See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/258953801

Flow-induced wall mechanics of patient-specific aneurysmal cerebral

arteries: Nonlinear isotropic versus anisotropic wall stress

Article in Proceedings of the Institution of Mechanical Engineers Part H Journal of Engineering in Medicine · April 2014
DOI: 10.1177/0954411913512283 · Source: PubMed

CITATIONS READS

5 270

5 authors, including:

Sergio Cornejo Alvaro A Valencia

University of Santiago, Chile University of Chile
4 PUBLICATIONS 10 CITATIONS 82 PUBLICATIONS 1,786 CITATIONS

SEE PROFILE SEE PROFILE

Jose F Rodriguez Ender A Finol

Politecnico di Milano University of Texas at San Antonio
170 PUBLICATIONS 3,286 CITATIONS 162 PUBLICATIONS 2,680 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Personalised Computational Cardiac Physiology for Diagnosis and Interventional Planning View project

Geometric Surrogates for Clinical Management of Abdominal Aortic Aneurysms View project

All content following this page was uploaded by Ender A Finol on 27 January 2014.

The user has requested enhancement of the downloaded file.

Original Article

Proc IMechE Part H:

J Engineering in Medicine
2014, Vol 228(1) 37–48
Flow-induced wall mechanics of Ó IMechE 2013
Reprints and permissions:
patient-specific aneurysmal cerebral sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0954411913512283

arteries: Nonlinear isotropic versus pih.sagepub.com

anisotropic wall stress

Sergio Cornejo1, Amador Guzmán2, Alvaro Valencia3,

Jose Rodrı́guez4 and Ender Finol5

Abstract
Fluid–structure interaction simulations of three patient-specific models of cerebral aneurysms were carried out with the
objective of quantifying the effects of non-Newtonian blood flow and the vessel mechanical behavior on the time-
dependent fluid shear and normal stresses, and structural stress and stretch. The average wall shear stress at peak sys-
tole was found to be approximately one order of magnitude smaller than the shear stresses in the proximal communicat-
ing arteries, regardless of the shape or size of the aneurysms. Spatial distributions of oscillatory shear index were
consistent with the reciprocal of wall shear stress distributions at peak systole for all aneurysm geometries, demonstrat-
ing that oscillatory shear index correlates inversely with wall shear at this time point in the cardiac cycle. An aneurysm
wall modeled with an isotropic material resulted in an underestimation of both the maximum principal stress and
stretch, compared to the anisotropic material model. For the three aneurysm geometries, anisotropic peak wall stresses
were approximately 50% higher than for an isotropic material. Regardless of the constitutive material, the maximum
stresses were consistently located at the aneurysm neck; stresses in the dome were 30% of those in the neck.

Keywords
Aneurysm, wall stress, anisotropy, soft tissue mechanics, computational modeling, cerebral vasculature, finite element
analysis

Date received: 20 June 2013; accepted: 17 October 2013

Introduction study reported a rupture rate of only 0.05% for unrup-

tured aneurysms smaller than 10 mm in diameter.
Subarachnoid hemorrhage (SAH) is a serious cerebro- Conversely, Orz et al.9 reported on a study involving
vascular injury with an annual incidence of 6–20 cases 737 aneurysms smaller than 6 mm in diameter and
per 100,000 individuals worldwide.1,2 In more than found that 64.3% of these experienced rupture, which
three-fourths of primary SAH cases, the cause of the
bleeding is rupture of an intracranial aneurysm (IA).3 1
Department of Mechanical Engineering, Universidad de Santiago de
It is estimated that in the United States, more than
Chile, Santiago, Chile
15,000 people may suffer an aneurysmal SAH each 2
Department of Mechanical and Metallurgical Engineering, Pontificia
year4 carrying a 40%250% mortality rate.5 The risk of Universidad Católica de Chile, Santiago, Chile
3
formation and subsequent growth of an IA increases Department of Mechanical Engineering, Universidad de Chile, Santiago,
for female gender, cigarette smoking, age, and hyper- Chile
4
Group of Structural Mechanics and Materials Modelling, Aragon Institute
tension.6 IAs are about 1.6 times more common in
of Engineering Research (I3A), Zaragoza, Spain
women,7 although this does not yield a higher rupture 5
Department of Biomedical Engineering, The University of Texas at San
risk for women, since growth rates do not differ by gen- Antonio, San Antonio, TX, USA
der and rupture risk increases with higher growth rate.6
The clinical management of unruptured aneurysms Corresponding author:
Amador Guzmán, Department of Mechanical and Metallurgical
is still a matter of ongoing debate, even more after the Engineering, Pontificia Universidad Católica de Chile, 7820436, Santiago,
controversial results of the International Study of Chile.
Unruptured Intracranial Aneurysms (ISUIA).8 This Email: [email protected]
38 Proc IMechE Part H: J Engineering in Medicine 228(1)

represents 38.6% of all aneurysms analyzed. Ujiie et have developed and applied characteristic impedance-
al.10 investigated the importance of aneurysm aspect based outflow boundary conditions for large arteries in
ratio (depth/neck width) to assess rupture risk. They the systemic circulation. The effect of blood viscosity
estimated a critical aspect ratio of 1.6, above which has also been studied with the outcome that Newtonian
there is a high risk of rupture. In a related study, fluids describe similar intra-aneurysmal flow patterns
Nader-Sepahi et al.11 estimated a critical size from a compared to non-Newtonian fluid models, but they fail
study of 75 ruptured and 107 unruptured aneurysms, in the estimation of hemodynamic forces acting on the
leading to a mean aspect ratio of 2.7 for ruptured aneurysmal wall.21 Valencia et al.28 reported variations
aneurysms and 1.8 for unruptured aneurysms. and instabilities in WSS distributions associated with
IAs are treated by either open surgical repair or differences in variable viscosity models. These image-
endovascular repair. For unruptured aneurysms, open based modeling techniques have led to the categoriza-
surgical repair is recommended for the majority tion of intra-aneurysmal flow patterns with respect to
(80%285%) of patients.12 In this type of treatment, their possible correlation with rupture risk. Cebral et
patient age and aneurysm size are used clinically to pre- al.29 found that unruptured aneurysms tend to have
dict patient outcome.13,14 simple and stable flow patterns and a large region of
Hemodynamic forces have a decisive role in the bio- flow impingement. Conversely, ruptured aneurysms
mechanical stability of arteries. Among them, forces tend to present complex and unstable flow patterns and
due to wall shear stress (WSS) have been linked to small impingement zones. Torii et al.30 compared WSS
atherosclerotic plaque formation where it has been distributions obtained from CFD and fluid–structure
observed that regions of plaque lesion correlate with interaction (FSI) simulations of patient-specific aneur-
low WSS. Fry15 reported an acute yield stress of 379 ysm models, indicating that the dynamic interaction of
dyn/cm2 in a canine model where shear stresses in blood flow and the arterial wall alters the hemody-
excess of this value lead to rapid deterioration of the namic forces to the extent that wall compliance model-
endothelium. Similarly, high WSS is suspected in the ing is necessary. In this study, the arterial wall was
early phase of aneurysm growth, low WSS has been modeled as an isotropic and linearly elastic material
suggested to facilitate the growing phase and may trig- with an elastic modulus of 1 MPa. Recently, Bazilevs
ger the rupture of cerebral aneurysms by causing degen- et al.31 presented a new methodology for FSI modeling
erative changes in the aneurysm wall.16 In addition, of cerebral aneurysms in which vessel wall tissue pres-
apoptosis of endothelial cells due to low blood flow tress is included and the constraint of fixing the inlet
and low WSS precedes significant arterial remodeling.17 and outlet branches is removed.
To this end, Ku et al.18 reported that the oscillatory In this study, we assess the wall mechanics of three
shear index (OSI) and the reciprocal of WSS strongly patient-specific models of the internal carotid artery
correlate with intimal thickening. Both OSI and low (ICA) presenting aneurysms in the ophthalmic region,
WSS lead to an increased residence time of fluid parti- with the objective of obtaining a better understanding
cles on the luminal surface, and the diffusive form of of the effect of time-dependent non-Newtonian shear
the mass transport phenomenon becomes significant. and pressure. This objective is achieved by performing
Thus, intima–media thickening is observed in sites fully coupled FSI simulations with both anisotropic
exposed to low WSS.19 and isotropic constitutive materials while implementing
Computational fluid dynamics (CFD) modeling a new method for deriving physiologically realistic out-
allows the user to quantify hemodynamic factors flow boundary conditions for the cerebral vasculature.
involved in the formation, growth, and rupture of The overall goal of this study is to quantify flow-
patient-specific cerebral aneurysm models. Steinman et induced biomechanical parameters sensitive to the indi-
al.20 reported on hemodynamic features computed from vidual aneurysm geometry and subject to realistic out-
image-based simulations describing the importance of flow conditions, thus allowing for future assessment of
this technique in clinical decision making as it relates to the rupture risk when correlated with large population
the selection of the most appropriate treatment strat- geometry-based indices.
egy. The study carried out by Cebral et al.21 is of partic-
ular significance in that they report that the most
important factor in predicting the intra-aneurysmal Methodology
flow patterns is the individual aneurysm geometry.
Other studies22,23 also support the geometric depen-
Subject population and image-based reconstruction
dence of hemodynamic factors. Three patient-specific models of arteries harboring cer-
Spatial distributions of WSS and OSI are highly ebral aneurysms in the ophthalmic region (labeled
dependent on the flow distribution in the vascular models 1–3) were generated for this study and are illu-
model,24 which indicates the importance of assigning strated in Figure 1. The subject population in this pilot
accurate outflow boundary conditions. Current efforts study consists of 21-year-old and 57-year-old females
to obtain realistic boundary conditions are focused in and a 42-year-old male. Image data of the ophthalmic
the coupling different dimensional models solved simul- region were obtained with a Phillips Allura three-
taneously.25,26 To this end, Vignon-Clementel et al.27 dimensional (3D) rotational angiograph and consisted
Cornejo et al. 39

Figure 1. Patient-specific models of aneurysmal cerebral arteries indicating inflow and outflow cross sections and communicating
arteries: (a) model 1, (b) model 2, and (c) model 3. The main arteries are internal carotid artery (ICA), ophthalmic artery (OA),
middle cerebral artery (MCA), and anterior cerebral artery (ACA).

of a set of two-dimensional (2D) equally spaced slices where D is the strain rate tensor. The relationship
obtained over 180° during intra-arterial injection of a between the scalar shear rate and strain rate tensor is
radio-opaque contrast agent. The data were imported given by
into ScanIP (Simpleware v2.0, Exeter, UK), where
luminal voxels were segmented with the combination of 1h i
D= r~ vÞ T
v + ðr~ ð3Þ
both manual and automatic techniques. The arterial 2
wall was obtained by dilating the outer luminal surface qffiffiffiffiffiffiffiffi
by a uniform wall thickness equivalent to 10% the dia- g_ = 2ID 2 ð4Þ
meter of the proximal ICA. Thus, the inner wall and
outer luminal surfaces coincide as the single FSI inter- To model the time-dependent non-Newtonian
face. For the three models, the ICA was extended incompressible blood flow, both mass and momentum
proximally by a length equivalent to five diameters conservation are considered
where the inflow boundary conditions were applied.
r  v=0 ð5aÞ
 
∂~v
Fluid flow modeling r v  r~
+~ v  r  s=0 ð5bÞ
∂t
While the viscosity of blood flow is nearly constant at a
high-shear rate regime, the instantaneous shear rate where v = vð~ r, tÞ = u^i + v^j + wk^ is the velocity vector;
over a cardiac cycle varies from 0 to approximately r = xi + yj + zk^ and t represent space and time, respec-
~ ^ ^
1000 s21.32 Thus, the shear thinning behavior of the tively; r is the density; and s is the Cauchy stress tensor
blood flow becomes relevant when it takes place at given by equation (6)
low-strain rate conditions. Under these conditions, red  
blood cells tend to aggregate and generate Rouleaux s =  pI + 2h ID
2 D ð6Þ
structures, increasing the viscosity of plasma.33 The
Carreau constitutive equation is proposed for the shear where the first term pI is related to the fluid dynamics
thinning behavior of blood, as described by Bird and pressure and the second term is related to the deviatoric
Armstrong34 viscous stress.
The flow rate waveform was measured in a transver-
h in1
2 sal section of 20.74 mm2 from a healthy portion of the
hðg_ Þ = h‘ + ðho  h‘ Þ 1 + ðlg_ Þ2 ð1Þ ICA and was normalized using mass conservation with
reference to a time-averaged flow rate of 4.14 mL/s.35
where g_ is the strain rate, the constitutive parameters The waveform was decomposed into discrete Fourier
l = 3:313 s and n = 0:3568, and asymptotic apparent modes and mapped as a Womersley profile at the inlet
viscosities ho = 0:056 Pa s and h‘ = 0:00345 Pa s at 0 of each fluid domain. A variable capacitance three-
and infinite shear rates, respectively.32 element Windkessel model was derived and implemented
The second invariant of the strain rate tensor ID2 is to estimate physiological outflow boundary conditions.
defined as This model represents an effort to impose quasi-patient-
 2 specific outflow boundary conditions for the intracranial
ID
2 = tr D ð2Þ vasculature in geometric models of cerebral aneurysms.
40 Proc IMechE Part H: J Engineering in Medicine 228(1)

Wall mechanics modeling where I4 and I6 represent the squares of the stretches
The aneurysmal wall is assumed to be an incompressi- along the two families of fibers. In addition, for the
SEF of equation (10), it is assumed that the anisotropic
ble hyperelastic material for which a strain energy func-
term only contributes when either I4 . 1 or I6 . 1. For
tion (SEF) W can be identified, of the form36,37
instance, if I4 . 1 and I6 \ 1, then only I4 contributes
 ðCÞ
W ð C Þ = U ð JÞ + W ð7Þ to W.
Using equation (8), the second Piola–Kirchhoff
where U and W  denote the volumetric and the isochoric stress tensor associated with the SEF of equation (10)
contributions, respectively. The right Cauchy–Green is
tensor is given by C = FT F and the modified Cauchy–
Green deformation tensor is given by C  = FT F,
 where S = JpC1 + 2c10 DEVðIÞ + 2J2=3 W
4
the deformation gradient F has been decomposed into a DEVðn0  n0 Þ + 2J2=3 W
 6 DEVðm0  m0 Þ ð12Þ
volumetric part J1=3 I and an isochoric part F = J1=3 F,
with J = detðFÞ . 0 and det(F)  = 1. With the SEF so The material parameters for the isotropic constitu-
defined, the second Piola–Kirchhoff stress tensor can tive model, equation (9), were obtained from Delfino et
be written as al.41 (a = 44.2 kPa and b = 16.7 kPa). For the para-
  meters of the anisotropic model, equation (10), a non-
1 2=3 ∂W linear regression fit was performed for the membrane
S = pC + 2J DEV ð8Þ
∂C solution using the experimental data from Delfino
et al.41 The regression procedure resulted in the follow-
where p = dU/dJ is the constitutive equation for the ing parameters: c10 = 71.5 kPa, k1 = k3 = 13.9 kPa,
hydrostatic pressure p, I is the second-order identity k2 = k4 = 13.2 kPa, with an orientation angle of
tensor, and DEV() = ()  1=3½() : CC1 is the u = 39°42 denoting the angle between the fiber reinfor-
deviatoric operator in the Lagrangian description. The cement and the circumferential direction of the wall
Cauchy stress tensor is obtained through the expression ( cos u = m0  eu = n0  eu ).
s = 1=JFSFT .
In this investigation, isotropic and anisotropic con-
stitutive models are used for the aneurysmal wall. For Finite element meshing
the case of isotropy, the material response of the aneur- The segmented data were meshed using an automatic
ysm is characterized by the SEF37 meshing protocol in ScanFE (Simpleware v2.0, Exeter,
a h bðI1 3Þ i UK) and exported by means of an in-house code into
W ð C Þ = U ð JÞ + e 1 ð9Þ ICEM CFD v10.0 (ANSYS, Canonsburg, PA, USA),
2b
where the structural domain (the artery wall) was
where a is a constant with dimension of stress, b is remeshed with 10-node tetrahedral elements. The fluid
dimensionless, and I1 = tr(C)  is the first invariant of domain was imported into Hypermesh v8.0 (Altair
the modified Cauchy–Green tensor. For the case of ani- Engineering Inc., Troy, MI, USA) and remeshed with
sotropy, the aneurysmal tissue is modeled as a hypere- four-node tetrahedral elements with a boundary layer
lastic material reinforced with two families of fibers region. The final meshes for the fluid and structural
aligned along two directions arranged in a double-helix domains were exported for compatibility with ADINA-
pattern. Therefore, the SEF W  is expressed as F and ADINA v8.4 (ADINA R&D Inc., Watertown,
W = W(  n0 , m0 ), with n0 and m0 denoting the direc-
 C, MA, USA), respectively, for the proper FSI analysis.
tions of collagen fibers within the tissue.38,39 In addi- Finite element meshes with an increasing density
tion, W is additively decomposed into an isotropic (expressed in number of elements per cubic millimeter)
contribution, corresponding to the matrix material, were used with the purpose of mesh sensitivity analyses,
and an anisotropic contribution, related to the (two combined with reasonable trade-offs in CPU time and
families of) collagen fibers,40 that is computational memory requirements. Table 1 shows
the maximum principal stress, relative error, memory
k1 h k2 ðI4 1Þ2 i
WðCÞ = UðJÞ + c10 ðI1  3Þ + e 1 requirement, and processing time for five different
2k2 meshes for model 3, where a given mesh density is
k3 h k4 ðI6 1Þ2 i
approximately double that of the previous mesh. The
+ e 1 ð10Þ
2k4 relative error is defined as the difference in the maxi-
mum principal stress obtained between two sequential
where c10 is a stress-like material parameter and meshes divided by the maximum principal stress
k1 . . . k4 are material parameters corresponding to the obtained with the coarser mesh. It is observed that the
fibers. The additional invariants appearing in equation relative error decreases exponentially with increasing
(10) are mesh density. The optimal mesh density for each model
is determined by considering the maximum principal
I4 = n0  C  n0
ð11Þ stress relative error, processing time and memory
I6 = m0  C  m0 requirements. The chosen discretization for model 3
Cornejo et al. 41

Table 1. Maximum principal stress, relative error, memory requirement, and processing time for increasing mesh densities
(elements/mm3) in the solid domain of model 3.

Mesh density (elements/mm3) Maximum principal stress (kPa) Relative error (%) Memory requirement (GB) CPU time (h)

97 532.57 – 0.43 0.80

227 674.78 26.70 1.04 2.18
484 763.07 13.08 2.59 6.15
920 817.66 7.15 6.20 18.63
2022 866.14 5.93 14.70 27.81

consists of mesh densities of 280 linear and 920 quadra- distribution, although time dependent, is mostly uni-
tic tetrahedral elements per cubic millimeter for the form within the aneurysm, with maximum wall pres-
fluid and structural domains, respectively. sures at the aneurysm dome of 98, 105, and 120 mm
Furthermore, the fluid domain of the three models was Hg, for models 1–3, respectively.
meshed with three layers of structured elements located Figure 5(a)–(c) shows the instantaneous WSS distri-
at the boundary layer region. bution for models 1–3, respectively, at peak systole.
Two different WSS patterns are observed: within the
aneurysm sac, the WSS is almost spatially uniform,
Stress analysis of patient-specific cerebral arteries
whereas the peripheral vasculature exhibits a wide
The hyperelastic anisotropic model is implemented in a range of stresses. The spatially averaged WSS at peak
material user subroutine within ADINA. The subrou- systole for the three aneurysms (15 dyn/cm2 (0.01125
tine was validated with the analytical solution of a pres- mm Hg) for model 1, 45 dyn/cm2 (0.03375 mm Hg) for
surized cylinder reinforced with two families of fibers, model 2, and 2.5 dyn/cm2 (0.001875 mm Hg) for model
as shown in Figure 2. The cylinder has an inner dia- 3) is approximately one order of magnitude less than
meter Di = 0.6 mm, outer diameter Do = 1.0 mm, and the maximum WSS obtained in the respective commu-
length L = 5 mm. The inner pressure is taken to be p nicating arteries. There is an evident association
= 4.0 kPa. The material properties are taken as c10 = between aneurysm diameter and average WSS, with the
0.5 kPa, k1 = k3 = 1.1 kPa, k2 = k4 = 5.0. Three lowest WSS obtained for model 3, which is the largest
cases were considered with fibers oriented at 2°, 45°, of the three aneurysms. In addition, the WSS distribu-
and 60° with respect to the circumferential direction. tions relate to the flow pathlines within the aneurysm
The numerical solution was compared with the semi- sac as the largest velocity gradients are obtained in
analytical solution given in Holzapfel et al.40 for the aneurysm model 2, which also yields the highest peak
three fiber orientations, as illustrated in Figure 2, show- WSS. Distributions of OSI at peak systole are dis-
ing excellent agreement of our numerical model. played in Figure 6 for the three vascular models.
Models 1 and 2 present a similar maximum OSI of
Results and discussion approximately 0.02 in the aneurysms, whereas model 3
shows a maximum OSI of about 0.2, that is, one order
Flow dynamics in the cerebral vasculature of magnitude greater. The overall distribution is consis-
Figure 3 shows flow pathlines—instantaneous stream- tent with the reciprocal of the WSS pattern developed
lines at peak systole—for patient-specific models 1–3. in the three models, as the OSI takes into account the
The three models develop a similar flow pattern charac- time-dependent oscillatory nature of the WSSs and the
terized by a recirculation region within each aneurysm velocity gradient fields. Similarly, Bazilevs et al.31,43
and relatively parallel pathlines in the peripheral vascu- performed FSI modeling of four patient-specific cere-
lature where the maximum velocities are obtained. At bral aneurysms with prescribed outflow resistance
the beginning of the cardiac cycle, the flow entering the boundary conditions, vessel wall tissue prestress, and a
aneurysm impinges on the wall originating vortices that hyperelastic isotropic material model, finding that FSI
remain inside the aneurysm throughout the entire pul- with flexible walls plays an important role toward the
satile period. Because of the time-dependent nature of accurate prediction of patient-specific hemodynamics.
the inflow boundary condition, the intra-aneurysmal While the differences in flow dynamics between the
flow patterns are characterized by vortices that are con- computed FSI and rigid wall models were not signifi-
tinuously increasing and decreasing in strength and size cant, the WSSs for rigid wall aneurysms were up to
and changing spatial locations within the sac. The 50% higher than those predicted with FSI. Torii et
instantaneous wall pressure distribution at peak systole al.30,44 numerically investigated the role of FSI on pre-
is illustrated in Figure 4 for the three patient-specific dicting the effect of aneurysm shape, pulsatile inlet
models, showing a low pressure on the aneurysm wall pressure (a hypertension condition), and incoming flow
(~94 mm Hg) compared to the relatively high inlet pres- type on the WSSs occurring in cerebral aneurysm
sure (~110 mm Hg) at the ICA. The wall pressure located at Y-shaped bifurcations and in a bifurcated
42 Proc IMechE Part H: J Engineering in Medicine 228(1)

Figure 2. Pressurized cylinder reinforced with two families of fibers. Comparison between analytical40 and finite element analysis
solutions.

artery. Their results showed that modeling impinging not enter the aneurysm and thus flows downstream
jet inlet conditions with FSI, instead of rigid vessel through the OA.
walls, greatly increases the understanding and precise
estimation of the WSSs occurring in the impinging jet
and nearby regions.
Wall mechanics of the aneurysm and cerebral
The maximum WSS for the three models is in the arteries
same order of magnitude as those computed by Cebral Figure 7 illustrates the maximum principal stress distri-
et al.21 and Castro et al.24 The high OSI in the aneur- bution on the luminal surface of model 1 for isotropic
ysm dome for each model coexists with low WSS val- and anisotropic constitutive material models. Figure
ues, when compared to the inlet and outlet arteries, 7(a) and (b) shows views of the entire surface and the
where OSI is smaller and WSS is higher. Fluid particles aneurysm dome (top view) for the constitutive isotropic
have a tendency to remain within the aneurysm sac material, whereas Figure 7(c) and (d) shows the same
because of the existing recirculation regions and vor- views for the anisotropic constitutive material. The
tices during the cardiac cycle. Thus, there is a higher maximum stresses are located mostly near the aneur-
residence time of fluid particles following this flow pat- ysm neck and to a lesser degree at the aneurysm dome,
tern, which tends to increase the transport of platelets regardless of the constitutive material used to model
and macrophages toward the wall, which increases their the arteries. However, the stresses for the anisotropic
likelihood of adhesion or deposition. In addition, the constitutive material are higher than those obtained
flow pattern developed in the three models can be clas- with the isotropic material and the peak stress is distrib-
sified as a type III flow pattern,29 which is represented uted over a larger surface area of the aneurysm dome.
by two flow characteristics: (1) the main flow impacts The isotropic peak stresses in the aneurysm neck and
the bottom of the aneurysm sac and returns through dome are about 0.25 and 0.09 MPa, respectively, that
the ICA, and therefore, it does not remain within the is, there is a 70% decrease in maximum principal stress
aneurysm sac; and (2) the other flow stream, which does at the dome compared to the neck.
Cornejo et al. 43

Figure 3. Instantaneous flow streamlines at peak systole for patient-specific (a) model 1, (b) model 2, and (c) model 3.

Figure 4. Wall pressure distribution at peak systole. The maximum pressures at the aneurysm dome for models 1–3 are (a) Pmax =
98 mm Hg, (b) Pmax = 105 mm Hg, and (c) Pmax = 120 mm Hg, respectively.

Figure 5. WSS distribution at peak systole. The WSS range for the aneurysm sac for models 1–3 is (a) WSS = 0–30 dyn/cm2,
(b) WSS = 50–70 dyn/cm2, and (c) WSS = 0–5 dyn/cm2, respectively.
WSS: wall shear stress.

Figure 6. OSI distribution at peak systole for (a) model 1, (b) model 2, and (c) model 3.
OSI: oscillatory shear index.
44 Proc IMechE Part H: J Engineering in Medicine 228(1)

Figure 7. Maximum principal wall stress for model 1 for isotropic and anisotropic constitutive material models: (a and b) views of
the vasculature and the aneurysm dome for the constitutive isotropic material and (c and d) the same views for the anisotropic
constitutive material.

Figure 8. Maximum principal wall stress for model 2 for isotropic and anisotropic constitutive material models: (a and b) views of
the vasculature and the aneurysm dome for the constitutive isotropic material and (c and d) same views for the anisotropic
constitutive material.
Cornejo et al. 45

Figure 9. Maximum principal wall stress for model 3 for isotropic and anisotropic constitutive material models: (a and b) views of
the vasculature and the aneurysm dome for the constitutive isotropic material and (c and d) the same views for the anisotropic
constitutive material.

The maximum principal stress for models 2 and 3 is anisotropic material model is computationally more
shown similarly in Figures 8 and 9, respectively, for expensive, it was observed that the anisotropic maxi-
both isotropic and anisotropic constitutive materials, mum principal stresses were approximately 1.5 times
exhibiting patterns similar to that of model 1. The peak larger than the isotropic counterparts for all three geo-
stress for model 2 is located at the aneurysm neck and metries and within the same order of magnitude as
is approximately 0.5 MPa for the anisotropic material, obtained by Ma et al.45 using finite element analysis.
which is higher than the isotropic material and extends Figure 10 shows the maximum principal stretch at
over a larger surface area. This pattern also develops in the aneurysm dome for models 1–3 obtained with iso-
the aneurysm dome, where the maximum principal tropic (Figure 10(a), (c), and (e)) and anisotropic
stress is about 0.1 MPa and its intensity higher and spa- (Figure 10(b), (d), and (f)) constitutive materials,
tial distribution larger for the anisotropic material. For respectively. Independent of the flow dynamics within
model 3, the highest stresses cover an extensive area of the aneurysm and the ensuing maximum principal
the body and dome of the aneurysm, with regions of stresses, maximum principal stretch is higher with the
high and medium intensity regardless the wall constitu- anisotropic wall material compared to the isotropic
tive material. The aneurysm body and dome peak stres- wall material, regardless of aneurysm geometry. For
ses are about 0.5 and 0.3 MPa, respectively, for the example, the peak anisotropic stretch is 1.16, 1.17, and
isotropic material model. As with models 1 and 2, the 1.32, for models 1–3, respectively, compared to the
maximum principal stress is higher with the anisotropic peak isotropic stretches of 1.14, 1.11, and 1.21. Upon
material model. While FSI modeling with an examination of Figures 7–10, it is evident that the dome
46 Proc IMechE Part H: J Engineering in Medicine 228(1)

Figure 10. Maximum principal stretch in the aneurysm dome (top view) for models 1–3: (a, c, and e) isotropic material model and
(b, d, and f) anisotropic material model.

regions with the highest stress coincide with those of displacement profiles were similar for all models, the
the highest stretch, for both wall material models. The maximum displacement using the hyperelastic model is
role of the cerebral aneurysm and communicating 36% smaller than the one with the linear elastic model
artery compositions, using linear elastic (with small with finite strain, and that the maximum WSS near the
strain and finite strain) and hyperelastic wall constitu- apex is 36% smaller for the hyperelastic model than for
tive models, has also been studied by Torii et al.46 by the linear elastic models.
performing FSI simulations of a patient-specific cere- The role of aneurysm wall thickness on the aneur-
bral aneurysm. They found that although the ysm hemodynamics and WSSs was investigated by
Cornejo et al. 47

Torii et al.47 by performing FSI simulations of patient- grant nos 1110008 and 1100238 at the Universidad de
specific cerebral aneurysms considering uniform and Chile, Pontificia Universidad Católica de Chile, and
pathological wall thickness estimations. Their results Universidad de Santiago de Chile.
showed that more accurate WSSs and hemodynamic
factors are predicted using a pathological wall thick- References
ness, and that uniform wall thickness simulations can
1. Wardlaw JM and White PM. The detection and manage-
underestimate rupture risk in cerebral aneurysms.
ment of unruptured intracranial aneurysms. Brain 2000;
Furthermore, the effect of the peripheral vasculature
123(2): 205–221.
(PVM) on the flow division occurring in the two outlets 2. Juvela S. Minor leak before rupture of an intracranial
of a cerebral aneurysm located at the middle cerebral aneurysm and subarachnoid hemorrhage of unknown
artery (MCA) was also numerically investigated by etiology. Neurosurgery 1992; 30(1): 7–11.
Torii et al.48 by means of FSI modeling. Their results 3. Zhang B, Fugleholm K, Day LB, et al. Molecular patho-
showed that the pressures at the two outlets with the genesis of subarachnoid haemorrhage. Int J Biochem Cell
PVM were almost identical, and that the flow division Biol 2003; 35: 1341–1360.
to the two branches with and without PVM is almost 4. Schievink W. Intracranial aneurysms. N Engl J Med
the same. 1997; 336: 28–40.
5. Fogelholm R, Hernesniemi J and Vapalahti M. Impact
of early surgery on outcome after aneurysmal subarach-
Conclusion noid hemorrhage: a population-based study. Stroke
1993; 24: 1649–1654.
Fully coupled FSI simulations of aneurysm geometries 6. Juvela S. Risk factors for multiple intracranial aneur-
in the ophthalmic region of three patient-specific mod- ysms. Stroke 2000; 31(2): 392–397.
els were performed to quantify the pulsatile wall shear, 7. Van Gijn J and Rinkel GJ. Subarachnoid hemorrhage:
pressure, and mechanical stress and stretch by consider- diagnosis, causes and management. Brain 2001; 124:
ing physiologically realistic outflow boundary condi- 249–278.
8. International Study of Unruptured Intracranial Aneur-
tions, non-Newtonian blood flow, and vascular wall
ysms Investigators. Unruptured intracranial
anisotropic and isotropic constitutive material models. aneurysms—risk of rupture and risks of surgical inter-
The average aneurysmal WSS at peak systole was found vention. N Engl J Med 1998; 339: 1725–1733.
to be an order of magnitude smaller than the shear 9. Orz Y, Kobayashi S, Osawa M, et al. Aneurysm size: a
stresses in the communicating arteries. Spatial distribu- prognostic factor for rupture. Br J Neurosurg 1997; 11:
tions of OSI were consistent with the reciprocal of the 144–149.
WSS at peak systole for the three aneurysm geometries, 10. Ujiie H, Tachibana H, Hiramatsu O, et al. Effects of size
demonstrating that OSI correlates inversely with wall and shape (aspect ratio) on the hemodynamic of saccular
shear at this time point in the cardiac cycle. The isotro- aneurysm: a possible index for surgical treatment of intra-
pic wall material underestimated the maximum princi- cranial aneurysm. Neurosurgery 1999; 45: 119–130.
11. Nader-Sepahi A, Casimiro M, Sen J, et al. Is aspect ratio
pal stress and stretch and their regional distributions at
a reliable predictor of intracranial aneurysm rupture?-
peak systole. On average, aneurysmal anisotropic peak
Neurosurgery 2004; 54: 1343–1348.
wall stress was approximately 50% higher compared to 12. Wolfe SQ, Basxkaya MK, Heros RC, et al. Cerebral
its isotropic equivalent. Regardless of the constitutive aneurysms: learning from the past and looking toward
material used, the peak stresses were located at the the future. Clin Neurosurg 2006; 53: 157–178.
aneurysm neck and were three times higher than those 13. Wiebers DO, Whisnant JP, Huston J III, et al.; Interna-
at the dome. Through the integration of a variable tional Study of Unruptured Intracranial Aneurysms
capacitance Windkessel outflow boundary condition Investigators. Unruptured intracranial aneurysms: natu-
model and a new anisotropic constitutive material for ral history, clinical outcome, and risks of surgical and
the cerebral vasculature, this investigation yielded an endovascular treatment. Lancet 2003; 362: 103–110.
improved understanding of the effect of pulsatile non- 14. Wiebers DO, Piepgras DG, Meyer FB, et al. Pathogen-
esis, natural history, and treatment of unruptured intra-
Newtonian blood flow on the cerebral aneurysm wall
cranial aneurysms. N Engl J Med 1997; 336: 28–40.
mechanics. 15. Fry DL. Acute vascular endothelial changes associated
with increased blood velocity gradients. Circ Res 1968;
Declaration of conflicting interests 22: 165–197.
16. Shojima M, Oshima M, Takagi K, et al. Magnitude and
The authors declare that there is no conflict of interest. role of wall shear stress on cerebral aneurysm computa-
tional fluid dynamic study of 20 middle cerebral artery
Funding aneurysms. Stroke 2004; 35: 2500–2505.
17. Sho E, Sho M, Singh TM, et al. Blood flow decrease
This study was supported by the Spanish Ministry of induces apoptosis of endothelial cells in previously dilated
Science and Technology through research grant no. arteries resulted from chronic high blood flow. Arterios-
DPI2004-07410-C03-01 at the University of Zaragoza cler Thromb Vasc Biol 2001; 21: 1139–1145.
and the Chilean National Fund for Scientific and 18. Ku DN, Giddens DP, Zarins CK, et al. Pulsatile flow and
Technological Development (FONDECYT) through atherosclerosis in the human carotid bifurcation. Positive
48 Proc IMechE Part H: J Engineering in Medicine 228(1)

correlation between plaque location and low oscillating 32. Pedley TJ. The fluid mechanics of large blood vessels.
shear stress. Arteriosclerosis 1985; 5: 293–302. Cambridge: Cambridge University Press, 1980.
19. Irace C, Cortese C, Fiaschi E, et al. Wall shear stress is 33. Cho YI and Kensey R. Effects of the non-Newtonian
associated with intima-media thickness and carotid ather- viscosity of blood flows in a diseased arterial vessel, part
osclerosis in subjects at low coronary heart disease risk. 1: steady flows. Biorheology 1991; 28: 241–262.
Stroke 2004; 35: 464–468. 34. Bird RB and Armstrong RC. Dynamics of polymeric
20. Steinman D, Milner J, Norley C, et al. Image-based com- liquids: fluid dynamics (eds RB Bird, RC Armstrong and
putational simulation of flow dynamics in a giant intra- O Hassager), vol. 1. 2nd ed.New York: Wiley, 1987.
cranial aneurysm. AJNR Am J Neuroradiol 2003; 24: 35. Marshall I, Papathanasopoulou P and Wartolowska K.
559–566. Carotid flow rates and flow division at the bifurcation in
21. Cebral J, Castro M, Appanaboyina S, et al. Efficient healthy volunteers. Physiol Meas 2004; 25: 691–697.
pipeline for image-based patient-specific analysis of cere- 36. Simo JC, Taylor RL and Pister KS. Variational and pro-
bral aneurysm hemodynamics: technique and sensitivity. jection methods for the volume constraint in finite defor-
IEEE Trans Med Imaging 2005; 24: 457–467. mation elastoplasticity. Comput Methods Appl Mech Eng
22. Hoi Y, Woodward SH, Kim M, et al. Validation of CFD 1985; 51: 177–208.
simulations of cerebral aneurysms with implication 37. Holzapfel GH. Nonlinear solid mechanics: a continuum
of geometric variations. J Biomech Eng 2006; 128: approach for engineering. Chichester: Wiley, 2000.
844–851. 38. Drangova M, Holdsworth DW, Boyd CJ, et al. Elasticity
23. Jou LD, Wong G, Dispensa B, et al. Correlation between and geometry measurements of vascular specimens using
luminal geometry changes and hemodynamics in fusiform a high-resolution laboratory CT scanner. Physiol Meas
intracranial aneurysms. AJNR Am J Neuroradiol 2005; 1993; 14: 277–290.
26: 2357–2363. 39. Demiray H. A note on the elasticity of soft biological tis-
24. Castro MA, Putman CM and Cebral JR. Patient-specific sues. J Biomech Eng 1972; 5: 309–311.
computational fluid dynamics modeling of anterior com- 40. Holzapfel GA, Gasser TC and Ogden RW. A new consti-
municating artery aneurysms: a study of the sensitivity of tutive framework for arterial wall mechanics and a com-
intra-aneurysmal flow patterns to flow conditions in the parative study of material models. J Elast 2000; 61: 1–48.
carotid arteries. AJNR Am J Neuroradiol 2006; 27: 2061– 41. Delfino A, Stergiopoulus N, Moore JE, et al. Residual
2068. strain effects on the stress field in a thick wall finite ele-
25. Formaggia L, Gerbean JF, Nobile F, et al. On the cou- ment model of the human carotid bifurcation. J Biomech
pling of 3D and 1D NS equations for flow problems in 1997; 30: 777–786.
compliant vessels. Comput Methods Appl Mech Eng 2001; 42. Hariton I, deBotton G, Gasser TC, et al. Stress driven
191: 561–582. collagen fiber remodeling in arterial walls. Biomech Model
26. Urquiza SA, Blanco PJ, Venere MJ, et al. Multidimen- Mechanobiol 2007; 6: 163–175.
sional modelling for the carotid artery blood flow. Com- 43. Bazilevs Y, Hsu M-C, Zhang Y, et al. A fully-coupled
put Methods Appl Mech Eng 2006; 195: 4002–4017. fluid–structure interaction simulation of cerebral aneur-
27. Vignon-Clementel IE, Figueroa CA, Jansen KE, et al. ysms. Comput Mech 2010; 46: 3–16.
Outflow boundary conditions for three-dimensional finite 44. Torii R, Oshima M, Kobayashi T, et al. Fluid–structure
element modeling of blood flow and pressure in arteries. interaction modeling of aneurysmal conditions with high
Comput Methods Appl Mech Eng 2006; 195: 3776–3796. and normal blood pressures. Comput Mech 2006; 38:
28. Valencia AA, Guzman AM, Finol EA, et al. Blood flow 482–490.
dynamics in saccular aneurysm models of the basilar 45. Ma B, Lu J, Harbaugh R, et al. Nonlinear anisotropic
artery. J Biomech Eng 2006; 128: 516–526. stress analysis of anatomically realistic cerebral aneur-
29. Cebral JR, Castro MA, Burgess JE, et al. Characteriza- ysms. J Biomech Eng 2007; 129: 88–96.
tion of cerebral aneurysms for assessing risk of rupture by 46. Torii R, Oshima M, Kobayashi T, et al. Fluid–structure
using patient-specific computational hemodynamics mod- interaction modeling of a patient-specific cerebral aneur-
els. AJNR Am J Neuroradiol 2005; 26: 2550–2559. ysm: influence of structural modeling. Comput Mech
30. Torii R, Oshima M, Kobayashi T, et al. Fluid–structure 2008; 43: 151–159.
interaction modeling of blood flow and cerebral aneur- 47. Torii R, Oshima M, Kobayashi T, et al. Influence of wall
ysm: significance of artery and aneurysm shapes. Comput thickness on fluid–structure interaction computations of
Methods Appl Mech Eng 2009; 198: 3613–3621. cerebral aneurysms. Commun Numer Methods Eng 2009;
31. Bazilevs Y, Hsu M, Zhang Y, et al. Computational vascu- 26: 336–347.
lar fluid–structure interaction: methodology and applica- 48. Torii R, Oshima M, Kobayashi T, et al. Role of 0D per-
tion to cerebral aneurysms. Biomech Model Mechanobiol ipheral vasculature model in fluid–structure interaction
2010; 9: 481–498. modeling of aneurysms. Comput Mech 2010; 46: 43–52.

View publication stats