Você está na página 1de 24

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN BIOMEDICAL ENGINEERING

Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586


Published online 14 January 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/cnm.2620

SPECIAL ISSUE PAPER - NUMERICAL METHODS AND APPLICATIONS OF


MULTI-PHYSICS IN BIOMECHANICAL MODELING

Modelling the influence of endothelial heterogeneity on the


progression of arterial disease: application to abdominal aortic
aneurysm evolution

P. Aparício 2 , A. Mandaltsi 1 , J. Boamah 1 , H. Chen 1 , A. Selimovic 1 , M. Bratby 3 ,


R. Uberoi 3 , Y. Ventikos 1 and P. N. Watton 4, * ,†
1 Departmentof Engineering Science, University of Oxford, Oxford, UK
2 Systems
Biology Doctoral Training Centre, University of Oxford, Oxford, UK
3 John Radcliffe Hospital, Oxford, UK
4 Department of Computer Science & INSIGNEO Institute of in silico Medicine, University of Sheffield, Sheffield, UK

SUMMARY
We sophisticate a fluid–solid growth computational framework for modelling aneurysm evolution. A realis-
tic structural model of the arterial wall is integrated into a patient-specific geometry of the vasculature. This
enables physiologically representative distributions of haemodynamic stimuli, obtained from a rigid-wall
computational fluid dynamics analysis, to be linked to growth and remodelling algorithms. Additionally, a
quasistatic structural analysis quantifies the cyclic deformation of the arterial wall so that collagen growth
and remodelling can be explicitly linked to the cyclic deformation of vascular cells. To simulate aneurysm
evolution, degradation of elastin is driven by reductions in wall shear stress (WSS) below homeostatic thresh-
olds. Given that the endothelium exhibits spatial and temporal heterogeneity, we propose a novel approach
to define the homeostatic WSS thresholds: We allow them to be spatially and temporally heterogeneous. We
illustrate the application of this novel fluid–solid growth framework to model abdominal aortic aneurysm
(AAA) evolution and to examine how the influence of the definition of the WSS homeostatic threshold
influences AAA progression. We conclude that improved understanding and modelling of the endothelial
heterogeneity is important for modelling aneurysm evolution and, more generally, other vascular diseases
where haemodynamic stimuli play an important role. Copyright © 2014 John Wiley & Sons, Ltd.

Received 27 January 2013; Revised 19 September 2013; Accepted 18 November 2013

KEY WORDS: abdominal aortic aneurysm; fluid–solid growth; growth and remodelling; wall shear stress;
endothelial cell; cyclic deformation; homeostasis

1. INTRODUCTION

The evolution of arterial diseases, for example, intracranial aneurysms, abdominal aortic aneurysms
(AAAs) and atherosclerosis, is generally accepted to be intimately linked to the mechanical envi-
ronment of the arterial wall. In particular, the wall shear stress (WSS) that acts on the endothelial
layer is hypothesised to play an influential role on disease progression. For instance, in response
to changing environmental conditions/demands on the artery, it is often hypothesised that devia-
tions of the magnitudes of WSS from homeostatic magnitudes may drive growth and remodelling
(G&R) processes that enable the artery to restore homeostasis or, alternatively, they may influence
degenerative processes that lead to disease progression.
Mathematical/computational models of arterial adaption often utilise conceptual (cylindrical)
geometries to represent the artery, and thus, the magnitude of the homeostatic WSS does not need

*Correspondence to: P. N. Watton, Department of Computer Science & INSIGNEO Institute of in silico Medicine,
University of Sheffield, Sheffield, UK.
† E-mail: P.Watton@sheffield.ac.uk

Copyright © 2014 John Wiley & Sons, Ltd.


564 P. APARÍCIO ET AL.

to take into account spatial heterogeneities in the WSS that arise in more complex geometries/flow
conditions [1–10]. However, recent vascular biology articles emphasise the heterogeneity of the
endothelium (e.g. see Regan and Aird [11]). More specifically, the properties of endothelial cells
(ECs) not only vary between different sites of the vasculature but they may also be able to adapt,
responding to the unique needs of the underlying tissue. Given that the vascular endothelium is spa-
tially and temporally heterogeneous, we propose that our definitions of homeostatic WSS thresholds
may need to incorporate concepts of spatial and temporal heterogeneity.
The focus of this article is twofold. The first focus is to sophisticate a fluid–solid growth (FSG)
computational framework to account for both the flow-induced stimuli and the cyclic pressure-
induced stimuli that act on the vascular cells and explicitly link them to arterial G&R. The second
focus is to bring to attention the need for improved understanding and modelling of how we define
homeostatic thresholds of WSS within the arterial tree. For an illustrative example, we consider
AAA evolution and illustrate how the choice of a homeostatic WSS influences predictions with
respect to AAA progression.
Abdominal aortic aneurysm is an excessive, localised dilatation of the abdominal aorta. Preva-
lence is relatively high (1–9% of the population [12]) and the mortality rates associated with rupture
are between 80% and 90% [13]. The disease accounts for 1–2% of all deaths in the UK each year
[14]. The precise aetiology of AAA is not known; hypertension [15], as well as environmental
influences, such as smoking [16] can contribute to enlargement and rupture. Treatment is available;
however, open surgery and endovascular intervention have 5% mortality rates [17], and thus, treat-
ment is delayed until the risk of rupture exceeds the risk of treatment. Unfortunately, the criterion for
predicting rupture is statistically based and only takes into account the diameter of the aneurysm: It
is observed that the risk of rupture exceeds the risk of treatment when the diameter exceeds 5.5 cm
[18]. Clearly, to achieve personalised healthcare, there is a need for more sophisticated and reli-
able diagnostic criteria. It is envisaged that models of AAA evolution [19–22] will lead to a greater
understanding of the disease and, in the longer term, may assist AAA management.
Watton et al. [19] proposed the first mathematical model of AAA evolution. A realistic struc-
tural model for the arterial wall [23] was sophisticated to incorporate variables that relate TO THE
normalised mass density (hereon referred to as concentration) of the elastinous and collagenous
constituents and the reference configurations in which the collagenous constituents are recruited to
load bearing. A key assumption of the model is that collagen fibres, which are in a continual state of
deposition and degradation [24], are configured to the artery in a state of stretch (in the loaded con-
figuration), which is independent of the current geometrical configuration of the tissue [25]. These
concepts enable the G&R of the collagen to be simulated as an aneurysm evolves. To simulate AAA
evolution, Watton et al. [19] prescribed a degradation of elastin and utilised a rate-based formu-
lation to simulate the G&R of collagen, that is, differential equations are employed to evolve the
collagen fibre reference configurations and fibre concentrations so that the stretch of the collagen
fabric remodels towards homeostatic magnitudes. The model predicts evolution of AAA mechanical
parameters and growth-rates consistent with clinical observations [20].
Local aortic haemodynamic conditions may influence the progression of AAA. For instance, WSS
affects the expression of elastase, an enzyme that breaks down elastin fibres [26]. Consequently,
there is a recognised need to connect models of aneurysm evolution to the evolving haemodynamic
environment. Indeed, Humphrey & Taylor [27] emphasised the need for a new class of models to
study aneurysm evolution and introduced the terminology FSG models. These describe the evolving
changes in vascular biology, geometry, wall properties, and haemodynamics. They proposed a full
fluid-structure interaction approach to couple the G&R of the tissue to the WSS and the pulsatile
fluid pressure on the vessel wall. Their approach has been applied to simulate the evolution of an
axisymmetric cerebral aneurysm on the basilar artery [5] and, more recently, AAA evolution on
patient-specific geometries [22].
An alternative FSG approach to couple G&R to haemodynamic stimuli to model aneurysm evo-
lution was proposed by Watton et al. [9]. A rigid-walled fluid analysis is used to define the haemo-
dynamic stimuli that act on the arterial wall that serve as input to G&R algorithms for the aneurysm
growth model. However, a limitation of the approach was that G&R was not linked to the pul-
satile pressure. Furthermore, simple conceptual geometries were utilised for the haemodynamic

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 565

Figure 1. Fluid–solid growth computational framework for modelling aneurysm evolution consisting of the
following: (i) structural analysis; (ii) computational fluid dynamics (CFD) analysis; (iii) local mechanical
environment of vascular cells; and (iv) growth and remodelling algorithms. Further details are provided in
Sections 2.1, 2.2 and 2.3, respectively.

domain. In this study, we apply the FSG model developed by Watton et al. [9] to model AAA
evolution and introduce several novel sophistications to the computational framework: (i) both the
growth and remodelling of collagen is linked to cyclic deformation; (ii) the structural model of
the arterial wall is integrated into patient-specific vascular geometries so that realistic patterns of
haemodynamic stimuli serve as input to G&R algorithms; and (iii) the novel concept of a spatially
heterogeneous and temporally adaptive homeostatic WSS threshold h .X, t / is proposed and mod-
elled. To drive aneurysm evolution, elastin degradation is linked to deviations of the WSS from
the homeostatic threshold. We consider three choices for h .X, t /: (i) spatially homogeneous; (ii)
spatially heterogeneous; and (iii) spatially heterogeneous and temporally adaptive. We illustrate the
influence of choice of the definition of homeostatic WSS on the evolution of the AAA and also
analyse the evolution of parameters associated with the local cyclic deformation of the tissue as the
AAA evolves.

2. FLUID–SOLID GROWTH MODEL FOR ANEURYSM EVOLUTION

We begin by overviewing the FSG computational framework. This broadly follows from Watton
et al. [9]; however, note that a novel update introduced is to link both the growth and the remodelling
of the collagen fabric to the cyclic deformation of the arterial wall. The modelling cycle starts with a
quasistatic structural analysis to solve the equilibrium deformation fields in the systolic and diastolic
configurations (Figure 1(i)). This quantifies the cyclic deformation and stress/stretch of the vascular
cells and components of the extracellular matrix (ECM). The systolic geometry of the aneurysm is
then (automatically) prepared for computational fluid dynamics (CFD) analysis (Figure 1(ii)): an
STL file of the structural domain geometry is exported from the structural solver and is integrated
into a physiological geometrical domain (encoded as a .tin file that has surface labels for inlets and
outlets) in a scripted manner using ANSYS ICEM; the labelled volumetric domain is meshed; phys-
iological flow rate and pressure boundary conditions are applied in a scripted manner; the flow is
solved with ANSYS CFX v14. Once the flow has been solved, Tecplot 360 is used to interpolate

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
566 P. APARÍCIO ET AL.

Figure 2. The Lagrangian coordinates, 1 2 Œ0, L1  and 2 2 Œ0, L2 /, of the computational domain.

Figure 3. Model of the healthy abdominal aorta in (a) reference configuration, (b) diastolic configuration
and (c) systolic configuration.

haemodynamic parameters evaluated on the surface of the fluid domain onto the nodes of the solid
domain mesh. Hence, each node of the structural mesh has associated information obtained from the
haemodynamic and structural analyses defining the mechanical stimuli acting at that point. Subse-
quently, G&R algorithms simulate cells responding to the mechanical stimuli and adapting the tissue
(Figure 1(iii)). Following G&R, the constitutive model of the aneurysmal tissue is updated, and the
structural analysis is re-executed to calculate the new systolic and diastolic deformation fields. The
updated geometry is exported for haemodynamic analysis. The cycle continues, and as the tissue
adapts, an aneurysm evolves. Sections 2.1, 2.2 and 2.3 describe the stages of the FSG framework in
more detail.

2.1. Structural model of aneurysm evolution


A geometric nonlinear membrane theory is adopted to model the steady deformation of the arterial
wall [19]. Briefly, the unloaded internal abdominal aorta is treated as a thin cylinder of undeformed
radius R, length L1 and thickness H . It is subject to a physiological pressure and physiological
axial prestretch. A body-fitted coordinate system is used to describe the cylindrical membrane with
axial and azimuthal Lagrangian coordinates 1 2 Œ0, L1  and 2 2 Œ0, 2R/, respectively (Figure 2).
Figure 3(a–c) illustrates the conceptual model of the abdominal aorta in the reference configuration,
diastolic configuration and systolic configuration, respectively.
The initial displacement field for mechanical equilibrium can be analytically determined (see
[19] for details). An aneurysm evolves as the material constituents adapt (via G&R). The principle
of stationary potential energy governs the displacement field for mechanical equilibrium, that is,

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 567

Figure 4. (a) The structural analysis solves (in parallel) the systolic and diastolic deformation fields as
the aneurysm evolves. The (b) diastolic and (c) systolic upstream and downstream displacement boundary
conditions are fixed during aneurysm evolution.

ı…int  ı…ext D 0, (1)

where ı…int represents the variation of the internal potential energy stored in the arterial wall and
ı…ext represents the variation of the external potential energy due to the pressure. Strain energy
density functions (SEDFs) are specified so that ı…int can be computed.
To quantify the cyclic deformation, we determine the geometry of the aneurysm at systolic and
diastolic pressures (see [10]). As the material constituents evolve, the systolic/diastolic deformation
fields to satisfy (1) are updated with a Newton–Raphson method using the systolic/diastolic defor-
mation fields from the previous time step as initial guesses; the positions of the boundaries of the
domain (1 D 0, L1 ) in the systolic/diastolic configuration are held fixed as the aneurysm evolves
(Figure 4).
In many aneurysms, diameter enlargement is asymmetric, with primarily anterior protrusion; the
posterior region is often constrained from radial expansion by the adjacent spinal column [28]. To
simulate this, we follow Watton et al. [19] and model the spine as a stiff spring-backed plate. Where
the aneurysm wall penetrates the plate, a penalty pressure acts normal to the aneurysm. The effec-
tive pressure acting on the membrane is given by the difference of internal physiological pressure
and the penalty pressure. Details of the theoretical formulation to describe the deformation of the
arterial wall and the numerical formulation to solve Equation (1) can be found in [19]. Briefly, for
the structural simulations, we used a Lagrangian mesh consisting of 60 x 60 bilinear elements; the
Fortran subroutine MA38 (available from the Harwell software library (www.hsl.ac.uk)) was

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
568 P. APARÍCIO ET AL.

employed to solve the linear system that arises in the Newton iteration, which is required to update
the deformation of the arterial wall at successive time steps.

2.1.1. Strain-energy functions for heterogeneous aneurysmal tissue. The structural model for the
abdominal aorta is identical to the one utilised in Watton et al. [19]. Briefly, the Holzapfel–Gasser–
Ogden constitutive model [23] was sophisticated for the following reasons: (i) to explicitly account
for the distinct reference configurations of the constituents and (ii) to incorporate normalised density
variables (concentration) so that the growth/atrophy of constituents can be simulated. The arte-
rial wall is modelled as two layers. The SEDFs for the elastinous contributions in the medial and
adventitia are multiplied by a normalised spatially and temporally dependent concentration function,
denoted mE .1 , 2 , t /. This is employed to prescribe the degradation of the elastinous constituents,
where mE .1 , 2 , t D 0/ D 1. Fields of spatially and temporally dependent fibre recruitment stretch
R C
Jp .1 , 2 , t / and fibre concentration mJp .1 , 2 , t / variables are defined throughout the arterial
wall, where the subindex J denotes the media M or the adventitia A. The fibres within each layer
are orientated at an angle of Jp to the azimuthal axis, where p denotes the pitch ˙J relative to
the azimuthal axis in the unloaded reference configuration. The SEDFs for the medial and advenal
layers are

‰J D mE KJE .E11 C E22 C E33 /


X    2  
C mC K C
exp AC
E C
 1 , J D M , A, (2)
Jp J Jp
pD˙, EJCp >0

where KJE (J D M , A) are material parameters for the neo-Hookean [29] elastinous constituents,
while KJC and AC are parameters that relate to the collagen fabric. The Green–Lagrange (GL)
strains of the elastin, that is, E11 , E22 and E33 , are defined relative to the unloaded configuration; in
the initial cylindrical configuration, these represent strains in the axial, azimuthal and radial direc-
tions, respectively. However, the GL strains in the collagen fibres are denoted by EJCp .1 , 2 , t / and
are defined relative to the configuration in which the collagen fibres are recruited to load bearing.
It is hypothesised that vascular cells configure the collagen fibres to achieve a maximum stretch,
denoted C AT , during the cardiac cycle. We refer to this as the attachment stretch. The initial values
of the recruitment stretches are determined so that the stretch of the medial and adventitial collagen
fibres at systole at t D 0 equals the attachment stretch C AT . This implies that the GL strains EJp
C

of the collagen fibres are a function of the GL strains of the elastin resolved in the directions of the
collagen fibres, denoted EJp . Thus,

EJp  EJRp
EJCp D , (3)
1 C 2EJRp
 2 
where EJRp D R
Jp  1 =2, R
Jp .1 , 2 , t / are the recruitment stretches and EJp D

E11 sin2 Jp C E22 cos2 Jp C 2E12 sin Jp cos Jp .
Table 1 summarises the geometric and material parameters for the model (for further details, see
[19, 20]).

2.1.2. Cyclic deformations of Aneurysmal tissue. Two indices of interest associated with the cyclic
deformation that are calculated are the cyclic areal stretch ACS and the biaxial stretch index (BSI)
BSI . The cyclic areal stretch describes the ratio of the differential areal stretch at systole to the
differential areal stretch at diastole. It is computed as the ratio of the area metrics. In the initial
geometrical configuration, the artery is subject to cyclic stretch that is one dimensional, that is, the
cyclic stretch is only in the circumferential direction. As the aneurysm evolves, regions of the artery
experience biaxial stretching. To characterise the evolution of the cyclic stretch environment, we
consider a novel biaxial stretch index (BSI) proposed by Watton et al. (see [10]), denoted BSI

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 569

Table I. Physiological data used for modelling the human abdominal aorta artery.
Radii
At systole rs 11 mm
Reference configuration R rs =s

Lengths
In vivo axial length L 88 mm
Reference configuration L1 L=´
Reference configuration L2 2R
Wall thickness
Total H R=5
Media HM 2H=3
Adventitia HA H=3
Fibre orientation
Media MC , M C30ı , 30ı
Adventitia AC , A C60ı , 60ı
Applied pressure; initial in vivo stretches
Systolic pressure ps 16 kPa
Diastolic pressure pd 10.67 kPa
Axial prestretch ´ 1.3
Systolic circumferential stretch at t D 0 s 1.25

Collagen
Attachment stretch C
AT 1.07
Recruitment stretch (media, t D 0) R R
M , M 1.18
C
Recruitment stretch (adventitia, t D 0) R R
A ,  A 1.20
C

where 0 6 BSI 6 1. This measure quantifies the degree of the biaxial distortion of the tissue
(during the cardiac cycle) and is independent of the magnitude of the strains. If BSI D 0, the
tissue is subject to one-dimensional cyclic stretching whereas if BSI D 1, the tissue is subject to
equibiaxial cyclic stretching. If the cyclic (linearised) strain in the direction of the first principal
stretch is a factor greater than the cyclic (linearised) strain in the direction of the second principal
stretch, then BSI D 1= .

2.2. Haemodynamic modelling


A patient-specific geometry of an advanced stage AAA is utilised to generate the pre-aneurysmal
geometry of the abdominal aorta; the process is depicted in Figure 5. A Digital Imaging and
Communications in Medicine data set of computed tomography angiography images depicting the
abdominal and pelvic regions was acquired (John Radcliffe Hospital, Oxford, UK) (Figure 5(a)).
Simpleware ScanIP (Simpleware, Exeter, UK) software was then used to segment the data set, pro-
cess the segmentation masks and generate the AAA three-dimensional geometry. The lumen region
was segmented by the confidence connected region growing method applied on all masks, with
the segmentation parameters fine-tuned for one initial slice and then kept constant throughout the
segmentation domain (Figure 5(b)); the holes within the segmentation domain were removed by
means of a closing morphological filter (Figure 5(c)), and a three-dimensional surface was automat-
ically generated (Figure 5(d)). A dilation morphological filter and a Gaussian smoothing algorithm
were then applied, improving surface smoothness while preserving the geometrical integrity of the
larger side branches (Figure 5(e)). The pre-aneurysmal geometry was subsequently generated from
the patient-specific AAA geometry using an adaptation of a similar methodology for reconstruct-
ing healthy sections of artery from vascular geometries with sidewall intracranial aneurysms [30]:
Upstream and downstream segments were clipped (Figure 5(f,g)); the centreline of the patient-
specific AAA and a reference anterior point were used to guide cylinder positioning (Figure 5(g)).
The cylinder radius was chosen to be equal to the average radius of the upstream section outlet, while

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
570 P. APARÍCIO ET AL.

Figure 5. Generation of patient-specific abdominal aortic geometry for fluid–solid growth study. (a) A com-
puted tomography angiography image of abdominal aorta in axial plane; (b) automatic lumen segmentation
(yellow mask) and (c) removal of holes; three-dimensional surface geometry (d) before and (e) after smooth-
ing ; (f, g) clipping of upstream and downstream segments; (h) computed centreline (pink line) and cylinder
deployment region (set of aligned blue circles); (i) downstream connectivity region; (j) final geometry for
fluid–solid growth simulations, showing united upstream (purple) and downstream (dark blue) segments and
growth domain (yellow).

the cylinder length (´ L) was chosen subjectively to be around 80% of the distance between the
outlet of the upstream section and the inlet of the downstream surface section (it needs to be shorter
so that it can be smoothly connected to the physiological geometry). The upstream and downstream
boundaries, the centreline points and the cylinder radius, length and neckpoint were then inputs for
a processing chain implemented in MATLAB R2012a (MathWorks, Natick, MA, USA) that auto-
matically generates smooth connection surfaces (e.g. see Figure 5(i) between the cylindrical section
and the upstream and downstream sections of (physiological) geometries). The resulting geometry
is depicted in Figure 5(j).
The methodological approach to solve the haemodynamics as the aneurysm evolves proceeds as
follows. The geometry of the aneurysmal section is exported from the structural solver to the mesh-
ing suite ANSYS ICEM (Figure 1). ANSYS ICEM automatically integrates the aneurysmal section
into the physiological geometrical domain, that is, attaches the upstream and downstream exten-
sions, and then automatically generates an unstructured tetrahedral mesh with three prism layers
lining the boundary (Figure 6). The volumetric mesh consisted of approximately 300K elements
for the upstream section, approximately 150K elements for the downstream section and between
170K and 300K for the aneurysm section. Boundary conditions are then generated, and the flow is
solved with ANSYS CFX v14 using a criterion for maximum residual of 0.0001 (incompressible
Navier–Stokes equations using a finite volume formulation [31]). Mesh independence for the rele-
vant haemodynamic quantities was studied, and this size mesh had convergence of WSS to within
5%. Blood is modelled as a Newtonian fluid with constant density
D 1069 kgm3 and constant
viscosity D 0.0035 Pa s. No slip, no flux conditions are applied at the arterial wall boundaries.
A steady flow analysis is adopted to reduce the cost of the computational simulations. The simi-
larity of the spatial WSS distributions for steady and pulsatile flow implies that this is a reasonable

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 571

Figure 6. Section of the mesh generated for the computational domain illustrating the circular boundary
between the upstream section and the aneurysm section.

Figure 7. Boundary Surfaces and prescribed boundary conditions. (a) Vessel wall (red) and boundary sur-
faces, inlet (blue, prescribed flow rate boundary condition apposed) and outlets (green); (b) upstream section
outlets (green) and corresponding prescribed pressure boundary conditions; (c) downstream section outlets
(green) and corresponding prescribed pressure boundary conditions.

approach for the purposes of investigating phenomenological hypotheses that explore the link
between G&R and deviations of the WSS from homeostatic levels (Section 4). Simulations were
performed on a 64BIT desktop PC (4 Intel Xeon CPU W3530 @2.8 GHZ, 12 GB RAM) run-
ning Windows 7 professional. Flow was solved by ANSYS CFX using four cores. A complete
FSG simulation consisting of 500 solid steps with the CFD performed every 10 steps takes around
2 days. The mean flow and pressure boundary conditions are taken from a one-dimensional model
of the arterial tree [32] that has been integrated into the software suite @neufuse [33]. It solves the
one-dimensional form of the Navier–Stokes equation in a distributed model of the human systemic
arteries, accounting for the ventricular–vascular interaction and wall viscoelasticity; it was recently
validated through a comparison with in vivo flow measurements [34]. The flow rate boundary con-
dition prescribed at the inlet and the pressure boundary conditions prescribed at each outlet are
detailed in Figure 7. It is of note that the prescribed boundary conditions resulted in an asymmetry
in the flow rates at the aortoiliac bifurcation, that is, a 60 : 40 split in flow rates between the right and
left iliac arteries, respectively; however, this is consistent with the range of flow rates splits (80 : 20
to 50 : 50) considered by Deplano et al. [35].

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
572 P. APARÍCIO ET AL.

2.3. Growth and remodelling


The G&R of the constituents of the arterial wall gives rise to the development of the aneurysm.
Remodelling is associated with changes in the natural reference configurations at which constituents
are recruited to load bearing. It is a natural consequence of constituents being configured to the
matrix in the loaded configuration. Growth is associated with changes in the mass of the constituents.

2.3.1. Elastin degradation. The dilation of an AAA is accompanied by loss of elastin because of
increased elastolysis [36]. It is also associated with increasing macrophage infiltration, secondary to
low WSS in the aneurysms recirculation region [37], and thus, the spatial and temporal changes in
the WSS distribution could influence the degradation of elastin and thus the rate at which an AAA
enlarges. In this study, we propose a phenomenological model to link the degradation of elastin to
the low WSS that acts on the vascular wall. We do not explicitly model the biological mechanisms
that link degradation of elastin to the low WSS stimuli, for example, representation of ECs and
associated signalling pathways leading to increased expression of matrix degrading enzymes. It is
also of note that we do not model the intraluminal thrombus (ILT) that is present in 75% of AAA
[38]. More specifically, we assume that no ILT is present during the AAA development and that the
ECs remain functional, so as to continue to transduce a signal to the medial and adventitial layers.
We model elastin degradation in two distinct stages. First, a degradation of elastin is prescribed
to create a small AAA. This creates a perturbation to the geometry and the haemodynamic environ-
ment so that the WSS distribution is perturbed from its original distribution. Subsequently, elastin
degradation is driven by deviations in WSS from homeostatic values (which needs to be defined).
The functional form for the initial prescribed loss of elastin is as follows:
"
2 #
E
 E
1
m .1 , 2 , t / D 1  1  mmin .t / exp !1 1 , (4)
Lmi n

where 1 , 2 are the Lagrangian coordinates of the computational domain, that is, 1 2 Œ0, L1 ,
2 2 Œ0, L2 /, and mE min represents the minimum concentration of elastin, at time t , which, by con-
struction, is located at .1 , 2 / D .Lmin , 2 /. The parameter !1 controls the degree of localisation of
the degradation function (we adopt !1 D 20, see [20] for influence of this parameter). We assume
that the minimum concentration decays exponentially. Thus,
˚  E  E t =T
mE min .t / D exp ln mT .t =T / D mT , for 0 6 t 6 T (5)

where T D 0.8, mE E
T D 0.5 and Lmin D 0.5L1 . Note that at t D 0, mmin .t D 0/ D 1.
The initial prescribed aneurysm is made to stabilise (0.8 < t 6 1) so that the stretch of the col-
lagen fabric is restored to homeostatic values throughout the domain (see [9] for methodology and
Figure 11 for illustration of the initial collagen strain distribution at t D 1, i.e. the start of the FSG
coupling). To assist the stabilisation process, the growth rate parameter ˇ in (13) that influences
the rate of evolution of the collagen density is increased by a factor of 10 for (0.8 < t < 1) and
reset to its default value for (t > 1). This is to prevent further enlargement of the aneurysm being a
consequence of the collagen fabric not initially being in homeostasis.
More specifically, the concentration of elastin mE evolves according to
@mE
D FD Dmax mE (6)
@t
where t is in years, Dmax specifies the maximum rate of degradation and FD .1 , 2 , t / W 0 6 FD 6 1
is a spatially dependent function of the haemodynamic quantities to be linked to elastin degradation.
Clearly, if FD D 1, elastin degrades at a maximum rate, while if FD D 0, no degradation occurs.
For the simulations in this study, we choose Dmax D 0.5.
In this study, low WSS drives elastin degradation, and assume that if the maximum value of the
WSS is greater than a homeostatic WSS threshold h , no degradation of elastin occurs, while lower
values of WSS give rise to degradation of elastin. Additionally, we suppose that there exists a value
of WSS, say X W 0 6 X < h , at which the maximum rate of degradation occurs. To simplify

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 573

Figure 8. Spatial distribution of wall shear stress for the model of the healthy abdominal aorta prior to the
simulation of aneurysm evolution.

the modelling, we assume a simple quadratic functional form for the degradation function FD that
describes the relation between the local WSS and the degree of the degradation of elastin is adopted,
that is,
8
< 0,
ˆ  2
 > h ,
FD . .1 , 2 , t // D h .X,t / .X,t /
,  X <  < h , (7)
:̂ h .X,t /X .X,t /
1,  6 X .

We set X D 0 Pa; however, the homeostatic WSS threshold h needs to be defined. We consider
three illustrative examples:

(i) h .X, t / D 0.5Pa (8)


(ii) h .X, t / D h .X, t D 0/  0 (9)
Zt
1
(iii) h .X, t / D  .X, t /dt (10)
TL
t TL

where we define that for t < 0,  .X, t / D  .X, t D 0/ D 0 , that is, the spatial distribution for
the model of the healthy abdominal aorta prior to aneurysm formation. In case (i), h is spatially
homogeneous (reference value of 0.5 Pa), whereas in case (ii), it is spatially heterogeneous with
magnitudes defined by the initial WSS distribution for the model of the healthy abdominal aorta
(Figure 8). Case (i) assumes that the biological mechanisms responsible for elastin degradation
have a hard-wired threshold value for a specific mechanical stimuli (WSS) below which those mech-
anisms become active or appreciable. However, the phenotypic response to WSS may vary spatially
heterogeneously if other genes also modulate the process, that is, case (ii). Alternatively, temporal
heterogeneity may be a natural consequence of the fact that ECs are in a continual state of turnover,
that is, they may recalibrate themselves to the local mechanical environment. Hence, we propose a
novel approach: Case (iii) h is assumed to be spatially heterogeneous and temporally adaptive; for
an illustrative example, an averaging period of 5 years (TL D 5) is adopted in conjunction with an
inception period of 1 year. NB, given the functional form of (10), the inception period (yielding the
small aneurysm) needs to be of shorter duration than the averaging period, otherwise h    0,
the elastin degrades at a very slow rate when elastin degradation is coupled to low WSS and no
aneurysm growth takes place.

2.3.2. Collagen adaption. The adaption of the collagen fabric during aneurysm evolution consists
of two distinct mechanisms: growth and remodelling.

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
574 P. APARÍCIO ET AL.

The attachment stretch C AT


Collagen fibres are in a continual state of deposition and degradation and attach to the artery in a
state of stretch. Watton et al. [19] hypothesised that the collagen fibres are configured to achieve
a maximum stretch during the cardiac cycle and introduced the terminology attachment stretch,
denoted C AT .
Remodelling the recruitment configuration of collagen
The concept of an attachment stretch implies that the collagen fibres have distinct natural refer-
ence configurations. A field of recruitment stretch variables is defined: These define the stretch
(relative to the unloaded reference configuration) that the tissue must be stretched in the direc-
tion of a fibre for it to begin to bear load. The magnitude of the recruitment stretch variables
RJp .1 , 2 , t / are defined such that at t D 0, the collagen stretches are equal to the attachment
stretch throughout the domain. As the geometry evolves, the recruitment stretches evolve so that
the maximum GL strain of the collagen during the cardiac cycle remodels towards the GL attach-
C
ment strain, denoted EAT . Initially, it was assumed that maximum strains are achieved in the
systolic configuration, and thus, the remodelling scheme was implemented computationally by
considering the steady deformation of the artery at systolic pressure; however, as the geometry
becomes more complex, this is not necessarily true [19]. Consequently, the methodology is now
updated to assume that the maximum strains occur in either the systolic or diastolic configuration,
that is,

@R
Jp EJCp jmax EAT
C
D˛ C
, (11)
@t EAT
where
 
EJCp jmax D max EJCp jsys , EJCp jdias (12)

and EJCp jsys and EJCp jdias denote the magnitude of the collagen strains evaluated in the systolic
and diastolic configurations, respectively; ˛ D 0.6 years1 .
Growth/atrophy of the collagen fabric In vascular homeostasis, the mass of the collagenous
constituents is constant even though the collagen fibres are in a continual state of deposition and
degradation. However, in response to perturbations to the mechanical environment, vascular cells
can respond by upregulating synthesis and downregulating enzymes that degrade the matrix [39];
this would lead to a net increase in mass. Conversely, a downregulation of synthesis coupled with
an upregulation matrix-degrading enzymes would lead to a local decrease in mass. Note that if there
is an upregulation of synthesis balanced by an upregulation of degrading enzymes (no net change
in mass), this can be simulated by increasing the magnitude of the remodelling rate parameter ˛ in
Equation (11).
We outline the algorithm proposed to simulate this proposed by Watton & Ventikos [40]. The key
assumptions are as follows:
 The reference configuration of the cells is equal to the reference configuration of the
constituents that they are maintaining.
 The number of cells is proportional to the mass of constituents they are maintaining.
 In vascular homeostasis, the mass of the constituents is constant.
From these assumptions, a simple (linear) differential equation for adapting the concentration of the
collagenous constituents can be derived to be

@mC
Jp EJCp jmax EAT
C
D ˇmC
Jp C
, (13)
@t EAT

where ˇ.ACS / is a phenomenological growth parameter that is a function of the magnitude of the
(differential) cyclic areal stretch .ACS /, that is, ratio of the differential areal stretch at systole to the

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 575

differential areal stretch at diastole [41]. Note that Equation (13) is derived by considering perturba-
tions of the strains of the vascular cells from homeostatic levels. The assumption that the reference
configuration of the cells is equal to that of the constituent that they are maintaining leads to it being
expressed as a function of the strains of the collagen fibres.
If the cyclic deformation of the artery is too large, it may be difficult for vascular cells to remain
attached to the matrix during the cardiac cycle. Hence, we hypothesise that there exists a homeo-
static range of magnitudes of cyclic strain for the vascular cells. Given that the magnitude of the
cyclic strain decreases as an artery ages and stiffens, for example, from around 13% to 2% [42],
we suppose that the magnitude of the cyclic strain at t D 0 represents a suitable upper bound. We
assume this implies a similar upper bound for the magnitude of cyclic areal stretch. Consequently,
we select a functional form for ˇ.ACS / that increases rates of production if the magnitude of the
cyclic areal stretch increases from its initial values for the healthy artery at t D 0:


CS CS ACS
ˇ.A / D ˇ0 exp ˇ max  1, 0 . (14)
ACS jt D0
This choice is somewhat arbitrary; however, it acts to prevent unrealistically large cyclic deforma-
tions that can occur in regions of the tissue as the aneurysm evolves (primarily the neck region). For
the analysis in this paper, we take ˇ0 D 0.7 years1 , ˇ CS D 3. Lastly, to simplify the presentation
of the results, we define the average fibre concentration mC of the medial and adventitial layers,
where
!
1 mCMC C mM
C
mCAC C mA
C
C
m D HM C HA . (15)
H 2 2

3. RESULTS

We present three examples of enlarging aneurysms. In all cases, the elastin degradation is initially
prescribed to create a small
axisymmetric aneurysm for which the collagen fabric is initially in
homeostasis C D C AT . For t > 1, evolution of the aneurysm is driven by the loss of elastin
being linked to deviations of WSS from homeostatic  levels. This leads to a transfer of load onto
the collagen, and the collagen stretch increases C > C AT . Subsequent collagen G&R attempts
 C C

restore collagen homeostasis  ! AT . The degradation of elastin is driven by reductions in
the magnitude of WSS below a homeostatic threshold, denoted h . Recall three definitions for
the WSS threshold h are considered: Case (i), it is constant, that is, h .X, t / D 0.5 Pa; case (ii)
h .X, t / D 0 .X/ where 0 is the mean WSS distribution for the healthy abdominal aorta (Figure 8).
Lastly, for case (iii), h D h .X, t ), that is, the WSS threshold is spatially heterogeneous and tem-
porally adaptive (see (10)). We first discuss and compare the evolution of the geometry, WSS,
degradation factor FD and elastin concentration mE for cases (i) and (ii). We then compare the
evolution of the same variables for case (iii) and, additionally, analyse the evolution of the GL
strains of elastin and collagen, the average collagen concentration and two indices associated with
the cyclic deformation, the cyclic areal stretch ACS and the biaxial stretch index BSI . The results
of the simulations are presented in Figures 9–12. All these images utilise identical viewpoints; the
viewpoint is depicted in more details in Figure 8.
The initial geometries at t D 0 for cases (i) and (ii) are identical (compare Figures 9(a) and
(b)). As the aneurysms enlarge, a preferential anterior bulging occurs because of the simulation of
contact with the spine (modelled as a spring-backed plate located at y D 0.12m). Interestingly, a
proximal–distal asymmetry in the geometries also develops. This is a consequence of the aneurysms
enlarging more quickly in the upstream direction.
At t D 0, the spatial distribution of WSS is identical for cases (i) and (ii), i.e. a localised decrease
in WSS occurs within the small (prescribed) axisymmetric aneurysm (see Figure 9(a and b)). As the
aneurysms enlarge, the proximal regions experience a lower WSS than the distal region, and a region
of elevated WSS appears distal of the downstream neck. The degradation factor ranges from 0 to 1
and relates to the rate at which the elastin is degrading. At t D 0, the distributions of FD appear

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
576 P. APARÍCIO ET AL.

Figure 9. Comparison of (top row) wall shear stress, (middle row) elastin degradation factor FD and (bot-
tom row) elastin concentration mE at t D 1, 2, 3, 4 and 5 years for (left) case (i) h D 0.5Pa and (right) case
(ii), h D 0 .X/ where the wall shear stress distribution for the model of the healthy abdominal aorta, 0 .X/,
is illustrated in Figure 8.

fairly similar for cases (i) and (ii), that is, maximum degradation occurs within the small prescribed
axisymmetric aneurysm (Figure 9(c and d)). However, some differences in the distributions of FD
are noticeable. For instance, for case (i), FD is elevated at the distal end of the abdominal aorta. This
is a consequence of the WSS in this region being below the threshold of 0.5 Pa required for elastin
degradation. In comparison, for case (ii) where the threshold is taken to be the heterogeneous spatial
distribution of WSS for the healthy abdominal aorta (Figure 8), FD .t D 0/ D 0 in the lower section
of the distal abdominal aorta (Figure 9(d)). However, note that at t D 3, elastin starts to degrade
in this region (Figure 9 t D 3) because of a reduction in the WSS below the WSS threshold as the
aneurysm enlarges that affects the downstream WSS distribution.
At t D 0, the spatial distribution of elastin concentration mE is identical for cases (i) and (ii).
Elastin subsequently degrades in the regions in which FD is nonzero; maximum rates of degrada-
tion occur when FD D 1. In both cases (i) and (ii) , the elastin evolves a complex spatial distribution
as a consequence of the evolving pattern of FD that is linked to the evolving pattern of the WSS.
However, note that for case (i), a region of elastin degrades at the distal region of the abdominal
aorta because of FD being elevated. This problem is not encountered for case (ii) where a spatially
heterogeneous threshold value is defined based on the initial WSS distribution.
Figure 10(a–d) illustrates the evolution of WSS, h , FD and mE for case (iii), respectively. As
with cases (i) and (ii), it can be seen that the geometry evolves more quickly in the upstream direc-
tion, and a preferential anterior bulging occurs. However, it is of note that this aneurysm took longer
to evolve. This is a consequence of the difference between the WSS and the homeostatic WSS h
being smaller. The WSS also follows a similar pattern to before, that is, reduced in the proximal
region of the aneurysm and elevated towards the distal end with a maximum just downstream of
the aneurysm. At t D 0, the homeostatic WSS threshold h is fairly uniform with a slight decrease
in the downstream direction. The spatial distribution of h loosely follows that for the WSS with a
timelag; however, the evolution of the spatial distribution of FD is more complex. Initially, FD is
maximally elevated in the central region of the aneurysm. However, as the aneurysms enlarge, the
region of maximum elevation shifts towards the proximal section of the aneurysm. Additionally, as
the aneurysm enlarges and alters the downstream WSS distribution, a local minima in WSS develops

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 577

Figure 10. Case (iii): spatially and temporally heterogeneous homeostatic wall shear stress (WSS) thresh-
old. Evolution of (a) WSS .X, t/, (b) homeostatic WSS threshold h .X, t/, (c) elastin degradation factor
FD and (d) elastin concentration mE at t D 1, 3, 5, 7, 9, 11 years.

that leads to a local elevation of FD and elastin being degraded in this region. The elastin concen-
tration mE evolves to a complex distribution with most of the elastin degraded in the proximal
region.
Figures 11(a and b) illustrate the evolution of the GL strains E11 and E22 of the elastin, respec-
tively. The strains increase from E11 D 0.345 and E22 D 0.281 for the healthy abdominal aorta to
E11 D 1.6 and E22 D 1.9 at t D 10. An asymmetry in the evolution of the spatial distribution is
apparent, that is, slightly larger magnitudes are observed towards the proximal end. It can also be
seen that the axial strains in the downstream section of the parent artery decrease as the aneurysm
enlarges. This is a consequence of the axial retraction of distal region of the abdominal aorta (due to
the axial prestretch of the healthy artery and remodelling of collagen) assisting the axial expansion
of the AAA.
The evolution of the collagen fibre GL strains differs substantially from that of the elastin strains
because of the evolution of the reference configurations that the fibres are recruited to load bearing;
C
Figure 11(c and d) depicts the evolution of the medial and adventitial collagen GL strain EM C and
C
EAC , respectively. At t D 0, even though the geometry has changed (and the elastin strains have
increased), the collagen fabric is in homeostasis, that is, the collagen strain is constant throughout the
C
domain with a magnitude of the attachment strain EAT D 0.073. This is a consequence of the fibre
recruitment stretch and concentration having evolved so that the collagen fabric is in homeostasis.

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
578 P. APARÍCIO ET AL.

Figure 11. Case (iii): spatially and temporally heterogeneous homeostatic wall shear stress threshold. Evo-
C
lution of Green–Lagrange (GL) strains of elastin, (a) E11 and (b) E22 , and GL strains of collagen, (c) EM C
C
and (d) EAC at t D 1, 3, 5, 7, 9, 11 years.

As the elastin degrades, the aneurysm enlarges (t > 0), and the load is shifted onto the collagen, and
hence, the magnitudes of the collagen strains increase. However, the fibre recruitment stretch and
concentration continue to evolve, and hence, the strains increase negligibly. In fact, interestingly,
it can be seen that at t D 4, the collagen strains have almost restored back to homeostatic levels.
However, as the aneurysm further enlarges, elevated collagen strains are observed at the proximal
end of the aneurysm as this region expands more rapidly. The average collagen concentration mC
increases to compensate for the loss of load borne by the elastin (Figure 12(a)). Differences can be
seen between the evolution of the medial and adventitial collagen strains. This is because the adven-
titial fibres have a preferential axial orientation, and thus, they are more affected by the retraction of
the distal region of the abdominal aorta.
Lastly, we consider the evolution of the cyclic deformation environment of the vascular wall.
Figure 12(b) illustrates the evolution of the cyclic areal stretch ACS . For the healthy artery,
ACS D 1.1 throughout the domain. However, at inception (12(b) t D 0), slightly elevated cyclic
areal stretches (ACS D 1.16) are present in the proximal and distal necks (Figure 12(b)). This is
a consequence of a slight axial pulsation of the aneurysm occurring as it expands and contracts
during the cardiac cycle. As the elastin degrades and collagen increasingly takes over the load
bearing, the tissue becomes stiffer and ACS reduces to around 1.04 within the aneurysm region.
Figure 12(c) illustrates the evolution of the biaxial stretch index BSI . The model of the healthy

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 579

Figure 12. Case (iii): spatially and temporally heterogeneous h . Evolution of (a) average collagen density
mC , (b) cyclic areal stretch ACS and (c) biaxial stretch index BSI at t D 1, 3, 5, 7, 9, 11 years.

artery is subject to one-dimensional cyclic circumferential stretching and initially, BSI D 0,


throughout the domain. It can be seen that as the aneurysm develops, the distribution of BSI
becomes more complex. In the central region of the aneurysm, maximum values of BSI D 1 are
achieved; these locations are subject to equibiaxial cyclic stretching. The distal region of the aorta
(where no aneurysm has developed) has an increased BSI because of axial pulsation of this region,
that is, this region contracts as the pressure increases and the aneurysm enlarges.

4. DISCUSSION

An FSG computational framework for modelling aneurysm evolution is presented. A realistic struc-
tural model of the arterial wall is integrated into a patient-specific geometry of the vasculature. This
enables physiologically representative distributions of haemodynamic stimuli, obtained from a rigid-
wall CFD analysis, to be linked to G&R algorithms. Additionally, a quasistatic structural analysis
quantifies the cyclic deformation of the arterial wall so that collagen growth and collagen remod-
elling can be explicitly linked to the cyclic deformation of vascular cells. The elastin degradation is
linked to deviations of WSS below homeostatic levels. The homeostatic threshold is allowed to be
spatially and temporally heterogeneous. Three distinct definitions of the WSS threshold, (h ), were
considered: constant h , spatially heterogeneous h , spatially heterogeneous and temporally adap-
tive h . We observed different spatial variations in G&R and the geometries of the aneurysms as they
evolved (Figure 13(a–c)). Hence, the spatial and temporal heterogeneity of endothelial homeostasis
may be an important feature to consider when developing predictive models of aneurysm evolution.
Computational fluid dynamic analyses of the vasculature illustrate that the WSS is spatially het-
erogeneous within the vasculature tree. Hence, it is perhaps not so surprising that EC phenotypes
display heterogeneity in both structure and function in the vascular tree [43, 44]. Several recent arti-
cles by Aird (see [11, 44, 45]) suggest that this heterogeneity is the result of both nature and nurture,
that is, a consequence of the epigenetics and environment. It represents a balance between stability

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
580 P. APARÍCIO ET AL.

Figure 13. In-plane profiles of aneurysms. Upper images depict the XY plane, and lower pictures depict the
XZ planes for the different cases: (a) spatially homogeneous h at t D 4, (b) spatially heterogeneous h at
t D 4 and (c) spatially heterogeneous and temporally adaptive h at t D 10.

and plasticity in gene expression and phenotype [45] and enables the ECs to adapt and function in
different environments within the vasculature [11]. Given the plasticity of ECs, we believe it may be
beneficial to keep in mind that homeostatic values could potentially be both spatially heterogeneous
and temporally adaptive.
Mathematical models of arterial adaption in response to altered flow/pressure often utilise concep-
tual cylindrical geometries to represent the artery, and thus, it is not necessary for the definition of
the homeostatic WSS to take into account spatial heterogeneities in the magnitude. For instance,
a mean WSS of 1.5 Pa is typically taken to be representative for large arteries [1–5], whereas
Valentin et al. [6–8] utilise 5.06 Pa for WSS homeostasis for a mature basilar artery. For more
complex geometries, the WSS distribution is nonuniform. Nevertheless, to date homeostatic WSS
magnitudes utilised in FSG models have been constant, for example, the following thresholds were
adopted for driving elastin degradation in recent FSG models of aneurysm evolution: WSS < 2 Pa
[9, 10]; WSS< 0.8 Pa [22]. In this article, we proposed three different approaches to defining the
homeostatic level of WSS (uniform, spatially heterogeneous, spatially heterogeneous and tempo-
rally adaptive) and illustrated that the choice of definition influences the predicted evolution of the
aneurysm. Of course, it should be recognised that this is a novel exploratory conceptual model to
highlight the need for improved understanding and modelling in this area. There is undoubtedly a
need for improved mathematical modelling, numerical parameter studies and guidance from exper-
iments. Moreover, an improved understanding and representation of how ECs sense WSS, how this
influences their functionality and how they communicate with other vascular cells is needed.
The vast majority of AAAs are at least partially lined by an ILT [46]. In fact, it can be seen from
the CT scan of the patient used in this study that a thick ILT is present (Figure 5). The ILT alters the
geometry of the lumen, and thus, the CFD stimuli that act on the luminal layer of the vessel wall.
Furthermore, it is biologically active because of its heterogeneous synthesis, storage and release of
relevant biomolecules [47]. Consequently, it plays a significant role in the aetiology, progression

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 581

and predicted outcome of the disease by altering the mechanical and biochemical condition of the
arterial wall [48]. The thrombus cascade consists of initiation through platelet activation, deposi-
tion and aggregation, majorally through fibrinogen polymerisation into fibrin fibres [49]. During the
platelet activation stage, the endothelial layer is locally disrupted and collagen fibres are exposed to
circulating platelets, which adhere directly to collagen with collagen-specific surface receptors [50].
Hence, ILT not only physically separates the flowing blood and resulting WSS from the wall but it
may also directly damage or even completely destroy the endothelial layer that is responsible for
transducing the WSS stimulus into biological activity. Conflictingly, while an increase in the oxida-
tive and proteolytic load on the wall may have a deleterious effect [51], the ILT may also act as a
buffer against wall stress offering a protective role for AAA against rupture [52]. We fully recognise
that a major limitation of our model of AAA evolution is that we do not account for the ILT. Never-
theless, our model has applicability to the early progression of AAAs prior to thrombus formation.
Furthermore, integration of a model for thrombus formation would be a natural development of our
simulation framework. However, incorporating a model for ILT evolution is challenging. It requires
integration of a model for thrombus evolution [53–56] in conjunction with a constitutive model to
describe its mechanical response [57–59]. From a modelling point of view, the modelling of ILT
during aneurysm evolution may merit investigation with conceptual mathematical models prior to
implementation into FSG approaches.
We modelled the healthy abdominal aorta as a cylindrical nonlinear elastic membrane subject to
an axial prestretch and uniform internal pressure. The distal and proximal ends of the abdominal
aorta are fixed to simulate vascular tethering by the renal and iliac arteries. Upstream and down-
stream sections of physiological geometry are utilised to obtain more realistic haemodynamics as
opposed to utilising straight inlets and outlets [9]; these sections are not subject to G&R. Mod-
elling the artery as a cylinder implies that the structure of the collagen and spatial distributions of
G&R variables are initially homogeneous and can be analytically determined. However, it should be
noted that the abdominal aorta is slightly tortuous and tapers; thus, representing it as a cylinder may
lead to inaccuracies in the distribution of WSS and consequently in predictions of model evolution.
Furthermore, we point out that as the aneurysm evolves, the boundaries will influence the evolving
geometry. For instance, it can be seen that for the patient AAA, the diameter of the proximal sec-
tion of the aorta enlarges (Figure 5(e)); however, such features cannot be captured with the current
modelling framework, and this is an inherent limitation. Interestingly, Zeinali-Davarani et al. [21]
applied a stress-mediated constrained mixture FEM approach [60] to model AAA evolution [22]
using patient-specific geometries with a nonlinear membrane formulation [61, 62]. A numerical
approach is implemented so that the constituents are initially in homeostasis on this more complex
geometry. While such sophistications are necessary for the development of patient-specific mod-
els, hypotheses can be explored and insight obtained with (simpler) nonpatient specific geometries.
For instance, the focus of our study was to illustrate competing hypotheses for how to define the
homeostatic WSS distribution.
We modelled the arterial wall as a nonlinear elastic membrane. The motivation for this is twofold.
Firstly, the residual strain that is present in the unloaded configuration gives rise to an approximately
uniform strain field through the thickness of the arterial wall at physiological pressures. Secondly, if
it is assumed that the physiological mechanism by which collagen fibres attach to the artery is inde-
pendent of both the current configuration of the artery and the radial position in the arterial wall,
then the remodelling process may naturally maintain a uniform strain field (in the collagen fibres)
through the thickness of the arterial wall as the AAA develops. These considerations thus support
the suitability of a membrane model to represent the development of a AAA at physiological pres-
sures. However, the distinction between the media and adventitia is significantly blurred in AAAs
as the smooth muscle and elastin in the media are lost. A limitation of a membrane model is that it
assumes constant, distinct properties in each layers. We point out, though, that the G&R framework
of Watton et al. [19] has recently been extended to consider transmural variations of G&R for a
thick-walled model of the artery [41, 63]. This provides the framework for studying the influence
of transmural variations in biochaemomechanical stimuli on G&R, simulating the blurring of the
medial/adventitial layers and modelling volumetric growth.

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
582 P. APARÍCIO ET AL.

In the model presented, the collagen fibre orientations are fixed relative to the unloaded reference
configuration of the artery. Consequently, in the in vivo configuration, they reorient in the directions
of increasing principal stretch. This approach is motivated by the assumption that new collagen
fibres in the arterial wall are laid down along the direction of existing fibres. Biologically, the jus-
tification for this is that collagen depositing and degrading fibroblasts move in the ECM along the
existing fibres [64], and thus, ECM G&R at future time points is dependent on the ECM structure at
the current time point. The fibres reorient (in the loaded configuration) towards directions of increas-
ing principal stretch as the aneurysm evolves. This approach is also adopted by Zeinali-Davarani
et al. [21], that is, in vivo configuration, newly deposited collagen fibres align along the direction
of the existing collagen fibres. Given that as an AAA develops the collagen takes over an increas-
ingly important load bearing role, accurately modelling how this structure adapts is important for
predicting disease evolution. Implementing more sophisticated models to represent the remodelling
of fibre alignment [65–67], collagen fibril distribution [68], fibre dispersion [69] or proteoglycan
cross-bridges [70] may prove useful in this respect. Alternatively, it would be of interest and value
to improve the representation of the mechanobiology of the fibroblasts by representation of the sig-
nalling pathways relating the mechanical stimuli acting on fibroblast cells to the biomolecules they
produce (growth factors, procollagen, zymogens), the existing enzymes (in particular those secreted
by immune cells) and the production/degradation of collagen, for example, implementation of a
model such as that proposed by Dale et al. [71]. Furthermore, given that vascular smooth muscle
cells (VSMCs) secrete connective tissue and matrix degrading enzymes [72] and are subject to
apoptosis during AAA evolution [73], explicitly modelling the VSMCs with a suitable constitutive
model [74–76] is needed.
The structural arrangement of vascular cells is influenced by their local mechanical environ-
ment. For instance, fibroblasts align perpendicular to the direction of interstitial flow [77], and their
orientation may be influenced by the direction of maximum principal stress or strain [78] or cyclic
stretch [79]. Recently, a novel parameter was proposed to characterise the biaxial cyclic stretch
environment [10], that is, the biaxial stretch index BSI : BSI D 0 denotes one-dimensional cyclic
stretch, and BSI D 1 denotes equibiaxial stretching. In this study, it was observed that as the
AAA evolved, the BSI distribution changed significantly: Within the aneurysm, the tissue experi-
ences almost equibiaxial cyclic stretching. We suggest analysing the evolving distributions of BSI
could be used to test competing hypotheses for remodelling of cell/fibre alignment and dispersion.
Elucidating the relationship between the cyclic mechanical environment of vascular cells and their
orientation and functionality is important to develop more sophisticated computational models.
We adopted a steady flow analysis to reduce the cost of the CFD simulations. The justifica-
tion for this is that the spatial distributions of WSS for steady flow is qualitatively similar to the
(time-averaged) mean WSS for pulsatile flow. Figure 14 illustrates the distribution of WSS and
time-averaged WSS from a pulsatile flow simulation (mean inlet flow and mean outlet pressures are
equal to those of the steady simulation; Figure 7) following the prescribed degradation of elastin
to create a small outpouching of the domain. It can be seen that while locations of maxima and

Figure 14. Spatial distribution of (left) wall shear stress magnitudes from steady flow and (right) time-
averaged WSS magnitude from pulsatile flow.

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 583

minima of WSS are similar locations, there are small differences in the spatial distributions. For
the purposes of investigating qualitative behaviour of a model that links elastin degradation to the
local magnitude of WSS, we suggest that a steady flow analysis is sufficient. However, if precise
magnitudes of WSS are of critical importance or if other haemodynamic measures that can only
be derived from a pulsatile flow analysis are required, then, of course, a pulsatile flow analysis is
necessary. For instance, the maximum WSS during the cardiac cycle acting on an EC may be an
influential stimuli or indices associated with the oscillatory nature of the WSS vectors.
We adopt a rigid-walled fluid analysis to quantify the haemodynamic stimuli acting on the vessel
wall and a quasistatic structural analysis to quantify the cyclic deformation of the arterial wall. This
is a simple and numerically efficient approach to quantify the mechanical environment of the vas-
cular cells because of the WSS and cyclic pressure. It only results in a doubling of computational
time for the structural analysis. This simplification affects the pressure and WSS spatial distribu-
tions: differences will arise in the WSS distributions between rigid wall and deformable wall models
[80, 81]; while we prescribe a spatially uniform systolic/diastolic pressures, for a full fluid-structure
interaction analysis the pressure distributions would be nonuniform.

5. CONCLUSION

We sophisticated a FSG computational framework and applied it to simulate AAA evolution.


A novel approach to simulate the spatial (and temporal) heterogeneity of the vascular endothelium
and link elastolysis to WSS is proposed, and the growth and the remodelling of vascular constituents
is explicitly linked to the local cyclic deformation of the wall. The growth model is integrated into
a patient-specific geometry so that physiologically realistic spatial distributions of haemodynamic
stimuli can be used as input to the G&R algorithms. We observe that to predict disease progression,
there is a need for improved understanding and modelling of the spatial and temporal heterogeneity
of vascular homeostasis and its influence on G&R processes. While we focused on AAA, this
general computational framework provides the basis for modelling a range of vascular diseases
where haemodynamic stimuli play an influential role on disease evolution.

ACKNOWLEDGEMENTS
P. Aparício is funded by the Engineering and Physical Sciences Research Council. We acknowledge The
Centre of Excellence in Personalised Healthcare (funded by the Wellcome Trust and EPSRC, grant number
WT 088877/Z/09/Z) providing some financial support towards Dr Paul Watton. Lastly, we acknowledge the
Harwell Software Library (www.hsl.ac.uk) for granting UK academics the free use of its FORTRAN
subroutines in noncommercial applications: MA38 was employed to solve the linear system that arises in the
Newton iteration, which is required to update the deformation of the arterial wall at successive time steps.

REFERENCES
1. Rachev A. A model of arterial adaptation to alterations in blood flow. Journal of Elasticity and the Physical Science
of Solids 2000; 61(1–3):83–111.
2. Gleason RL, Taber LA, Humphrey JD. A 2-D model of flow-induced alterations in the geometry, structure and
properties of carotid arteries. Journal of Biomechanical Engineering 2004; 126:371–381.
3. Gleason RL, Humphrey JD. A mixture model of arterial growth and remodeling in hypertension: altered muscle tone
and tissue turnover. Journal of Vascular Research 2004; 41:352–363.
4. Gleason RL, Humphrey JD. Effects of a sustained extension on arterial growth and remodeling: a theoretical study.
Journal of Biomechanical Engineering 2007; 38:1255–1261.
5. Figueroa CA, Baek S, Taylor CA, Humphrey JD. A computational framework for fluid-solid-growth model-
ing in cardiovascular simulations. Computational Methods in Applied Mechanics and Engineering 2009; 198:
3583–3602.
6. Valentín A, Humphrey JD. Modeling effects of axial extension on arterial growth and remodeling. Medical and
Biological Engineering and Computing 2009; 47:979–987.
7. Valentín A, Humphrey JD. Evaluation of fundamental hypotheses underlying constrained mixture models of arterial
growth and remodelling. Philosophical Transactions of the Royal Society of London, Series: A 2009; 367:3585–3606.
8. Valentín A, Humphrey JD. Parameter sensitivity study of a constrained mixture model of arterial growth and
remodeling. Journal of Biomechanical Engineering 2009; 10:101006.

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
584 P. APARÍCIO ET AL.

9. Watton PN, Raberger NB, Holzapfel GA, Ventikos Y. Coupling the hemodynamic environment to the evolu-
tion of cerebral aneurysms: computational framework and numerical examples. ASME Journal of Biomechanical
Engineering 2009; 131:101003.
10. Watton PN, Selimovic A, Raberger NB, Huang P, Holzapfel GA, Ventikos Y. Modelling evolution and the evolv-
ing mechanical environment of saccular cerebral aneurysms. Biomechanics and Modeling in Mechanobiology 2011;
11:109–132.
11. Regan ER, Aird WC. Dynamical systems approach to endothelial heterogeneity. Circulation Research June 2012;
111(1):110–30.
12. Sakalihasan N, Kuivaniemi H, Nusgens B, Durieux R, Defraigne JG. Aneurysm: Epidemiology Aetiology and
Pathophysiology, Biomechanics and Mechanobiology of Aneurysms. Springer-Verlag: Heidelberg, 2011.
13. Wilmink WBM, Quick CRG, Hubbard CS, Day NE. The influence of screening on the incidence of ruptured
abdominal aortic aneurysms. Journal of Vascular Surgery 1999; 30:203–208.
14. Carrell TWG, Smith A, Burnand KG. Experimental techniques and models in the study of the development and
treatment of abdominal aortic aneurysms. Journal of Surgery 1999; 86:305–312.
15. Davies MJ. Aortic aneurysm formation: Lessons from human studies and experimental models. Circulation 1998;
98:193–195.
16. Baxter BT, Terrin MC, Dalman RL. Medical management of small abdominal aortic aneurysms. Circulation 2008;
117:1883–1889. DOI: 10.1161/CIRCULATIONAHA.107.735274.
17. Raghavan ML, Vorp DA. 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 Biomechanics 2000;
33:475–482.
18. Powell JT, Brady AR. Detection, management and prospects for the medical treatment of small abdominal aortic
aneurysms. Arteriosclerosis, Thrombosis and Vascular Biology 2004; 24:241–245.
19. Watton PN, Hill NA, Heil M. A mathematical model for the growth of the abdominal aortic aneurysm. Biomechanics
and Modeling in Mechanobiology 2004; 3:98–113.
20. Watton PN, Hill NA. Evolving mechanical properties of a model of abdominal aortic aneurysm. Biomechanics and
Modeling in Mechanobiology 2009; 8:25–42.
21. Zeinali-Davarani S, Sheidaei A, Baek S. A finite element model of stress-mediated vascular adaptation: application
to abdominal aortic aneurysms. Computer Methods in Biomechanics and Biomedical Engineering 2011; 14:803–817.
22. Sheidaei A, Hunley SC, Zeinali-Davarani S, Raguin LG, Baek S. Simulation of abdominal aortic aneurysm growth
with updating hemodynamic loads using a realistic geometry. Medical Engineering and Physics 2011; 33:80–88.
23. Holzapfel GA, Gasser TC, Ogden RW. A new constitutive framework for arterial wall mechanics and a comparative
study of material models. Journal of Elasticity 2000; 61:1–48.
24. Nissen R, Cardinale GJ, Udenfriend S. Increased turnover of arterial collagen in hypertensive rats. Proceedings of
the National Academy of Sciences of the United States of America. Medical Sciences 1978; 75:451–453.
25. Humphrey JD. Remodelling of a collagenous tissue at fixed lengths. Journal of Biomechanical Engineering 1999;
121:591–597.
26. Hennig T, Mogensen C, Kirsch J, Pohl U, Gloe T. Shear stress induces the release of an endothelial elastase: role in
integrin (v)(3)-mediated fgf-2 release. CORD Conference Proceedings January 2011; 48(6):453–464.
27. Humphrey JD, Taylor CA. Intracranial and abdominal aortic aneurysms: Similarities, differences, and need for a new
class of computational models. Annual Review of Biomedical Engineering 2008; 10:221–246.
28. Dua MM, Dalman RL. Hemodynamic influences on abdominal aortic aneurysm disease: application of biomechanics
to aneurysm pathophysiology. Vascular Pharmacology 2010; 53:11–21.
29. Watton PN, Ventikos Y, Holzapfel GA. Modelling the mechanical response of elastin for arterial tissue. Journal of
Biomechanics 2009; 42:1320–1325.
30. Chen H, Selimovic A, Thompson H, Chiarini A, Penrose J, Ventikos Y, Watton PN. Investigating the influ-
ence of haemodynamic stimuli on intracranial aneurysm inception. Annals of Biomedical Engineering 2013;
41(7):1492–1504.
31. Patakar HS. Numerical Heat Transfer and Fluid Flow (Hemisphere Series on Computational Methods in Mechanics
and Thermal Science). Taylor & Francis: London, 1980.
32. Reymond P, Merenda F, Perren F, Rufenacht D, Stergiopulos N. Validation of a one-dimensional model of the
systemic arterial tree. American Journal of Physiology: Heart and Circulatory Physiology 2009; 297:H208–H222.
33. Villa-Uriol MC, Berti G, Hose DR, Marzo A, Chiarini A, Penrose J, Pozo J, Schmidt JG, Singh P, Lycett R, Larra-
bide I, Frangi AF. @neurist complex information processing toolchain for the integrated management of cerebral
aneurysms. Interface Focus 2011; 1:308–319.
34. Reymond P, Bohraus Y, Perren F, Lazeyras F, Stergiopulos N. Validation of a patient-specific one-dimensional
model of the systemic arterial tree. American Journal of Physiology: Heart and Circulatory Physiology 2011;
301:H1173–H1182.
35. Deplano V, Meyer C, Guivier-Curien C, Bertrand E. New insights into the understanding of flow dynamics in an in
vitro model for abdominal aortic aneurysms. Medical Engineering and Physics 2013; 35(6):800–809.
36. Shimizu K, Mitchell RN, Libby P. Inflammation and cellular immune responses in abdominal aortic aneurysms.
Arteriosclerosis, Thrombosis and Vascular Biology 2006; 26:987–994.
37. Sho E, Sho M, Hoshina K, Kimura H, Nakahashi TK, Dalman RL. Hemodynamic forces regulate mural macrophage
infiltration in experimental aortic aneurysms. Experimental and Molecular Pathology 2004; 72:108–116.

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
MODELLING ENDOTHELIAL HETEROGENEITY 585

38. Wang DHJ, Makaroun MS, Webster MW, Vorp DA. Effect of intraluminal thrombus on wall stress in patient-specific
models of abdominal aortic aneurysms. Journal of Vascular Surgery 2002; 36:598–604.
39. Chiquet M. Regulation of extracellular matrix gene expression by mechanical stress. Matrix Biology 1999;
18:417–426.
40. Watton PN, Ventikos Y. Modelling evolution of saccular cerebral aneurysms. Journal of Strain Analysis 2009;
44:375–389.
41. Schmid H, Grytsan A, Postan E, Watton PN, Itskov M. Influence of differing material properties in media and adven-
titia on arterial adaption: application to aneurysm formation and rupture. Computer Methods in Biomechanics and
Biomedical Engineering 2013; 16(1):33–53.
42. Länne T, Sonesson B, Bergqvist D, Bengtsson H, Gustafsson D. Diameter and compliance in the male human
abdominal aorta: Influence of age and aortic aneurysm. European Journal of Vascular Surgery 1992; 6:178–184.
43. Davies PF. Hemodynamic shear stress and the endothelium in cardiovascular pathophysiology. Nature Clinical
Practice. Cardiovascular Medicine 2009; 6(1):16–26.
44. Aird WC. Endothelium in health and disease. Pharmacological Reports : PR 2008; 60(1):139–43.
45. Aird WC. Mechanisms of endothelial cell heterogeneity in health and disease. Circulation Research 2006;
98(2):159–62.
46. McGloughlin T. Biomechanics and Mechanobiology of Aneurysms, Studies in mechanobiology, Tissue Engineering
and Biomaterials 7. Springer: Heidelberg, 2011.
47. Wilson JS, Virag L, Di Achille P, Karaj I, Humphrey JD. Biochemomechanics of intraluminal thrombus in abdominal
aortic aneurysms. Journal of Biomechanical Engineering 2013; 135(2):021011. DOI: 10.1115/1.4023437.
48. Humphrey JD, Holzapfel GA. Mechanics, mechanobiology, and modeling of human abdominal aorta and aneurysms.
Journal of Biomechanics 2012; 45(5):805–814.
49. Moiseyev G, Givli S, Bar-Yoseph PZ. Fibrin polymerization in blood coagulationa statistical model. Journal of
Biomechanics 2013; 46(1):26–30.
50. Pallister CJ, Watson M. Haematology, second edition. Scion Publishing Ltd.: Banbury, UK, 2010.
51. Michel JB, Martin-Ventura JL, Egido J, Sakalihasan N, Treska V, Lindholt J, Allaire E, Thorsteinsdottir U, Cockerill
G, Swedenborg J. Novel aspects of the pathogenesis of aneurysms of the abdominal aorta in humans. Cardiovascular
Research 2011; 90:18–27.
52. Martufi G, Gasser TC. Review: The role of biomechanical modeling in the rupture risk assessment for abdominal
aortic aneurysms. Journal of Biomechanical Engineering 2013; 135(2):021010. DOI: 10.1115/1.4023254.
53. Xu Z, Chen N, Kamocka MM, Rosen ED, Alber MS. Multiscale model of thrombus development. Journal of Royal
Society Interface 2008; 5:705–722.
54. Xu Z, Kamocka M, Alber M, Rosen ED. Computational approaches to studying thrombus development. Arterioscle-
rosis, Thrombosis, and Vascular Biology 2011; 31:500–505.
55. Biasetti J, Hussain F, Gasser TC. Blood flow and coherent vortices in the normal and aneurysmatic aortas: a fluid
dynamical approach to intra-luminal thrombus formation. Journal of the Royal Society Interface 2011; 8:1449–1461.
56. G M, Bar-Yoseph P Z. Computational modeling of thrombosis as a tool in the design and optimization of vascular
implants. Journal of Biomechanics 2013; 46(2):248–252.
57. Van de Geest JP, Sacks MS, Vorp DA. A planar biaxial constitutive relation for the luminal layer of intra-luminal
thrombus in abdominal aortic aneurysms. Journal of Biomechanics 2006; 39:2347–2354.
58. van Dam EA, Dams SD, Peters GWM, Rutten MCM, Schurink GWH, Buth J, van de Vosse FN. Non-linear vis-
coelastic behavior of abdominal aortic aneurysm thrombus. Biomechanics and Modeling in Mechanobiology 2008;
7:127–137.
59. Tong J, Sommer G, Regitnig P, Holzapfel GA. Dissection properties and mechanical strength of tissue components
in human carotid bifurcations. Annals of Biomedical Engineering 2011; 39:1703–1719.
60. Baek S, Rajagopal KR, Humphrey JD. A theoretical model of enlarging intracranial fusiform aneurysms. Journal of
Biomechanical Engineering 2006; 128:142–149.
61. Zeinali-Davarani S, Raguin LG, Baek S. An inverse optimization approach toward testing different hypotheses of
vascular homeostasis using image-based models. International Journal of Structural Changes in Solids - Mechanics
and Applications 2011; 3:33–34.
62. Zeinali-Davarani S, Raguin LG, Vorp DA, Baek S. Identification of in vivo material and geometric parameters of
a human aorta: toward patient-specific modeling of abdominal aortic aneurysm. Biomechanics and Modeling in
Mechanobiology 2010; 10:689–699.
63. Schmid H, Watton PN, Maurer MM, Wimmer J, Winkler P, Wang YK, Roehrle O, Itskov M. Impact of trans-
mural heterogeneities on arterial adaptation: application to aneurysm formation. Biomechanics and Modeling in
Mechanobiology 2010; 9:295–315.
64. Dallon J, Sherratt JA. A mathematical model for fibroblast and collagen orientation. Bulletin of Mathematical Biology
1998; 60:101–130.
65. Driessen NJB, Wilson W, Bouten CVC, Baaijens FPT. A computational model for collagen fibre remodelling in the
arterial wall. Journal of Theoretical Biology 2004; 226:53–64.
66. Baaijens F, Bouten C, Driessen N. Modeling collagen remodeling. Journal of Biomechanics 2010; 43:166–175.
67. Creane A, Maher E, Sultan S, Hynes N, Kelly DJ, Lally C. A remodelling metric for angular fibre distributions and its
application to diseased carotid bifurcations. Biomechanics and Modeling in Mechanobiology 2012; 11(6):869–882.
DOI: 10.1007/s10237-011-0358-3.

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm
586 P. APARÍCIO ET AL.

68. Zulliger MA, Fridez P, Hayashi K, Stergiopulos N. A strain energy function for arteries accounting for wall
composition and structure. Journal of Biomechanics 2004; 37:989–1000.
69. Gasser TC, Ogden RW, Holzapfel GA. Hyperelastic modelling of arterial layers with distributed collagen fibre
orientations. Journal of the Royal Society Interface 2006; 3:15–35.
70. Gasser TC. An irreversible constitutive model for fibrous soft biological tissue: a 3-D microfiber approach with
demonstrative application to abdominal aortic aneurysms. Acta Biomaterialia 2011; 7:2457–2466.
71. Dale P, Sherratt J, Maini P. A mathematical model for collagen fibre formation during foetal and adult dermal wound
healing. Proceedings of the Royal Society of London 1996; 263(B):653–660.
72. Asanuma K, Magid R, Johnson C, Nerem RM, Galis ZS. Uniaxial strain regulates matrix-degrading enzymes pro-
duced by human vascular smooth muscle cells. American Journal of Physiology – Heart and Circulatory Physiology
2003; 284:H1778–H1784.
73. Zhang J, Schmidt J, Ryschich E, Schumacher H, Allenberg JR. Increased apoptosis and decreased density of medial
smooth muscle cells in human abdominal aortic aneurysms. Chinese Medical Journal 2003; 116:1549–1552.
74. Zulliger MA, Rachev A, Stergiopulos N. A constitutive formulation of arterial mechanics including vascular smooth
muscle tone. American Journal of Physiology – Heart and Circulatory Physiology 2004; 287:H1335–1343.
75. Baek S, Valentín A, Humphrey JD. Biochemomechanics of cerebral vasospasm and its resolution: II constitutive
relastions and model simulations. Annals of Biomedical Engineering 2007; 35:1498–1509.
76. Murtada S, Kroon M, Holzapfel GA. A calcium-driven mechanochemical model for prediction of force generation
in smooth muscle. Biomechanics and Modeling in Mechanobiology 2010; 9:749–762.
77. Ng CP, Schwartz MA. Mechanisms of interstitial flow-induced remodeling of fibroblast collagen cultures. Annals of
Biomedical Engineering 2006; 34:446–454.
78. Wagenseil JE. Cell orientation influences the biaxial mechanical properties of fibroblast populated collagen vessels.
Annals of Biomedical Engineering 2004; 32:720–731.
79. Neidlinger-Wilke C, Grood E, Claes L, Brand R. Fibroblast orientation to stretch begins within three hours. Journal
of Orthopaedic Research 2002; 20:953–956.
80. Lantz J, Renner J, Karlsson M. Wall shear stress in a subject specific human aorta influence of fluid-structure
interaction. International Journal of Applied Mechanics 2011; 03(04):759–778.
81. Reymond P, Crosetto P, Deparis S, Quarteroni A, Stergiopulos N. Physiological simulation of blood flow in the aorta:
Comparison of hemodynamic indices as predicted by 3-d fsi, 3-d rigid wall and 1-d models. Medical Engineering
and Physics 2013; 35:784–791.

SUPPORTING INFORMATION

Additional supporting information may be found in the online version of this article at the
publisher’s web site.

Copyright © 2014 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Biomed. Engng. 2014; 30:563–586
DOI: 10.1002/cnm

Você também pode gostar