood loss are reduced,
and the need for cross clamping of the aorta
IN VIVO BEHAVIOR OF
is eliminated.1,5,8–14 Early mortality, length of
STENT-GRAFTS AND
hospital stay, and recovery time are lower
EXCLUDED ANEURYSMS
after EVAR than after OR.6,10,15 After EVAR, most aneurysms tend to show a
Reservations have been expressed about decrease in maximum diameter, cross-sec-
the long-term ability of endovascular grafts to tional area, AAA volume, and thrombus
prevent aneurysm rupture.16,17 Furthermore, volume.25,26 The neck angle of the aneurysm
problems such as endoleak, graft migration, can change dramatically after the proce-
and fabric tears, which are not seen after dure.27 Sternbergh et al.27 found that preop-
open repair, have been widely document- erative neck angle was associated with post-
ed.18–24 As long-term results of second- and operative adverse events, so it is reasonable
third-generation devices are currently becom- to assume that postoperative changes in neck
ing available, this study sought to identify the angulations could also cause these adverse
most serious failure mechanisms and their events. White et al.26 reported neck angle
frequency. changes of 5u to 25u after EVAR, which could
alter the drag forces acting on the graft and be
partly responsible for migration of a stent-
LITERATURE REVIEW graft.
The reference databases of all the authors The wall of the excluded aneurysm thins
were searched initially for relevant articles on after EVAR, possibly due to a continuation of
EVAR, then the Elsevier and Science Direct the disease process or to enzymatic activity in
electronic databases were searched for the the thrombus within the sac.28 In an open
period January 2000 to January 2008 to repair, this thrombus is removed, and so the
identify English-language articles containing wall does not thin.28 In the event of an

J ENDOVASC THER POSTOPERATIVE PERFORMANCE OF STENT-GRAFTS 0
2008;15:000–000 Corbett et al.
endoleak (blood flow external to the implant- repercussions for the endovascular approach
ed graft, but within the aneurysm), this because a subset of patients may experience
thinned wall could be weakened enough to inadequate long-term proximal seal regard-
rupture even at pressures below the systemic less of advances in available stent-grafts.
blood pressure. Singh-Ranger and Adise-
shiah29 also reported that the length of the
aneurysm increases after EVAR, possibly due
COMPLICATIONS OF EVAR
to straightening of tortuous arteries after Migration and Type I Endoleak
stent-graft implantation. Contradicting this
assumption, Resch et al.30 found no increase Migration poses a serious threat to the
in renal artery-to-bifurcation distance after sealing zones between the stent-graft and
EVAR. Increase in the length of the aneurysm the artery. If it develops to a point where the
could cause displacement forces at the prox- seal fails, allowing blood at systemic pressure
imal and distal attachment zones, precipitat- into the sac (type I endoleak), the aneurysm is
ing delayed failures in these areas. Further- at high risk for rupture.33 Late (.1 year after
more, it could cause separation between implantation) migration of stent-grafts,38–41
modular components in some stent-grafts. which is the main reason for postoperative
Singh-Ranger and Adiseshiah29 found that reinterventions,42,43 is frequently seen when
aneurysm morphology changes depended the dilating proximal neck exceeds the max-
on the type of stent-graft used, and Kramer imum diameter of the proximal stent-graft.
et al.31 saw no relationship between the Sternbergh et al.44 reported that excessive
original aneurysm morphology and the oversizing of the stent-graft increases the risk
long-term ‘‘attitude’’ of the stent-graft. Vos of migration 14 fold. The most frequent type
et al.32 observed an increased longitudinal of migration is distal movement of the stent-
movement of the AAA sac after EVAR; they graft at the proximal fixation site; however,
concluded that this was due to lack of proximal movement of the distal limbs in the
compliance, angulations, and tapering in the iliac arteries has also been reported.45–48
stent-graft, which also proved that there were Differing approaches have been used to
downward displacement forces at the proxi- prevent migration of stent-grafts.49 Early
mal fixation site. devices, such as the Ancure, used hooks to
One of the most important morphological penetrate the internal layers of the aorta and
changes after EVAR is the dilation of the provide protection from migration. The Ze-
proximal aortic neck. Stent-grafts are over- nith and the Excluder stent-grafts currently on
sized by 10% to 20% to ensure adequate the market use barbs or anchors, respectively,
radial force to make a seal against the aorta.33 which work in the same manner. The barbs of
Due to this radial force, the diameter of the the Zenith are placed at different levels on a
neck increases postoperatively.26 However, bare stent that is placed in the suprarenal
continued dilatation has been reported in a aorta, while the proximal edge of the Excluder
number of studies.26,29,30,34,35 Palombo et has a scalloped sealing cuff in which the
al.35 concluded that long-term dilatation of nitinol anchors are embedded.
the proximal neck was mainly due to the The relatively stiff and fully stented AneuRx
continual radial force of the stent-graft acting and Powerlink devices rely on a combination
on the artery wall. However, continued neck of radial force from the stent along with
enlargment has also been reported in open columnar support from the metal frame to
surgical repair in the absence of radial force prevent migration. The devices differ in that
from a stent.35 A number of studies have the Powerlink sits on the aortic bifurcation,
reported that continued dilatation occurs only while the AneuRx bifurcates closer to the top
in a subset of patients after EVAR,26,30,36 so of the device. It is proposed that sitting the
neck enlargement might be dependent on Powerlink device on the aortic bifurcation
patient-specific factors, such as continuation further increases columnar support of this
of the disease process, as Schlensak et al.37 device.8,50 The Talent device has a bare
hypothesized. If this is the case, it has major ‘‘spring’’ stent that flares out at the proximal

Corbett et al. 2008;15:000–000
end of the device to provide fixation based on are also important factors in preventing
radial force. All the devices with the exception migration. As the iliac bifurcation is relatively
of the Powerlink device, which possesses a fixed in place in the pelvis, this would seem
completely internal metal frame, achieve a especially relevant for devices that rely on
seal using the radial force of the stent columnar support to prevent migration.61–63 A
pressing into the artery wall51; however, lower incidence of migration has been report-
migration has been reported with all of these ed with balloon-expandable stent-based
models.52 grafts, which purportedly break the internal
Another possible reason for late stent-graft elastic lamina causing greater incorporation
migration is that healing between the aorta than with self-expanding stent-based
and device (incorporation) does not happen grafts.54,64 However, the study by Malas et
even after a long period of time.41,48,53–56 al.54 was performed using aortomonoiliac
Resch et al.57 performed distraction experi- stent-grafts, which may be subjected to 40%
ments using stent-grafts placed in cadaver greater drag forces than bifurcated stent-
aortas. They reported that a hand-sewn grafts in vivo and are therefore more prone
anastomosis used in OR could withstand a to migration than bifurcated devices.65 Thus,
force of 150 Newtons (N) compared to a in the presence of these forces, it is notable
median 24 N for the Zenith stent-graft and that no migration was recorded. The promis-
only 9 N for the Vanguard stent-graft. While ing outcomes from these trials underline the
the Vanguard has been withdrawn from the case for further research and development of
market, its proximal attachment is somewhat stent-grafts based on balloon-expandable
like the AneuRx and Powerlink devices. stents.
Using a stent-graft somewhat like the Stent-grafts have also been developed with
Excluder, Malina et al.58 reported that grafts fenestrations for the renal and visceral arter-
with hooks and barbs had a median distrac- ies. The graft fabric of these grafts can be
tion force of 7.8 N. Veerapen et al.59 measured placed in the suprarenal aorta, which should
dislodgement forces for the Talent and Ex- improve migration resistance and reduce
cluder devices at 8 N and 6.5 N, respectively. type I endoleak due to a greater sealing
Using a mathematical model with a high length in a healthier section of aorta.66
bifurcation angle (80u), Morris et al.60 found Additional stents are often deployed through
that the drag force on a bifurcated graft could the fenestrations into the target arteries,
be as high as 14N, which would seem to put aiding migration resistance.67 Stent-grafts
most grafts at very high risk of migration. that bifurcate at the iliac bifurcation could be
Using a more modest angle of 40u and a neck useful in AAAs where adequate distal fixation
diameter of 28 mm, they found that the drag would normally require covering and occlud-
force could be almost 8 N, which would place ing the internal iliac artery.68 Interestingly,
all but the Zenith device in danger of type II endoleak may be a protective factor
migration. However, the experiments carried against migration due to a lower pressure
out by Resch et al.57 and Malina et al.58 did gradient between the stent graft and the
not consider the columnar support of the sac.49
devices, so the distraction forces for some
stent-grafts may be underestimated.
Type II Endoleak
Correct placement of the device is crucial to
prevent migration. Zarins et al.41 reported Leakage of blood from patent side branches
that each millimeter increase from the lowest into the aneurysm sac after EVAR occurs in
renal artery to the top of the AneuRx stent- 10% to 25% of cases.24 This type II endoleak
graft increased the risk of migration by 5.8%, may occur with one or more patent branches
confirming that adequate proximal fixation is and can establish an inflow/outflow circuit in
vital to achieving successful aneurysm exclu- the aneurysm sac.24 Unlike other complica-
sion. tions of EVAR where intervention must be
Iliac fixation length and the distance from aggressive, there is no consensus on when to
the iliac bifurcation to the ends of the device intervene in type II endoleaks,24,69,70 primarily

2008;15:000–000 Corbett et al.
because a good percentage of them (30% to cause a graft tear, leading to type III endo-
100%) seal spontaneously, most likely due to leak,82 or they could perforate the aorta.
the development of intraluminal thrombus in Ghanim et al.86 reported a very serious case
post-EVAR aneurysms.24,69 Aneurysm sac of suture breakage in a Zenith device causing
pressure can be increased due to type II complete separation of the uncovered proxi-
endoleak in vivo, and aneurysm rupture has mal stent with a type I endoleak. Barb and
been reported in the setting of type II hook failures at the proximal fixation sites
endoleak.71,72,73 have also been reported, which may put
Intervention is usually required if an AAA stent-grafts at risk of migration. The response
with type II endoleak enlarges or does not to these fatigue failures has been to increase
contract in the long term.18 If a type II the number of sutures or barbs to afford
endoleak is to spontaneously seal, it will greater redundancy to the designs. Compo-
usually do so within 12 months after EVAR.47 nent separation leading to type III endoleak
The strategy for persistent type II endoleak has been reported with modular devices and
that does not cause sac enlargement remains puts the aneurysm at very serious risk of
the subject of debate.47,74,75 Treatment of type rupture.5,9,87
II endoleak is usually by coil embolization of
the patent arteries.24,46,74,76 Successful embo-
Kinks, Thrombosis, and Limb Occlusions
lization of the patent branches in an AAA
with an unsealed type II endoleak has been Occlusion of stent-graft limbs due to kinks
successful in reducing sac pressure and the or thrombosis is a relatively infrequent but
risk of rupture.72 nonetheless important problem that occurs
The Powerlink stent-graft is designed so as a delayed event in 2.2% to 3.9% of
that the graft material ‘‘balloons’’ off the patients.8,11,45,88 Hobo et al.88 found that the
endoskeleton in concert with pulsatile blood use of iliac extensions during the endovas-
flow, supposedly reducing the incidence of cular procedure was related to the develop-
type II endoleak by filling more of the volume ment of kinks in the stent-graft. These exten-
of the aneurysm sac, thereby sealing any sions, which may be applied in ~23% of
patent branches.8,50 Initial results did indeed patients,89 are usually used in iliac arteries
show lower rates of type II endoleak than with that are aneurysmal or have suboptimal
some other grafts; however, the rate of type II landing zones. It would seem reasonable to
endoleak seems to have normalized in further conclude that preventing stent-graft kinks
studies.8,50 Uflacker et al.77 do not agree with relies on satisfactory distal fixation. It is likely
the hypothesis of reduced type II endoleak that patients who have iliac aneurysms are
due to ballooning of the graft; they propose more vulnerable to stent-graft kinks than
that unsupported grafts are more prone to patients with healthy iliac arteries.
type II endoleak due to changes in sac volume Morphological changes in the aneurysm
and increased flexibility of the graft wall. that cause geometric changes in the stent-
graft are another major cause of stent-graft
kinks.31 Blood flow through a kinked stent-
Graft Defects and Component Separation
graft may be irregular and could lead to
Defects, such as graft fabric tears, suture localized thrombosis and stent-graft occlu-
breaks, and stent/hook/barb fractures, have sion. Hypothetically, kinks or occlusions that
been widely reported5,53,56,78–81 and are most inhibit the flow of blood through the stent-
likely due to fatigue loading from pulsatile graft substantially increase drag force, possi-
blood flow. Zarins et al.82 reported stent bly causing migration of the graft. Fransen et
fractures and suture breaks in the AneuRx al.87 reported that kinks were significantly
stent-graft (Fig. 2), but these defects have associated with type I endoleak, but they were
also been reported in the body of Zenith, unable to pinpoint whether the migration and
Talent, and Excluder devices.44,46,78,84,85 The subsequent failure of the proximal seal was a
movement of these unconstrained metallic result of the kinks or the kinks were a result of
components against the graft fabric could the migrated graft.

Corbett et al. 2008;15:000–000
Endotension dated in vivo and in vitro; it is repeatable and

could be an invaluable tool in deciding the
This term refers to continued pressurization
best course of action for aneurysms that
and expansion of the aneurysm sac after EVAR
remain unchanged in diameter or expand
when no endoleak can be identified on com-
after EVAR.94
puted tomography (CT). Endotension, which
Recently, the FDA has approved a wireless
occurs in up to 15% of patients,89 can cause
AAA pressure sensor that can monitor sac
significant increase in aneurysm diameter and
pressures over the life of the stent-graft. This
increase the risk of rupture. Endotension
device may produce completely noninvasive
usually requires secondary intervention.89
measurements similar to Sonesson et al.93
Several theories have been put forward to
and Dias et al.72 and could lead to a more in-
explain endotension, including transfer of
depth understanding of the endotension
pressure through a thrombotic seal at the
enigma, facilitating corrective action prior to
attachment sites, very low flow endoleak, and
the aneurysm being at risk of rupture.95 The
graft fabric tears sealed with thrombus. Lin et
two methods measure pressure locally and
al.90 stated that aneurysm enlargement can-
then assume it as the global pressure inside
not occur without expansion of intrasac the aneurysm sac. Unfortunately, pressure
volume and hypothesized that this in turn distribution in the thrombus within the ex-
means continued thrombus generation from cluded aneurysm sac may not be uniform,96
endoleak, supporting the low flow theory. which could lead to the underestimating the
Another possible explanation is that serous intrasac pressure and the rupture risk of
blood components could transude through certain aneurysms. These pressure differenc-
the walls of the stent-grafts and be responsi- es are most likely due to the microstructure of
ble for this phenomenon.28,54,82,91 Endoten- the intraluminal thrombus, which is dis-
sion is more frequently associated with stent- cussed in detail further on.
grafts that are fabricated from expanded
polytetrafluoroethylene (ePTFE).92
Until recently, noninvasive monitoring of ANALYSIS OF ANEURYSM WALL
excluded aneurysm sac pressures had not AND STENT-GRAFT FAILURES
been possible. Sonesson et al.93 described a
Experimental Analysis
catheter-based method that could identify
aneurysms subjected to high pressure when In vitro AAA flow circuits have been developed
no endoleak is visible on the completion by a number of authors to investigate flow in
angiogram. Using this method, they conclud- aneurysms and stent-grafts in a laboratory
ed that successful EVAR reduced sac pressure environment. Many studies focus on preinter-
to 20% of systemic pressure.93 They also vention wall stresses, which may not be rele-
utilized this method to investigate the differ- vant when a functioning stent-graft is in situ.
ence in sac pressures in shrinking, un- However, if there is a high pressure endoleak,
changed, and expanding aneurysms.72 In the sac may be prone to systemic pressure, in
patients without endoleaks, the investigators which case, these studies become relevant.
found that the mean pressure index (MPI: Morris et al.97 used the photoelastic meth-
percentage of mean aneurysm pressure rela- od of stress analysis to identify high-stress
tive to mean systemic pressure) for shrinking sites on the artery wall of an ideal aneurysm
aneurysms was 19% versus 59% for expand- without a stent-graft. They found that the site
ing aneurysms. In aneurysms that did not of the largest diameter was in a state of low
change in size, they identified 2 groups: stress and that the highest stresses were
patients with an MPI similar to the expand- located close to inflection points on the
ing-aneurysm group, who subsequently aneurysm. The AAA model used in this study
showed expansion of their aneurysm, and had a 2-mm wall thickness and an idealized
patients with lower MPIs, who displayed geometry taken from EUROSTAR data.
shrinking aneurysms.72 This method of aneu- Realistic models of AAAs have also been
rysm pressure measurement has been vali- developed.98,99 Chong et al.100 used a pulsa-

2008;15:000–000 Corbett et al.
tile flow circuit and a symmetrical AAA model Albertini et al.109 found that greater neck
to investigate the flow patterns in an implant- angles increased the rate of type I endoleak.
ed stent-graft by means of laser Doppler Moreover, in angulated necks, the longitudi-
anemometry. Using endo- and exostent con- nal stiffness of the proximal stents may lift the
figurations, they found that the stent struts of graft from the artery wall, creating gaps
endostent devices caused local flow alter- between the stent-graft and the artery. It has
ations. More recently, they used a modified also been reported that the anterior abdom-
AAA model with an out-of-plane neck having inal wall undergoes strains 4 times greater
an angle of 60u; they concluded that the flow than the posterior wall. It is possible that this
patterns in the stent-graft were complex and nonuniform straining of the aortic wall could
influenced by the geometries of the stent- lead to similar attachment issues for implant-
graft and iliac vessels. They identified loca- ed stent-grafts.110
tions of flow separation, low velocity, and Investigating dislocation of components
stagnation where thrombus could possibly ofmodular stent grafts, Liffman et al.111
form. foundthat components were most likely to
Types I and II endoleaks were also investi- separate when the pressure difference be-
gated. A stent-graft clotted with gelatin was tween the stent-graft and aneurysm sac
used in an ideal AAA model. Negligible was large, that is, when there was no endo-
sac pressures were noted with no endoleak, leak.
while systemic pressures were found when Volodos et al.,112 who studied the factors
atype I endoleak was simulated. In the case of affecting the displacement force acting on a
a type II endoleak with no outflow from the tube stent-graft, concluded that displacement
sac, they found that that sac pressure ap- force increased as sac pressure rose. Small
proached mean aortic pressure; as outflow increases in systemic blood pressure could
from the sac increased, the sac pressure markedly increase the risk of migration. In a
decreased.101–103 later study,113 the authors found that a
Mehta et al.104 used an in vitro model to retrograde displacement force exists on the
investigate whether pulsatile aneurysms were distal end of the stent-graft, which could
an indication of high pressure after EVAR. cause proximal migration and kinking. They
They found that an aneurysm may not be also determined that PTFE grafts increase in
pulsatile even when a type III endoleak is length under internal pressure, which could
present and the aneurysm sac is subjected to contribute to kink formation.
high pressure. They determined that pulsati- Arko et al.114 investigated the role of iliac
lity depends only on outflow from the sac fixation length in preventing migration using
through patent side branches. Gawenda et an in vivo ovine model. In their study, stent-
al.105,106 reported that sac pressure depends grafts with greater iliac fixation length
on the thickness and stiffness of the stent- (31 mm) could tolerate 67% higher displace-
graft itself and the compliance of the aneu- ment forces over stent-grafts with minimal
rysm wall. Latex aneurysm models with iliac fixation length (11 mm).
thicker Dacron-based grafts had lower sac Chaudhuri et al.115 used a thrombus ana-
pressures than aneurysm models with thin- logue (gelatin) to study transmission through
ner ePTFE grafts. More compliant AAA mod- pulsatile waveforms; they claimed that type I
els were found to have lower excluded sac endoleak was easily identifiable by analyzing
pressures and therefore could be less prone these waveforms. However, Hinnen et al.116
to rupture than stiffer AAAs. concluded that gelatin was not a realistic
Skillern et al.107 and Pacanowski et al.108 thrombus analogue. The range of Young’s
developed a bench top model to analyze modulus (E) for thrombus was 1.3 to 5.9 N/
endotension and found that thrombus in the cm2, while mean value of E for gelatin was
sac could reduce sac pressurization They also 0.4 N/cm2. They also described a more real-
determined that attachment failure corrected istic polymer-based thrombus analogue. Di
with a proximal cuff with no endoleak could Martino et al.117 found the range of Young’s
be responsible for endotension. modulus to be 5 to 20 N/cm2.

Corbett et al. 2008;15:000–000
Wang et al.118 determined that intraluminal 52% and 97%. It is likely that the voids in the
thrombus may be an isotropic material. It is thrombus matrix are filled with fluid in vivo,
composed of 3 layers (luminal, medial, and making thrombus a ‘‘semi-fluid’’ material. The
abluminal) that have differences in micro- pressure transmitted through intraluminal
structure and mechanical properties. The thrombus has been shown to be dependent
luminal layer consisted of fibrin fibers ar- on the matrix density, with low-porosity
ranged into thick primary bundles, cross- thrombus specimens transmitting less pres-
linked with finer secondary structures; the sure through them than specimens with high
Young’s modulus values for this layer were levels of porosity.96
57 and 54 N/cm2 in the circumferential and Vallabhaneni et al.121 and Raghavan et al.122
longitudinal directions, respectively. The me- investigated the tensile strength and failure
dial layer, which showed some degeneration properties of AAA walls. The failure strength
of these fibrin fibers, had E values of 27 and varied from 33.6 N/cm2 in aneurysmal loca-
33 N/cm2 in the circumferential and longitu- tions to 235.1 N/cm2 at the proximal neck; the
dinal directions, respectively. The abluminal stiffness of the aneurysm tissue was 180 N/
layer had little or no structure, as all the fibrin cm2. Biaxial testing on normal aortic and
fibers had degenerated, but this layer was too aneurysm tissue has also shown that AAA
degenerated to allow mechanical testing. tissue is stiffer in the circumferential orienta-
Vande Geest et al.119 performed planar biaxial tion than normal aortic tissue, indicating that
testing on the luminal layer of intraluminal it is more anisotropic.123 This variation in wall
thrombus, concluding that the thrombus properties is not easily recreated using latex
behaves as an isotropic material, which is in or silicone rubber bench-top AAA models and
agreement with Wang et al.118 Their maxi- could have relevance to rupture or migration
mum tangential modulus (equivalent to studies.
Young’s modulus below the proportional Singland et al.110 carried out mechanical
limit) was 20.1 N/cm2 in the circumferential testing on a custom-made stent-graft to deter-
direction and 23.1 N/cm2 in the longitudinal mine whether it could be damaged during
direction. The published range of Young’s deployment. They tested the radial strength of
modulus for intraluminal thrombus is thus stents used in the device and performed
very broad (1.3 to 57 N/cm2), which indicates destructive testing of stents and graft and
that the properties of intraluminal thrombus suture materials. They concluded that the
vary widely and could be dependent on the forces placed on the stent-graft during deploy-
testing method used (tensile or compression). ment were unlikely to damage the device.
When designing a thrombus analogue for From these results, it would seem that all
bench-top testing, careful consideration must suture breaks and stent fractures that occur in a
be given to the stiffness required because the stent-graft do so after the device is implanted.
pressure transferred through a thrombus
analogue from an implanted stent-graft is
Computational Modeling
related to the stiffness of the thrombus
analogue.108,120 It has also been shown both The finite element method (FEM) has been
in vivo and ex vivo that intraluminal thrombus used to investigate wall stress of ruptured
from the same aneurysms can dissipate or and non-ruptured AAAs and to examine the
amplify pressure with marked variation, mak- impact of intraluminal thrombus on aneu-
ing the design of thrombus analogues even rysm wall stress. Venkatasubramaniam et
more challenging. It is possible for the mean al.125 found that peak wall stress was signif-
pressure in intraluminal thrombus to exceed icantly higher in ruptured versus non-rup-
the mean systemic pressure for reasons that tured AAAs, with the area of peak wall stress
are not fully understood; however, it is pro- correlating with rupture. They also deter-
posed that the intraluminal thrombus could mined that wall stress was dependent on
‘‘trap’’ pressure, preventing a decline to dia- geometry. Mower et al.126 found that intralu-
stolic levels. Intraluminal thrombus is a porous minal thrombus could reduce wall stress by
material, with the matrix density varying from 51% in an ideal AAA model. Raghavan et

2008;15:000–000 Corbett et al.
al.127 used realistic AAA geometry to show wall stress in an ideal aneurysm without an
that wall stress is complexly distributed; the implanted stent-graft was close to the yield
area of highest stress always occurs on the stress of the wall tissue, further underlining
posterior surface, so basing treatment on the grave nature of this failure mechanism.133
maximum diameter and the patient’s blood Papaharilaou et al.136 also investigated wall
pressure was an inaccurate method. This was stress using a decoupled fluid structure
in agreement with the observations of Fill- approach; they found that a static structural
inger et al,128 who also found that patients analysis underestimates the magnitude of
with higher blood pressure were more prone peak wall stress by as much as 12.5%, which
to rupture and that elevated wall stress was casts doubt on using FEM alone to predict the
not just an acute event near time of rupture. maximum stress in an AAA.
Computational fluid dynamics (CFD) has
been used to investigate the flow in simulated
Mathematical Approaches
aneurysms after EVAR. Morris et al.129 inves-
tigated different inlet flow profiles on realistic Liffman et al.137 used mathematical model-
and idealized stent-grafts. They found that ing and CFD with steady flow and mean blood
flow in the iliac limbs of realistic stent-grafts pressure (100 mmHg) to investigate drag
occurred in a helical motion; the drag force forces on stent-graft models. They found that
was 26% higher in realistic than in idealized larger graft diameter and iliac angle increased
models. They hypothesized that this skewing drag force, which agreed with others.60,135
of the flow could be responsible for throm- Morris et al.60 compared drag force obtained
bosis in the stent-graft limbs. Frauenfelder et using a mathematical model to the drag force
al.130 also reached the same conclusions obtained using CFD to study bifurcated stent-
using CFD and an experimental model. grafts using a pulsatile inflow condition.
Fluid-structure interaction is a relatively Mathematical modeling was used by Canic
new approach that allows investigation of et al.138 to show that shear stress was
wall deformations and their effects on fluid elevated in iliac limb diameters ,14 mm,
flow be investigated. Li and Klein- which could lead to limb thrombosis. Canic139
streuer2,65,131–134 have used this method to derived a set of equations that described wall
investigate many different aspects of post deformations at the anchoring sites of stent-
EVAR AAAs. In one such study they showed grafts; these equations were then used to
that high blood pressure and proximal diam- compute the discontinuity of wall motion of a
eter substantially increased drag force on the stent-graft placed in an AAA. The deforma-
graft. Sac pressures were non-zero after EVAR tions in the stent were dramatic at the
due to fluid structure interactions between boundaries where the stent was supported
pulsatile blood and the stent-graft. Sac pres- by the proximal neck and where the stent was
sure was found to decrease drag force due to unsupported in the aneurysm sac. These
sac pressure acting normal to lumen pres- sudden deformations could lead to fractures
sure. Drag force also increased non-linearly of stent-graft components in vivo.
with increasing iliac angle, which is in
agreement with Morris et al.60 Moreover, the
body-to-leg length ratio and the body-to-leg
DISCUSSION
diameter ratio affect drag force: a shorter Abdominal aortic aneurysm is a significant
body with longer legs was deemed to be the cause of mortality and morbidity in the
most favorable design. To the contrary, Western world. The detection of AAA is
Mohan et al.135 recommended a stent-graft steadily increasing, and the frequency of life-
with a long body as the optimum design. threatening ruptured AAA is rising. Traditional
Li and Kleinstreuer132 determined that sac surgical repair with the associated long recov-
pressure could come close to systemic pres- ery times has excellent long-term reliability
sure in the face of types I or II endoleak; a type and is an appropriate procedure in the treat-
I endoleak rendered the stent-graft useless in ment of AAA patients who have low surgical
preventing rupture. They showed that the risk. However, open repair is a high-risk

Corbett et al. 2008;15:000–000
procedure in patients with significant comor- the removal of sac thrombus, the use of
bidities, for whom the mortality rate is high. thicker graft material, and the absence of
The low incidence of perioperative deaths metallic components. The net effect of these
has made EVAR an increasingly attractive events is an increased cost for the EVAR
therapeutic alternative over the last 15 years. procedure due to reinterventions and surveil-
However, the reservations expressed after lance.17,144,145
EVAR-2 have raised doubts as to whether this As the interest in EVAR has soared in the
procedure is safe in patients who are unfit for last two decades, a vast amount of clinical, in
open surgery.17 Critics have questioned the vitro, and computational research has been
health of the patients in the EVAR-2 trial, undertaken. Since the beginning of this
noting that inclusion of these patients in any century, this research has moved from prov-
trial may bias the results. Although it is ing that EVAR was feasible to investigating
probable that there exists a subset of patients why and when the applied technologies fail.
who are too unwell for any AAA repair, be it Mechanical structures that may be subjected
open or endovascular, promising outcomes to movement are usually designed with the
have been reported in trials where EVAR was limits of that motion clearly defined. In vivo,
undertaken on carefully selected patients who there is no certainty as to how or when a
were unfit for open repair.140–142 Therefore, stent-graft may be stressed or forced to
EVAR is a justifiable solution in patients who change shape. Dramatic changes in neck
are considered well enough for endovascular angle, neck diameter, and/or aneurysm length
surgery but unfit for open aneurysm repair. have effects on an implanted stent- graft that
The long-term clinical outcomes for EVAR are are not yet fully understood. Using stent-graft
less satisfactory than for open repair due models reconstructed from postoperative CT
largely to the mechanical problems that arise scans could allow FEA and CFD analysis to
in vivo. Interestingly, patients who have determine any negative consequences of this
undergone open repair have a higher risk of morphological change on the long-term du-
laparotomy-related reinterventions.143 Unfor- rability of the stent-graft.
tunately, the mortality and cost implications It seems that there is still controversy as to
owing to these reinterventions have not yet whether long-term neck dilatation is a result
been quantified, so it is difficult to accurately of continued aneurysmal disease or due to
compare the long-term results and cost of self-expanding endografts continually exert-
EVAR versus open repair. ing radial force against the aortic wall. The
With EVAR, migration and types I and III absence of neck dilatation when using bal-
endoleaks can have serious consequences loon-expandable stent–based devices is a
and must be treated quickly and aggressively. promising outcome.54,64 These trials were
Endovascular techniques are usually effective small in comparison to the longitudinal
in remedying these problems, but if open multicenter trials of many of the marketed
surgery was necessary, it could be high devices, so larger studies are required. The
risk in many EVAR patients. Other problems, effect of material mismatch at the junction of
such as change in aneurysm morphology or the stent-graft and the aortic wall is still not
kinks/occlusions, may be the first step in completely understood. It is highly possible
the development of migration or the separa- that there are high wall shear stress gradients
tion of a modular limb (type IIIa endoleak). that could cause intimal hyperplasia and lead
Likewise, stent/suture breaks are usually to further local degradation of the wall.
the starting point for a graft tear (type IIIb Migration with type I endoleak, along with
endoleak). type III endoleak, constitute complete failure
The potential for mechanical problems of the stent-graft to exclude the aneurysm
necessitates costly ongoing monitoring of and are still the two most alarming forms of
patients with periodic CT scans. These me- EVAR failure. Methods for comparing the
chanical issues are rarely present after open migration risk of different stent-grafts in the
surgery due to the superior hand-sewn anas- literature are biased against devices based on
tomosis, the elimination of side branch flow, columnar support (AneuRx, Powerlink) and

2008;15:000–000 Corbett et al.
favor devices that rely mainly on proximal research. New alloys and suture materials
fixation (Zenith, Talent). The reason for this is need to be developed to reduce or eliminate
that these pullout tests are based on devices these problems.
implanted within a neck section only. In Stent-graft kink is an infrequent problem;
devices with columnar support, the attach- however, it could be the first step in device
ment of the device in the iliac arteries could migration. It is possible that these kinks could
be a major factor in preventing migration. In change local hemodynamics within an iliac
addition, a pullout test conducted under static graft limb, leading to the development a
conditions (no pulsatile fluid in the system) is partially occluded leg; the resultant higher
invalid because it ignores the potential for drag force on the graft could increase the
hooks and barbs to become dislodged during potential for migration. However, the effect of
wall motion at the proximal neck. A method partially occluded legs on stent-graft drag has
has been developed in our institution where- not been ascertained as of yet.
by stent-grafts are implanted in ideal aneu- It has been shown experimentally that
rysm models and pullout tests are undertaken endotension can be caused by treating a type
with pulsatile fluid in the system.146 We feel I endoleak with a proximal cuff. By extension,
that this represents a more realistic set of it is highly probable that any graft tear that
conditions to compare stent-grafts in terms of spontaneously seals could cause endoleak.
protection against migration. Experimental This is not to say that there are not other
and computational studies have earmarked reasons for endotension. It is likely that
preoperative neck angle, hypertension, large ‘‘endotension’’ is an umbrella term for a
graft inlet diameter, large iliac leg angle, and number of phenomena, such as pressure
use of an aortomonoiliac stent-graft as the transfer through the graft wall, low flow
greatest risks to the durability of a stent- endoleak, and microfiltration of serous com-
graft’s proximal seal. An optimum body-to- ponents through the graft.
leg length ratio has not yet been derived While newer-generation endografts have
for AAA stent-grafts, with some authors improved the outcome after EVAR, all the
arguing that a short body is preferable and failure modes discussed are still relevant; as
vice versa. long as these problems persist, reservations
Type II endoleak is purely a clinical issue, will remain over the success of the endovas-
and decisions on whether to treat or not are cular method.147,148
usually made on a case by case basis. Unlike
type I or III endoleak, type II endoleak may not
be a danger in terms of aneurysm rupture. Conclusion
However, type II endoleak can allow the EVAR technology is still evolving, and as
aneurysm to experience systemic pressure, the design of the devices and delivery sys-
and a growing aneurysm is reason for tems improves, it is likely that the incidence
reintervention. Aneurysms that have only an of mechanical problems will be reduced.
inflow branch, as opposed to an inflow/ Careful clinical analysis of the long-term
outflow circuit, have been experimentally EVAR performance must continue to gain a
and computationally shown to be subjected detailed understanding of existing clinical
to higher pressures and thus may be at higher issues and identify any new issues that may
rupture risk. arise. The research community must identify
The problem of stent or graft breakages has and analyze all the forces placed on the graft
been offset by incorporating greater redun- in vivo to ensure that future stent-grafts can
dancy in the devices, i.e., more sutures, be designed for greater durability with a low
barbs, etc. This represents a stop gap ap- incidence of reinterventions.
proach; presumably, it will not eliminate
device failures but will mean that failures will
Acknowledgments: The authors would like to gratefully
take longer to manifest themselves. If EVAR is acknowledge the support of the Irish Research Council for
to become the gold standard of AAA surgery, Science, Engineering and Technology and Enterprise
then this is an area that requires substantial Ireland.

Corbett et al. 2008;15:000–000
REFERENCES 13. Chahwan S, Comerota AJ, Pigott JP, et al.

Elective treatment of abdominal aortic aneu-
1. Sakalihasan N, Limet R, Defawe OD. Abdomi- rysm with endovascular or open repair: the
nal aortic aneurysm. Lancet. 2005;365:1577– first decade. J Vasc Surg. 2007;45:258–262.
1589.
14. Lederle FA, Kane RL, MacDonald R, et al.
2. Li Z, Kleinstreuer C. Fluid-structure interaction Systematic review: repair of unruptured ab-
effects on sac-blood pressure and wall stress in
dominal aortic aneurysm. Ann Intern Med.
a stented aneurysm. J Biomech. 2005;127:
2007;146:735–741.
662–671.
15. Greenhalgh RM, Brown LC, Kwong GP, et al.
3. Office of Population Censuses and Surveys.
Comparison of endovascular aneurysm repair
Mortality Statistics, England and Wales. Lon-
with open repair in patients with abdominal
don HMSO, 1989.
aortic aneurysm (EVAR trial 1), 30-day opera-
4. Gouëffic Y, Becquemin JP, Desgranges P, et al.
tive mortality results: randomised controlled
Midterm survival after endovascular versus
trial. Lancet. 2004;364:843–848.
open repair of infrarenal aortic aneurysms.
16. EVAR trial participants. Endovascular aneu-
J Endovasc Ther. 2005;12:47–57.
rysm repair versus open repair in patients with
5. Garcia-Madrid C, Josa M, Riambau V, et al.
abdominal aortic aneurysm (EVAR trial 1):
Endovascular versus open surgical of abdom-
randomised controlled trial. Lancet. 2005;
inal aortic aneurysm: a comparison of early
365:2179–2186.
and intermediate results in patients suitable for
17. EVAR trial participants. Endovascular aneu-
both techniques. Eur J Vasc Endovasc Surg.
rysm repair and outcome in patients unfit for
2004;28:365–372.
open repair of abdominal aortic aneurysm
6. Vogel TR, Nackman GB, Crowley JG, et al.
(EVAR trial 2): randomised controlled trial.
Factors impacting functional health and re-
source utilization following abdominal aortic Lancet. 2005;365(9478):2187–2192.
aneurysm repair by open and endovascular 18. Sampram ES, Karafa MT, Mascha EJ, et al.
techniques. Ann Vasc Surg. 2005;19:641–647. Nature, frequency, and predictors of secondary
7. Volodos NL, Karpovich IP, Troyan VI, et al. procedures after endovascular repair of ab-
Clinical experience of the use of self-fixing dominal aortic aneurysm. J Vasc Surg. 2003;
synthetic prostheses for remote endopros- 37:930–937.
thetics of the thoracic and the abdominal aorta 19. Hobo R, Buth J. Secondary interventions fol-
and iliac arteries through the femoral artery lowing endovascular abdominal aortic aneu-
and as intraoperative endoprosthesis for aorta rysm repair using current endografts. A EURO-
reconstruction. Vasa Suppl. 1991;33:93–95. STAR report. J Vasc Surg. 2006;43:896–902.
8. Carpenter JP, Endologix Investigators. Mid- 20. Walschot LHB, Laheij RJF, Verbeek ALM. Out-
term results of the multicenter trial of the come after endovascular abdominal aortic an-
Powerlink bifurcated system for endovascular eurysm repair: a meta-analysis. J Endovasc
aortic aneurysm repair. J Vasc Surg. 2004;40: Ther. 2002;9:82–89.
849–859. 21. Verhoeven EL, Tielliu IF, Prins TR, et al.
9. Aarts F, van Sterkenburg S, Blankensteijn JD. Frequency and outcome of re-interventions
Endovascular aneurysm repair versus open after endovascular repair for abdominal aortic
aneurysm repair: comparison of treatment aneurysm: a prospective cohort study.
outcome and procedure-related reintervention Eur J Vasc Endovasc Surg. 2004;28:357–364.
rate. Ann Vasc Surg. 2005;19:699–704. 22. Thomas SM, Beard JD, Ireland M, et al. Results
10. Mehta M, Roddy SP, Darling RC, et al. Infrare- from the prospective registry of endovascular
nal abdominal aortic aneurysm repair via treatment of abdominal aortic aneurysms
endovascular versus open retroperitoneal ap- (RETA): midterm results to 5 years. Eur J Vasc
proach. Ann Vasc Surg. 2005;19:374–378. Endovasc Surg. 2005;29:563–570.
11. Drury D, Michaels JA, Jones L, et al. System- 23. Franks SC, Sutton AJ, Bown MJ, et al. System-
atic review of recent evidence for the safety atic review and meta-analysis of 12 years of
and efficacy of elective endovascular repair in endovascular abdominal aortic aneurysm re-
the management of infrarenal abdominal aor- pair. Eur J Vasc Endovasc Surg. 2007;33:154–
tic aneurysm. Br J Surg. 2005;92:937–946. 171.
12. Zarins CK, White RA, Moll FL, et al. The 24. Veith FJ, Baum RA, Ohki T, et al. Nature and
AneuRx stent graft: four-year results and significance of endoleaks and endotension: sum-
worldwide experience 2000. J Vasc Surg. mary of opinions expressed at an international
2001;33:S135–145. conference. J Vasc Surg. 2002;35:1029–1038.

2008;15:000–000 Corbett et al.
25. Bargellini I, Cioni R, Petruzzi P, et al. Endovas- 38. Waasdorp EV, Vries JP, Hobo R, et al.
cular repair of abdominal aortic aneurysms: Aneurysm diameter and proximal aortic neck
analysis of aneurysm volumetric changes at diameter influence clinical outcome of endo-
mid-term follow up. Cardiovasc Intervent vascular abdominal aortic repair: a 4-year
Radiol. 2005;28:426–433. EUROSTAR experience. Ann Vasc Surg. 2005;
26. White RA, Donayre CE, Walot I, et al. Comput- 19:755–761.
ed tomography assessment of abdominal 39. Conners MS, Sternbergh WC, Carter G, et al.
aortic aneurysm morphology after endograft Endograft migration one to four years after
exclusion. J Vasc Surg. 2001;33:S1–10. endovascular abdominal aortic aneurysm re-
27. Sternbergh WC, Carter G, York JW, et al. Aortic pair with the AneuRx device: a cautionary note.
neck angulation predicts adverse outcome J Vasc Surg. 2002;36:476–484.
with endovascular abdominal aortic aneurysm 40. Tonnessen BH, Sternbergh WC, Money SR.
repair. J Vasc Surg. 2002;35:482–486. Mid- and long-term device migration after
28. White GH, May J, Petrasek P, et al. Endoten- endovascular abdominal aortic aneurysm re-
sion: an explanation for continued AAA growth pair: a comparison of AneuRx and Zenith
after successful endoluminal repair. J Endo- endografts. J Vasc Surg. 2005;42:392–401.
vasc Surg. 1999;6:308–315. 41. Zarins CK, Bloch DA, Crabtree T, et al. Stent
29. Singh-Ranger R, Adiseshiah M. Differing mor- graft migration after endovascular aneurysm
phological changes following endovascular repair: importance of proximal fixation. J Vasc
AAA repair using balloon-expandable or self- Surg. 2003;38:1264–1272.
expanding endografts. J Endovasc Ther. 2000; 42. Cao P, Verzini F, Zannetti S, et al. Device
7:479–485. migration after endoluminal abdominal aortic
30. Resch T, Ivancev J, Brunkwall J, et al. Midterm aneurysm repair: analysis of 113 cases with a
changes in aortic aneurysm morphology after minimum follow-up period of 2 years. J Vasc
endovascular repair. J Endovasc Ther. 2000;7: Surg. 2002;35:229–235.
279–285. 43. van Herwaarden JA, van de Pavoordt ED,
31. Krämer SC, Seifarth H, Pamler R, et al. Geo- Waasdorp EJ, et al. Long-term single-center
metric changes in aortic endografts over a 2- results with AneuRx endografts for endovas-
year observation period. J Endovasc Ther. cular abdominal aortic aneurysm repair.
2001;8:34–38. J Endovasc Ther. 2007;14:307–317.
32. Vos AW, Wisselink W, Marcus JT, et al. Cine 44. Sternbergh WC, Money SR, Greenberg RK,
MRI assessment of aortic aneurysm dynamics et al. Influence of endograft oversizing on
before and after endovascular repair. J Endo- device migration, endoleak, aneurysm shrink-
vasc Ther. 2003;10:433–439. age, and aortic neck dilation: results from the
33. Chuter TA. Stent-graft design: the good, the Zenith Multicenter Trial. J Vasc Surg. 2004;39:
bad, and the ugly. Cardiovasc Surg. 2002;10: 20–26.
7–13. 45. Ouriel K, Clair DG, Greenberg RK, et al.
34. Prinssen M, Wever JJ, Mali WP, et al. Concerns Endovascular repair of abdominal aortic aneu-
for the durability of the proximal abdominal rysms: device-specific outcome. J Vasc Surg.
aortic aneurysm endograft fixation from a 2003;37:991–998.
2-year and 3-year longitudinal computed to- 46. Greenberg RK, Chuter TA, Sternbergh WC,
mography angiography study. J Vasc Surg. et al. Zenith AAA endovascular graft: interme-
2001;33:S64–69. diate-term results of the US multicenter trial.
35. Palombo D, Valenti D, Ferri M, et al. Changes in J Vasc Surg. 2004;39:1209–1218.
the proximal neck of abdominal aortic aneu- 47. Parent FN, Meier GH, Godziachville V, et al. The
rysms early after endovascular treatment. Ann incidence and natural history of type I and II
Vasc Surg. 2003;17:408–410. endoleak: a 5-year follow-up assessment with
36. Prinssen M, Buskens E, Blankensteijn JD. color duplex ultrasound scan. J Vasc Surg.
Quality of life after endovascular and open 2002;35:474–481.
AAA repair. Results of a randomized trial. 48. Zarins CK, White RA, Fogarty TJ. Aneurysm
Eur J Vasc Endovasc Surg. 2004;27:121–127. rupture after endovascular repair using the
37. Schlensak C, Doenst T, Hauer M, et al. Serious AneuRx stent graft. J Vasc Surg. 2000;31:960–970.
complications that require surgical interven- 49. Chuter TA, Parodi JC, Lawrence-Brown M.
tions after endoluminal stent-graft placement Management of abdominal aortic aneurysm:
for the treatment of infrarenal aortic aneu- a decade of progress. J Endovasc Ther. 2004;
rysms. J Vasc Surg. 2001;34:198–203. 11(Suppl II):II-82–II-95.

Corbett et al. 2008;15:000–000
50. Carpenter JP, Endologix Investigators. Multi- 62. Heikkinen MA, Alsac JM, Arko FR, et al. The
center trial of the PowerLink bifurcated system importance of iliac fixation in prevention of
for endovascular aortic aneurysm repair. J Vasc stent graft migration. J Vasc Surg. 2006;43:
Surg. 2002;36:1129–1137. 1130–1137.
51. Katzen BT. Update on devices for endovascular 63. Murphy EH, Johnson ED, Arko FR. Device-
therapy of AAA’s. Endovasc Today. January specific resistance to in vivo displacement of
2004:42–46. stent-grafts implanted with maximum iliac
52. van Marrewijk CJ, Leurs LJ, Vallabhaneni fixation. J Endovasc Ther. 2007;14:585–592.
SR, et al. Risk-adjusted outcome analysis of 64. Dalainas I, Nano G, Casana R, et al. Mid-term
endovascular abdominal aortic aneurysm re- results after endovascular repair of abdominal
pair in a large population: how do stent- aortic aneurysms: a four-year experience.
grafts compare? J Endovasc Ther. 2005;12: Eur J Vasc Endovasc Surg. 2004;27:319–323.
417–429. 65. Li Z, Kleinstreuer C, Farber M. Computational
53. Major A, Guidoin R, Soulez G, et al. Implant analysis of biomechanical contributors to pos-
degradation and poor healing after endovas- sible endovascular graft failure. Biomech Mod-
cular repair of abdominal aortic aneurysms: an el Mechanobiol. 2005;4:221–234.
analysis of explanted stent-grafts. J Endovasc 66. Semmens JB, Lawrence-Brown MM, Hartley
Ther. 2006;13:457–467. DE, et al. Outcomes of fenestrated endografts
54. Malas MB, Ohki T, Veith FJ, et al. Absence of in the treatment of abdominal aortic aneurysm
proximal neck dilatation and graft migration in Western Australia (1997-2004). J Endovasc
after endovascular aneurysm repair with bal- Ther. 2006;13:320–329.
loon-expandable stent-based endografts. J 67. Zhou SS, How TV, Vallabhaneni SR, et al.
Vasc Surg. 2005;42:639–644. Comparison of the fixation strength of stan-
55. Gross NR. Workshop on preclinical testing dard and fenestrated stent-grafts for endovas-
for endovascular grafts. Session 1, animal cular abdominal aortic aneurysm repair.
studies: a retrospective and prospective evalu- J Endovasc Ther. 2007;14:168–175.
ation. Food and Drug Administration. July 28, 68. Greenberg RK, West K, Pfaff K, et al. Be-
2004. yond the aortic bifurcation: branched endo-
56. Guidoin R, Marois Y, Douville Y, et al. First- vascular grafts for thoracoabdominal and
generation aortic endografts: analysis of ex- aortoiliac aneurysms. J Vasc Surg. 2006;43:
planted Stentor devices from the EUROSTAR 879–887.
registry. J Endovasc Ther. 2000;7:105–122. 69. Sanfelippo P. Abdominal aortic aneurysm –
57. Resch T, Malina M, Lindblad B, et al. The 2003: what we know, what we don’t know – a
impact of stent design on proximal stent-graft review. Int J Angiol. 2003;12:145–152.
fixation in the abdominal aorta: an experimen- 70. Rose J. Stent-grafts for unruptured abdominal
tal study. Eur J Vasc Endovasc Surg. 2000;20: aortic aneurysms: current status. Cardiovasc
190–195. Intervent Radiol. 2006;29:332–343.
58. Malina M, Lindblad B, Ivancev K, et al. En- 71. Fransen GA, Vallabhaneni SR, van Marrewijk
dovascular AAA exclusion: will stents with CJ, et al. Rupture of infrarenal aortic aneurysm
hooks and barbs prevent stent-graft migration? after endovascular repair: a series from EURO-
J Endovasc Surg. 1998;5:310–317. STAR registry. Eur J Vasc Endovasc Surg.
59. Veerapen R, Dorandeu A, Serre I, et al. Im- 2003;26:487–493.
provement in proximal aortic endograft fixa- 72. Dias NV, Ivancev K, Malina M, et al. Intra-
tion: an experimental study using different aneurysm sac pressure measurements after
stent-grafts in human cadaveric aortas. endovascular aneurysm repair: differences
J Endovasc Ther. 2003;10:1101–1109. between shrinking, unchanged, and expanding
60. Morris L, Delassus P, Walsh M, et al. A aneurysms with and without endoleaks. J Vasc
mathematical model to predict the in vivo Surg. 2004;39:1229–1235.
pulsatile drag forces acting on bifurcated stent 73. Abraham CZ, Chuter TA, Reilly LM, et al.
grafts used in endovascular treatment of Abdominal aortic aneurysm repair with the
abdominal aortic aneurysm (AAA). J Biomech. Zenith stent graft: short to midterm results.
2004;37:1087–1095. J Vasc Surg. 2002;36:217–225.
61. Benharash P, Lee JT, Abilez OJ, et al. Iliac 74. Carpenter JP, Anderson WN, Brewster C, et al.
fixation inhibits migration of both suprarenal Multicenter pivotal trial results of the Lifepath
and infrarenal aortic endografts. J Vasc Surg. system for endovascular aortic aneurysm re-
2007;45:250–257. pair. J Vasc Surg. 2004;39:34–43.

2008;15:000–000 Corbett et al.
75. Silverberg D, Baril DT, Ellozy SH, et al. 88. Hobo R, Laheij RJ, Buth J, et al. The influ-
An 8-year experience with type II endoleaks: ence of aortic cuffs and iliac limb extensions
natural history suggests selective intervention on the outcome of endovascular abdominal
is a safe approach. J Vasc Surg. 2006;44: aortic aneurysm repair. J Vasc Surg. 2007;45:
453–459. 79–85.
76. Faries PL, Brener BJ, Connelly TL, et al. A 89. Imamura A, Koike Y, Iwaki R, et al. Infrarenal
multicenter experience with the Talent endo- abdominal aortic aneurysm complicated by
vascular graft for the treatment of abdominal persistent endotension after endovascular re-
aortic aneurysms. J Vasc Surg. 2002;35: pair: report of a case. Surg Today. 2005;35:
1123–1128. 879–882.
77. Uflacker R, Robison J. Endovascular treatment 90. Lin PH, Bush RL, Katzman JB, et al. Delayed
of abdominal aortic aneurysms: a review. Eur aortic aneurysm enlargement due to endoten-
Radiol. 2001;11:739–753. sion after endovascular abdominal aortic an-
78. Jacobs TS, Won J, Gravereaux RC, et al. eurysm repair. J Vasc Surg. 2003;38:840–842.
Mechanical failure of prosthetic human im- 91. Risberg B, Delle M, Eriksson E, et al. Aneurysm
plants: a 10-year experience with aortic stent sac hygroma: a cause of endotension.
graft devices. J Vasc Surg. 2003;37:16–26. J Endovasc Ther. 2001;8:447–453.
79. Beebe HG, Cronenwett JL, Katzen BT, et al. 92. Tricciola SM, Dayal R, Chaer RA, et al. The
Results of an aortic endograft trial: impact of development of endotension is associated with
device failure beyond 12 months. J Vasc Surg. increased transmission of pressure and serous
2001;33:S55–63. components in porous polytetrafluoroethylene
80. Najibi S, Steinberg J, Katzen BT, et al. Detec- stent-grafts: characterization using a canine
tion of isolated hook fractures 36 months after model. J Vasc Surg. 2006;43:109–116.
implantation of the Ancure endograft: a cau- 93. Sonesson B, Dias N, Malina M, et al. Intra-
tionary note. J Vasc Surg. 2001;34:353–356. aneurysm pressure measurements in success-
81. Heintz C, Riepe G, Birken L, et al. Corroded fully excluded abdominal aortic aneurysm
nitinol wires in explanted aortic endografts: an after endovascular repair. J Vasc Surg. 2003;
important mechanism of failure? J Endovasc 37:733–738.
Ther. 2001;8:248–253. 94. Dias NV, Ivancev K, Malina M, et al. Direct
82. Zarins CK, Arko FR, Crabtree T, et al. Explant intra-aneurysm sac pressure measurement
analysis of AneuRx stent grafts: relationship using tip-pressure sensors: in vivo and in vitro
between structural findings and clinical out- evaluation. J Vasc Surg. 2004;40:711–716.
come. J Vasc Surg. 2004;40:1–11. 95. Ohki T, Ouriel K, Galvagni Silveira P, et al.
83. Cosin O, Rousseau H, Otal P, et al. Late Initial results of wireless pressure sensing for
perforation of a thoracic aortic Dacron graft endovascular aneurysm repair: the APEX Trial
by a metallic stent-graft component. J Endo- (Acute Pressure Measurement to Confirm
vasc Ther. 2006;13:676–680. Aneurysm Sac EXclusion). J Vasc Surg. 2007;
84. Rumball-Smith A, Wright IA, Buckenham TM. 45:236–242.
Strut failure in the body of the Zenith abdom- 96. Vallabhaneni SR, Gilling-Smith GL, Brennan
inal endoprosthesis. Eur J Vasc Endovasc JA, et al. Can intrasac pressure monitoring
Surg. 2006;32:136–139. reliably predict failure of endovascular aneu-
85. Matsumura JS, Brewster DC, Makaroun MS, rysm repair? J Endovasc Ther. 2003;10:
et al. A multicenter controlled clinical trial of 524–530.
open versus endovascular treatment of ab- 97. Morris L, O’Donnell P, Delassus P, et al.
dominal aortic aneurysm. J Vasc Surg. Experimental assessment of stress patterns in
2003;37:262–271. abdominal aortic aneurysms using the photo-
86. Ghanim K, Mwipatayi BP, Abbas M, et al. Late elastic method. Strain. 2004;40:165–172.
stent-graft migration secondary to separation 98. Morris LG. Numerical and experimental inves-
of the uncovered segment from the main body tigation of mechanical factors in the treatment
of a Zenith endoluminal graft. J Endovasc of abdominal aortic aneurysms. University of
Ther. 2006;13:346–349. Limerick, Doctoral Thesis; 2004.
87. Fransen GA, Desgranges P, Harris PL, et al. 99. O’Brien T, Morris L, O’Donnell M, et al. Injec-
Frequency, predictive factors, and conse- tion-moulded models of major and minor
quences of stent-graft kink following endovas- arteries: the variability of model wall thickness
cular AAA repair. J Endovasc Ther. 2003;10: owing to casting technique. Proc Inst Mech
913–918. Eng [H]. 2005;219:381–386.

Corbett et al. 2008;15:000–000
100. Chong CK, How TV, Black RA, et al. Develop- 113. Volodos SM, Sayers RD, Gostelow JP, et al.
ment of a simulator for endovascular repair of An investigation into the cause of distal
abdominal aortic aneurysms. Ann Biomed endoleaks: role of displacement force on the
Eng. 1998;26:798–802. distal end of a stent-graft. J Endovasc Ther.
101. Chong CK, How TV, Gilling-Smith GL, et al. 2005;12:115–120.
Modeling endoleaks and collateral reperfu- 114. Arko FR, Heikkinen MA, Lee ES, et al. Iliac
sion following endovascular AAA exclusion. J fixation length and resistance to in-vivo stent
Endovasc Ther. 2003;10:424–432. graft displacement. J Vasc Surg. 2005;41:
102. Chong CK, How TV. Flow patterns in an 664–671.
endovascular stent-graft for abdominal aor- 115. Chaudhuri A, Ansdell LE, Grass AJ, et al.
tic aneurysm repair. J Biomech. 2004;37: Intrasac pressure waveforms after endovas-
89–97. cular aneurysm repair (EVAR) are a reliable
103. Chong CK, How TV, Harris PL. Flow visualiza- marker of type I endoleaks, but not type
tion in a model of a bifurcated stent-graft. II or combined types: an experimental study.
J Endovasc Ther. 2005;12:435–445. Eur J Vasc Endovasc Surg. 2004;28:373–
104. Mehta M, Veith FJ, Ohki T, et al. Significance 378.
of endotension, endoleak, and aneurysm 116. Hinnen JW, Rixen DJ, Koning OH, et al.
pulsatility after endovascular repair. J Vasc Development of fibrinous thrombus analogue
Surg. 2003;37:842–846. for in vitro abdominal aortic aneurysm stud-
105. Gawenda M, Jaschke G, Winter S, et al. ies. J Biomech. 2007;40:289–295.
Endotension as a result of pressure transmis- 117. Di Martino E, Mantero S, Inzoli F, et al.
sion over the graft following endovascular Biomechanics of abdominal aortic aneurysm
aneurysm repair - an in vitro study. Eur J Vasc in the presence of endoluminal thrombus:
Endovasc Surg. 2003;26:501–505. experimental characterization and structural
106. Gawenda M, Knez P, Winter S, et al. Endoten- static computational analysis. Eur J Vasc En-
sion is influenced by wall compliance in a dovasc Surg. 1998;15:290–299.
latex aneurysm model. Eur J Vasc Endovasc 118. Wang DH, Makaroun M, Vorp DA. Mechanical
Surg. 2004;27:45–50. properties and microstructure of intraluminal
107. Skillern CS, Stevens SL, Piercy KT, et al. thrombus from abdominal aortic aneurysm.
Endotension in an experimental aneurysm J Biomech Eng. 2001;123:536–539.
model. J Vasc Surg. 2002;36:814–817. 119. Vande Geest J, Sacks M, Vorp D. A planar
108. Pacanowski JP, Stevens SL, Freeman MB, biaxial constitutive relation for the luminal
et al. Endotension distribution and the role of layer of intraluminal thrombus in abdominal
thrombus following endovascular AAA exclu- aortic aneurysms. J Biomech. 2006;39:2347–
sion. J Endovasc Ther. 2002;9:639–651. 2354.
109. Albertini JN, Macierewicz JA, Yusuf SW, et al. 120. Hinnen JW, Koning OH, Visser MJ, et al.
Pathophysiology of proximal perigraft endoleak Effect of intraluminal thrombus on pressure
following endovascular repair of abdominal transmission in the abdominal aortic aneu-
aortic aneurysms: a study using a flow model. rysm. J Vasc Surg. 2005;42:1176–1182.
Eur J Vasc Endovasc Surg. 2001;22:53–56. 121. Vallabhaneni SR, Gilling-Smith GL, How TV,
110. Goergen CJ, Johnson BL, Greve JM, et al. et al. Heterogeneity of tensile strength and
Increased anterior abdominal aortic wall matrix metalloproteinase activity in the wall
motion: possible role in aneurysm pathogen- of abdominal aortic aneurysms. J Endovasc
esis and design of endovascular devices. Surg. 2004;11:494–502.
J Endovasc Ther. 2007;14:574–584. 122. Raghavan ML, Kratzberg J, Castro de Tolosa
111. Liffman K, Sutalo ID, Lawrence-Brown MM, EM, et al. Regional distribution of wall thick-
et al. Movement and dislocation of modular ness and failure properties of human abdom-
stent-grafts due to pulsatile flow and the inal aortic aneurysm. J Biomech. 2005;39:
pressure difference between the stent-graft 3010–3016.
and the aneurysm sac. J Endovasc Ther. 123. Vande Geest J, Sacks M, Vorp D. The effects
2006;13:51–61. of aneurysm on the biaxial mechanical be-
112. Volodos SM, Sayers RD, Gostelow JP, et al. havior of human abdominal aorta. J Biomech.
Factors affecting the displacement force ex- 2006;39:1324–1334.
erted on a stent graft after AAA repair–an in 124. Singland JD, Voulgre J, Picard G, et al. Static
vitro study. Eur J Vasc Endovasc Surg. mechanical properties of custom-made aortic
2003;26:596–601. endografts. Ann Vasc Surg. 2005;19:1–9.

2008;15:000–000 Corbett et al.
125. Venkatasubramaniam AK, Fagan MJ, Mehta T, 137. Liffman K, Lawrence-Brown MMD, Semmens
et al. A comparative study of aortic wall stress JB, et al. Analytical modeling and numerical
using finite element analysis for ruptured and simulation of forces in an endoluminal graft.
non-ruptured abdominal aortic aneurysms. J Endovasc Ther. 2001;8:358–371.
Eur J Vasc Endovasc Surg. 2004;28:168–176. 138. Canic S. Krajcer Z, Lapin S. Design of optimal
126. Mower WR, Quinones WJ, Gambhir SS. Effect endoprosthesis using mathematical model-
of intraluminal thrombus on abdominal aortic ling. Endovasc Today. May 2006:48–50.
aneurysm wall stress. J Vasc Surg. 1997;26: 139. Canic S. Blood flow through compliant vessels
602–608. after endovascular repair: wall deformations
127. Raghavan ML, Vorp DA, Federle MP, et al. induced by the discontinuous wall properties.
Wall stress distribution on three dimension- Comput Visual Sci. 2002;4:147–155.
ally reconstructed models of human abdom- 140. Sicard GA, Zwolak RM, Sidawy AN, et al.
inal aortic aneurysm. J Vasc Surg. 2000;31: Endovascular abdominal aortic aneurysm
760–769. repair: long-term outcome measures in pa-
128. Fillinger MF, Marra SP, Raghavan ML, et al. tients at high-risk for open surgery. J Vasc
Prediction of rupture risk in abdominal aortic Surg. 2006;44:229–236.
aneurysm during observation: wall stress 141. Hynes N, Sultan S. A prospective clinical,
versus diameter. J Vasc Surg. 2003;37: economic, and quality-of-life analysis compar-
724–732. ing endovascular aneurysm repair (EVAR),
129. Morris L, Delassus P, Grace P, et al. Effect of open repair, and best medical treatment in
flat, parabolic and realistic steady flow inlet high-risk patients with abdominal aortic aneu-
profiles on idealised and realistic stent graft rysms suitable for EVAR: the Irish patient trial.
J Endovasc Ther. 2007;14:763–776.
fits through abdominal aortic aneurysms
142. Bush RL, Johnson ML, Hedayati N, et al.
(AAA). Med Eng Phys. 2006;28:19–26.
Performance of endovascular aortic aneu-
130. Frauenfelder T, Lotfey M, Boehm T, et al.
rysm repair in high-risk patients: results from
Computational fluid dynamics: hemodynamic
the Veterans Affairs National Surgical Quality
changes in abdominal aortic aneurysms after
Improvement Program. J Vasc Surg. 2007;45:
stent-graft implantation. Cardiovasc Intervent
227–235.
Radiol. 2006;29:613–623.
143. Schermerhorn M, O’Malley J, Jhaveri A, et al.
131. Li Z, Kleinstreuer C. Blood flow and structure
Endovascular vs. open repair of abdominal
interactions in a stented abdominal aortic
aortic aneurysms in the Medicare population.
aneurysm model. Med Eng Phys. 2005;27: N Engl J Med. 2008;358:464–474.
369–382. 144. Michaels JA, Drury D, Thomas SM. Cost-
132. Li Z, Kleinstreuer C. Effects of major endo- effectiveness of endovascular abdominal aor-
leaks on a stented abdominal aortic aneu- tic aneurysm repair. Br J Surg. 2005;92:
rysm. J Biomech Eng. 2006;128:59–68. 960–967.
133. Li Z, Kleinstreuer C. Analysis of biomechan- 145. Angle N, Dorafshar AH, Moore WS, et al.
ical factors affecting stent-graft migration in Open versus endovascular repair of abdomi-
an abdominal aortic aneurysm model. nal aortic aneurysms: what does each really
J Biomech. 2006;39:2264–2273. cost? Ann Vasc Surg. 2004;18:612–618.
134. Li Z, Kleinstreuer C. Computational analysis 146. Corbett T, Callanan A, Morris L, et al. Con-
of type II endoleaks in a stented abdominal ventional methods overestimate the efficacy
aortic aneurysm model. J Biomech. 2006;39: of proximal fixation in abdominal aortic aneu-
2573–2582. rysm stent-grafts. Ir J Med Sci. 2007;176(S1):
135. Mohan IV, Harris PL, van Marrewijk CJ, et al. S48.
Factors and forces influencing stent-graft 147. Torella F. Effect of improved endograft design
migration after endovascular aortic aneurysm on outcome of endovascular aneurysm re-
repair. J Endovasc Ther. 2002;9:748–755. pair. J Vasc Surg. 2004;40:216–221.
136. Papaharilaou Y, Ekterinaris JA, Manousaki E, 148. Cao P, Verzini F, Parlani G, et al. Clinical effect
et al. A decoupled fluid structure approach for of abdominal aortic aneurysm endografting:
estimating wall stress in abdominal aortic 7-year concurrent comparison with open
aneurysms. J Biomech. 2006;40:367–377. repair. J Vasc Surg. 2004;40:841–848.

3D Reconstruction and Manufacture of utilizing computer aided design/computer aided manufacture
techniques and combined with the injection-moulding method.
Real Abdominal Aortic Aneurysms: Silicone rubber forms the basis of the resulting AAA model. As-
sessment of wall thickness and overall percentage difference from
From CT Scan to Silicone the final silicone model to that of the computer-generated model
Model was performed. In these realistic AAA models, wall thickness was
found to vary by an average of 9.21%. The percentage difference
in wall thickness recorded can be attributed to the contraction of
B. J. Doyle the casting wax and the expansion of the silicone during model
Centre for Applied Biomedical Engineering Research manufacture. This method may be used in conjunction with wall
stress studies using the photoelastic method or in fluid dynamic
共CABER兲, studies using a laser-Doppler anemometry. In conclusion, these
and Materials and Surface Science Institute, patient-specific rubber AAA models can be used in experimental
University of Limerick, investigations, but should be assessed for wall thickness variabil-
Limerick, Ireland ity once manufactured. 关DOI: 10.1115/1.2907765兴
L. G. Morris Keywords: abdominal aortic aneurysm (AAA), 3D reconstruction,

silicone
Galway Medical Technology Centre,
Galway Mayo Institute of Technology,
Galway, Ireland
Introduction
An abdominal aortic aneurysm 共AAA兲 can be defined as a per-
A. Callanan manent and irreversible localized dilation of the infrarenal aorta
关1兴. It has been proposed that an AAA is an aorta with a diameter
P. Kelly 1.5 times that of the normal infrarenal aorta 关2兴. Currently, the
timing of surgical intervention is determined by the maximum
Centre for Applied Biomedical Engineering Research diameter of the AAA, with an AAA diameter greater than 5 cm
共CABER兲, deemed to be at high risk of rupture. Much work has been aimed
at the rupture prediction of these aneurysms, in particular, the use
and Materials and Surface Science Institute, of finite element analysis to determine wall stress 关3–8兴. Although
University of Limerick, the use of numerical studies to gain an insight into the stress
Limerick, Ireland acting on the AAA wall is of obvious benefit to the particular
AAA case, validation of these techniques is of equal importance.
The ability to manufacture patient-specific AAA models for ex-
D. A. Vorp perimental studies could extend the use of preoperative wall stress
Department of Surgery, techniques. These realistic silicone models could be employed,
Department of Bioengineering, not only for stress analyses, similar to the photoelastic work of
Morris et al. 关9兴, but also for fluid dynamics studies and postop-
McGowan Institute for Regenerative Medicine, erative experimental testing, such as stent graft distraction testing.
and Centre for Vascular Remodelling and Regeneration, The models are created by first reconstructing a virtual AAA
University of Pittsburgh, model, leading to mould design, and then to manufacturing via the
Pittsburgh, PA injection-moulding technique. Previous research has examined the
use of rapid prototyping as a method of producing elastomeric
T. M. McGloughlin1 replicas of arterial vessels 关10兴. This method, although quick and
Centre for Applied Biomedical Engineering Research effective, does not produce the surface finish that can be achieved
using the injection-moulding process. Surface finish is of para-
共CABER兲, mount importance when using arterial models for experimental
and Materials and Surface Science Institute, testing using the photoelastic method, such as that previously con-
University of Limerick, ducted at our laboratory 关9兴. Other techniques have been em-
Limerick, Ireland ployed in order to make models for use in laser-Doppler anemom-
etry 共LDA兲关11兴 and particle image velocimetry 共PIV兲 flow
e-mail: tim.mcgloughlin@ul.ie
studies, where surface finish was of lesser importance than when
conducting wall stress studies.
The principal objective of this study is to describe the modeling
Abdominal aortic aneurysm (AAA) can be defined as a permanent and manufacturing process used and to determine the effective-
and irreversible dilation of the infrarenal aorta. AAAs are often ness of the technique. This process of converting a standard com-
considered to be an aorta with a diameter 1.5 times the normal puted tomography 共CT兲 scan to a patient-specific silicone model is
infrarenal aorta diameter. This paper describes a technique to of value to many researchers in this field.
manufacture realistic silicone AAA models for use with experi-
mental studies. This paper is concerned with the reconstruction Methods
and manufacturing process of patient-specific AAAs. 3D recon-
struction from computed tomography scan data allows the AAA to 3D Reconstruction. Four patients were chosen from our AAA
be created. Mould sets are then designed for these AAA models database. The CT scans of each patient were then imported into
the software package MIMICS 共Materialise, Belgium兲. This soft-
ware allows the transformation of 2D CT scans into realistic 3D
1
Corresponding author. models of exact geometry. This software uses a marching squares
Contributed by the Bioengineering Division of ASME for publication in the JOUR-
NAL OF BIOMECHANICAL ENGINEERING. Manuscript received February 7, 2007; final
algorithm to threshold and segment the regions of interest of the
manuscript received September 11, 2007; published online April 28, 2008. Review CT scan according to a predetermined grayscale value. Once seg-
conducted by B. Barry Lieber. mented, the software generates polylines around the segmented
Journal of Biomechanical Engineering Copyright © 2008 by ASME JUNE 2008, Vol. 130 / 034501-1
Downloaded 02 Apr 2009 to 193.1.104.8. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 1 Segmentation and polyline generation of CT scan. „a…
shows the full CT scan, while „b… is a close-up of the region of
interest. For the design of moulds, the AAA was regarded to be
the full volume of the lumen and intraluminal thrombus „ILT….
regions, to a user-controlled level of smoothing. An example of

this image segmentation and polyline generation can be seen in Fig. 3 Example machined mould piece of inner AAA model
Fig. 1. In this study, polylines were created with approximately 20
control points per scan, allowing optimum smoothing without the
loss of model accuracy. These polylines can then be exported as package AlphaCAM in order to generate the toolpath commands
initial graphics exchange specification 共IGES兲 format. Previous used to control the milling machine. Each mould is set up with the
work has utilized various other forms of reconstruction software, same reference points so as that each mould piece fits exactly
such as SCION IMAGE 共Release Beta 4.0.2, Scion Corporation, Fre- together, ensuring that the resulting model has an almost negli-
derick, MD兲关12兴. Validation of MIMICS against this work has been gible seamline.
performed, with a percentage difference of 1.2% between the re- Machining is performed by a three-axis computer numerical
construction methods. control 共CNC兲 milling machine. Moulds are machined from solid
Computer Aided Design (CAD). Polylines created in MIMICS aluminum blocks and are finished by hand to remove any un-
are imported into PROENGINEER WILDFIRE 2.0 共PTC, Parametric wanted burrs that result from the milling process. Figure 3 shows
Technology兲. Surfaces are then recreated along these polylines. an example machined mould piece. This illustration shows regions
These surfaces are then exactly split into two halves, thus creating where extensions have been included into the design in the proxi-
a two-piece mould set used in the manufacturing technique. Each mal and distal regions of the AAA, and also the inlet through
patient-specific mould design consists of two sets of moulds. The which the wax is poured. Necessary holes and vents were added to
first mould is designed to produce the casting wax model of the each model after machining.
AAA, and the second set to produce the outer silicone model. The Model Manufacture. All mould pieces were cleaned using ac-
outer mould is approximately 2 mm larger in all regions than that etone prior to use. The wax moulds were preheated to 40° C to
of the wax mould, so as to produce a silicone model with a 2 mm minimize the contraction of the wax upon pouring. A casting wax
thick wall. As the wall of an AAA can range in thickness of 共Castylene B581, REMET Corporation兲 was used to make the
0.23– 4.33 mm 关13兴, a wall thickness of 2 mm is a reasonable lumen casts. The wax lumen cast was then coated with Wacker
assumption and has been used in previous studies 关14兴. protective film SF18 共Wacker-Chemie GmbH兲. The lumen casts
Example mould designs can be seen in Fig. 2. Each outer were then placed into the outer moulds, which were coated with
mould design includes supports for the inner wax cast to ensure releasing agent 共Wacker Mould Release兲 and then clamped. The
location of the wax model inside the larger outer mould. Of the silicone rubber 共Wacker RT601兲 was then prepared and slowly
four AAAs used in this study, three AAAs were modeled without injected into the preheated outer mould. Silicone rubber was em-
the iliac arteries 共Patients A, B, and C兲, and one AAA with the ployed as the material due to its nonlinear behavior when sub-
iliac arteries included 共Patient D兲. For experimental studies injected to large strains 关15兴 and is believed to be a good arterial
volving stress analyses, the iliac arteries are believed to be unim- analog. The mould is then placed into an oven at a temperature of
portant, whereas for fluid dynamic and stent graft testing, the iliac 50° C and cured for 24 h. Once cured, the model is removed and
arteries are of paramount importance. Moulds designed without the temperature is increased to 100° C in order to melt the wax
the iliac arteries have cylindrical sections included both in the from the mould. The resulting silicone model is then thoroughly
proximal and distal regions of the AAA, to allow attachment to cleaned, dried, and inspected for defects. The full procedure can
experimental test rigs. be seen in the Appendix.
Computer Aided Manufacture (CAM). Once the mould sets
have been designed in PROENGINEER, the designs are exported Results
again in IGES format. These files are imported into the software
Sectioning the Model. Each model was sectioned at regular
intervals to assess the dimensional accuracy of the resulting sili-
cone model compared to the CAD model. Each silicone AAA
model was carefully split using a scalpel longitudinally along the
left and right sides, thus leaving each model in two halves. Each
half-model was then axially sliced at 10 mm intervals along the
length of the model, leaving a series of cross-sectional slices for
each patient-specific model.
Wall Thickness Measurements. For each cross-sectional slice
of the silicone model, the wall thickness was measured at four
90 deg equispaced locations around the edge. Therefore, wall
thickness was measured along the left, right, anterior, and poste-
rior walls of the full AAA model. Measurements were obtained
Fig. 2 Example mould designs of patient-specific AAAs. „a… is using a digital micrometer. Measurement readings ranged from 40
a mould design including iliac arteries and „b… without iliac to 60 readings per AAA model, depending on the patient. Mea-
arteries. surement results were then averaged for each patient-tailored sili-
034501-2 / Vol. 130, JUNE 2008 Transactions of the ASME
Table 1 Averaged wall thickness measurements at four locations on the AAA wall
Axial position
Anterior Posterior Right Left
Patient A Average wall thickness 共mm兲 1.87 1.97 2.09 2.55

Standard deviation 0.276 0.314 0.173 0.327
Percentage difference 6.95 1.78 4.23 21.47
Patient B Average wall thickness 共mm兲 2.12 2.29 2.09 2.31

Patient C Average wall thickness 共mm兲 2.18 2.17 2.53 2.30

Patient D Average wall thickness 共mm兲 2.65 2.11 2.38 1.97

cone model with the percentage difference between the actual Discussion
silicone model and the 2 mm wall CAD model noted. The stan-
This study describes a procedure for manufacturing patient-
dard deviation was also included in the results. The measurement
specific rubber models of AAAs both with and without the iliac
results can be seen in Table 1 and are summarized as a total model
wall thickness for each patient in Table 2. Percentage difference arteries. A 3D reconstruction technique using commercially avail-
refers to the difference between the silicone model wall thickness able software is coupled to a CAD/CAM technique to achieve the
and the original 2 mm wall thickness in the mould design. The desired mould designs capable of forming realistic AAAs using
AAA model of Patient D included the iliac arteries. the injection-moulding method. Previous studies 关14,16兴 have
used similar techniques to produce rubber models of vascular ves-
Wall Stress Distribution. Figure 4 shows the resulting von sels. The models developed in this study are of greater complex-
Mises wall stress distribution for Patient A and the region of re- ity. These reproducible silicone models can be utilized for experi-
sulting peak stress. The results show how the peak stress in the mental testing of vessel hemodynamics, wall stress analyses, and
AAA model was 0.533 MPa and was located on the anterior wall stent graft studies, all of which may contribute to experimental
of the AAA. The combination of finite element analysis 共FEA兲 validation of numerical work. This technique may allow other
results with validated experimental wall stress studies could fur- researchers to begin manufacturing realistic AAA silicone models
ther the use of numerical studies in the field of AAA rupture for use in their experimental work. In recent years, emphasis has
prediction. A more detailed study of the wall stress experienced in been placed on the use of numerical studies to attempt to predict
AAAs is currently in progress, and will expand on the preliminary AAA rupture. The use of experimental research into this area is
FEA results presented here. also of importance. Not only could these silicone AAA models
help in validation purposes but may also become an important
asset in AAA rupture prediction.
For each of the patient-specific models created, wall thickness
Table 2 Averaged wall thickness measurements for each is the most variable factor. It has been reported 关13兴 how the AAA
patient-specific AAA model wall can range in thickness from 0.23 mm to 4.33 mm, with aor-
tic wall thickness reported to range from 1.1 mm to 3.4 mm
Average wall thickness 共mm兲 Average percentage difference 关7,17,18兴. Average wall thickness over the four models was noted
to be 2.26⫾ 0.39 mm. In this study, wall thickness lies within the
Patient A 2.12 4.24%
range recorded by previous researchers 关7,17,18兴, and so can be
Patient B 2.20 9.01%
Patient C 2.29 12.51% deemed as acceptable. Wall thickness results also favorably com-
Patient D 2.22 11.09% pare with those found for a realistic aorta by O’Brien et al. 关14兴,
who recorded a wall thickness in realistic aortas of
2.39⫾ 0.32 mm. Although wall thickness appears to be within an
acceptable range, the moulds were designed to have a wall thick-
ness of 2 mm. The resulting silicone models differ from the mould
designs by an average of 9.21% in wall thickness, which can be
attributed to the contraction of the casting wax during solidifica-
tion and the thermal expansion of the silicone during the curing
process. These limitations of the technique were also observed by
O’Brien et al. 关14兴. This previous work recorded percentage dif-
ferences in wall thickness to mould design ranging from 20% for
a realistic straight section of an aorta to 58% for a section of a
saphenous vein. The results observed in this study are deemed
Fig. 4 Example FEA von Mises wall stress distribution for Pa- acceptable in that percentage differences are considerably less
tient A showing a region of peak stress on the anterior wall. The than those previously reported 关14兴. It should also be mentioned
corresponding mould piece and resulting silicone model of the that the models produced here are full AAA models and not
same patient can be seen on the right of this figure. straight vessel sections and, therefore, one would expect the per-
Journal of Biomechanical Engineering JUNE 2008, Vol. 130 / 034501-3
centage differences to be higher than those previously reported merical methods through photoelastic validation 关9兴 or also
关14兴. Therefore, confidence has been established in the method through experimental testing such as LDA or PIV.
used to produce these models. In conclusion, 3D reconstruction and CAD/CAM techniques
The issue of uniformity in wall thickness should also be ad- proved to be successful in the replication of patient-specific rub-
dressed. In the mould designs, the wall thickness was set at 2 mm, ber AAA models and may help contribute to the use of patient-
and therefore, the resulting silicone models should also have a specific AAA models in experimental testing. Therefore, the use
uniform wall thickness. Due to reasons mentioned above, that is, of silicone AAA models with uniformly thick walls will help to-
wax contraction and silicone expansion, the wall thickness varies ward the validation of numerical work, for both wall stress studies
in each AAA model. These differences in wall thickness can be and flow dynamics.
attributed to the complex and tortuous geometry of these patient-
specific AAA models. This limitation was also noted by O’Brien Acknowledgment
et al. 关14兴. Both O’Brien et al. 关14兴 and Chong et al. 关16兴 produced
idealized vascular models, and in these much simpler models, the The authors would like to thank 共i兲 the Irish Research Council
wall uniformity issue was easily overcome, thus highlighting the for Science, Engineering and Technology 共IRCSET兲 Grant No.
fact that realistic geometries increase the difficulty of not only the RS/2005/340, 共ii兲 Grant No. R01-HL-060670 from the US Na-
CAD/CAM process but also the model manufacture itself. tional Heart Lung and Blood Institute, 共iii兲 Dr. Eamon Kavanagh,
The use of a uniform wall in these experimental models may be an Endovascular surgeon at Midwestern Regional Hospital, Lim-
considered inappropriate, since it is known that the actual AAA erick, for his help in collecting patient data and background infor-
wall tissue can include various forms of arterial tissue 共calcified mation, and 共iv兲 Samarth Shah from the Centre for Vascular Re-
deposits and thrombus兲 and thus is usually nonuniform. While modelling and Regeneration.
these regions of both calcified tissue and thrombus can be detected
from the CT scan, their incorporation into the wall of the model Appendix
introduces additional complexity. First, incorporating regions of
varying material properties into the AAA wall results in greatly Realistic Model Manufacture
increased computations as the numerical equations used to solve
In creating the male wax models, the following steps are given.
for wall stress at these locations become extremely complex. Also,
the primary purpose of producing realistic silicone models of 1. The first sets of moulds are used to create the wax models.
known uniform wall is to experimentally validate numerical stud- 2. Clean the surface of the moulds with acetone; make sure it is
ies of the same AAA model. Most previous work regarding nu- free of loose debris and coat with mould release 共Ambersil
merical stress analyses of AAAs 关3–8,19–22兴 has conducted test- Formula 8, Chemcraft Industries Ltd. Dublin, Ireland兲.
ing using uniform walls. Our earlier work 关9兴 on idealized AAA 3. Bolt the two mould pieces together tightly.
models of known uniform wall thickness proved quite successful 4. Melt the casting wax 共Castylene B581, REMET Corpora-
and has paved the way for the introduction of realistic AAA mod- tion兲 on a hot plate at a temperature of around 150° C.
els to be tested using the same technique. This experimental pho- 5. Preheat moulds in oven to 40° C.
toelastic work was later numerically validated on an idealized 6. Position the mould at an angle of 45 deg to aid the liquid
AAA model by our group 关23兴, which confirmed the locations of wax to flow into the mould and minimize the risk of trapping
peak stress on the model. It is planned to revisit the concept of air, thus creating voids and bubbles.
nonuniformity of the silicone models using this described proce- 7. Pour the wax into the mould as slowly as possible to prevent
dure. Work has also begun in this laboratory on the inclusion of splashing, which can also create voids. As the cavity is
thrombus in the AAA models. filled, the mould is returned to the vertical position to finish
It has been suggested 关6兴 that the use of a fluid-structure inter- pouring the wax.
action 共FSI兲 approach to stress analysis may yield more accurate 8. The wax is then left to solidify at room temperature for up to
wall stress results than FEA alone. Some researchers have shown 4 h. During this cooling period, the mould is gently tapped
that the wall stress is increased by amounts ⬍1% 关21兴, whereas with a mallet to allow any trapped air to rise to the surface.
others have reported increases ranging from 12.5% 关6兴 to 20.5% 9. As the wax cools and solidifies, additional wax is added to
关20兴. As there are conflicting results in the benefits of FSI in wall the mould to ensure a complete wax model.
stress studies, these uniform-walled AAA models can help toward 10. Open the moulds and carefully remove the wax model from
validating both methods of stress studies. Work has begun within the mould.
our laboratory on the use of FSI, employing mesh-based parallel
code coupling interface 共MPCCI 3.0.6, Fraunhofer SCAI, Germany兲 In creating the silicone model, the following steps are given.
software, which couples both ABAQUS and FLUENT together to ob-
tain realistic wall stress values. Notably, the use of a uniform wall 1. Mix the silicone in the required ratio of silicone and curing
is widespread among researchers conducting FSI studies agent 共9:1兲.
关5,6,19–22兴, and so this future work will allow the numerical and 2. Hand mixing the material components for a period of 2 min
experimental validation of wall stress in realistic AAA models is sufficient. This liquid silicone will represent the aorta
based on the uniform wall thickness. wall. The liquid silicone contains air bubbles once hand
mixed which must be removed. The working time for the
mixed, components at room temperature is around 90 min.
Conclusion 3. In order to remove the trapped bubbles, place the container
The procedure for the manufacture of patient-specific AAA of liquid silicone into a freezer until all bubbles have been
models has been described. Confidence has been established in the naturally removed. Duration of time in the freezer depends
reproducibility of the rubber models, and the limitations noted. In on the viscosity of the liquid silicone and can vary from
general, rubber models of good geometrical accuracy can be pro- 1 h to 3 h.
duced by sensible mould design and use of controlled parameters 4. Once all the air has been removed, suck all the liquid sili-
in the silicone production. Models showed a maximum percentage cone into a 60 ml syringe.
difference of 9.21% between the designed moulds and the result- 5. Clean the second set of aluminum moulds with acetone.
ing silicone models. Uniformity of wall thickness proved to be the 6. Spray on a coat of silicone mould release 共Ambersil Formula
most difficult parameter to control, with finished silicone models 8兲 onto both the aluminum moulds.
always inspected and assessed before commencement of experi- 7. Carefully remove any excess material and flash from the
mental testing. This technique may aid in the validation of nu- wax model.
034501-4 / Vol. 130, JUNE 2008 Transactions of the ASME
8. Coat Wacker protective film SF18 共Wacker-Chemie GmbH兲 mental Assessment of Stress Patterns in Abdominal Aortic Aneurysms Using
the Photoelastic Method,” Strain, 40共4兲, pp. 165–172.
onto the wax model. Once applied, leave it for 2 min. 关10兴 Seong, J., Sadasivan, C., Onizuka, M., Gounis, M. J., Christian, F., Miskolczi,
9. Position the inner wax model inside the outer aluminum L., Wakhloo, A. K., and Lieber, B. B., 2005, “Morphology of Elastase-Induced
mould ensuring that the wax model sits in a way that will Cerebral Aneurysm Model in Rabbit and Rapid Prototyping of Elastomeric
allow uniform wall thickness. Transparent Replicas,” Biorheology, 42共5兲, pp. 345–361.
10. Bolt the two aluminum moulds together tightly. 关11兴 Loth, F., Jones, S. A., Giddens, D. P., Bassiouny, H. S., Glagov, S., and Zarins,
C. K., 1997, “Measurements of Velocity and Wall Shear Stress Inside a PTFE
11. Seal around the edges of the mould with sealant to avoid Vascular Graft Model Under Steady Flow Conditions,” ASME J. Biomech.
any unwanted leakage from the mould. Eng., 119共2兲, pp. 187–194.
12. Slowly inject the liquid silicone into the aluminum cast 关12兴 Morris, L., Delassus, P., Callanan, A., Walsh, M., Wallis, F., Grace, P., and
using the filled 60 ml syringe. McGloughlin, T., 2005, “3D Numerical Simulation of Blood Flow Through
13. Once the liquid silicone is injected, place the mould in an Models of the Human Aorta,” ASME J. Biomech. Eng., 127, pp. 767–775.
关13兴 Raghavan, M. L., Kratzberg, J., de Tolosa, E. M. C., Hanaoka, M. M., Walker,
oven at a temperature of 50° C for a period of 24 h. P., and de Silva, E. S., 2005, “Regional Distribution of Wall Thickness and
14. Once cured, open the cast and carefully remove the silicone Failure Properties of Human Abdominal Aortic Aneurysm,” J. Biomech.,
and wax model. 39共16兲, pp. 3010–3016.
15. Place the silicone and wax model into the oven at a tem- 关14兴 O’Brien, T., Morris, L., O’Donnell, M., Walsh, M., and McGloughlin, T.,
2005, “Injection-Moulded Models of Major and Minor Arteries: The Variabil-
perature of 100° C to melt the wax from the mould. ity of Model Wall Thickness Owing to Casting Technique,” Proc. Inst. Mech.
Eng., Part H: J. Eng. Med., 219, pp. 381–386.
References 关15兴 Marins, P. A. L. S., Natal Jorge, R. M., and Ferreira, A. J. M., 2006, “A
Comparative Study of Several Material Models for Prediction of Hyperelastic
关1兴 Sakalihasan, N., Limet, R., and Defawe, O. D., 2005, “Abdominal Aortic Properties: Application to Silicone-Rubber and Soft Tissues,” Strain, 42共3兲,
Aneurysm,” Lancet, 365共9470兲, pp. 1577–1589. pp. 135–147.
关2兴 Johnston, K. W., Rutherford, R. B., Tilson, M. D., Shah, D. M., Hollier, L., 关16兴 Chong, C. K., Rowe, C. S., Sivanesan, S., Rattray, A., Black, R. A., Shortland,
and Stanley, J. C., 1991, “Suggested Standards for Reporting on Arterial An- A. P., and How, T. V., 1999, “Computer Aided Design and Fabrication of
eurysms. Subcommittee on Reporting Standards for Arterial Aneurysms, Ad Models for In Vitro Studies of Vascular Fluid Dynamics,” Proc. Inst. Mech.
Hoc Committee on Reporting Standards, Society for Vascular Surgery and
Eng., Part H: J. Eng. Med., 213, pp. 1–4.
North American Chapter, International Society for Cardiovascular Surgery,” J.
关17兴 Nicholls, S. C., Gardner, J. B., Meissner, M. H., and Johansen, H. K., 1998,
Vasc. Surg., 13, pp. 452–458.
“Rupture in Small Abdominal Aortic Aneurysms,” J. Vasc. Surg., 28, pp.
关3兴 Fillinger, M. F., Raghavan, M. L., Marra, S. P., Cronenwett, J. L., and
884–888.
Kennedy, F. E., 2002, “In Vivo Analysis of Mechanical Wall Stress and Ab-
关18兴 Gullace, G., and Ruffa, F., 1995, “Effect of Pravastatin on Abdominal Aorta
dominal Aortic Aneurysm Rupture Risk,” J. Vasc. Surg., 36, pp. 589–597.
关4兴 Fillinger, M. F., Marra, S. P., Raghavan, M. L., and Kennedy, F. E., 2003, and Carotid Wall Thickness in Dyslipidemia,” J. Am. Coll. Cardiol., 25共2兲, pp.
“Prediction of Rupture Risk in Abdominal Aortic Aneurysm During Observa- 369A–370A.
tion: Wall Stress Versus Diameter,” J. Vasc. Surg., 37, pp. 724–732. 关19兴 Li, Z., and Kleinstreuer, C., 2005, “Blood Flow and Structure Interactions in a
关5兴 Di Martino, E. S., Guadagni, G., Fumero, A., Ballerini, G., Spirito, R., Big- Stented Abdominal Aortic Aneurysm Model,” Med. Eng. Phys., 27, pp. 369–
lioli, P., and Redaelli, A., 2001, “Fluid-Structure Interaction Within Realistic 382.
3D Models of the Aneurysmatic Aorta as a Guidance to Assess the Risk of 关20兴 Scotti, C. M., and Finol, E. A., 2007, “Compliant Biomechanics of Abdominal
Rupture of the Aneurysm,” Med. Eng. Phys., 23, pp. 647–655. Aortic Aneurysms: A Fluid-Structure Interaction Study,” Comput. Struct., 85,
关6兴 Papaharilaou, Y., Ekaterinaris, J. A., Manousaki, E., and Katsamouris, A. N., pp. 1097–1113.
2007, “A Decoupled Fluid Structure Approach for Estimating Wall Stress in 关21兴 Leung, J. H., Wright, A. R., Cheshire, N., Crane, J., Thom, S. A., Hughes, A.
Abdominal Aortic Aneurysms,” J. Biomech., 40共2兲, pp. 367–377. D., and Xu, Y., 2006, “Fluid Structure Interaction of Patient Specific Abdomi-
关7兴 Raghavan, M. L., Vorp, D. A., Federle, M. P., Makaroun, M. S., and Webster, nal Aortic Aneurysms: A Comparison With Solid Stress Models,” Biomed.
M. W., 2000, “Wall Stress Distribution on Three-Dimensionally Reconstructed Eng. Online, 5;33.
Models of Human Abdominal Aortic Aneurysm,” J. Vasc. Surg., 31, pp. 760– 关22兴 Gao, F., Watanabe, M., and Matsuzawa, T., 2006, “Stress Analysis in a Lay-
769. ered Aortic Arch Model Under Pulsatile Blood Flow,” Biomed. Eng. Online,
关8兴 Vorp, D. A., Raghavan, M. L., and Webster, M. W., 1998, “Mechanical Wall 5;25.
Stress in Abdominal Aortic Aneurysm: Influence of Diameter and Asymme- 关23兴 Callanan, A., Morris, L., and McGloughlin, T., 2004, “Numerical and Experi-
try,” J. Vasc. Surg., 27共4兲, pp. 632–639. mental Analysis of an Idealized Abdominal Aortic Aneurysm,” European So-
关9兴 Morris, L., O’Donnell, P., Delassus, P., and McGloughlin, T., 2004, “Experi- ciety of Biomechanics, S-Hertogenbosch, Netherlands.
Journal of Biomechanical Engineering JUNE 2008, Vol. 130 / 034501-5
BioMedical Engineering OnLine BioMed Central
Research Open Access

A comparison of modelling techniques for computing wall stress in
abdominal aortic aneurysms
Barry J Doyle*, Anthony Callanan and Timothy M McGloughlin
Address: Centre for Applied Biomedical Engineering Research (CABER), Department of Mechanical and Aeronautical Engineering and Materials
and Surface Science Institute, University of Limerick, Ireland
Email: Barry J Doyle* - barry.doyle@ul.ie; Anthony Callanan - anthony.callanan@ul.ie; Timothy M McGloughlin - tim.mcgloughlin@ul.ie
* Corresponding author
Published: 19 October 2007 Received: 28 February 2007

Accepted: 19 October 2007
BioMedical Engineering OnLine 2007, 6:38 doi:10.1186/1475-925X-6-38
This article is available from: http://www.biomedical-engineering-online.com/content/6/1/38
© 2007 Doyle et al; licensee BioMed Central Ltd.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0),
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
Background: Aneurysms, in particular abdominal aortic aneurysms (AAA), form a significant
portion of cardiovascular related deaths. There is much debate as to the most suitable tool for
rupture prediction and interventional surgery of AAAs, and currently maximum diameter is used
clinically as the determining factor for surgical intervention. Stress analysis techniques, such as finite
element analysis (FEA) to compute the wall stress in patient-specific AAAs, have been regarded by
some authors to be more clinically important than the use of a "one-size-fits-all" maximum
diameter criterion, since some small AAAs have been shown to have higher wall stress than larger
AAAs and have been known to rupture.
Methods: A patient-specific AAA was selected from our AAA database and 3D reconstruction
was performed. The AAA was then modelled in this study using three different approaches, namely,
AAA(SIMP), AAA(MOD) and AAA(COMP), with each model examined using linear and non-linear
material properties. All models were analysed using the finite element method for wall stress
distributions.
Results: Wall stress results show marked differences in peak wall stress results between the three
methods. Peak wall stress was shown to reduce when more realistic parameters were utilised. It
was also noted that wall stress was shown to reduce by 59% when modelled using the most
accurate non-linear complex approach, compared to the same model without intraluminal
thrombus.
Conclusion: The results here show that using more realistic parameters affect resulting wall
stress. The use of simplified computational modelling methods can lead to inaccurate stress
distributions. Care should be taken when examining stress results found using simplified
techniques, in particular, if the wall stress results are to have clinical importance.
Background comes. Aneurysm, meaning widening, can be defined as a

Cardiovascular disease is the leading cause of morbidity permanent and irreversible localised dilatation of a vessel
and premature deaths of modern era medicine and aneu- [1]. Although this disease can form in any blood vessel,
rysms are a major contributor to the poor clinical out- artery or vein, the more serious aneurysms occur in the
Page 1 of 12
(page number not for citation purposes)
BioMedical Engineering OnLine 2007, 6:38 http://www.biomedical-engineering-online.com/content/6/1/38
abdominal aorta, the brain arteries, and the heart and tho- material properties of the vessel as linearly elastic and
racic aorta. The vast majority of these aneurysms occur in excluding the intraluminal thrombus (ILT) from the
the abdominal aorta, and are termed Abdominal Aortic model. One patient-specific AAA was examined in this
Aneurysm (AAA). There is much debate amongst research- study. The model was examined from various aspects of
ers and surgeons as to the most appropriate criteria for complexity, starting with a simplified approach,
surgical intervention of unruptured AAAs [2-6]. Currently, AAA(SIMP), followed by a moderate method,
the maximum diameter of the aneurysm is the most com- AAA(MOD), and then a complex technique, AAA(COMP),
monly used determinant of surgical intervention, with cli- with results and differences contrasted and compared. In
nicians opting for surgery when the AAA exceeds 5 cm in this study, simplified refers to the use of common simpli-
transverse diameter [4,5,7]. Although this maximum fying assumptions. Moderate refers to the use of a limited
diameter criterion can be justified, as the rupture risk for amount of simplifying assumptions, with complex infer-
an AAA is clearly related to its maximal diameter [5,8], it ring the inclusion of ILT. All models were examined using
is also known that small AAAs can rupture [5,9-11]. both linear and non-linear material properties.
Nicholls et al. [11] reported in one study that 10% to 24%
of ruptured aneurysms were 5 cm or less in diameter, and Methods
therefore there is clearly a need for an improved predictor A patient suffering from an AAA was selected from our
of rupture. AAA database which included modest levels of intralumi-
nal thrombus (ILT). The study subject was male, and the
It is believed by many researchers that stress analyses are AAA maximum transverse diameter was 5.1 cm, and so
more accurate indicators than diameter for predicting rup- had exceeded the 5 cm threshold for surgical intervention.
ture [2,5-7,12-14], and it is clear that a reliable method of The total length of the AAA was 13.2 cm, had a total vol-
predicting AAA rupture has definite clinical importance ume of 176.7 cm3, of which 67.5 cm3 was ILT. For clini-
[2,3,6,7,13]. cally meaningful stress analysis of AAAs, patient-specific
computed tomography (CT) scans were obtained for the
It has been previously documented how AAA formation is patient. This CT data was then reconstructed using com-
accompanied by an increase in wall stress [2,7], and a mercially available software (Mimics v10.0, Materialise,
decrease in wall strength [15-17]. It has been identified Belgium). In total, six models were created and analysed.
that aneurysm wall stress does not follow the traditional Three different methods of reconstruction were utilised to
Law of Laplace, in that AAAs with equivalent diameters create the AAA, with each model examined using both lin-
and pressures could have largely different stress distribu- ear and non-linear material properties. For the purpose of
tions [2,3,7,13]. Researchers have been using stress analy- this study, the AAAs modelled using the first set of simpli-
sis techniques, such as finite element analysis (FEA), to fying assumptions will be referred to as AAA(SIMP)L and
compute stresses within AAAs for some time now AAA(SIMP)NL, with the subscripts L and NL referring to
[2,5,7,13], but this work has yet to be validated experi- linear and non-linear material properties, respectively.
mentally using realistic AAA models. Morris et al. [18] The second set of AAAs will be referred to as AAA(MOD)L
showed experimentally the stress contours in an idealised and AAA(MOD)NL, due to the use of moderate simplifying
AAA using the photoelastic method, which was later vali- assumptions. The third set of AAAs, which used a more
dated numerically by Callanan et al. [19]. Previous complex method of analysis will be referred to as
numerical studies have shown how asymmetry and diam- AAA(COMP)L and AAA(COMP)NL. All stress analyses were
eter affect wall stress [2], and also how this wall stress is performed using the finite element analysis software
distributed within a realistic AAA [5,7,13,20]. Dobrin [21] ABAQUS v6.6-2 (Dassault Systemes, SIMULIA, Rhode
suggested that the presence of ILT neither reduces the Island, USA) on a standard desktop with a 3.2 GHz proc-
luminal pressure exerted on the wall nor offers a retractive essor and 4GB of RAM.
force and thus has no effect on the wall stress. This ILT was
later reported to significantly reduce the stress acting on 3D reconstruction
the AAA wall [13] and to act as a mechanical cushion Spiral CT data was then used to reconstruct the infrarenal
[13,17], although some researchers dispute this [22]. section of the aorta. As CT scanning is routinely per-
formed on AAA patients scheduled for repair, collection of
The purpose of this study is to analyse and compare the this information involved no extra participation by the
differences associated with finite element analysis (FEA) study subject. Digital files in Digital Imaging and Com-
modelling of AAAs. Many researchers have used simpli- munications in Medicine (DICOM) file format, contain-
fied AAA models in their studies [2,23-25], compared to ing cross-sectional information was then imported to
others who have used much more complex, realistic Mimics for reconstruction. A Hounsfield unit (HU)
methods [2,3,7,13,20,26-28]. The more common simpli- thresholding technique was then applied to each CT slice
fying assumptions include treating the biomechanical in order to identify the region of interest. A HU value of
Page 2 of 12
226 is sufficient to identify the lumen region of the AAA ness throughout the model and was set to 1.5 mm
due to the non-ionic contrast. This contrast dye is an [2,26,29].
organic solution that makes internal bodily structures vis-
ible. The ILT regions of the AAA must be assigned a lower AAA(MOD)
HU value, as the material has a pixel intensity that is closer When reconstructing the AAA(MOD) models, the surface
to fat than bone. A HU value of 0 is sufficient to distin- polylines created using Mimics v10.0, were exported to
guish the ILT from the surrounding tissue. For all models ProEngineer Wildfire 3.0 (Parametric Technology Corpo-
reconstructed, the iliac arteries have been omitted as with ration, Needham, M.A., USA) in order to create a 3D wall.
previous research [5]. Figure 1 shows the conversion from The original polylines were offset by 1.5 mm [2,26,29] so
CT scan to 3D model. On the left the thresholding and as to create a second artificial surface offset from the orig-
segmentation process highlights both the ILT and lumen inal surface. These surfaces were then connected so as to
of the AAA, with the software generating 3D reconstruc- form a realistic uniform 3D aortic wall.
tions, shown on the right. For the AAA(SIMP) and
AAA(MOD) models, the ILT was omitted, and so the AAA(COMP)
reconstruction consisted of a single layer wall of uniform For this AAA, the ILT region was included into the models.
thickness. Therefore, the ILT is a completely different entity to that of
the AAA wall, with the two sections tied together using
AAA(SIMP) realistic constraints. This is a more realistic representation
For this set of AAAs, the ILT was omitted from the recon- of the actual AAA with the ILT. An artificial wall was cre-
struction, and was instead incorporated into the full AAA ated to represent the AAA wall and, like earlier, was set to
model as lumen. Previous work [5,6] has also taken this be a uniform 1.5 mm in thickness [2,26,29].
approach to AAA modelling. The thickness of the aortic
wall is not easily identifiable from CT scans. Previous Model smoothing
research [2,5-7,18,19,25,26,29] has used wall thickness Each of these 3D models were initially quite rough con-
varying from 1–2 mm. As no information was available taining sharp edges and surface artefacts, and required
from the CT scans about the wall thickness of each AAA, smoothing to ensure that the model could be readily
in this study the wall was assumed to be of uniform thick- meshed and that stress analyses could be performed. The
Figure CT
Typical 1 scan and 3D reconstruction of patient suffering from an AAA
Typical CT scan and 3D reconstruction of patient suffering from an AAA. On the left shows a typical CT scan after
segmentation using Mimics v10.0, with the ILT (red) and lumen (yellow) clearly distinguishable. 3D reconstruction of examined
AAA is shown on right.
Page 3 of 12
method of reconstruction and smoothing was validated ference between the resulting rough and smooth models.
from previous work by our group, [30,31] and it was These AAA surfaces were then exported in IGES file format
shown that there are negligible differences between the for further work.
two methods. In order to determine the optimum level of
smoothing for these reconstructions, four degrees of Biomechanical material properties
smoothing were examined. These four smoothing levels AAA(SIMP)
were based on axial smoothing of individual polyline Firstly, the AAA material was assumed to be homogenous
slices created from the CT scans. Four models were then and isotropic with linear elastic material properties. This
reconstructed. Model 1 contained 7 control points per model is referred to as AAA(SIMP)L. Although human
polyline slice, model 2 had 20 control points, model 3 arterial tissue acts like a non-linear material, at pressures
had 50 control points, and model 4 had 70 control points above 80 mmHg (10.67 KPa) the aorta behaves more like
per slice. The more control points per polyline decreases a linearly elastic material [23,24,32,33]. The modulus of
the surface smoothness of the resulting AAA. Figure 2 elasticity applied to the AAA wall was 4.66 MPa, as used
shows the difference in polylines between the various in previous research [24,27]. Many researchers have mod-
degrees of smoothing. Non-linear stress analyses were elled the aorta as linearly elastic in order to examine the
then performed on each of the four models. Resulting wall influence of geometry and various other parameters on
stress distributions are shown in Figure 3, with stress wall stress [2,4,23-25,34].
results normalised to the peak stress experienced in the
roughest model, namely model 4. Mesh independence The same model was then analysed using non-linear elas-
was performed on all models. From these results, model 2 tic material properties, and is referred to as AAA(SIMP)NL.
was deemed to be the optimum level of smoothing as The AAA wall was modelled as a homogenous isotropic
unwanted surface detail is removed without causing hyperelastic material using the finite strain constitutive
unnecessary over-smoothing. Figure 4 illustrates the dif- model proposed by Raghavan and Vorp [3] in Eqn. 1.
These material properties have been utilised for many
stress analysis studies [5-7,13,20,26-28].
W = C1(IB - 3) + C2(IB - 3)2 (1)
Where, W is the strain energy and IB is the first invariant of

the left Cauchy-Green tensor B (IB = tr B). The constants
were set to the population mean values C1 = 174,000 Pa
(0.174 MPa) and C2 = 1,881,000 Pa (1.881 MPa). Aortic
tissue is also known to be nearly incompressible with a
Poisson's ratio of approximately 0.49 [2,24,35].
AAA(MOD)
For this set of AAAs, the wall was again assigned both lin-
ear and non-linear material properties, referred to as
AAA(MOD)L and AAA(MOD)NL, respectively.
AAA(MOD)L was assigned an Young's modulus of 4.66
MPa, like earlier, and AAA(MOD)NL utilized the non-lin-
ear model proposed by Raghavan and Vorp [3].
AAA(COMP)
In the AAA(COMP) models, the ILT was included as a sep-
arate entity in the AAA model, and therefore required its
own material property. Previous work has used the theory
Figure
Effect
the CTofscans
2axial smoothing on the polylines constructed from of linear elasticity to model the ILT [4,28], and so for the
Effect of axial smoothing on the polylines con- AAA(COMP)L the ILT was assigned a Young's modulus of
structed from the CT scans. Model 1 consists of 7 con- 0.11 MPa, and a Poisson's ratio of 0.45. The AAA wall was
trol points, Model 2 has 20 control points, Model 3 has 50 also assumed to be linearly elastic with the same material
control points, and Model 4 has 70 control points per properties as described earlier. For the AAA(COMP)NL
polylines slice. Green line is the original polyline from the CT model, the ILT was modelled as a hyperelastic material
scan slice, with the red line being the new smoothed polyline. using the material constants derived from 50 ILT speci-
mens from 14 patients performed by Wang et al. [36]. The
Page 4 of 12
Figure 3 wall stress distributions of the various degrees of smoothing examined

Resulting
Resulting wall stress distributions of the various degrees of smoothing examined. Wall stress results are normal-
ised to the peak stress found in Model 4. Black mark indicates abnormal locations of elevated wall stress in Models 3 and 4.
AAA wall was again analysed using the non-linear model

proposed by Raghavan and Vorp [3]. Therefore, the model
consisted of two non-linear materials, both with realistic
material properties.
Mesh generation
AAA(SIMP)
Once the AAA was imported into ABAQUS v.6.6-2 for
stress analysis, a mesh was generated on the AAA model.
The elements chosen for the application were linear shell
elements [2,3,23]. The surface was partitioned, by divid-
ing the model into different regions in order to optimise
the mesh creation. Each shell element was assigned a uni-
form thickness of 1.5 mm [2,26,29]. Mesh independence
was performed in order to determine the optimum
number of elements. In order to gain confidence in the
mesh size of each AAA model, the number of elements
was incrementally increased and the peak wall stress com-
puted. The optimum mesh size was determined once the
Figureof4smoothing on the reconstructed AAA
Effect peak stress did not increase by more than 2%. This
Effect of smoothing on the reconstructed AAA. The method of convergence has been used in previous studies
optimum smoothing factor removes unwanted surface detail [37,38]. It was noted that the locations of the peak stress
without over-smoothing the AAA. (A) Shows the rough
did not significantly alter when examining each model for
model without smoothing and (B) shows the effect of a 20
control point axial smoothing factor. Model shown in the lat- mesh convergence. Both the AAA(SIMP)L and the
eral view. AAA(SIMP)NL models were meshed using shell elements,
Page 5 of 12
with element numbers for each model in the order of stress results for the AAA(SIMP), AAA(MOD) and
17000 and 30000, respectively. AAA(COMP) models can be seen in Figures 5, 6 and 7. In
these figures, wall stress results were normalised by using
AAA(MOD) the peak stress experienced in the linearly elastic model of
In order to mesh the 3D wall of the AAA(MOD) models, each case. In each case, the linearly elastic model returned
3D stress elements were required in order to mesh a higher peak stress than the corresponding non-linear
throughout the thickness of the AAA wall. The elements model. The location of peak stress remained the same in
used for this part of the study were quadratic, 10-noded, all models except the AAA(SIMP) models, where the loca-
tetrahedral elements. Mesh convergence was achieved tion of peak stress shifted from a centred anterior location
using the same approach as described earlier. In this case, to a left lateral location. For both linearly elastic and non-
the AAA(MOD)L and the AAA(MOD)NL models required linearly elastic AAA(MOD) and AAA(COMP) models, the
29000 elements. The location of peak stress did not signif- peak stress region was located on the inner surface of the
icantly alter in location during the convergence study. AAA wall. When using 3D stress elements the stress is not
interpolated through the wall thickness, as with shell ele-
AAA(COMP) ments, but rather at the individual integration points of
As with the AAA(MOD) models, the elements employed the element. For the AAA(COMP) models the peak stress
here were quadratic tetrahedral 3D stress elements, with was located at regions between the intersection of the ILT
10-nodes per element. The AAA wall and ILT regions were region and the AAA wall.
meshed using a master-slave contact scenario, where the
slave region (ILT) has a larger number of elements than In both the AAA(SIMP) models, the peak stress occurred
the master region (AAA wall). Both the AAA(COMP)L and at regions of relatively maximum diameter, rather than at
the AAA(COMP)NL models converged at 13000 elements inflection points along the surface of the AAA sac. For all
using the same convergence criteria as described previ- AAA(MOD) and AAA(COMP) cases studied, the peak
ously. stress occurred at an inflection point on the inner surface
of the AAA. An inflection point is defined as points on the
Forces and boundary conditions AAA surface at which the local AAA wall shape changes
The blood pressure within the AAA acts on the AAA inner
wall, therefore, the pressure was applied to the inner sur-
face of all the virtual AAA models studied. A static peak
pressure of 120 mmHg (16 KPa) was used. Ideally,
patient-specific blood pressures would be applied to each
particular AAA, by measuring blood pressure at the time
of CT scan, but for this study the standard peak pressure
of 120 mmHg was felt to be sufficient. The shear stress
induced by blood flow was neglected in this study [7,24],
although the effects of blood flow have been shown to
reduce wall stress by 10% in uniformly thick walled ideal
models and by up to 30% in variable wall thickness mod-
els [34]. In order to simulate the attachment of the AAA to
the aorta at the renal junction and iliac bifurcation, the
AAA model was fully constrained in the proximal and dis-
tal regions. Residual stresses that may exist within the aor-
tic wall in vivo and tethering forces on the posterior surface
caused by the lumbar arteries were neglected. Bulging is
on the anterior surface due to the constraint the spinal
cord places on posterior dilation.
Results
In order to easily observe and visualise the resulting wall mmHg
Figure
and
von AAA(SIMP)
5 wallNL
Mises models
stress at an internal
distributions pressure
for both of 120
the AAA(SIMP)L
von Mises wall stress distributions for both the
stress of each AAA model, contours of the von Mises stress
AAA(SIMP)L and AAA(SIMP)NL models at an internal
were plotted on the surface of each AAA model. The von pressure of 120 mmHg. Wall stress results are normal-
Mises stress is a stress index especially suited for failure ised to the peak stress found AAA(SIMP)L. The black mark
analysis, as stress is a tensor quantity with nine compo- indicates the region of peak wall stress. Models are shown in
nents, with the von Mises stress being a combination of the anterior view.
these components [7]. The normalised computed wall
Page 6 of 12
AAA(MOD)
von
of 120
Figure
Mises
mmHg
6 wall
L andstress
AAA(MOD)
distributions
NL models
foratthe
an both
internal
thepressure mmHg
Figure
von AAA(COMP)
and Mises
7 wall stress
NL models
distributions
at an internal
for both
pressure
the AAA(COMP)
of 120 L
von Mises wall stress distributions for the both the von Mises wall stress distributions for both the
AAA(MOD)L and AAA(MOD)NL models at an internal AAA(COMP)L and AAA(COMP)NL models at an inter-
pressure of 120 mmHg. Wall stress results are normal- nal pressure of 120 mmHg. Wall stress results are nor-
ised to the peak stress found AAA(MOD)L. The black mark malised to the peak stress found AAA(COMP)L. The black
indicates the region of peak wall stress. Models are shown in mark indicates the region of peak wall stress. Models are
the anterior view. shown in the anterior view.
from convex to concave [2]. Peak stress located at inflec- mented. This drop in peak stress when modelled from a
tion points has been previously reported in idealised non-linear approach followed through to the second case
models, both numerically [2,19,34] and experimentally also, that is, AAA(MOD). In this comparison of AAAs with
[18]. In general, wall stress reduces across the entire sur- 3D 1.5 mm thick uniform walls, the peak stress reduced
face of the AAA when the material properties become non- from 2.232 MPa to 1.034 MPa. In the third case,
linear. The recorded peak stresses for each case are also AAA(COMP), again peak stress was reduced when realistic
noticeably different. The peak stress results, along with material properties were utilised. In this case the peak
location of peak stress and overall computational time stress reduced from 0.8036 MPa to 0.4219 MPa. These
can be seen in Table 1. Computational time here is large reductions in peak stress for each comparison can be
rounded to the nearest minute. Peak stresses varied con- attributed to the use of realistic material properties, as the
siderably depending on the modelling approach. By com- geometry of the models are identical. The peak stress in
paring the first approach of AAA reconstruction consisting the AAA(SIMP), AAA(MOD) and AAA(COMP) models
of a single layer of shell elements, AAA(SIMP), the peak reduced by 34%, 55% and 47%, respectively, when
stress reduced from 0.8833 MPa to 0.5814 MPa when accounting for material non-linearity. This results in an
more realistic non-linear material properties were imple- average reduction in peak stress of 45% in all models.
Table 1: Comparison of peak wall stress, location of peak wall stress and FEA computational time for the AAA(SIMP), AAA(MOD) and
the AAA(COMP) models.
Peak Wall Stress (MPa) Location of Peak Stress CPU Time (min)
AAA(SIMP)L 0.8833 Centre of anterior region 1

AAA(SIMP)NL 0.5814 Left lateral region 9
AAA(MOD)L 2.282 Inflection point on inner surface of anterior region 4
AAA(MOD)NL 1.034 Inflection point on inner surface of anterior region 19
AAA(COMP)L 0.8036 Inflection point on inner surface of anterior region 8
AAA(COMP)NL 0.4291 Inflection point on inner surface of anterior region 82
Page 7 of 12
The effect of geometrical modelling parameters can be

seen in Figure 8 for the linearly elastic models, and in Fig-
ure 9 for the non-linearly elastic models. The location of
peak stress is also indicated in these figures. The wall
stresses were normalised using the peak wall stress experi-
enced in the linear elastic study (Figure 8) and the peak
stress obtained in the non-linear elastic study (Figure 9).
By comparing the three models examined from a linearly
elastic approach, the AAA(MOD)L returned the highest
peak stress of 2.282 MPa. Both the AAA(SIMP)L and
AAA(COMP)L returned similar peak stresses of 0.8833
MPa and 0.8036 MPa, respectively.
The "cushioning effect" of the ILT which has been previ-

ous documented by other researchers [13,17,35-40] can
be observed by comparing the linear and non-linear cases Figurein9all the
Normalised
stress wallnon-linearly
stress results
elastic
and models
locationsexamined
of peak wall
of both the AAA(MOD) models and the AAA(COMP) Normalised wall stress results and locations of peak
models. The AAA(MOD)L and the AAA(COMP)L models wall stress in all the non-linearly elastic models
examined. Wall stress results are normalised to the peak
are similar in material properties except for the inclusion
stress found AAA(MOD)NL. The figure shows the effect of
of thrombus. There is a 65% reduction in stress when modelling parameters and inclusion of the ILT on peak wall
including the ILT in the linear case. This "cushioning stress. The black mark indicates the region of peak wall
effect" is also evident in the non-linear cases, stress. Location of peak stress for the AAA(MOD)NL and
AAA(MOD)NL and AAA(COMP)NL. Here, inclusion of ILT AAA(COMP)NL models are on the inner surface of the AAA
reduces peak stress by 59%. wall. All models are shown in the anterior view.
Discussion
In this study, one patient-specific AAA was selected for rig- niques, with each technique studied using both linearly
orous finite element analysis. The 3D reconstruction and elastic and non-linearly elastic material properties. This
stress analysis was performed using three different tech- study is believed to be the first paper to contrast and com-
pare established and widely used techniques of comput-
ing AAA wall stress distributions. The first technique is
classed as AAA(SIMP), the second as AAA(MOD), and the
third as AAA(COMP). Each model was then thoroughly
analysed and compared to determine any differences in
both peak wall stress and also wall stress distribution. The
FEA technique used in this study has been validated both
experimentally by Morris et al. [18] and numerically by
Callanan et al. [19] in idealised AAA models in previous
research by our group. Briefly, these previous studies used
an idealised AAA model designed by our group using real-
istic dimensions from a large population study [41]. Alu-
minium moulds were then created using a CAD/CAM
procedure, and experimental models manufactured using
a technique described in previous research by both Morris
et al. [18] and Doyle et al. [31]. The photoelastic method
Figurein8all the
Normalised
stress walllinearly
stress results
elastic models
and locations
examined
of peak wall of determining stress distributions was then employed. A
Normalised wall stress results and locations of peak numerical model of the same AAA was then analysed
wall stress in all the linearly elastic models examined.
using the finite element method, with good agreement
Wall stress results are normalised to the peak stress found in
AAA(MOD)L. The figure shows the effect of modelling param- between the two. Peak stresses were found to occur at
eters and inclusion of the ILT on peak wall stress. The black regions of inflection on the surface of the AAA rather than
mark indicates the region of peak wall stress. Location of at regions of maximum diameter. This result was also
peak stress for the AAA(MOD)L and AAA(COMP)L models are reported by other authors using both FEA [2] and fluid-
on the inner surface of the AAA wall. All models are shown structure interaction (FSI) [34]. 3D solid stress elements
in the anterior view. were used in the numerical aspect of this previous work as
shell elements resulted in erroneous stress distributions.
Page 8 of 12
3D reconstruction rial properties are believed to resemble the behaviour of

The reconstruction process used here compared favoura- the actual AAA wall more closely than those of a linearly
bly to previous methods used by our group [30,31]. For elastic material. When utilising the AAA(COMP) method,
the AAA(SIMP) and AAA(MOD) models, reconstruction the ILT is an entirely different structure than that of the
involved the omission of the ILT, and so the AAA was AAA wall, and therefore required its own material proper-
modelled as a uniform walled vessel, consisting of one ties. The ILT was modelled as linearly elastic in the
structure. For the AAA(SIMP) models, the single layer sur- AAA(COMP)L using previously published data [4,28,35].
faces of these 3D reconstructions were then exported for The non-linear material properties derived by Wang et al.
stress analyses. The AAA(MOD) required the generation of [36] have been implemented to model the ILT region of
a uniform artificial wall 1.5 mm thick in ProEngineer the AAA(COMP)NL in a more realistic manner.
Wildfire 3.0, before being exported to the FEA solver. For
the AAA(COMP), the ILT was incorporated into the vessel Computational effort
wall and so the complex AAA had an artificial uniform The computational time for each stress analysis increases
wall and ILT region of varying thickness. The full model with the inclusion of non-linear material properties and
was then exported for stress analyses. The iliac arteries complexity of the model, and so a compromise is usually
were omitted from these models, as in previous studies desirable between computational time and accuracy. For
[5]. The effect of smoothing on the 3D reconstruction was this study the computational time for each analysis was
also examined. Four degrees of smoothing were studied in recorded and compared. These results can be seen in Table
order to determine the effect smoothing plays on wall 1. The differences in CPU times can be clearly seen, and
stress distributions. Normalised wall stress for the four highlight how the inclusion of non-linearity increases
models is shown in Figure 3. For model 1, the AAA was computational effort. Inclusion of the non-linear behav-
over-smoothed, and resulted in low wall stress. Smooth- iour of the AAA wall resulted in a 9-fold increase in com-
ing was then reduced for model 2, and resulted in higher putational time for the AAA(SIMP) models, 5-fold
stress and less gradual changes in stress along the AAA sur- increase for AAA(MOD) models, and a 10-fold increase
face. Smoothing was then reduced further for models 3 for the AAA(COMP) models. Accounting for the "nonslip"
and 4, which both resulted in elevated stresses and also contact between the AAA wall and the ILT region of the
more surface detail. This surface detail results in rapid AAA(COMP) cases also increased the computational
changes from regions of higher and lower stresses, show- effort.
ing jagged wall stress distributions. Also, if the stress dis-
tributions of models 3 and 4 were used in the study, they Wall stress
would indicate possible failure sites in abnormal regions The modelling techniques used in this study all return sig-
of the AAA, as elevated stresses are shown in the healthy nificantly different wall stress results. Peak wall stress was
proximal neck. This observation is shown in Figure 3 by shown to reduce when taking into account the realistic
the black marks indicating these irregular locations of non-linear behaviour of AAA tissue in all three modelling
high stress. From this study the optimum level of smooth- approaches. All wall stress results can be seen in Table 1.
ing was determined to be that of model 2, 20 control When modelled using non-linear material properties,
points per axial slice, and was used in all further recon- wall stress reduced by 34% in the AAA(SIMP) case, 55%
structions of this study. in the AAA(MOD) case, and by 47% in the AAA(COMP)
case. These reductions in wall stress can be observed in
Biomechanical material properties Figures 5, 6, 7 for each of the three modelling techniques.
In each of the three geometrically different models, a lin- From these figures it can be seen how the change in mate-
early elastic model was examined. It has been argued in rial properties does not significantly alter the stress distri-
previous AAA wall stress studies [23,24] that the aorta bution patterns on the surface of the AAA, instead it
behaves more like a linearly elastic material at pressures simply reduces the magnitude of these acting stresses.
above 80 mmHg (10.67 KPa). Therefore, many research-
ers utilise the much simpler technique of linear elasticity The location of peak stress only differed when modelled
in their studies [2,4,23-25]. Other researchers however, using the AAA(SIMP) approach. This shift in peak stress
have reported that the use of simplified linearly elastic location in the AAA(SIMP) models may indicate that the
material properties or other inappropriate tissue constitu- use of shell elements results in inaccurate stress distribu-
tive models can lead to erroneous stress distributions tions. Previous research undertaken by our group [19]
[2,36-40]. For each of the non-linear models, the material also noted that shell elements can lead to erroneous stress
properties of 69 AAA specimens form the basis of the distributions, in particular when using a linearly elastic
finite strain constitutive model for AAA wall tissue pro- material. In this previous numerical work [19], the ideal
posed by Raghavan and Vorp [3] and has been utilised by AAA examined showed peak stresses along the region of
many previous researchers [5-7,13,20,26-29]. These mate- maximum diameter, rather than at the proximal and dis-
Page 9 of 12
tal inflection points, as observed experimentally [18]. Cal- went routine CT scanning, thus detecting the aneurysm.
lanan et al. [19] then used 3D stress elements to Previous work has identified that AAA wall strength has an
numerically analyse the ideal AAA model, and found that average failure stress of 0.942 MPa [16]. If the AAA was
the stress distributions matched those found experimen- analysed for wall stress without inclusion of the ILT, such
tally, with good correlation between the two sets of as cases AAA(MOD)L and AAA(MOD)NL, results would
results. This background may indicate that the use of shell suggest that rupture has already occurred as local stress
elements could possibly lead to inaccurate results in real- had exceeded local strength. By implementing the ILT into
istic AAA models also. For both the AAA(MOD) and the model, peak stress was reduced to a level believed to
AAA(COMP) models, 3D stress elements were used, and more accurately represent that in vivo. The reported failure
the location of peak stress did not significantly shift when strength of 0.942 MPa [16] is higher than the predicted
the model accounted for non-linearity of the material. As wall stress of 0.4291 MPa, indicating that the rupture
the location of peak stress did not significantly alter in potential of the AAA may be equalised. If wall stress
these models, it is believed that the use of 3D elements is results are to be incorporated into the clinical decision-
more accurate than the use of shell elements in FE analy- making process, inclusion of non-linear material proper-
ses of realistic AAAs. As mentioned earlier, this observa- ties and the ILT may be important to the accuracy of the
tion was reported in previous research by our group [19]. results.
Figure 8 compares the wall stress results of the three line- Limitations
arly elastic models examined. Both the AAA(MOD) and The study presented here is not without limitations.
AAA(COMP) models showed peak stress regions at inflec- Firstly, it is known that calcifications occur in almost all
tion points on the inner surface at the proximal region of AAAs. These calcium deposits are believed to act as stress
the AAA sac. These findings are consistent with previous raisers within the wall and so incorporation into AAA
experimental [18] and numerical work [19]. The stress studies may have significance [29]. Also, patient-
AAA(MOD)L model returned the highest peak stress in all specific blood pressures should be recorded at the time of
three cases. The peak stress in AAA(MOD)L is over 2.5 CT scan. This would allow the accurate stress analysis of
times the peak stress in both AAA(SIMP)L and patient-specific AAAs, and as most patients exhibit ele-
AAA(COMP)L. Both the AAA(SIMP)L and AAA(COMP)L vated blood pressures, this could have an impact on peak
models returned similar peak stresses of 0.8833 MPa and wall stress. It is known that the AAA wall thickness is non-
0.8036 MPa, respectively. In comparison, the wall stress uniform [42], and therefore, incorporating a non-uniform
results for the three non-linear cases are presented in Fig- wall into the modelling process may also alter stress dis-
ure 9. Again, the AAA(MOD)NL model experienced the tributions and peak stresses. The use of pulsatile pressures,
highest wall stress, with a peak stress 44% higher than the possibly leading to fatigue testing, could also be incorpo-
AAA(SIMP)NLand 59% higher than that experienced in rated into the modelling process. The relatively new com-
the AAA(COMP)NL model. By including the ILT into the putational field of fluid-structure interaction (FSI) could
models, peak wall stress was reduced by 65% in the linear also lead to more accurate wall stress results [26-28,34].
case and 59% in the non-linear case. This results in an Material properties could be improved. It is known that
average reduction of 62% in peak wall stress simply by AAA tissue is non-homogenous and anisotropic, factors of
including the ILT. The ILT absorbs the internal pressure which have been ignored for this study. The inclusion of
induced by blood flow, and transfers the load to the AAA more accurate material properties could lead to more pre-
wall in a reduced amount. From the stress results found cise wall stress distributions. This comparison also only
for all the various models analysed, the AAA(COMP)NL accounts for one AAA model. Each individual AAA is
model is deemed to be the most accurate method of pre- unique, and a larger number of cases may give a better
dicting wall stress in AAAs. This model incorporates the understanding of the role of certain modelling parameters
realistic non-linear behaviour of both the AAA wall and and assumptions. Another limitation to this work, and
the ILT, and therefore produces the most accurate wall also other AAA stress analyses, is that stress results are not
stress results. intuitive to the clinician [5]. Use of the maximum diame-
ter criterion is easy and clear to the surgeon, and therefore,
Significance of results wall stress results should be converted into a more clini-
In order to determine the significance of the results cian-friendly index. There is ongoing research on the rup-
reported here, other factors must be considered. It is ture prediction of AAAs with the view to develop an easy
known that AAA rupture occurs when the locally acting to use tool that the clinician can implement into the deci-
wall stress exceeds the locally acting wall strength. To sion-making process.
make valid conclusions from the wall stress results pre-
sented here, wall stress was examined with respect to AAA
wall strength. The AAA case examined in this study under-
Page 10 of 12
Conclusion Acknowledgements
This paper has examined the effect of modelling on the We would like to thank (i) the Irish Research Council for Science, Engineer-
resulting wall stress distributions of a realistic AAA. Three ing and Technology (IRCSET) Grant no. RS/2005/340 (ii) NIH Grant R01-
different modelling techniques were implemented and HL-060670-05A1 (iii) Prof. David A. Vorp of the Centre for Vascular
Remodelling and Regeneration, McGowan Institute for Regenerative Medi-
each studied using both linear and non-linear material
cine, University of Pittsburgh (iv) Mr. Eamon Kavanagh, a Consultant Vas-
properties. It was found from the resulting stress distribu- cular Surgeon at Midwestern Regional Hospital, Limerick, for his help in
tions that inclusion of the ILT and implementing non-lin- collecting patient data and background information.
ear material properties may be important if accurate stress
distributions are to be obtained. Peak stresses were shown References
to significantly vary depending on the modelling tech- 1. Sakalihasan N, Limet R, Defawe OD: Abdominal aortic aneu-
nique, with the most accurate model, namely rysm. Lancet 365(9470):1577-89. Apr 30–May 6 2005
2. Vorp DA, Raghavan ML, Webster MW: Mechanical wall stress in
AAA(COMP)NL, returning the lowest peak stress. Shell ele- abdominal aortic aneurysm: influence of diameter and asym-
ments were shown to yield poor stress results, as there was metry. J Vasc Surg 1998, 27(4):632-639.
3. Raghavan ML, Vorp DA: Toward a biomechanical tool to evalu-
a significant shift in peak stress location when non-linear ate rupture potential of abdominal aortic aneurysm: identi-
material properties were implemented into the model. fication of a finite strain constitutive model and evaluation of
Also, in these AAA(SIMP) models the location of peak its applicability. J Biomech 2000, 33(4):475-482.
4. Di Martino ES, Guadagni G, Fumero A, Ballerini G, Spirito R, Biglioli
stress was shown to occur at regions of maximum diame- P, Redaelli A: Fluid-structure interaction within realistic 3D
ter, which previous research [18,19] has shown not to be models of the aneurysmatic aorta as a guidance to assess the
the case. Simplifying assumptions used in predicting wall risk of rupture of the aneurysm. Med Eng Phys 2001, 23:647-655.
5. Fillinger MF, Raghavan ML, Marra SP, Cronenwett JL, Kennedy FE: In
stress with FEA should be limited. vivo analysis of mechanical wall stress and abdominal aortic
aneurysm rupture risk. J Vasc Surg 2002, 36:589-597.
6. Fillinger MF, Marra SP, Raghavan ML, Kennedy FE: Prediction of
The results reported here suggest that even patients with rupture risk in abdominal aortic aneurysm during observa-
very modest levels of ILT, should be analysed using the tion: wall stress versus diameter. J Vasc Surg 2003, 37:724-732.
more complex, time-consuming methods, such as includ- 7. Raghavan ML, Vorp DA, Federle MP, Makaroun MS, Webster MW:
Wall stress distribution on three-dimensionally recon-
ing the ILT into FEA models, and the use of realistic, non- structed models of human abdominal aortic aneurysm. J Vasc
linear material properties. The effect of omitting these Surg 2000, 31:760-769.
8. Conway KP, Byrne J, Townsend M, Lane IF: Prognosis of patients
important parameters may lead to erroneous peak stress turned down for conventional abdominal aortic aneurysm
results, and ultimately to incorrect surgical decision-mak- repair in the endovascular and sonographic era: Szilagyi
ing. The principal idea behind reconstruction and stress revisited? J Vasc Surg 2001, 33:752-757.
9. Darling RC, Messina CR, Brewster DC, Ottinger LW: Autopsy
analysis of AAAs is to determine the wall stress distribu- study of unoperated abdominal aortic aneurysms. The case
tions of the AAA, and the potential of rupture. As rupture for early resection. Circulation 1977, 56(3 suppl):II161-II164.
will occur when the local wall stress exceeds the local wall 10. Cronenwett JL, Murphy TF, Zelenock GB, Whitehouse WM Jr, Linde-
nauer SM, Graham LM: Actuarial analysis of variables associ-
strength, and AAA have been shown to have variations in ated with rupture of small abdominal aortic aneurysms.
local wall strength [42], the need for improved prediction Surgery 1985, 98:472-483.
11. Nicholls SC, Gardner JB, Meissner MH, Johansen KH: Rupture in
of peak wall stress location may also have clinical impor- small abdominal aortic aneurysms. J Vasc Surg 1998,
tance. 28:884-888.
12. Vande Geest JP, Wang DHJ, Wisniewski SR, Makaroun MS, Vorp DA:
Towards a non-invasive method for determination of
To conclude, the modelling technique employed in patient-specific wall strength distribution in abdominal aor-
patient-specific FEA of AAAs, may be a valuable additional tic aneurysms. Ann Biomed Eng 2006, 34(7):1098-1106.
tool for clinical use. 13. Wang DHJ, Makaroun MS, Webster MW, Vorp DA: Effect of intra-
luminal thrombus on wall stress in patient-specific models of
abdominal aortic aneurysm. J Vasc Surg 2002, 36:598-604.
Declaration of competing interests 14. Sacks MS, Vorp DA, Raghavan ML, Federle MP, Webster MW: In-vivo
3D surface geometry of abdominal aortic aneurysms. Ann
The author(s) declare that they have no competing inter- Biomed Eng 1999, 27:469-479.
ests. 15. Vorp DA, Lee PC, Wang DHJ, Makaroun MS, Nemoto EM, Ogawa S,
Webster MW: Association of intraluminal thrombus in
abdominal aortic aneurysm with local hypoxia and wall
Authors' contributions weakening. J Vasc Surg 2001, 34:291-9.
BJD designed the study, reconstructed the AAA models, 16. Raghavan ML, Webster MW, Vorp DA: Ex-vivo biomechanical
behavior of abdominal aortic aneurysm: assessment using a
conducted the wall stress simulations, analysed the results new mathematical model. Ann Biomed Eng 1996, 24(5):573-582.
and prepared the manuscript. AC contributed to the wall 17. Vorp DA, Mandarino WA, Webster MW, Gorcsan J III: Potential
stress studies and analysis of the results. TMM supervised influence of intraluminal thrombus on abdominal aortic
aneurysm as assessed by a new non-invasive method. Cardio-
the study, revised and gave the final approval of the man- vasc Surg 1996, 4:732-739.
uscript. All authors read and approved the final manu- 18. Morris L, O'Donnell P, Delassus P, McGloughlin T: Experimental
script. assessment of stress patterns in abdominal aortic aneurysms
using the photoelastic method. Strain 2004, 40:165-172.
19. Callanan A, Morris L, McGloughlin T: Numerical and Experimen-
tal Analysis of an Idealised Abdominal Aortic Aneurysm. In
Page 11 of 12
European Society of Biomechanics S-Hertogenbosch, Netherlands; 41. EUROSTAR Data Registry Centre, Laheij R, van Marrewijk C, Buth J:
2004. Progress report on the procedural and follow up results of
20. Venkatasubramaniam AK, Fagan MJ, Mehta T, Mylankal KJ, Ray B, 3413 patients who received stent graft treatment for infrar-
Kuhan G, Chetter IC, McCollum PT: A comparative study of aor- enal aortic aneurysms for a period of 6 years. 2001.
tic wall stress using finite element analysis for ruptured and 42. Raghavan ML, Kratzberg J, de Tolosa EMC, Hanaoka MM, Walker P,
non-ruptured abdominal aortic aneurysms. Eur J Vasc Endovasc da Silva ES: Regional distribution of wall thickness and failure
Surg 2004, 28(2):168-176. properties of human abdominal aortic aneurysm. J Biomech
21. Dobrin PB: Pathophysiology and pathogensis of aortic aneu- 2006, 39(16):3010-3016.
rysms. Current concepts. Surg Clin N Am 1989, 69:687-703.
22. Schurink GW, van Baalen JM, Visser MJ, van Bockel JH: Thrombus
within an aortic aneurysm does not reduce pressure on the
aneurysmal wall. J Vasc Surg 2000, 31:501-506.
23. Giannoglu G, Giannakoulas G, Soulis J, Chatzizisis Y, Perdikides T,
Melas N, Parcharidis G, Louridas G: Predicting the risk of rupture
of abdominal aortic aneurysms by utilizing various geomet-
rical parameters: Revisiting the diameter criterion. Angiology
2006, 57(4):487-494.
24. Thubrikar MJ, Al-Soudi J, Robicsek F: Wall stress studies of
abdominal aortic aneurysm in a clinical model. Ann Vasc Surg
2001, 15:355-366.
25. Hua J, Mower WR: Simple geometric characteristics fail to
reliably predict abdominal aortic aneurysm wall stress. J Vasc
Surg 2001, 34:308-315.
26. Leung JH, Wright AR, Cheshire N, Crane J, Thom SA, Hughes AD, Xu
Y: Fluid structure interaction of patient specific abdominal
aortic aneurysms: a comparison with solid stress models.
Biomed Eng Online 2006, 5:33.
27. Li Z, Kleinstreuer C: Blood flow and structure interactions in a
stented abdominal aortic aneurysm model. Med Eng Phys 2005,
27:369-382.
28. Papaharilaou Y, Ekaterinaris JA, Manousaki E, Katsamouris AN: A
decoupled fluid structure approach for estimating wall stress
in abdominal aortic aneurysms. J Biomech 2007, 40:367-377.
29. Speelman L, Bohra A, Bosboom EMH, Schurink GWH, van de Vosse
FN, Makaroun MS, Vorp DA: Effects of wall calcifications in
patient-specific wall stress analyses of abdominal aortic
aneurysms. J Biomech Eng 2007, 129:1-5.
30. Morris L, Delassus P, Callanan A, Walsh M, Wallis F, Grace P,
McGloughlin T: 3D numerical simulation of blood flow through
models of the human aorta. J Biomech Eng 2005, 127:767-775.
31. Doyle BJ, Morris LG, Callanan A, Kelly P, Vorp DA, McGloughlin TM:
3D reconstruction and manufacture of real abdominal aortic
aneurysms: From CT scan to silicone model. J Biomech Eng
2007 in press.
32. Langewouters GJ, Wesseling KH, Goedhard WJA: The static elas-
tic properties of 45 human thoracic and 20 abdominal aortas
in vitro and the parameters of a new model. J Biomech 1984,
17:425-435.
33. McDonald DA: The elastic properties of the arterial wall", Blood flow in
arteries Baltimore, Williams and Wilkins; 1974:238-282.
34. Scotti CM, Shkolnik AD, Muluk SC, Finol E: Fluid-structure inter-
action in abdominal aortic aneurysms: effect of asymmetry
and wall thickness. Biomed Eng Online 2005, 4:64.
35. DiMartino E, Mantero S, Inzoli F, Melissano G, Astore D, Chiesa R,
Fumero R: Biomechanics of abdominal aortic aneurysm in the
presence of endoluminal thrombus: experimental character-
isation and structural static computational analysis. Eur J Vasc
Endovasc Surg 1998, 15:290-299.
36. Wang DHJ, Makaroun M, Webster MW, Vorp DA: Mechanical
properties and microstructure of intraluminal thrombus
from abdominal aortic aneurysm. J Biomech Eng 2001, Publish with Bio Med Central and every
123:536-539.
37. Wang DHJ, Makaroun MS, Webster MW, Vorp DA: Effect of intra- scientist can read your work free of charge
luminal thrombus on wall stress in patient-specific models of "BioMed Central will be the most significant development for
abdominal aortic aneurysm. J Vasc Surg 2002, 36:598-604. disseminating the results of biomedical researc h in our lifetime."
38. Truijers M, Pol JA, SchultzeKool LJ, van Sterkenberg SM, Fillinger MF,
Blankensteijn JD: Wall stress analysis in small asymptomatic, Sir Paul Nurse, Cancer Research UK
symptomatic and ruptured abdominal aortic aneurysms. Eur Your research papers will be:
J Vasc Endovasc Surg 2007, 33:401-407.
39. Mower WR, Quinones WJ, Gambhir SS: Effect of intraluminal available free of charge to the entire biomedical community
thrombus on abdominal aortic aneurysm wall stress. J Vasc peer reviewed and published immediately upon acceptance
Surg 1997, 26:602-608.
40. Vorp DA, Vande Geest J: Biomechanical determinants of cited in PubMed and archived on PubMed Central
abdominal aortic aneurysm rupture. Arth Throm Vasc Bio 2005, yours — you keep the copyright
25:1558-1566.
Submit your manuscript here: BioMedcentral
http://www.biomedcentral.com/info/publishing_adv.asp
Page 12 of 12
APPENDIX B
Conference Publications
Appendix B
Conference Publications
Doyle, B.J., A. Callanan, T.J. Corbett, D.A. Vorp and T.M. McGloughlin, In Vitro
Modelling of an Abdominal Aortic Aneurysm to Determine Rupture Locations,
Sylvester O’Halloran Surgical Scientific Meeting, Limerick, Ireland, March 6-7, 2009.
Irish Journal of Medical Science, 2009: 178(2); S50.
Doyle, B.J., T.J. Corbett, M.R. O’Donnell, D.A. Vorp and T.M. McGloughlin, Design
and Development of a Range of Silicone Elastomers for Use in Experimental
Studies, Bioengineering… in Ireland 15; Proceedings of the Fifteenth Annual of the
Section of Bioengineering of the Royal Academy of Medicine in Ireland, p.112.
Limerick, Ireland, January 30-31, 2009.
Doyle, B.J., A. Callanan, P.A. Grace and T.M. McGloughlin, A Finite Element
Analysis Rupture Index (FEARI) as an Additional Tool in Abdominal Aortic
Aneurysm Assessment, Bioengineering ’08, Imperial College, London, England, Sept
18-19 2008.
Doyle, B.J., A. Callanan, T.J. Corbett, A.J. Cloonan, M.R. O’Donnell, D.A. Vorp and
T.M. McGloughlin, The Use of Silicone Materials to Model Abdominal Aortic
Aneurysm Behaviour, Society of Plastics Engineers, SPE European Conference on
Medical Polymers, Belfast, Sept 3-5 2008, p.115-120.
Doyle, B.J., A. Callanan, P.A. Grace, D.A. Vorp, and T.M. McGloughlin, Vessel
Asymmetry as an Additional Tool for Aneurysm Rupture Risk, 16th Congress of the
European Society of Biomechanics, Lucerne, Switzerland, July 6-9, 2008. Journal of
Biomechanics, 2008: 41(1); S11.
Molony, D., A. Callanan, B. Doyle, M. Walsh and T. McGloughlin, Influence of

Modelling Parameters on Abdominal Aortic Anuerysm Stent-Grafts, 16th Congress
of the European Society of Biomechanics, Lucerne, Switzerland, July 6-9, 2008.
Journal of Biomechanics 2008: 41(1); S395.
Doyle, B.J., A. Callanan, M.T. Walsh, D.A. Vorp, and T.M. McGloughlin, Assessment
of Abdominal Aortic Aneurysm Risk – Asymmetry as an Additional Diagnostic
Appendix B
Tool, Summer Bioengineering Conference 2008, American Society of Mechanical

Engineers (ASME), Marco Island, Florida, USA, June 26-29, 2008.
Molony, D.S., A. Callanan, B.J. Doyle, L.G. Morris, M.T. Walsh and T.M.
McGloughlin, Affect of Abdominal Aortic Aneurysm Stent-Graft Design on
Arterial Haemodynamics, Summer Bioengineering Conference 2008, American
Society of Mechanical Engineers (ASME), Marco Island, Florida, USA, June 26-29,
2008.
Doyle, B.J., A. Callanan, P.A. Grace, D.A. Vorp, and T.M. McGloughlin, The Use of
an Asymmetry Parameter as an Additional Tool for Abdominal Aortic Aneurysm
Rupture Risk, Sylvester O’Halloran Surgical Scientific Meeting, Limerick, Ireland,
February 29- March 1, 2008. Irish Journal of Medical Science, 2008: 177(1); S21.
Doyle, B.J., A. Callanan, P.E. Burke, P.A. Grace, M.T. Walsh, D.A. Vorp, and T.M.
McGloughlin, Vessel Asymmetry as an Additional Tool for Rupture Prediction of
Abdominal Aortic Aneurysms, Bioengineering…in Ireland 14, Proceedings of the
Fourteenth Annual of the Section of Bioengineering of the Royal Academy of Medicine
in Ireland. Sligo, Ireland, January 25-26, 2008.
Doyle, B.J., A. Cloonan, A. Callanan, D.A. Vorp, and T.M. McGloughlin,

Development of a New Pseudomaterial for Rupture Studies of Abdominal Aortic
Aneurysms, Bioengineering…in Ireland 14, Proceedings of the Fourteenth Annual of
the Section of Bioengineering of the Royal Academy of Medicine in Ireland. Sligo,
Ireland, January 25-26, 2008.
Doyle, B.J., A. Callanan, D. A. Vorp and T.M. McGloughlin, A Finite Element

Analysis Rupture Index (FEARI): An Additional Tool for Abdominal Aortic
Aneurysm Burst Prediction? The Biomedical Engineering Society, BMES 2007
Annual Fall Meeting, Los Angeles, California, USA, September 26-29, 2007.
Doyle, B.J., A. Callanan, D. A. Vorp and T.M. McGloughlin, A Finite Element

Analysis Rupture Index (FEARI) as an Additional Tool for Abdominal Aortic
Aneurysm Burst Prediction, European Society of Biomechanics Workshop 2007,
Appendix B
Finite Element Modelling in Biomechanics and Mechanobiology, Trinity College,

Dublin, August 26-28, 2007.
Doyle, B.J., L.G. Morris, A. Callanan, P. Kelly, D. A. Vorp and T.M. McGloughlin, 3D
Reconstruction of Patient-Specific Abdominal Aortic Aneurysms: From CT Scan
to Silicone Model, Summer Bioengineering Conference 2007, American Society of
Mechanical Engineers (ASME), Keystone, Colorado, USA, June 20-24, 2007.
Doyle, B.J., L.G. Morris, P. Kelly, D. A. Vorp and T.M. McGloughlin, 3D

to Silicone Model, Bernard Crossland Symposium, NUI Galway, Ireland, March 28-29,
2007.
Doyle, B.J., L.G. Morris, P. Kelly, D. A. Vorp and T.M. McGloughlin, 3D

to Silicone Model, Bioengineering… in Ireland 13; Proceedings of the Thirteenth
Annual of the Section of Bioengineering of the Royal Academy of Medicine in Ireland,
p.21. Enniskillen, Northern Ireland, January 29-30, 2007.
Doyle, B.J., A. Callanan, L.G. Morris, T.M. McGloughlin, P. O’Donnell and P.

Delassus, Comparison of the Photoelastic Method and the Finite Element Method
to Predict the Stress Concentrations on an Abdominal Aortic Aneurysm Model,
Bioengineering… in Ireland 12; Proceedings of the Twelfth Annual of the Section of
Bioengineering of the Royal Academy of Medicine in Ireland, p.20. Galway, Ireland,
January 27-28, 2006.
THE USE OF SILICONE MATERIALS TO MODEL ABDOMINAL
AORTIC ANEURYSM BEHAVIOUR
Barry J. Doyle,1 Anthony Callanan,1 Timothy J. Corbett,1 Aidan J. Cloonan,1

Michael R. O’Donnell,1 David A. Vorp2 and Timothy M. McGloughlin1
1. Centre for Applied Biomedical Engineering Research (CABER), Department of

Mechanical and Aeronautical Engineering, and Materials and Surface Science
Institute, University of Limerick, Ireland.
2. Centre for Vascular Remodelling and Regeneration, McGowan Institute for

Regenerative Medicine, University of Pittsburgh, USA.
ABSTRACT
This paper aims to identify the rupture locations of abdominal aortic aneurysms. Dow Corning
Sylgard 184 was mechanically characterised. Five idealised models were then manufactured using this
silicone rubber which were subsequently inflated to rupture with the images recorded using a high
speed camera. Four of the five models tested ruptured at inflection points in the proximal and distal
regions of the aneurysm sac, and not at regions of maximum diameter.
INTRODUCTION rubber material to be used as the wall analogue

of ideal AAA models, and to perform in vitro
Aneurysms form a significant portion rupture studies to determine the site of rupture.
of cardiovascular related deaths in the Western Numerical modelling will attempt to quantify
world. Aneurysms are permanent and these experimental results.
irreversible localised dilations [1], and although
can form in any blood vessel, artery or vein, the METHODS
more serious aneurysms occur in the abdominal
aorta, the brain arteries, and the thoracic aorta. Material Characterisation
Approximately 500,000 AAAs are diagnosed Dow Corning Sylgard 184 silicone
worldwide each year [2] resulting in 15,000 rubber was used as the material for this study.
deaths per year in the USA alone [3]. Aluminium moulds were designed and
Typically, an AAA is surgically repaired once manufactured to be used with the injection-
shown to have reached or exceeded a diameter method to create the silicone rubber samples.
of 5cm. There have been reports that this The mould cavity conforms to a Type 2 tensile
threshold may lead to inaccurate surgical- test specimen as outlined in BS ISO 37. The test
decision making as not only can smaller AAAs apparatus used was the Tinius Olsen (Surrey
rupture [4-6], but also, large AAAs can remain RH1 5DZ, England). For the tensile testing, a
stable [6]. It is known that by applying the strain rate of 500mm/min was applied, as
definition of material failure to AAAs, the recommended in BS ISO 37 for Type 2
AAA will rupture when the locally acting wall specimens. Preconditioning of each sample was
stress exceeds the locally acting wall strength. also performed. An initial stretch-relax program
Currently, much research is aimed at examining of ten cycles to a 20% strain rate was carried out
the wall stress distribution within the AAA wall on each sample.
[7-14]. Researchers have also examined
methods to determine the strength of the AAA Rupture Modelling
wall, both invasively from excised tissue [15- In order to study the rupture of an
17] and non-invasively [18]. While there is aneurysm in vitro, the idealised AAA model
much focus on attempting to numerically developed and used extensively in previous
understand the mechanics of AAA rupture, only research by our group [19-23] was utilised. Five
limited work has focussed on the development AAA models to be used in the rupture study
of experimental methods of determining rupture were manufactured using this technique. The
potential. The purpose of this study is to experimental rig consisted of a mirrored-wall
mechanically characterise a type of silicone arrangement, pneumatic airline, pressure
regulator, pressure manometer and high speed
camera. The model was clamped to a retort
stand and connected to the air supply with
silicone tubing. A high speed camera
(Olympus i-Speed, Olympus Corporation) was
used to capture the point and location of
rupture. The camera is capable of recording
images at rates up to 33,000 frames per second
(fps). A pixel resolution of 800 x 600 at
1000fps was deemed adequate for this
application, and images were recorded using a Figure 1: Engineering stress and engineering
monochrome lens. Once the AAA model was strain experimental data fit of 5th order
attached to the test rig, the high speed camera polynomial curve for Sylgard 184.
was adjusted to ensure optimum focus and
angle. To determine the accuracy of the
experimental tests, the evaluated material was
implemented in the FEA software ABAQUS
v6.7 (SIMULIA, R.I., USA) as the wall of the
ideal AAA. Boundary conditions similar to the
experimental set-up were implemented to the
virtual AAA model.
RESULTS
Material Characterisation Figure 2: Comparison of results for the tensile

Tensile testing revealed that the tests and the numerical simulation of Sylgard
average UTS of Sylgard 184 is 7.7361 ± 1.6597 184.
MPa, compared with 8.1 MPa on the
specification sheet. In order to mechanically Rupture Modelling
characterise the material, the experimental data Of the five silicone models ruptured in
from the tensile tests were converted to true vitro, four models experienced rupture at a
stress and true strain, and then a 5th order region of inflection on the surface of the model.
polynomial curve was fitted to the data to An inflection point is defined as points on the
obtain a mean experimental data curve, as AAA surface at which the local AAA wall shape
shown in Figure 1. changes from concave outward to concave
inward [25]. This finding is consistent with
The basic Ogden stain energy function previous reports by our group [13,19,21] and
[24] can be seen in Eqn.1, with the resulting others [25] that noted peak stresses at these
material coefficients shown in Table 1. regions instead of at the maximum diameter of
the AAA. A summary of the rupture results can
(1) be seen in Table 2. Burst pressures were also
recorded. One silicone model ruptured at the
iliac bifurcation.
Table 1: Material coefficients for the third order
Ogden model for Sylgard 184. D=0 for all Table 2: Summary of experimental rupture
constants. results
Mu Alpha Rupture Rupture Pressure
Test
1 -304.235 1.2667 Location (mmHg)
2 148.232 1.5962 Proximal
1 254.7
3 157.156 0.9075 Inflection Point
Proximal
2 278.6
The results of the FEA and the Inflection Point
uniaxial tensile tests compared favourably, as Proximal
3 466.2
shown in Figure 2. Therefore, confidence was Inflection Point
established in the material characterisation of Distal
4 278.7
Sylgard 184. Inflection Point
Iliac
5 544.6
Bifurcation
The sequence of events leading to

rupture for Test 1 can be seen in Figure 3. This
illustration shows the frame where material
failure initiates, leading to rupture of the model,
and ultimately complete failure of the silicone DISCUSSION
model. It should be noted that no models failed
along the ‘seam’ line of the silicone model from The focus of this paper was to
the manufacturing process. In Figure 3, (A) experimentally rupture rubber ideal AAA
shows the inflated model, (B) the initial point of models in order to observe burst locations.
rupture, (C) propagation of the failure zone, and There has been little reported on the in vitro
(D) complete failure of the silicone model. A rupture of abdominal aortic aneurysms, with
similar sequence was observed for all four much focus on the computational analysis of
models that ruptured at regions of inflection. these aneurysms [7-14]. Morris et al. [19]
observed rupture locations at the inflection
regions during the use of the photoelastic
method. Five models were manufactured using
a technique developed by our group [22,23].
The commercially available silicone rubber
Sylgard 184 from Dow Corning was used as an
analogue for the AAA wall. Silicone rubber is
an appropriate analogue for arterial tissue and
has been used in previous studies [22,23,26,27].
Tensile testing revealed that the UTS of the
material (mean ± standard deviation) is 7.7361 ±
1.6597 MPa. By fitting a polynomial curve to
Figure 3: Sequence of events leading to model the experimental data and evaluating the
rupture of Test 1. The rupture location is material using ABAQUS v6.7, material
highlighted in (B) and (C). A similar sequence coefficients were determined. The optimum
was observed for all four models that ruptured strain-energy function for this particular material
at regions of inflection. is a third-order Ogden function as shown in Eqn.
1, and allows the material to remain stable at all
By implementing the material stresses and strains.
constants derived earlier, it was possible to
simulate the experimental rupture study The use of a high-speed camera to
numerically. Stress distributions on the record the point of rupture proved to be a very
surfaces of the virtual AAA model reveal that powerful experimental tool. The optimum
high stresses occur at the regions of inflection image resolution was found to be 1000fps. Air
and not at regions of maximum diameter. This pressure was increased incrementally until the
has been proven by our group both numerically point of rupture. Rupture pressures varied
using the finite element method [21] and significantly for two of the five models, as
experimentally using the photoelastic method shown in Table 2. Mean rupture pressure was
[19]. 364.56 ± 131.89 mmHg. Three models
experienced rupture at the proximal inflection
region, one at the distal inflection point, and one
model ruptured at the iliac bifurcation. Rupture
studies were then replicated using the finite
element method. Similar boundary conditions
were used to those in vitro. Comparing the
experimental rupture locations with the high
stress regions found using FEA showed good
correlation, as shown in Figure 6.
CONCLUSION
Ideal AAA models ruptured at regions
of inflection and not at areas of maximum
diameter. This is consistent with predicted
Figure 4: In vitro rupture locations and FEA results. The use of a high speed camera is a
stress patterns. High stress occurs at inflection useful experimental tool in monitoring rubber
points which correlate with rupture locations in AAA rupture locations. To improve the method
experimental models. described, more suitable arterial analogues that
mimic arterial properties more closely are
required, thus possibly leading to an improved
understanding of AAA rupture.
ACKNOWLEDGEMENTS and Endovascular Surgery. 33:401-407
(2007).
The authors would like to thank (i) the 13. WR Mower, WJ Quinones, SS
Irish Research Council for Science, Gambhir. Journal of Vascular Surgery.
Engineering and Technology (IRCSET) Grant 26:602-608 (1997).
RS/2005/340 (ii) Grant #R01-HL-060670 from 14. BJ Doyle, A Callanan, TM
the US National Heart Lung and Blood Institute McGloughlin. Biomedical Engineering
(iii) Dr. Liam Morris from the Galway Medical Online. 6:38 (2007).
Technology Centre, GMIT, Ireland (iv) Kevin 15. JF Rodriguez, C Ruiz, M Doblare, et al.
O’Flanagan from the Department of Journal of Biomechanical Engineering.
Manufacturing and Operation Engineering, 130:021023-1-10 (2008)
University of Limerick, Ireland, for his 16. MJ Thubrikar, M Labrosse, F Robicsek,
assistance with the high-speed imaging (v) et al. Journal of Medical Engineering
Maria Ryan for her help with the experimental and Technology. 25(4):133-142 (2001).
rupture modelling (vi) the Department of 17. ML Raghavan, MW Webster, DA
Vascular Surgery in the Midwestern Regional Vorp. Annals of Biomedical
Hospital, Ireland, in particular, Mr. Eamon Engineering . 24:573-582 (1996).
Kavanagh, Mr. Paul Burke and Prof. Pierce 18. ML Raghavan, J Kratzberg, EMC de
Grace (vii) Samarth Shah from the Centre for Tolosa, et al. Journal of Biomechanics.
Vascular Remodelling and Regeneration, 39(16):3010-6 (2006).
University of Pittsburgh, USA and (viii) Michel 19. JP Vande Geest, DHJ Wang, SR
S. Makaroun, MD, Department of Surgery, Wisniewski, et al. Annals of
University of Pittsburgh, USA. Biomedical Engineering. 34:7:1098-
1106 (2006).
REFERENCES 20. L Morris, P O’Donnell, P Delassus, et
al. Strain. 40:165-172 (2005).
1. N Sakalihasan, R Limet and OD 21. EUROSTAR Data Registry Centre,
Defawe. Lancet. 365(9470):1577-89 2001. R Laheij, C van Marrewijk, J
(2005). Buth.
2. JP Vande Geest, MS Sacks, DA Vorp. 22. A Callanan, LG Morris, TM
Journal of Biomechanics. 39:2347- McGloughlin. European Society of
2354 (2006). Biomechanics. S-Hertogenbosch,
3. C Kleinstreuer, Z Li. Biomedical Netherlands (2004).
Engineering Online. 5:19 (2005). 23. HS Flora, B Talie-Faz, L Ansdell, et al.
4. SC Nicholls, JB Gardner, MH Journal of Endovascular Therapy.
Meissner et al. Journal of Vascular 9:665-675 (2002).
Surgery. 28:884-888 (1998). 24. BJ Doyle, LG Morris, A Callanan, et al.
5. RC Darling, CR Messina, DC Journal of Biomechanical Engineering.
Brewster, et al. Circulation. 130:034501-5 (2008).
56(II):161-164 (1997). 25. T O’Brien, L Morris, P O’Donnell, et
6. DA Vorp. Journal of Biomechanics. al. Pro. IMechE. (H) Journal of
40:1887-1902 (2007). Engineering in Medicine. 219. (2005)
7. ML Raghavan, DA Vorp, MP Federle, 26. RW Ogden. Non-linear elastic
et al. Journal of Vascular Surgery. deformations. Mineola, New York:
31:760-769 (2000). Dover Publication Inc. (1984).
8. AK Venkatasubramaniam, MJ Fagan, 27. DA Vorp, ML Raghavan, MW
T Mehta, et al. European Journal of Webster. Journal of Vascular Surgery.
Vascular and Endovascular Surgery. 27(4):632-639 (1998).
28:168-176 (2004). 28. PALS Marins, RM Natal Jorge, AJM
9. MF Fillinger, SP Marra, ML Ferreira. Strain. 42:135-147 (2006).
Raghavan, et al. Journal of Vascular 29. OA Shergold, NA Fleck, D Radford.
Surgery. 37:724-732 (2003). International Journal of Impact
10. MF Fillinger, ML Raghavan, SP Engineering. 32:1384-1402 (2006).
Marra, et al. Journal of Vascular 30. DHJ Wang, MS Makaroun, MW
Surgery. 36:589-597 (2002). Webster, et al. Journal of Vascular
11. MF Fillinger, ML Raghavan, SP Surgery. 36:598-604 (2002).
Marra, et al. Journal of Vascular 31. GW Schurink, JM van Baalen, MJ
Surgery. 36:589-597 (2002). Visser, et al. Journal of Vascular
12. M Truijers, JA Pol, LJ SchultzeKool, Surgery. 31:501-506 (2000).
et al. European Journal of Vascular
APPENDIX C
AAA Moulds
Appendix C
AAA Moulds
As described in Chapter V, experimental models were created using aluminium moulds

designed with CAD/CAM techniques. This appendix presents the moulds manufactured
and refers to the Patient ID and Chapter they appear in. Several additional moulds were
also machined. The following figures are not shown to scale.
Patient A (Chapter V) & Patient 1 (Chapter VIII)
Figure C-1: Moulds for inner wax model on left and for silicone model on right
Patient B (Chapter V)
Appendix C
Patient C (Chapter V) & Patient 2 (Chapter VIII)
Patient D (Chapter V)
Appendix C
Patient 3 (Chapter VIII)
Patient 4 (Chapter VIII)
Appendix C
Additional AAA Moulds
Appendix C
APPENDIX D
Honours and Awards

Appendix D
Honours and Awards
Finalist in Sylvester O’Halloran Prize 2009

Doyle, B.J., A. Callanan, T.J. Corbett, D.A. Vorp and T.M. McGloughlin, In Vitro
Modelling of an Abdominal Aortic Aneurysm to Determine Rupture Locations,
Sylvester O’Halloran Surgical Scientific Meeting, Limerick, Ireland, March 6-7, 2009.
Irish Journal of Medical Science, 2009: 178(2); p.20
Finalist in ASME SBC 2008 PhD Student Paper Award – 3rd place
Doyle, B.J., A. Callanan, M.T. Walsh, D.A. Vorp, and T.M. McGloughlin, Assessment
of Abdominal Aortic Aneurysm Risk – Asymmetry as an Additional Diagnostic
tool, Summer Bioengineering Conference 2008, American Society of Mechanical
Engineers (ASME), Marco Island, Florida, USA, June 26-29, 2008.
Finalist in Mimics Innovation Award 2008

Doyle, B.J., L.G. Morris, A. Callanan, P. Kelly, D.A. Vorp and T.M. McGloughlin, 3D
Reconstruction and Manufacture of Real Abdominal Aortic Aneurysms: From CT
Scan to Silicone Model, Mimics Innovation Awards, Vienna, Austria, May 30-31,
2008.
Finalist in Engineers Ireland Biomedical Research Medal 2008 – Runner Up

Doyle, B.J., A. Callanan, P.A. Grace, P.E. Burke, D.A. Vorp, M.T. Walsh and T.M.
McGloughlin, The Use of Asymmetry in Abdominal Aortic Aneurysm Rupture
Risk Assessment: A Useful Diagnostic Tool?, Engineers Ireland Biomedical Research
Medal, Royal College of Surgeons Ireland, Dublin, Ireland, May 19, 2008.
Awarded Irish Research Council for Science, Engineering and Technology

(IRCSET) Postgraduate Scholarship 2005

BDoyle Thesis

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

BDoyle Thesis

Enviado por

Direitos autorais:

Formatos disponíveis

Rupture Behaviour of Abdominal Aortic Aneurysms:

A Computational and Experimental Investigation

A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

SUBMITTED TO THE UNIVERSITY OF LIMERICK, APRIL 2009.

University of Limerick. Tel: +353-61-202700 Fax: +353-61-330316

This thesis is dedicated to my loving parents for their continued

Also in loving memory of

PART I Computational Investigation

Chapter II A comparison of modelling techniques for computing wall 28

Chapter III A finite element analysis rupture index (FEARI) as an 51

Chapter IV Vessel asymmetry as an additional diagnostic tool in the 67

PART II Experimental Investigation

Chapter VI An experimental and numerical comparison of the rupture 103

Chapter VII Experimental modeling of aortic aneurysms: Novel 127

Chapter VIII Identification of the rupture locations in patient-specific 155

Summary and Conclusions 191

2 Biomedical Engineering Online Biomedical 1.58

3 Vascular Disease Prevention Medical N/A*

4 Journal of Vascular Surgery Medical 3.272

5 Journal of Biomechanical Engineering Biomedical 1.591

6 Journal of Endovascular Therapy Medical 2.392

7 Medical Engineering & Physics Biomedical 1.471

8 Journal of Biomechanics Biomedical 2.897

Aneurysm A permanent swelling of a blood vessel caused by disease or

With recent advancements in medical imaging technology, the recorded incidence of

approach is often careful and frequent observation using either ultrasonography or

In Part I, Chapter I addresses the issue of 3D reconstructions of computed tomography

experimental models. 3D reconstruction of AAA cases allows further computer-aided

In order to successfully model aneurysms in vitro, material analogues are required.

Barry J. Doyle,1 Pierce A. Grace,2 Eamon G. Kavanagh,2 Paul E.

2. Department of Vascular Surgery, HSE Midwestern Regional Hospital, Limerick, Ireland.

3. Department of Radiology, HSE Midwestern Regional Hospital, Limerick, Ireland

Irish Journal of Medical Science (in press)

The conversion of computed tomography (CT) scans into 3D reconstructions of patient-

Finite Element Analysis (FEA)

Figure 4: Example reconstruction showing measurements used to size the stent-graft.

The patient-specific reconstruction shown in Figure 6 was a particularly extreme case.

Figure 5: 3D reconstruction revealed extremely tortuous proximal neck which can be

Figure 6: 3D reconstruction showing thoracic aortic aneurysm (TAA), abdominal aortic

Finite Element Analysis (FEA)

3D reconstructions also allow the clinical problem to be examined using various

Figure 8: Post-operative EVAR illustration showing deployed TAA stent-graft

Barry J. Doyle, Anthony Callanan and Timothy M. McGloughlin

Biomedical Engineering Online 2007; 6: 38

It has been previously documented how AAA formation is accompanied by an increase

will be referred to as AAA(COMP)L and AAA(COMP)NL. All stress analyses were

Figure 1: Typical CT scan and 3D reconstruction of patient suffering from an AAA. On

Figure 3: Resulting wall stress distributions of the various degrees of smoothing

Figure 4: Effect of smoothing on the reconstructed AAA. The optimum smoothing

Biomechanical Material Properties

W = C1(IB − 3) + C2 (IB − 3)2 Eqn. (1)

Forces and Boundary Conditions

AAA(SIMP)NL 0.5814 Left lateral region 9

Biomechanical Material Properties

To conclude, the modelling technique employed in patient-specific FEA of AAAs, may

Barry J. Doyle,1 Anthony Callanan,1 Michael T. Walsh,1 Pierce A.

2. Department of Vascular Surgery, HSE Midwestern Regional Hospital, Limerick, Ireland.

Vascular Disease Prevention 2009; 6: 114-121

Currently, the rupture risk of AAAs is regarded as a continuous function of aneurysm

MATERIALS AND METHODS

Wall Stress Analysis

FEA Wall Stress Eqn. (2)

By combining all the previously published experimental data [13,24,25], average

Table 2: Regional UTS values obtained by combining and averaging previous