A Thick-Walled

Fluid-Solid-Growth Model
Andrii Grytsan
Department of Solid Mechanics,
Royal Institute of Technology (KTH),
Teknikringen 8d,
Stockholm 10044, Sweden
Paul N. W atton
Department of Computer Science,
University of Sheffield,
Sheffield, UK;
INSIGNEO Institute of In Silico Medicine,
University of Sheffield,
Sheffield, UK
Gerhard A. Holzapfel 1
Institute of Biomechanics,
Graz University of Technology,
Kronesgasse 5-I,
Graz 8010, Austria
e-mail: holzapfel@tugraz.at
of Abdominal Aortic Aneurysm
Evolution: Application to a
Patient-Specific Geometry
We propose a novel thick-walled fluid-solid-growth (FSG) computational framework fo r
modeling vascular disease evolution. The arterial wall is modeled as a thick-walled non-
linearly elastic cylindrical tube consisting o f two layers corresponding to the media-
intima and adventitia, where each layer is treated as a fiber-reinforced material with the
fibers corresponding to the collagenous component. Blood is modeled as a Newtonian
fluid with constant density and viscosity; no slip and no-flux conditions are applied at the
arterial wall. Disease progression is simulated by growth and remodeling (G&R) o f the
load bearing constituents o f the wall. Adaptions o f the natural reference configurations
and mass densities o f constituents are driven by deviations o f mechanical stimuli from
homeostatic levels. We apply the novel framework to mode! abdominal aortic aneurysm
(AAA) evolution. Elastin degradation is initially prescribed to create a perturbation to
the geometry which results in a local decrease in wall shear stress (1VSS). Subsequent
degradation o f elastin is driven by low MASS and an aneurysm evolves as the elastin
degrades and the collagen adapts. The influence o f transmural G&R o f constituents on
the aneurysm development is analyzed. We observe that elastin and collagen strains
evolve to be transmurally heterogeneous and this may facilitate the development of
tortuosity. This multiphysics framework provides the basis fo r exploring the influence o f
transmural metabolic activity on the progression o f vascular disease.
[DOI: 10.1115/1.4029279]

Keywords: aneurysm, three-dimensional, elastin degradation, growth, remodeling,

fluid-solid-growth model

1 Introduction posite (inner) and the adventitia (outer). As the artery ages or
becomes diseased, the intima is thickened. Hence, modeling the
An AAA is a localized permanent dilation of the abdominal
intima may be relevant for simulating AAA evolution. However,
aorta. If an AAA ruptures, it is a surgical emergency with reported
to the best of our knowledge, there are no experimental data found
in-hospital mortality rates up to 49% [1]. Currently, no medication
in the literature that quantify the temporal progression of intimal
is able to prevent AAA growth or rupture. While intervention is
thickening. Hence, in the present work, we consider the media-
available, i.e., open surgery or endovascular aneurysm repair,
intima composite as one layer. The main load bearing structural
both are associated with 30-day mortality rates of up to 4% [2].
elements in the arterial wall are the isotropic noncollagenous
Consequently, once an AAA is detected, patients undergo regular
matrix material (i.e., ground substance, elastin, nonactived cells,
screening until it is deemed that the risk of aneurysm rupture
etc.) and the two families of collagen fibers. The collagen network
exceeds the risk of intervention. Unfortunately, statistically based
is continuously remodeled.
criteria such as maximum AAA diameter or AAA expansion rate
Structurally motivated hyperelastic material models are com­
are used as an indication for rupture. These fail to identify small
monly used to model the mechanical response of arterial walls
aneurysms with high risk of rupture and those larger aneurysms
[3]. These are suitable for developing models that simulate the
which are at lower risk. Models of AAA evolution may lead to an
changing structure and composition of the artery during vascular
improved understanding of the disease pathogenesis; this may be
disease evolution. For instance, Watton et al. [4] proposed the first
important for developing therapeutic interventions. Moreover,
mathematical model of AAA evolution. A realistic structural
from a clinical perspective, it is envisaged that they may have the
model for the arterial wall [5] was sophisticated to incorporate
potential to assist personalized management of the disease by
(temporal and spatially heterogeneous) variables which relate to
guiding frequency of follow-up monitoring prior to intervention.
the normalized (volumetric) mass density of the elastinous and
In a young healthy human artery the wall consists of two
collagenous constituents and the reference configurations in which
mechanically significant layers, namely, the media-intima com-
the collagenous constituents are recruited to load bearing. Schmid
et al. [6,7] extended this approach to a thick-walled finite element
(FE) model and Eriksson et al. [8] have recently sophisticated the
theoretical formulation to model volumetric adaption. Valentin
'Corresponding author.
Manuscript received March 20, 2014; final manuscript received December 1, et al. [9] proposed a novel 3D FE G&R framework using an
2014; published online January 29, 2015. Assoc. Editor: Jonathan Vande Geest. integral-based approach to simulate G&R [10-12],

Journal of Biomechanical Engineering Copyright ©2015 by ASME MARCH 2015 , Vol. 137 / 031008-1
 Local hemodynamic factors influence the risk of aneurysm
development and progression, in particular, aneurysm enlarge­
ment is believed to be linked to low WSS, see, e.g., [13]. Hence,
spatial distributions of WSS distribution could influence G&R of
arterial tissue and disease progression. Humphrey and Taylor [14]
emphasized 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 hemodynamics.” Several FSG models are
documented in the literature; they treat aneurysmal walls as a
membrane [15-19] and adopt an integral or rate based approach to
address the G&R of the arterial wall. However, the aorta is a
thick-walled structure and there is a potential for transmurally
nonuniform arterial G&R. This motivates the need for thick-
walled FSG frameworks; to our knowledge, there are none yet
available in the literature.
The focus of this study is to couple the thick-walled FE aneu­
rysm evolution model developed by Schmid et al. [6,7] to a FSG
framework [17,19] and to investigate the influence of transmurally
nonuniform elastin degradation on AAA evolution. In order to
illustrate the proposed methodology, one numerical example is
analyzed in detail by using a patient-specific geometry of an
abdominal aorta to obtain realistic flow patterns as input to G&R
algorithms.
2 Methods
In this section, we briefly describe the methodology for this
study. We start with the structural model of the arterial wall and
continue with some aspects of hemodynamic modeling. Next, the
models for elastin degradation and collagen G&R are described.
Finally, the FSG framework is overviewed.
2.1 Structural Model of the Arterial Wall. Here we
describe the nonlinear solid mechanics required to capture the
steady deformation of arterial walls. The artery is modeled as a
two-layered thick-walled cylindrical structure where the inner
layer is the mcdia-intima composite and the outer layer is the
adventitia. Each layer is composed of an isotropic noncollagenous
matrix material (i.e., ground substance, elastin, nonactive cells,
etc.), subsequently called elastinous matrix, and two families of
collagen fibers.
2.1 A Large Strain Kinematics. Since we introduce compress­
ible constitutive formulations for the tissues with a penalty term
that is used to accommodate the incompressibility condition
numerically, we base the kinematic formulation on a multipli­
cative split of the deformation gradient F = J 1^ F. Herein, ./I/,I
is associated with volume-changing deformations, where
./ = detF > 0 defines the volume ratio, and the modified deforma­
tion gradient F, with detF = 1 , is associated with volume­
preserving deformations of the material. The right Cauchy-Green
tensor is C = F 1 F, while the modified right Cauchy-Green tensor
T-
is given by C = F F with the three invariants of C according to
!\ = trC, /i = trC and E = detC = 1. The modified
Green-Lagrange (GL) strain tensor is denoted as E = (C - I) /2 .
A cylindrical coordinate system is adopted to describe the arte­
rial geometry with axial, azimuthal, and radial coordinates z, 0,
and r, respectively. The collagen fibers in both arterial layers are
assumed to be oriented in a double helical pitch around the lumen,
forming angles <p‘ with the azimuthal coordinate 0, where i'= 1, 2
denotes the fiber family within one layer. The fiber angles in the
inner and outer layers can be different. Let [M‘] = [sin </>',
cos (p‘. 0] be a unit vector representing the direction of the fiber
family i in the reference configuration, and 7.^(>0) be the stretch
of the elastin in the direction M'. By defining the modified elastin
stretch 7.J, = ./_l^7.!e we may calculate the square of /” as
( K ) 2= iVI' ■cm' = r 4( c , m;) ( 1)
031008-2 / Vol. 137, MARCH 2015
where }\ is a pseudo-invariant of € and M'. For subsequent use
we also introduce the modified GL strain E[ = [ (!') - 1]/2 of the
elastin.
The natural reference configuration of a collagen fiber might
not be the same as the unloaded configuration of the artery due to
the fact that the collagen fibers are wavy when the artery is
unloaded. We assume that the collagen fibers have no compres­
sive stiffness, and hence are unloaded until the modified elastin
stretch T.j, reaches a certain value, say when the fiber is
recruited to bear load. Thus, the modified collagen fiber stretch,
say 7.J., is modeled by a piecewise linear function of 7.^, and the
modified recruitment stretch 7.[ec as
<-(£■ ^ y 1, else
2./.2 Equilibrium and Constitutive Equations. The equilib­
rium equation without any body forces reads
Div(FS) = 0 (3)
where Div( ) denotes the material divergence of (•), and S is the
second Piola-Kirchhoff stress tensor. The solution of the equilib­
rium Eq. (3) requires the computation of S. For a hyperelastic
material S may be derived from a strain-energy function (SEF),
= <f,voi + lf,isoi as S = 2<CV/dC, where 4/jso is the isochoric
strain energy for the tissue per unit reference volume and 4 'vo! (J)
is a given scalar-valued function o f ./ describing the volumetric
(dilatational) elastic response of the tissue. For numerical pur­
poses, we use here
Tvoi(T) = k G, G = ~(J2 - l ~ 2 \ n j ) (4)
where k is a (positive) penalty parameter and Q serves only as a
penalty function introduced to accommodate incompressibility.
The penalty method for incompressibility is the basis for the nu­
merical approach; see also Sec. 8.3 in Ref. [5],
The mechanical response of arterial tissue is modeled with con­
tributions from the (isotropic) elastinous matrix and the two fami­
lies of collagen fibers with orientations tp' to the azimuthal axis. A
neo-Hookean law with the SEF VFCis used to describe the mechan­
ical response of the elastinous matrix [20] according to
= *e(/| - 3) (5)
where kc is the related material parameter. The mechanical
response of each collagen fiber family is modeled with a constitu­
tive equation following [4,21]
{^:{exp<('1/(4 )2- )2 c i (6 )
U else
where ], k'C2 are the material parameters of the i-th collagen fiber
family. Hence, the total isochoric SEF *Fiso of each layer is
4'is„ = /e 'E e + £ £ 4 ' ' (7)
i=l.2
w h ere/c G (0, l ] , / c' e (0 ,oo) are the normalized (volumetric)
mass densities of the constituents [4,6,22], Initially, the normal­
ized densities are equal to 1, i.e.,/c = 1 and f t = 1, however, they
can increase/decrease to simulate growth/atrophy of individual
constituents.
T ransactions of th e A SM E
 24 cm3s 1
Fig. 1 Patient-specific attachments to the artery model and
prescribed boundary conditions: (a) upstream section, inlet
with flow rate boundary condition shown aside, outlets with
corresponding pressure boundary conditions; (b) downstream
section, outlets with corresponding pressure boundary
conditions.
2.2 Hemodynamic Modeling. We use the computational
fluid dynamics (CFD) approach as detailed in Ref. [19]. Briefly,
during the FSG simulation, the flow domain is assembled and
meshed, boundary conditions applied and the hemodynamic simu­
lation is executed in an automated manner; this is achieved using
(standard) ansys icem and cfx scripts and a perl wrapper. The
boundary of the flow domain is comprised of the inner surface of
the structural model together with upstream/downstream sections
to represent the geometry of the healthy abdominal aorta
(obtained from computer tomography scans, refer to Ref. [19] for
a detailed description).
ansys icem generates an unstructured tetrahedral mesh with
prism layers lining the boundary of the fluid domain. This is the
starting point for the creation of an unstructured tetrahedral mesh
throughout the remaining volume domain. After meshing, the
boundary conditions are generated. Blood is modeled as a
Newtonian fluid with constant density p = 1069 kg m-3 and con­
stant viscosity i/= 0.0035 Pa s. No slip and no-flux conditions are
applied at the arterial wall. The flow rate at the inlet and the pres­
sure boundary conditions at the outlets are prescribed, see Fig. 1.
The governing equations for the fluid are the incompressible
Navier-Stokes equations, which are solved by ansys cfx using a
finite volume formulation [23,24].
2.3 Metabolic Activity in the Arterial Wall. This section is
concerned with the modeling of the metabolic activity in the arte­
rial wall. We start by describing the functional forms of elastin
degradation and continue to provide the link to WSSs. Finally, the
methodology for collagen G&R is detailed. This follows [17],
however given that we are using a thick-walled model of the
artery within a FSG framework, we sophisticate the approach to
account for transmural degradation of elastin.
Biochemical evidence suggests that only a small fraction of
elastin is left in aneurysmal tissue when compared to normal arte­
rial tissue [25], Thus we assume that aneurysm evolution is
(partly) driven by the degradation of the elastinous matrix. Ini­
tially, we prescribe a loss of elastin to create a perturbation to the
geometry of the domain. This alters the WSS distribution and sub­
sequent elastin degradation is driven by deviations of WSS from
(illustrative) homeostatic levels.
2.3.1 Prescribed Elastin Degradation. The prescribed elastin
degradation function does not depend on the WSS distribution.
Here we consider a multiplicative decomposition of the normal­
ized elastin density/c according to
/e(x,f) =/e(x,r = 0) - f :(:)fi(0)f,.(r,t) (8)
Journal of Biomechanical Engineering
where x indicates space and t time. The degradation functions
describing the variations in the axial, the circumferential and the
radial directions, and the time are given by /)(z), /«(0), and/,.(/•, r),
respectively. In this study the prescribed elastin degradation is
considered to be independent of the circumferential coordinate 0,
i.e.,/,)(()) s 1. The functional form for/l(z) is according to [4]
/:(z) = exp = exp^-nt|(2z - l ) 2j (9)
where L is the length of the considered artery segment, z = z/L,
with z e [0, 1], is the normalized axial coordinate and mt controls
the “sharpness” of the elastin degradation function.
The thick-walled model allows the exploration of the trans­
mural variations in elastin degradation. However, there is no
experimental evidence of how elastin degradation propagates
through the wall. Hence, we propose a functional form for trans­
mural elastin degradation/,.(/•, t), which can be replaced later with
an experimentally motivated one. For illustration, we consider a
simple functional form to model the transmural profile of elastin,
i.e., we assume a linear dependence in the radial direction. The
elastin degradation is prescribed as a function of time at the inner
and the outer boundaries of each layer, denoted as c[ (t) and c:,n(t)
(y denotes the layer; M for media-intima composite and A for
adventitia), respectively. The prescribed functions have to be
compatible at the interface between the layers so that cJJ1 = cA.
The elastin content is considered to be low in the adventitia and it
is the collagen that plays the main load bearing role in this layer.
Taking this into account, we assume that the elastin degrades with
a constant rate in the adventitia so that i f = i f .
Next, we assume that the degradation starts at the lumen and
there is a time lag d before the degradation reaches the
media-adventitia interface. Thus, c f = c(t) and cj1 = c(t - d). It
is important to note that (/ —d) may become negative when t < d.
Hence, the functional form of c(t) is chosen as an exponential
decay that filters negative arguments, i.e.
c(t) = | c™i" ’ f - ° (10)
[ 1, else
Following Eq. (10), c,nin of elastin is left after the degradation
time T. The elastin degradation functionf.(r, t) describing the var­
iation in the radial direction r, and time t is interpolated linearly
throughout each layer. For the media-intima composite and for
the adventitia, we have
f M(h, l) = {[1 - < « ] 0 - / < ) + [ ! - c(t - i?)]/7}
f A( h , t ) = c ( t - d )
where li is the normalized thickness of each layer
/; = (;• - Ri)/(/?[),r £ [/?•',/?£] with R'-.R‘„ being the inner and
outer radii of the layer y, respectively. It can be seen, that with
d = 0 Eq. (11) reduces to the case of transmurally uniform degra­
dation, i.e.,/M(/t,r) = / A(/i,r) = c(r).
2.3.2 Elastin Degradation Linked to VF5S. We suppose that
the rate of elastin degradation d fjd t is a function of space x, time
t and WSS distribution r; here, the mean WSS z during the cardiac
cycle is used. In order to couple elastin degradation with the mag­
nitude of WSS, we follow Ref. [17] and choose the functional
form for the elastin degradation rate as
^ = - T D(z (x,t))£W e(x,f) (12)
where Dmax is the maximum degradation rate (how much of exist­
ing elastin can be degraded per year) and T o € [0; 1] is a spatially
MARCH 2015, Vol. 137 / 031008-3
dependent normalized function of the WSS to be linked to elastin
degradation. The elastin degradation rate reaches a maximum at
To = 1 while there is no elastin degradation at To = 0.
We assume that there is no elastin degradation if the WSS value
is greater than a critical value, say Tcrj,. Subsequently, elastin
degrades for lower values of the WSS. In addition, we suppose
that if a WSS value is lower than a certain value, say ix:
0<Tx<Tem. the maximum elastin degradation occurs. Finally,
we assume a simple quadratic form for the normalized elastin deg­
radation T q that describes the relation between the local WSS
and the degree of elastin degradation [17], i.e.,
Fig. 2 Cross section of the initial configuration of an artery:
T > Tcrit complete cross section (left), half (right)

•^D(t(x ,r)) = < TX < T < Tcri, (13)

and OE. Only half of the cylinder is used in this study, as shown
1, T < TX on the right side of Fig. 2.
First, the cylinder is closed by pulling CD (see Fig. 2) so that
2.3.3 Collagen Growth and Remodeling. We now concisely it is located on the same line as AB, while AB cannot move in the
overview the approach for simulating G&R of collagen; for fur­ y-direction. Point A is displaced in the .v-direction such that
ther details on the motivation of this approach, see Refs. [4,17] |OC| = |OA| = Ri, where /?, is the inner radius of the unloaded
and [22], The collagen fabric of arterial walls is continuously artery. In addition, the displacement of the bisector of angle AOC
remodeling via a process of fiber deposition and degradation. The is prescribed such that it is located in the plane .v= 0 after the cyl­
standard assumption adopted in all constrained mixture models of inder closure, and the inner radius matches R,. The displacement
arterial collagen remodeling is that new collagen fibers are config­ in the z-direction is prohibited during this procedure. The closure
ured to the matrix in a state of stretch: we refer to this as the of the cylinder allows recreating circumferential residual stresses.
attachment stretch Aa [4]; note terminology includes deposition Second, an axial prestretch and an internal pressure are applied to
stretch or collagen prestretch. The related modified (GL) attach­ the arterial tube. The displacement boundary conditions are
ment strain is then computed as Eid = [(A !,)2 - l]/2. The conse­ applied at the end cross sections: axial displacement is equal to
quence of fiber deposition/degradation and new fibers being zero at one end, and equal to the value of prestretch at the other
configured in a state of stretch is that the collagen fabric continu­ end. The pressure boundary condition is applied at the inner sur­
ously remodels to maintain a homeostatic level of strain, i.e., the face of the artery. Third, the artery is remodeled following
attachment strain. This process is simulated by a differential Eq. (14) until homeostasis is reached. The displacement boundary
equation that evolves the recruitment stretch field throughout the condition is applied at the ends of the tube: axial displacement is
artery, i.e., set to zero.
Next, the displacement boundary conditions are applied: all
^ -re c gc~g, three normal displacements at the edges of the inner surface of the
(14)
dt EL
Table 1 Parameters used to generate a prescribed initial bulge
so that the modified (GL) collagen strain E[ = [(Aj .)2 - l]/2 in the arterial wall, and used for the FSG model
remodels toward the attachment strain £!,; a is a rate parameter
related to the collagen fiber half-life. P aram eter P rescribed FSG
A simple strain rate-based evolution law for the normalized
collagen density is used according to T im e step for solid 0.02 years 0.02 years
T im e step for fluid — 0.2 years
R ate pa ra m e ter i 0.6 y e a rs- 1 0.6 y e a rs- 1
R ate pa ra m e ter /f 2.0 y e a rs - 1
(15) 2.0 y e a rs -1
dt ,/c el D egradation p aram eter c mm 0.5 __
D egradation tim e 7' 4 years __

where /( is a growth-rate parameter. T im e lag d 0 and 2 years __

C ritical W SS r tri, — 0.5 Pa
M axim um deg rad atio n rate D mlu — 0.75
2.4 The FSG Framework. This section briefly describes the
coupled framework. The structural model of the arterial wall is
loosely coupled to the CFD analysis, which informs the G&R
framework about changes of the hemodynamic conditions. This Table 2 Geometrical parameters for the stress-free configura­
approach allows to account for very different time scales associ­ tion of the arterial wall model and the physiological loads; free
ated with G&R (weeks to years) and the cardiac cycle (about 1s). parameters are marked by an asterisk

P aram eter S tress-free H om eostasis

2.4.1 Solid Model: Initial Setup. The structural model of an
artery, with given inner radius and length in the physiological Inner radius 13.4 m m * 11.1 m m
state, is required. The length of the artery in a stress-free configu­ A rtery length 73.7 m m 88.4 m m
ration can be computed using a (known) value of axial prestretch. M edia-intim a thickness 1.2 m m 0.8 m m *
We use an initial guess of the inner radius of the stress-free artery A d v entitia thickness 0.6 mm 0.4 m m *
and then iterate to reach a given inner radius in the physiological F ib er orien tatio n a ngle </>'
M edia-intim a ± 3 0 deg
state, which matches the (known) radii of up and downstream
A dventitia ± 6 0 dee __
extensions. The model of a healthy artery is developed following O pen in g angle 0 0 __
120 d e c
Ref. [7], The initial stress-free configuration is an opened-up A xial prestretch — 1.2
thick-walled cylinder with the cross section as shown on the left S ystolic pressure — 16 kPa
side of Fig. 2. The opening angle <I>(>is defined by the rays OC

031008-4 / Vol. 137, MARCH 2015 Transactions of the ASME

 Table 3 Material parameters for the arterial wall model Eqs. (14) and (15). Table 1 lists the parameters that we adopted
from Refs. [17] and [19] to generate a prescribed initial bulge in
Parameter Media-intima Adventitia the arterial wall. In the last preparatory step, we temporarily
increase the growth-rate parameter /} tenfold and halt the pre­
Elastin parameter ke l43.2kPa l4.32kPa scribed elastin degradation. Within a year of such a remodeling
Col lagen parameter G, 3.84 kPa 0.96 kPa
procedure, the modified GL strain E'c of the collagen is close to
Collagen parameter kC2 40 40
the (homeostatic) attachment strain. This approach is to ensure
that a subsequent aneurysm enlargement during the FSG coupling
is not attributed to the collagen fabric not initially being in home­
ostasis. This configuration is the starting point for coupling the
aneurysm evolution to blood flow.

2.4.2 Coupling Between Solid, Fluid, and Growth. The

coupling between the solid, the fluid, and the growth is performed
in a repetitive loop. First, mechanical-equilibrium for the solid
model is obtained, which determines the geometry of the arterial
lumen and quantifies the strained state of the constituents/cells of
the arterial wall, which drives the collagen G&R. Then, the
patient-specific up and downstream extensions are attached to the
geometry of the lumen. These collective parts together define the
fluid domain. The fluid analysis with the arterial walls being
I treated as rigid is then carried out in order to quantify the mechan­
ical environment of the endothelial cells (i.e., obtain the WSS dis­
tribution), which is then used to update the elastin degradation
rate according to Eq. (12). Subsequently, the elastin is degraded
for a period of time, and the collagen adapts. Finally, the loop
/=0 yrs =5 yrs /=!() yrs
starts over again, using the updated arterial geometry to quantify
the mechanical environment of the cells and to update the elastin
Fig. 3 Evolution of the WSS distribution r during FSG degradation rates.

artery are set to zero. A small bulge is developed in order to create 3 A Patient-Specific Example
a localized area of low WSS. This is done by prescribing the elas­
tin degradation following Eq. (8) in combination with Eqs. (9) and 3.1 Geometry and Modeling Parameters. The cylindrical
(11), and letting collagen grow and remodel according to model of the healthy arterial wall is integrated into a

(a) Normalized elastin degradation rate

I
0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

r= 0 yrs t= 8 yrs t = l() yrs

I 0

Fig. 4 Evolution of the normalized elastin degradation T o (a), and the normalized elastin density fe (b) in the aneurysm
region

Journal of Biomechanical Engineering MARCH 2015, Vol. 137 / 031008-5

 configuration are listed in Table 2. The material parameters for the
artery model are summarized in Table 3; we have assumed the
same mechanical properties for the two collagen fiber families. The
parameters used for FSG model are given in Table 1.

3.2 Numerical Tools and Finite Element Mesh. The

balance Eq. (3) is solved using the Galerkin finite element method
with the help of the fortran based FE-solver CMISS [26]. The
displacement field is interpolated linearly and the pressure field is
approximated using constant values by enforcing the incompressi­
bility condition element-wise. The thick-walled model of half of
the aorta is meshed with 12 elements in the circumferential direc­
tion, 60 elements in the axial direction, and 12 elements through
the wall thickness (eight elements in the media-intima composite
and four in the adventitia).

Fig. 5 Solid model mesh before (left) and after FSG (right) in a cy­
lindrical coordinate system. Locations Au A2, A3, Aa are shown, at
which the transmural distribution of the quantities are analyzed.
patient-specific geometry of the abdominal aorta. The methodol­
ogy to achieve this is detailed in Ref. [19],
We take advantage of nearly symmetric (left-right) WSS distri­
bution in the aneuiysm region and model only half of the solid
domain. This allows to capture the major asymmetry in the patient-
specific WSS distribution, saves computational time, and keeps the
results relatively simple for the analysis. The geometrical parame­
ters of the artery in homeostasis, the physiological loads and the
known and fitted geometrical parameters of the stress-free
(a) A |: posterior wall,
middle of aneurysm region
4 Numerical Results
Here, we present the numerical results of the FSG simulation.
Figure 3 shows the aneurysm shape evolution during the FSG
and the evolution of the WSS distribution x. The plots exhibit a
proximal-distal asymmetry: the aneurysm grows toward the
upstream direction. In addition, there is a posterior-anterior asym­
metry: the elastin degrades more in the posterior aneurysm region.
No left-right asymmetry evolves because the solid model consid­
ers only the right half of the aneurysm region.
Figure 4 shows the evolution of two FSG parameters in the
aneurysm domain from 0 to 10 years in steps of 2 years. The dis­
tribution of the normalized elastin degradation To is computed on
the basis of a WSS distribution and is shown in Fig. 4(a). Accord­
ing to the model, elastin should degrade faster in the regions with
high To value, while no degradation happens where To = 0.
(b) At: anterior wall,
middle of aneurysm region

Legend:

---- t = 2.5
1= 5
• ••• / = 10

(c) A3 : posterior wall, (d) A4 : posterior wall,

distal aneuiysm neck proximal aneurysm neck
Imima-Media Adventitia
----fl = 0
----d = 2

--

-------- 1------- I--------- ---- 1-------

0 0.2 0.4 0.6 0.8 1.0
Normalized thickness h Normalized thickness h
Fig. 6 Evolution of the transmural profiles of the normalized elastin density fe over
time t at (a) the posterior and (b) the anterior aneurysm wall at z = U2\ (c) the posterior
distal and (d) the proximal aneurysm neck. Plots are shown for a time lag i9 = 0 and 2.

031008-6 / Vol. 137, MARCH 2015 Transactions of the ASME

This behavior can be seen in Fig. 4(h): elastin loss compared to
previous time snapshot is more dramatic in the regions where J-q
is high and there is no degradation if To is low or equal to zero.
The thick-walled model allows the investigation of the trans­
mural variations in the aneurysm G&R. For further analysis we
choose four locations on the inner wall, see Fig. 5. The points
and A2 are located at the middle of the aneurysm region at the
reference state, on the posterior and anterior wall, respectively.
Later, the aneurysm grows upstream and pushes the lower part
downstream. The points A2 and A4 are located at the posterior dis­
tal and the proximal aneurysm neck, respectively.
Figure 6 shows the evolution of the transmural profile of the
normalized volumetric mass density f c of elastin at the four loca­
tions /41,__ A4, see Fig. 5. Elastin is degrading continuously at A|
and A2 due to the low levels of WSS (see Figs. 6(a) and 6(h)).
One can observe a considerable level of asymmetry. Elastin deg­
radation is slow at the distal aneurysm neck A3, see Fig. 6(c), due
to the high levels of WSS. Initially, intact elastin at the proximal
aneurysm neck A4 degrades fast (see Fig. 6(d)), because the aneu­
rysm and the region of low WSS levels propagate upstream.
Figure 7 shows the evolution of the transmural profile of the
circumferential modified GL strain Ee of the elastinous matrix at
the four different locations. Due to the loss of elastin during FSG
the circumferential modified GL strain is gradually and asymmet­
rically increasing at the middle of the aneurysm region, see
Figs. 1(a) and 1(b). The circumferential strain does not change
significantly at the distal aneurysm neck, see Fig. 7(c), while it
increases significantly at the proximal aneurysm neck due to the
aneurysm propagation in the upstream direction, see Fig. 1(d).
(a) A): posterior wall,
middle of aneurysm region
(c) A3: posterior wall,
distal aneurysm neck
Intima-Media Adventitia
Normalized thickness h
Fig. 7 Evolution of _the transmural profiles of the circumferential modified
Green-Lagrange strain Ee of elastin over time t at (a) the posterior (b) the anterior
aneurysm wall at z = U2\ (c) the posterior distal and (d) the proximal aneurysm neck.
Plots are shown for a time lag il = 0 and 2.
Journal of Biomechanical Engineering
The circumferential modified GL strain distribution becomes
more elevated toward the lumen during FSG.
The solid model is set up such that at t = 0 years the modified
GL strain £c of the collagen is close to uniform, see Sec. 2.4.1.
The evolution of the transmural distribution of this quantity is
shown in Fig. 8. The magnitude of the collagen strain remains
lower in the adventitia due to more axially aligned fibers, which
are then less influenced by the increase in the circumferential
strain than the collagen fibers in the media-intima. The magnitude
of collagen strain remains almost uniform in each layer through
FSG. However, rapid changes in the mechanical environment
cause the elevation of the modified GL collagen strain toward the
lumen at t = 2.5 years, see Figs. 8(r/)-8(c). The effect of collagen
G&R can be observed in the Figs. 8(n) and 8(b). Until ~2.5 years,
the elastin loss is more rapid than the collagen G&R, which results
in an elevation of the modified GL collagen strain following the
increase in the circumferential strain. Then, the elastin loss slows
down, see Figs. 6(a) and 6(b) due to the higher levels of WSS,
and collagen G&R lowers the collagen strain after r = 5 years.
The collagen strain value is close to the homeostatic value at
t= 10 years, see Figs. 8(a) and 8(b).
Finally. Fig. 9 shows the evolution of the distribution of the
normalized collagen density f c through the arterial wall thickness
at the four different locations. The distributions off c at both poste­
rior and anterior walls in the middle of the aneurysm region are
quantitatively similar to each other with some level of asymmetry,
see Figs. 9(a) and 9(b). The collagen amount at the proximal neck
increases more than twofold from t = 5 to t = 10 years due to the
aneurysm propagation upstream, see Fig. 9(d). The normalized
(b) A2: anterior wall,
middle of aneurysm region
Legend:
----- I = 2.5
1= 5
• ••• / = 10
(d) A4: posterior wall,
proximal aneurysm neck
Normalized thickness h
MARCH 2015, Vol. 137 / 031008-7
 (a) A i: posterior wall, (b) A2: anterior wall,
middle o f aneurysm region middle of aneurysm region

Legend:

--- t = 2.5
t =5
----- t = 10

(c) A3: posterior wall, (d) A4: posterior wall,

distal aneurysm neck proximal aneurysm neck

o. Adventitia Intima-Media Adventitia

T 0.088
o. -- -- fl = 0 0.086 — fl = 0
N°o.
<£>

0.084 — d= 2
II
I

03 0.074 0.082
u.
t/5 0..072 rwmtr 0.080
**•*«kV4 kV< W A V | | W w i , w #w

O ' .070 ;■
0.078 ■■■■■SSijg-g. n

g ° .068 0.076
.066 0.074
oo..064 0.072
U
0. 0.070
----- \----- 1
----- J— --- t
----- 1 0.068
0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 i.O
Normalized thickness Ti Normalized thickness Ti

Fig. 8 Evolution of the transmural profiles of the modified Green-Lagange strain E c of

collagen over time fa t (a) the posterior (to) the anterior aneurysm wall at z = U2\ (c) the dis­
tal and (d) the proximal aneurysm neck at the proximal aneurysm wall. Plots are shown for
a time lag ti = 0 and 2.

collagen density f c decreases at the distal aneurysm neck due to
the retraction of the aortic tube end. see Fig. 9(c ).
5 D is c u s s io n
We present a thick-walled FSG framework and illustrate its
application for modeling AAA evolution. The improvement over
existing FSG models is that it incorporates a thick-walled model
of the artery as opposed to a membrane approach. This represents
a more realistic approach to model the abdominal aorta, and thus
AAA evolution. Interestingly, we observe that the magnitude of
the circumferential GL strain is elevated toward the lumen
throughout the aneurysm enlargement. This observation may be
relevant for modeling elastin damage/fragmentation due to
mechanical damage (over-stretching) and thus has implications
for modeling aneurysm evolution, which a membrane model
would be unable to capture.
The limitations of the numerical model concern the geometry,
the boundary conditions, reconstruction of the residual stresses
and the finite element mesh. A conceptual model of the healthy
abdominal aorta is utilized. The healthy aorta is modeled as a
straight cylinder without tapering to the patient-specific up and
downstream extensions. Asymmetry of the aneurysm may develop
due to the contact with the spine; however, this was omitted for
simplicity in the present study. The opening angle approach was
used to reconstruct the circumferential residual stresses in the
arteries. However, a 3D residual stress state is observed experi­
mentally [27]. In addition, the opening angle approach is applica­
ble only to cylindrical geometries and not to (the more realistic)
3D geometries obtained from medical images.
There are substantial mass and volume changes of tissue con­
stituents during aneurysm evolution. The mass changes are
explicitly modeled in the present framework via adapting constitu­
ent densities to simulate collagen growth and elastin degradation;
however, volumetric changes are not modeled. Lack of considera­
tion of the volumetric growth may lead to the computation of
unrealistic wall stresses as the aneurysm evolves (unless a remod­
eled thickness is postprocessed). However, the present thick-
walled arterial model allows the implementation of a volumetric
growth model, which is not possible with available membrane
models. The recent studies [8,9] have proposed how to model
volumetric adaption using integral- and rate-based G&R
approaches, respectively. Predicting volumetric growth will ena­
ble the computation of more realistic stresses and the investigation
of the stress-driven evolution laws versus the strain-driven ones.
The active response of smooth muscle cells (SMCs) contributes
significantly to the mechanical response of a healthy arterial wall,
while the adaption of the basal tone of SMC and the SMC apopto­
sis can influence the aneurysm evolution. The active response of
SMC is omitted in the present model for sake of simplicity. The
collagen fibers are produced by SMC in the media-intima compos­
ite and by fibroblasts in the adventitia. Hence, the SMC apoptosis
should decrease the collagen production in the media-intima com­
posite. Therefore, different growth laws might be considered for
collagen fibers in the media-intima composite and the adventitia.
The intraluminal thrombus and the calcifications are not consid­
ered in the present framework. The blood flow is considered to be
steady, while it is pulsatile in vivo. In addition, the arterial wall is
assumed to be rigid for CFD, and thus, the framework does not
account for fluid-structure interaction. These simplifications may
influence the spatial distribution and magnitudes of the WSS.
However, these assumptions are reasonable given they do not
affect the qualitative behavior of the model; see Ref. [19] for fur­
ther discussion. We modeled only half of the arterial domain and

0 3 1 0 0 8 -8 / Vol. 137, M AR C H 2015 T r a n s a c t io n s o f t h e A S M E

 (a) Ai: posterior wall, (b) A2 : anterior wall,
middle of aneurysm region middle of aneurysm region
Intima-Media Adventitia Intima-Media Adventitia Legend:

— fl= o p. d=0 - - - r = 2.5

c ----0 = 2
o — *=2 ... ,=5
I0 - . . . . / = 10
.......
8-
.......
6-
tr n iT .-.i

o
U
----------------I--------h- — I------- 1 0 -— ----- ---------1
0 0.2 0.4 0.6 0.8 1.0

(c) A3 : posterior wall, (d) A4 : posterior wall,

distal aneurysm neck proximal aneurysm neck
In tim a - M e d ia A d v e n titia I n tim a - M e d ia A d v e n titia
1.4 <£> <£> 6

0 (N
11
II II
•4
1.2 - 5 -
C — 9 = 2
O
d 1.0 -
4 -
d
0.8 -
0
3
0 0 .6 -
. . . . . . . .
s 2 - ...........................
CfJ 0 .4 -

1 -
U
0 ------------------------ 1----------- ------- 1------------1 0 ------------1------------1--------------- ------- 1------------1
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0

Normalized thickness Ti Normalized thickness h

Fig. 9 Evolution of the transmural profiles of the normalized collagen density fc over
time fat (a) the posterior and (b) the anterior aneurysm wall at z = U2\ (c) the posterior
distal and (d) the proximal aneurysm neck. Plots are shown for a time lag t? = 0 and 2.

assumed left-right symmetry. This was to increase the computa­

tional efficiency of the structural solver and is reasonable given that
the focus of the work is to extend the previous FSG framework
[17] to a thick-walled approach. However, we point out that the
WSS distribution is not symmetric, aneurysms may evolve to be
asymmetric and thus a more general approach should be adopted
for more realistic simulations. We assumed that the elastin degrada­
Ec
tion is driven by deviations from the magnitude of the WSS from a
I 0.100 constant homeostatic value. However, the endothelium is heteroge­
0.095
- 0.090
neous. A recent sophistication to the modeling approach is to sup­
0.085 pose that homeostatic values of mechanical stimuli are spatially
0.080 and temporally heterogenous; this influences the qualitative behav­
;---- 0.075 ior of the model, see, e.g., Aparicio et al. [19].
0.070 Increased tortuosity of the abdominal aorta correlates with the
0.065
0.060
vessel disease or aging [28], Moreover, it is suggested that AAA
tortuosity may be quantified and used to predict aneurysm rupture
along with the maximum AAA diameter [29]. Hence a model that
has the capabilities of predicting the evolution of AAA tortuosity
may be of clinical value. In the present simulation it can be seen
that the distal neck of the aneurysm has increased tortuosity after
several years of FSG. In fact, the model predicts that the distal
neck can become severely tortuous after a longer period of aneu­
rysm evolution, see Fig. 10. While this may be clinically unrealis­
tic due to the prescribed boundary conditions, i.e., the artery
model is fixed at the ends, it does illustrate the capabilities of this
growth model for simulating the evolution of tortuosity.

6 Conclusion
Fig. 10 Distribution of the modified GL strain Ec of collagen To the authors’ knowledge, this is the first thick-walled FSG
after 11.5 years of FSG using /i = 1.4 framework for modeling the AAA evolution. Although the

Journal of Biomechanical Engineering MARCH 2015 , Vol. 137 / 031008-9

framework has a number of limitations, it is able to predict the
typical AAA features observed in clinics such as aneurysm propa­
gation in the upstream direction and the asymmetry in aneurysm
evolution. Importance of the thick-walled approach is underlined
by observation of transmurally nonuniform elastin and collagen
strains. In addition, the thick-walled model can develop tortuous
shapes which are not possible with a membrane model. Further
development of the framework will help to better understand the
pathology of abdominal aneurysmal disease on a patient-specific
basis.
Acknowledgment
Paul Watton acknowledges the Wellcome Trust/EPSRC, Centre
of Excellence in Personalised Healthcare (Grant No. WT 088877/
Z/09/Z), for providing support.
References
[1] Chen, C.-K., Chang, H.-T., Chen, Y.-C., Chen, T.-J., Chen, I.-M., and Shih,
C.-C., 2013, “Surgeon Elective Abdominal Aortic Aneurysm Repair Volume
and Outcomes of Ruptured Abdominal Aortic Aneurysm Repair: A 12-Year
Nationwide Study," World J. Surg., 37(10), pp. 2360-2371.
[2] Greenhalgh, R. M., Brown, L. C., Powell, J. T., Thompson, S. G., Epstein, D.,
Sculpher, M. J., and United Kingdom EVAR Trial Investigators, 2010,
“Endovascular Versus Open Repair of Abdominal Aortic Aneurysm,” N. Engl.
J. Med., 362(20), pp. 1863-1871.
[3] Holzapfel, G. A., and Ogden, R. W.. 2010, “Constitutive Modelling of
Arteries,” Proc. R. Soc. London Ser. A, 466(2118). pp. 1551-1597.
[4] Watton, P. N., Hill, N. A., and Heil, M., 2004, “A Mathematical Model for the
Growth of the Abdominal Aortic Aneurysm." Biomech. Model. Mechanobiol.,
3(2), pp. 98-113.
[5] Holzapfel, G. A.. 2000, Nonlinear Solid Mechanics. A Continuum Approach for
Engineering, Wiley, Chichester, UK.
[6] Schmid, H.. Watton, P. N., Maurer, M. M., Wimmer. J., Winkler, P., Wang,
Y. K.. Rdhrle, O.. and Itskov, M., 2010, “Impact of Transmural Heterogeneities
on Arterial Adaptation: Application to Aneurysm Formation,” Biomech. Model.
Mechanobiol., 9(3), pp. 295-315.
[7] Schmid. H.. Grytsan, A., Poshtan, E., Watton, P. N., and Itskov, M., 2013,
“Influence of Differing Material Properties in Media and Adventitia on Arterial
Adaptation—Application to Aneurysm Formation and Rupture,” Comput.
Methods Biomech. Biomed. Eng., 16( 1), pp. 33-53.
[8] Eriksson, T., Watton, P. N., Luo, X. Y., and Ventikos, Y., 2014, “Modelling
Volumetric Growth in a Thick-Walled Fibre Composite,” J. Mech. Phys. Solids,
73, pp. 134-150.
19] Valentin, A.. Humphrey, J. D., and Holzapfel, G. A., 2013, “A Finite Element
Based Constrained Mixture Implementation for Arterial Growth, Remodeling,
and Adaptation: Theory and Numerical Verification,” Int. J. Numer. Method
Biomed. Eng., 29(8), pp. 822-849.
[101 Humphrey, J. D., and Rajagopal, K. R„ 2002, “A Constrained Mixture Model
for Growth and Remodeling of Soft Tissues,” Math. Models Methods Appl.
Sci., 12(3), pp. 407-430.
| l l ] Humphrey. J. D.. and Holzapfel, G. A.. 2012, “Mechanics, Mechanobiology,
and Modeling of Human Abdominal Aorta and Aneurysms,” J. Biomech.,
45(5), pp. 805-814.
031008-10 / Vol. 137, MARCH 2015
[12] Valentin, A., and Holzapfel, G. A., 2012, “Constrained Mixture Models as
Tools for Testing Competing Hypotheses in Arterial Biomechanics: A Brief
Survey,” Mech. Res. Commun., 42, pp. 126-133.
[13] Dua, M. M., and Dalman, R. L., 2010, "Hemodynamic Influences on
Abdominal Aortic Aneurysm Disease: Application of Biomechanics to Aneu­
rysm Pathophysiology,” Vase. Pharmacol., 53(1-2), pp. 11-21.
[14] Humphrey, J. D., and Taylor, C. A., 2008, “Intracranial and Abdominal Aortic
Aneurysms: Similarities, Differences, and Need for a New Class of Computa­
tional Models,” Annu. Rev. Biomed. Eng., 10(1), pp. 221-246.
[ 15] Figueroa, C. A., Baek, S., Taylor, C. A., and Humphrey, J. D., 2009, "A Com­
putational Framework for Fluid-Solid-Growth Modeling in Cardiovascular
Simulations,” Comput. Methods Appl. Mech. Eng., 198(45-46), pp
3583-3602.
[ 16] Sheidaei, A., Hunley, S. C., Zeinali-Davarani, S., Raguin, L. G., and Baek, S.,
2011, “Simulation of Abdominal Aortic Aneurysm Growth With Updating
Hemodynamic Loads Using a Realistic Geometry,” Med. Eng. Phys., 33( 1), pp.
80-88.
[17] Watton, P. N., Raberger, N. B., Holzapfel, G. A., and Ventikos, Y., 2009,
“Coupling the Haemodynamic Environment to the Growth of Cerebral Aneur­
ysms: Computational Framework and Numerical Examples,” ASME J. Bio­
mech. Eng., 131(10), p. 101003.
1181 Watton, P. N., Selimovic, A., Raberger, N. B., Huang, P., Holzapfel, G. A., and
Ventikos, Y., 2011, “Modelling Evolution and the Evolving Mechanical Envi­
ronment of Saccular Cerebral Aneurysms,” Biomech. Model. Mechanobiol.,
10(1 ),pp. 109-132.
[19] Aparicio, P., Mandaltsi, A., Boamah, J., Chen, H., Selimovic, A., Bratby, M.,
Uberoi, R., Ventikos, Y., and Watton, P. N., 2014, "Modelling the Influence of
Endothelial Heterogeneity on the Progression of Arterial Disease: Application
to Abdominal Aortic Aneurysm Evolution,” Int. J. Numer. Method Biomed.
Eng., 30(5), pp. 563-586.
[20] Watton, P. N., Ventikos, Y., and Holzapfel, G. A., 2009, "Modelling the
Mechanical Response of Elastin for Arterial Tissue,” J. Biomech., 42(9), pp
1320-1325.
[21] Holzapfel, G. A., Gasser, T. C., and Ogden, R. W., 2000, “A New Constitutive
Framework for Arterial Wall Mechanics and a Comparative Study of Material
Models." J. Elasticity, 61(1-3), pp. 1-48.
[22] Watton, P. N., Ventikos, Y., and Holzapfel, G. A., 2009, “Modelling the
Growth and Stabilisation of Cerebral Aneurysms,” Math. Med. Biol., 26(2), pp.
133-164.
[23] Patankar, S. V., 1980, Numerical Heat Transfer and Fluid Flow, McGraw-Hill,
New York.
[24] Ferziger, J. H., and Peric, M., 2002, Computational Methods for Fluid Dynam­
ics, 3rd ed.. Springer, New York.
[25] Tong, J., Cohnert, T., Regitnig, P., Kohlbacher, J., Birner-Gruenberger, R.,
Schriefl, A. J., Sommer, G., and Holzapfel, G. A., 2014, “Variations of Dissec­
tion Properties and Mass Fractions With Thrombus Age in Human Abdominal
Aortic Aneurysms,” J. Biomech., 47(1), pp. 14-23.
[26] "Official Website of CMISS Solver,” www.cmiss.org
[27] Holzapfel, G. A., Sommer, G., Auer, M., Regitnig, P., and Ogden, R. W.,
2007, “Layer-Specific 3D Residual Deformations of Human Aortas With
Non-Atherosclerotic Intimal Thickening,” Ann. Biomed. Eng., 35(4), pp.
530-545.
[28] Martin Rodriguez, Z., Kenny, P., and Gaynor, L., 2011, “Improved Character­
isation of Aortic Tortuosity,” Med. Eng. Phys., 33(6), pp. 712-719.
[29] Pappu, S., Dardik, A., Tagare, H., and Gusberg, R. J., 2007, “Beyond Fusiform
and Saccular: A Novel Quantitative Tortuosity Index May Help Classify Aneu­
rysm Shape and Predict Aneurysm Rupture Potential,” Ann. Vase. Surg., 22(1),
pp. 88-97.
Transactions of the ASME
Copyright of Journal of Biomechanical Engineering is the property of American Society of
Mechanical Engineers and its content may not be copied or emailed to multiple sites or posted
to a listserv without the copyright holder's express written permission. However, users may
print, download, or email articles for individual use.