Computers and Structures 80 (2002) 1651–1665

www.elsevier.com/locate/compstruc

Simulating cyclic artery compression using a 3D

unsteady model with fluid–structure interactions
a,*
Dalin Tang , Chun Yang b, Homer Walker a, Shunichi Kobayashi c,
David N. Ku d
a
Mathematical Sciences Department, Worcester Polytechnic Institute, 100 Institute Road, Worcester, MA 01609-2280, USA
b
Mathematics Department, Beijing Normal University, Beijing, China 100875
c
Department of Functional Machinery and Mechanics, Shinshu University, 3-15-1 Tokida Ueda, 386-8567 Nagano, Japan
d
Georgia Institute of Technology, School of Mechanical Engineering, Atlanta, GA 30322, USA
Received 11 October 2001; accepted 24 April 2002

Abstract
High pulsating blood pressure and severe stenosis make fluid–structure interaction (FSI) an important role in
simulating blood flow in stenotic arteries. A three-dimensional nonlinear model with FSI and a numerical method using
GFD are introduced to study unsteady viscous flow in stenotic tubes with cyclic wall collapse simulating blood flow in
stenotic carotid arteries. The Navier–Stokes equations are used as the governing equations for the fluid. A thin-shell
model is used for the tube wall. Interaction between fluid and tube wall is treated by an incremental boundary iteration
method. Elastic properties of the tube wall are determined experimentally using a polyvinyl alcohol hydrogel artery
stenosis model. Cyclic tube compression and collapse, negative pressure and high shear stress at the throat of the
stenosis, flow recirculation and low shear stress just distal to the stenoses were observed under physiological conditions.
These critical flow and mechanical conditions may be related to platelet aggregation, thrombus formation, excessive
artery fatigue and possible plaque cap rupture. Computational and experimental results are compared and reasonable
agreement is found.
Ó 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Generalized finite difference; Free moving boundary; Stenosis; Blood flow; Artery

1. Introduction in blood vessels) which lowers flow pressure. If the ste-

nosis is severe enough, blood pressure may become
A nonlinear three-dimensional (3D) model with negative and cause artery compression or even collapse
fluid–structure interactions (FSI) and an iterative nu- leading to serious clinical consequences such as stroke or
merical method based on generalized finite difference heart attack [1,3,41,44]. The mechanism for the whole
(GFD) are introduced to model unsteady blood flow in collapse process is not fully understood. Since blood
stenotic arteries and simulate cyclic wall compression flow is pulsatile and tube collapse is fully 3D, a 3D
and unsteady viscous flow in a stenotic elastic tube with unsteady model is necessary even though the resting
large strain and large deformation. Flow velocity ac- shape of the stenotic tube is axisymmetric in this simu-
celerates when passing through a stenosis (a constriction lation. The incompressible Navier–Stokes (N–S) equa-
tions are used for the fluid model while a thin-shell
model is used for the tube wall so that cyclic wall col-
*
Corresponding author. Tel.: +1-508-831-5332; fax: +1-508- lapse can be simulated [10]. Pressure–cross-section area
831-5824. relationship of the tube wall (known as the tube law,
E-mail address: [email protected] (D. Tang). which is essentially a bi-axial stress/strain relation) is

0045-7949/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 5 - 7 9 4 9 ( 0 2 ) 0 0 1 1 1 - 6
1652 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665

measured at three locations of the tube experimentally stroke, etc. [5,7,8,15,30]. Experiment-based computa-
using a PVA hydrogel stenotic tube whose mechanical tional models with strong FSI are needed to better un-
properties are close to that of bovine carotid arteries derstand these processes and to quantify physiological
[23,24]. The classic tube law introduced for a uniform conditions under which artery collapse may occur.
collapsible tube [21,38] is extended to include axial po- Extensive experimental research have been conducted
sition, longitudinal tension and axial curvature changes to quantify mechanical properties of arteries [13,14,18].
so that it is more suitable for a stenotic compliant tube. However, most of the data obtained are for arteries
A physiological 36.5% axial pre-stretch is applied to the under expansion with positive pressures. For collapsible
stenotic tube model. Physical parameters and geomet- tubes, pressure–tube cross-section area relationship
rical dimensions corresponding to blood flow in human (tube law) has been widely used to describe the elastic
carotid arteries are used to make the model physiologi- properties of the tube wall under both positive and
cally relevant. In the computational wall model, tube negative pressure conditions. Considerable work for
expansion (which is not assumed to be axisymmetric) flow in collapsible tubes of uniform diameter has been
under positive pressure is obtained pointwisely using a reported in the last 25 years and many interesting phe-
pressure–radius relationship derived from the tube law. nomena such as flow choking, flow limitation and dy-
For tube deformation under negative pressure, the cir- namic behavior (flutter) have been identified and
cumferential arc length is assumed to be inextensible investigated (for a review, see [9,32]). For flow in ste-
[10], and tube compression and collapse are determined notic collapsible tubes and arteries, research has been
by solving the thin-shell equilibrium equations using focused on the effect of stenosis severities on the flow
flow pressure and shear stress distributions on the tube and wall motions under various pressure conditions.
wall. Mechanical parameters such as the Young’s Different stenoses were used in several investigations to
modulus and the bending stiffness coefficient are deter- quantify pressure–area and pressure–flow relations and
mined by experimental measurements. The wall model collapse conditions [4,20,23,24,34,39,41].
and fluid model are solved iteratively using an incre- Various computational models (from 1D to 3D, with
mental boundary iteration method [37]. Effects of ste- rigid or compliant tubes) have been used to quantify
nosis severity and pressure conditions on cyclic wall flow and wall mechanical behaviors [1,3,9,11,21,27,
bending and compression, flow velocity, pressure, and 44,48]. It has been found that artery stiffness, stenosis
shear stress are investigated to quantify possible wall geometry and severity and imposed pressure conditions
collapse conditions and flow characteristics which may are the dominating factors affecting blood flow and ar-
be related to artery fatigue and plaque cap rupture. tery motion. However, except the 1D models, higher
Details of the model, method and results are given in the dimensional models with FSI simulating cyclic wall
following sections. compression and collapse are still lacking in the litera-
ture for the following reasons:

2. Background (i) Mechanical properties of arteries under compres-

sion are not readily available. Without the experi-
FSI play important roles in many biological pro- mental data, modeling for artery compression will
cesses, especially for blood flow in stenotic arteries. have no supporting basis and no validation.
Blood vessels are highly compliant. Under pulsatile (ii) Stenotic artery wall behavior under pulsatile pres-
blood pressure, high grade stenoses cause critical flow sure is very complex. It is fully 3D, dynamic, in-
and mechanical conditions such as high flow velocity, volves large strain, large deformation, and cyclic
high shear stress, flow recirculation, negative pressure tube collapse and expansion. Severe stenosis makes
and cyclic artery compression which may be related to FSI very strong with small change in one causing
platelet aggregation, thrombus formation, and excessive large change in another. Regular boundary itera-
artery fatigue (for reviews, see [25,47]). Changes in blood tion methods may fail to converge because of that
pressure causes artery deformation which affects stenosis [37].
severity (percentage of artery diameter narrowing). A (iii) Severe stenosis causes critical flow conditions which
small change in stenosis severity leads to considerable are computationally difficult to handle. Algorithms
changes of flow velocity, shear stress and pressure in the that converge for normal pressure and flow condi-
stenotic region, which in turn affects artery deformation. tions may not converge under these critical condi-
The FSI continues and may lead to artery compression tions. For the case we are considering, flow
or even collapse when stenosis becomes severe enough pressure drops from 150 to )12.5 mmHg just across
(severity 70–80% in diameter). There has been increasing a 80% stenosis. That is from about 199,800 to
evidence that stenotic plaque may rupture under physi- )16,650 dyn/cm2 in less than 1 cm. That leads to
ological conditions and cause fatal subsequential athero- very large pressure gradient (105 ) in the flow field.
sclerotic diseases such as myocardial infarction, cerebral The flow speed at the throat of the stenosis is about
 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665 1653

400 cm/s just 0.01 cm away from the wall which At the inlet and outlet of the tube, we set:
leads to very high flow shear stress on the tube wall.
pjz¼0 ¼ pin ðtÞ; pjz¼‘ ¼ pout ðtÞ; ð7Þ
These critical conditions require special handling, at
least much finer mesh should be used near the tube
wall and stenosis to get enough resolution. ou
¼ 0: ð8Þ
oz z¼0;‘
The above leads to the introduction of our compu-
We start the computation from zero pressure and zero
tational model which provides a first order approxima-
flow state and gradually raise the pressure to the pre-
tion of the complex artery cyclic collapse process with
scribed conditions.
available experimental data.
3.2. The wall model
3. The Computational model
Mechanical properties of arteries under compression
3.1. The fluid model and pulsatile conditions are very hard to obtain. It is
hard to simulate dynamic wall deformation and collapse
We consider viscous flow in a compliant tube simu- correctly without this information. However, to measure
lating pulsatile blood flow in stenotic carotid arteries. the 3D nonlinear dynamic wall mechanical properties
The flow is assumed to be laminar, Newtonian, viscous and solve the complete nonlinear wall model [13] involving
and incompressible. Using the arbitrary Lagrangian– large strain and large deformation in the collapse process
Eulerian (ALE) formulation [19,35], the N–S equations is a forbidding task. Most existing linear or nonlinear
become: wall models for arteries are applicable only to normal
positive pressure conditions and are no longer valid when
qðou=ot þ ððu  ug Þ  rÞuÞ ¼ rp þ lr2 u; ð1Þ pressure becomes negative and the artery is under com-
pression [13,40].
r  u ¼ 0; ð2Þ In this paper, we use tube law and a thin-shell model
[10] to determine the wall motion under both positive
where u and p are flow velocity and pressure, o=ot is t- and negative pressure conditions. To determine the elastic
derivative with mesh points fixed, ug is the mesh velocity. properties of the stenotic tube, the pressure–area rela-
Assuming the fluid and wall move together at the tube tionship (tube law)
wall, we have:
ðp  pe ÞjZ¼Zi ¼ pi ðaÞ; a ¼ A=A0 ; i ¼ 1; 2; 3 ð9Þ
ox
u ¼ ðu; v; wÞjC ¼ : ð3Þ
ot is measured for a tube made of PVA hydrogel with a
80% thick-walled stenosis under 36.5% axial stretch. The
Here C stands for the inner tube wall (Fig. 1), x ¼
measurements are taken at three locations of the tube
ðr; h; zÞ is the position vectors of the deformed tube wall,
(straight, shoulder and throat) to take the stenotic effect
u, v and w are the radial, angular and axial components
into consideration. The experimental data is given by
of the flow velocity. The undeformed inner tube wall
Fig. 1. A and A0 are the deformed and undeformed
radius with a symmetric stenosis is given by
cross-section areas of the tube respectively. The external
RðZÞ ¼ R0  SðZÞ; ð4Þ pressure pe is set to zero in this paper. For computa-
8 tional convenience (as well as when conducting the ex-

2
>
< 2pðZ  Z1 Þ periments), the inverse of (9) is used to determine a when
1  cos p is given (noting pe ¼ 0),
SðZÞ ¼ ðZ2  Z1 Þ
> SR ; Z1 6 Z 6 Z2 ;
: 0 0 4 a ¼ ai ðpÞ; i ¼ 1; 2; 3: ð10Þ
0; otherwise:
ð5Þ Mathematical interpolation is used to connect the
where X ¼ ðR; H; ZÞ is the position vector of the unde- three experimentally measured tube laws to cover the
formed tube wall, R0 is the radius of the uniform part of entire tube
X
the tube, SðZÞ specifies the shape of the stenosis, S0 is the ag ðz; pÞ ¼ Ci ðzÞai ðpÞ; ag ðzi ; pÞ ¼ ai ðpÞ; ð11Þ
stenosis severity by diameter, i.e., reduction of the tube i
diameter caused by the stenosis, Z1 and Z2 specify the
beginning and ending of the stenosis. Stenosis severity is where Ci ðzÞ are properly chosen to reflect the influence
commonly defined as of the shape and stiffness of the stenosis.
Wall deformation is determined using two different
ðR0  Rmin Þ methods depending on whether the tube is under ex-
S0 ¼ 100%: ð6Þ
R0 pansion or compression. For the portion where the
1654 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665

Fig. 1. The stenotic tube and the tube law measurements.

pressure is positive, the tube is under expansion and set of z-rings and a set of h-lines (longitudinal lines on
axisymmetric (or is nearly axisymmetric), (11) is used to the tube with fixed h angles). Following Flaherty [10],
derive the radius–pressure relation: the rings are assumed to be inextensible (so it will change
shape under negative pressure or stress) and the bending
r ¼ rðz; pÞ: ð12Þ
moment M is assumed to be proportional to the cur-
which is used to determine tube radius pointwisely. Eq. vature change:
(12) is equivalent to a stress/strain relation for the tube M ¼ Kp ð1=R  jc Þ; ð13Þ
under positive pressure conditions.
For the post-buckling stage, tube law (11) is no where R is the undeformed tube radius, jc is the de-
longer adequate to determine the tube deformation be- formed ring curvature, and Kp is a stiffness coefficient
cause the tube is no longer nearly axisymmetric. Actu- determined by the wall material and geometry. For a
ally, for a < 1, our experimental measurements provide thin-wall tube, Kp is determined by [9]:
additional information for tube wall deformation under Eh3w
compression (see Fig. 1 for collapsed shape). We use a Kp ¼ ; ð14Þ
12ð1  m2 Þr3
thin-shell model [10] to determine wall deformation
under compressed or collapsed conditions. Motivated by where E is the Young’s modulus of the tube determined
Peskin’s fiber idea [33], we discretize the tube wall by a from experiment [44], hw is wall thickness, m is Poisson
 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665 1655

ratio, r is the mean tube radius. Following the derivation Remark 1. Due to lack of information about the con-
in Flaherty’s paper [10] with some adjustments, use the stitutive laws of arteries under compression, the thin-
natural coordinates and neglecting the inertia force shell model and tube law (11) combined together
(<1% of tension forces) and the circumferential shear provide a direct method to determine tube deformation
stress from the fluid, the equilibrium equations for the under collapsed condition. While the wall model is very
tube wall are given by simplified, the wall deformation obtained is controlled
directly by experimental data, and the results obtained
oTc
þ jc N ¼ 0; ð15Þ provide a good approximation for wall and flow be-
os1
haviors in the physiological artery collapse process.
Better wall models and experimental data for tube ma-
oN terial under compression are needed to get detailed
 jc Tc ¼ ðp þ Kp TL ðjL  jL0 ÞÞ cos b; ð16Þ
os1 stress/strain distributions in the tube wall which cannot
be obtained from the current model [43,44].
oM
þ N ¼ 0; ð17Þ
os1
4. The numerical method
oTL
¼ s; ð18Þ
os2 A numerical method using the conventional ALE-
based staggered GFD over an irregular grid with up-
where Tc and N are the tangential and outward normal wind differencing and an incremental boundary iteration
components of the resultant stress on the z-rings re- technique for FSI is introduced to solve the model
spectively, TL is the longitudinal tension along the h-lines [19,22,29,31,37]. Use of ALE formulation enables us to
and s is the fluid shear stress, jL and jL0 are curvatures choose the mesh properly to avoid large mesh distortion
of the deformed and undeformed h-lines, b is the angle and eliminates the needs of interpolating the flow vari-
between the normal directions of the ring and the tube ables for previous steps at the new grids. GFD makes it
wall surface, s1 and s2 are the arc lengths along the z- possible for us to use finer mesh near the tube wall and
rings and h-lines respectively. The axial tension and in the stenotic region to handle the critical flow con-
strain relationship is measured experimentally, ditions involved in the problem (Fig. 2). The incre-
L ¼ L ðTL Þ; ð19Þ mental boundary iteration method is essentially a
relaxation technique which is used to handle ‘‘pressure
which is needed to determine the axial displacement of over-shooting’’ and ‘‘boundary over-shooting’’ [2,37]
the grid points when TL is obtained from (18). The inlet and improve on the regular boundary iteration method
and outlet of the tube are not allowed to move in the to get convergence for the FSI model with large strain
axial direction to prevent the entire tube from being and large deformations. Details of the numerical method
pushed away by the flow. This completes the FSI model. are outlined below.

Fig. 2. The nonuniform mesh used in the computation.

1656 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665

4.1. The generalized finite difference concept

The advantage of the GFD method is that GFD

schemes can be derived using arbitrary irregular grids
[29]. To derive the second order GFD schemes for the
derivatives fr ; fh ; fz ; frr ; . . . ; fzz at a given point x0 , let
xi ðri ; hi ; zi Þ, ði ¼ 1; . . . ; n; n P 9Þ be n neighboring points
of x0 and,
ai ¼ ri  r0 ; bi ¼ hi  h0 ; ci ¼ zi  z0 ;
qi ¼ ½a2i þ b2i þ c2i 1=2 ; fi ¼ f ðri ; hi ; zi Þ;

use the Taylor expansion of f at x0 and omitting higher

order terms, we have,
fi ¼ f0 þ ai fr0 þ bi fh0 þ ci fz0 þ 12ða2i frr0 þ b2i fhh0
þ c2i fzz0 þ 2ai bi frh0 þ 2bi ci fhz0 þ 2ai ci frz0 Þ: ð20Þ

The GFD schemes for the nine derivatives can be

obtained from these equations using least-squares ap-
proximations with proper weight functions. Other GFD
schemes can be derived similarly. Because there are
virtually no restrictions on the selection of the points xi
and the GFD schemes can be obtained automatically in
the computer program each time the domain and mesh
are adjusted, the GFD method is a suitable tool to
handle the irregular geometry with nonuniform grids
and a free moving boundary which requires frequent
Fig. 3. The staggered grids and numbering of the neighboring
remeshing of the domain.
points.

4.2. Discretization of N–S equations over the irregular

geometry with a nonuniform mesh
can be discretized in the same way. Using vector nota-
tion, we have:
Rewriting the N–S equations into the form (use the u-
equation as an example), X
18 X
18

 u0 þ kiu ui þ k19
u
þ u
k19þi pi ¼ 0;
qðv  vg Þ i¼1 i¼1
qut þ qðu  ug Þ; ; qðw  wg Þ ður ; uh ; uz ÞT
r ðequation of motionÞ ð22Þ
l l
þ pr þ  ; 0; 0;  l;  2 ;  l ður ; uh ; uz ; urr ; uhh ; uzz ÞT
r r X
12 X
12 X
12
q lu 2l kic ui þ c
k12þi vi þ c
k24þi c
wi þ k37 ¼ 0;
 v2  2 þ 2 vh ¼ 0; ð21Þ
r r r i¼1 i¼1 i¼1

ðequation of continuityÞ ð23Þ

for each grid point x0 , using a three-point backward
difference scheme ut ð3uðnþ1Þ  4un þ uðn1Þ Þ=2Dt for where notations kiu ¼ ðkiu ; kiv ; kiw ÞT and kiu ui ¼ ðkiu ui ; kiv vi ;
the t-derivative, ðu; v; wÞ values from last iteration in the kiw wi ÞT were used for simplicity.
convection terms to linearize the equation, the GFD
schemes for the space derivatives with the neighboring
points with staggered grids and upwind techniques (Fig. 4.3. The SIMPLER algorithm
3), and rearranging the terms and omitting the super-
script (n þ 1), we have the discretized equation: Because the fluid–wall interaction model is complex,
we start from a well-tested SIMPLER algorithm [31] to
X
18 X
18
solve (22) and (23). Let the residuals of the equations of
u0 þ kiu ui þ k19
u
þ u
k19þi pi ¼ 0;
motion by um , pm be:
i¼1 i¼1
X
18 X
18
where the nonhomogeneous term k19 u
came from quan- um
0 þ kiu um u
i þ k19 þ
u
k19þi pim ¼ Ru ðum ; pm Þ; ð24Þ
tities considered known at this step. The other equations i¼1 i¼1
 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665 1657

where notations kiu , kiu um

i and Ru were used in a similar 5. Results
way as before. Assuming um þ du and pm þ dp satisfy
(22) and using (24) and (22) becomes, Computations are conducted for various pressure
and stenosis conditions to quantify cyclic wall com-
X
18 X
18
pression and collapse phenomena and related flow be-
ðduÞ0 þ kiu dui þ u
k19þi dpi þ Ru ðum ; pm Þ ¼ 0; haviors. Inlet (upstream) pressure is set to 70–110
i¼1 i¼1
mmHg and 90–150 mmHg respectively representing
normal and high blood pressures. As for the distal
neglecting dui terms leads to:
pressure, Hafner has reported a carotid stump pressure
X
18 between 0 and 25 mmHg for 78 out of 418 patients [17].
ðduÞ0 ¼  u
k19þi dpi  Ru ðum ; pm Þ: ð25Þ So outlet (downstream) pressure is set to 20, 10 or 0
i¼1 mmHg for different cases in our computations and ex-
periments. Cyclic wall compression, tube collapse and
Substituting um þ du into (23) and using (25), dp can be flow characteristics for an 80% stenosis (by diameter)
determined. Then du follows from (25) and u and p can with pin ¼ 90–150 mmHg and pout ¼ 20 mmHg are more
be updated accordingly. critical and results are presented in more details. Results
of other cases are reported as appropriate.
4.4. Incremental boundary iteration method
5.1. Parameters, geometry, mesh and convergence
Boundary iteration methods have become popular
The ranges of parameters and the geometry of the
for solving problems with FSI recently where the fluid
tube used in the computations are chosen to match the
and solid models are solved iteratively until convergence
experimental set-up [23,26]:
is obtained. However, it has been known that the
boundary iteration method may fail to converge if the R0 ¼ 0:4 cm; l ¼ 8 cm ðtube lengthÞ; Z1 ¼ 3:2 cm;
tube wall is considerably compliant [37]. We use an in-
Z2 ¼ 4:8 cm; m ¼ 0:04 cm2 =s; q ¼ 1 g=cm3 ; m ¼ l=q:
cremental iterative method to reduce the ‘‘displacement
over-shooting’’ and improve the convergence. Displace- Reynolds number Re is defined as D  U =m where D is the
ment over-shooting causes velocity over-adjustment at entrance tube diameter (2R0 ) and U is the entrance av-
the boundary which affects the convergence of the fluid erage velocity. The Reynolds number for a 80% stenosis
model. For a given wall adjustment Drðh; zÞ, if the fluid with inlet pressure 100 mmHg and outlet pressure 20
model fails to converge, we reduce Dr to half and try to mmHg is about 300. The units given above are used
solve the fluid model again. This is repeated until con- throughout the paper.
vergence is reached. A similar relaxation technique can Nonuniform meshes (Fig. 2) are used in the compu-
also be used to reduce ‘‘pressure over-shooting’’ which is tation to handle the critical flow conditions involved in
the cause of tube wall over-adjustment. When du and dp the collapse process. We look for solutions which are
are obtained, we update u and p with symmetric with respect to the horizontal plane and the
model is solved over the ðr; h; zÞ-domain ½0; H ½0; p
unew ¼ uold þ xu du; pnew ¼ pold þ xp dp; ð26Þ
½0; l . The step sizes in r- and z-directions are reduced by
fixed ratios towards the wall and the middle of the tube
where xu and xp can be chosen between 0 and 1 to length to get better resolution there. The ratios are
achieve best convergence.
driþ1 dziþ1
qr ¼ ¼ 0:92; qz ¼ ¼ 0:95: ð27Þ
dri dzi
4.5. Numerical method for the wall model
For a ð10r 20h 120z Þ mesh, we have dr1 ¼ 0:0566,
We introduce an iterative method to solve the thin- dr10 ¼ 0:0267 at the inlet of the tube where the step size
shell model. At each boundary iteration, the material at the wall dr10 is less than half of the starting step size
coefficients are determined for each grid point with up- dr1 . For z, we have dz1 ¼ 0:2097 at the inlet, dz60 ¼
dated pressure information using the tube laws. Then 0:01017 at the middle of the tube length which is about
Eqs. (15)–(17) are solved on each ring with updated 1/20 of dz1 . Uniform step size is used for h variable. By
pressure and longitudinal tension to determine the cir- using nonuniform mesh, much better resolution is
cumferential and radial displacements [10]. With these achieved near the wall and stenotic region with fewer
adjustments, (18) is solved to update the longitudinal grid points which lead to considerable savings of mem-
tension which determines the axial displacement of each ory and CPU time.
grid point. Then (15)–(17) are solved again. This is re- Since analytic solutions for flow in a compliant ste-
peated until convergence is reached. notic tube are not available, numerical solutions are
1658 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665

compared with the exact solution for flow in a rigid 5.2. Wall deformation, cyclic wall compression and
straight tube [12], collapse
1 2 Fig. 4 shows tube deformation when the inlet pres-
w¼ ðR  r2 Þpz ; u ¼ v ¼ 0; pz ¼ ðpout  pin Þ=l:
4m 0 sure is at its minimum and maximum respectively. The
ð28Þ tube is collapsed at the distal side of the stenosis when
inlet pressure is at its maximum. The location of the
and Table 1 gives a summary of the errors which show collapse is about one diameter distal to the stenosis,
that the algorithm converges reasonably well. The ac- consistent with experimental observations. The cyclic
curacy of the numerical solutions may be better than compression repeats as the inlet pressure changes peri-
what Table 1 shows because the exact solution is as- odically.
sumed to be z-independent while the actual tube length Fig. 5 gives ultrasound images of the vertical views of
is always finite and the numerical solutions are slightly the tube with pressure conditions set at pin ¼ 70–130
z-dependent. mmHg and pout-avg ¼ 0 mmHg. The tube is fully col-
To check the accuracy and convergence of the algo- lapsed when pin is at its maximum. pout-avg was set to 0
rithm for the complaint model with stenosis, three me- mmHg in the experiment so that full tube collapse could
shes were tested and the results are given in Table 2. As be observed. pout ¼ 20 mmHg is more physiologically
the mesh gets finer, errors decrease roughly in a linear relevant and is used in our numerical simulation as the
fashion indicating that the method is of first-order ac- main case.
curacy. Mesh ð10r 20h 120z Þ is used in our general
computations in this paper. The tolerance for the N–S
solver (SIMPLER iteration) is set to 1:0 107 , i.e. 5.3. Negative pressure in the stenotic tube
solutions of (22) and (23) are considered obtained if the
relative errors (corrections) of velocity and pressure are Pressure imposed at inlet is given by Fig. 6(a). Since
less than the specified tolerance (TOL1). The tolerance pressure is directly related to tube compression and
for the boundary iteration is set to 1:0 104 (TOL2), collapse, minimum pressure for the entire fluid domain
i.e., the solution for the tube wall, flow velocity and as a function of time is plotted in Fig. 6(b). The slight
pressure were considered ‘‘converged’’ for a given time waviness is due to the fact that minimum pressure may
step if the relative errors became less than the tolerance occur at different neighboring points for different times.
specified. Periodic solutions were considered obtained Transmural pressure at the tube wall at t ¼ 0:5 is plotted
when the solutions started to repeat itself within 1% in Fig. 6(c). Transmural pressure at the tube wall at
tolerance (TOL3). Our calculations indicate that three h ¼ 90° changing with time is plotted by Fig. 7. Pressure
periods are needed for the solutions to become periodic. distributions inside the tube were reported in our pre-

Table 1
Comparison of 3D numerical solutions with exact solution for flow in a rigid straight tubea
r h z en ðuÞ en ðvÞ en ðwÞ en ðpÞ
20 16 60 5:7 107 4:4 109 0.098 3:7 106
30 20 80 3:1 107 4:2 109 0.031 2:3 106
40 24 100 1:4 107 3:8 109 0.018 1:1 106
a
pin ¼ 100 mmHg, p2 ¼ 99:8 mmHg, umax ðexactÞ ¼ 32:98 cm/s, dt ¼ 0:005, time step computed ¼ 1600. Relative errors are defined
as en ðf Þ ¼ kfn  fexact k2 =kfn k2 , n ¼ time step.

Table 2
Order of accuracy of the numerical methoda
r h z ðdr; dh; dzÞend en ðuÞ en ðvÞ en ðwÞ en ðpÞ en ðHÞ
8 16 80 (0.0367, 0.196, 3.918E3 1.283E10 1.369E3 9.922E6 4.683E6
0.0310)
10 20 100 (0.0267, 0.157, 2.398E3 1.343E11 0.874E3 5.981E6 2.584E6
0.0175)
12 24 120 (0.0202, 0.131, 1.527E3 2.473E12 0.538E3 3.791E6 1.576E6
0.0102)
a
p1 ¼ 130 mmHg, p2 ¼ 40 mmHg, S0 ¼ 80%, dt ¼ 0:001, time steps computed ¼ 800. Mesh step size reduction ratios are 0.92 for r
and 0.95 for z. Step sizes given in the table are the minimum r-step size at the inlet of the tube and the minimum z-step at the middle of
the tube length.
 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665 1659

Fig. 4. Numerical results for tube wall deformation under maximum and minimum inlet pressure showing cyclic wall compression.
pin ¼ 90–150 mmHg, pout ¼ 20 mmHg, S0 ¼ 80%, portion of the tube plotted: z ¼ 1:5–6.5 cm.

Fig. 5. Ultrasound image of cross-section during flow collapse. pin ¼ 70–130 mmHg, pout-avg ¼ 0 mmHg.

vious papers [45,46] and it was found that minimum inlet under 100 mmHg). However, negative pressure
pressure occurs at the throat of stenosis. Because the may occur at and distal to the stenosis if the stenosis is
prescribed inlet and outlet pressures are positive, the severe enough and pressure drop is large enough. Min-
pressure inside the tube is, in general, positive and imum pressure occurs at the throat of the stenosis as
the tube is inflated almost everywhere (about 60% at the expected. Fig. 6(c) and Fig. 7 show clearly that the
1660 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665

reaches its minimum at the throat, then recovers to 20

mmHg, the specified downstream pressure. Minimum
pressure decreases as the prescribed inlet pressure in-
creases, reaches its lowest value ()12.5 mmHg) when
inlet pressure reaches its maximum 150 mmHg, then
recovers to about 2.5 mmHg, corresponding to mini-
mum inlet pressure 90 mmHg. While the minimum
pressure occurs at the throat, the tube wall does not
collapse there because the thick-wall stenosis is very stiff.

5.4. Maximum velocity and flow recirculation

Fig. 8 gives plots of the maximum axial velocity as a

function of time and the axial velocity profiles at several
horizontal cross-sections of the tube with S0 ¼ 80% at
t ¼ 0:5 s when pin ¼ 150 mmHg, pout ¼ 20 mmHg. Dif-
ferent scales are used for the profiles at different loca-
tions to show details of the flow field. The recirculation
region and the flow in the negative axial direction are
clearly visible.

5.5. Maximum shear stress

Fig. 6. Transmural pressure changes drastically across the Shear stresses at the tube wall corresponding to
stenosis: (a) the inlet pressure; (b) minima of the corresponding maximum and minimum inlet pressure are plotted in
pressure field; (c) transmural pressure at t ¼ 0:5 s along the tube Fig. 9 for severity S0 ¼ 80% with inlet pressure pin ¼
wall. pin ¼ 90–150 mmHg, pout ¼ 20 mmHg, S0 ¼ 80%. 150, 90 mmHg and outlet pressure pout ¼ 20 mmHg. The
peak shear stress reaches 3270 (dyn/cm2 ) which may
pressure does not change much on the upstream side. It cause damage to the vessel wall and platelet activation
decreases drastically across the throat of the stenosis, [28].

Fig. 7. Transmural pressure along the tube wall pðz; h ¼ 90°; tÞ. pin ¼ 90–150 mmHg, pout ¼ 20 mmHg, S0 ¼ 80%.
 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665 1661

Fig. 8. Behavior of velocity under unsteady pressure and across the stenosis. (a) Maximum axial velocity occurs at the throat of
stenosis (centerline) and changes with in-let pressure accordingly; (b) velocity profiles at different axial positions, horizontal cross-
section. Different scales are used at different z-locations to show details. pin ¼ 90–150 mmHg, pout ¼ 20 mmHg, S0 ¼ 80%.

Fig. 10. Comparison of tube radius between numerical and

experimental results at cross-section z ¼ 2:0 cm. S0 ¼ 80%,
pin ¼ 70–130 mmHg, pout ¼ 60–80 mmHg.

Fig. 9. Shear stress distribution along h ¼ 90° line under The inlet pressure was set to 70–130 mmHg, and outlet
maximum and minimum inlet pressure. pressure was set to 60–80 mmHg (average 70 mmHg) in
the experiment. The radius was measured at z ¼ 2 cm.
5.6. Comparison between numerical and experimental The curves show that there is a reasonable agreement
results between computational and experimental results.
Fig. 11 compares computational and experimental
Fig. 10 plots computational and experimental tube flow rates under unsteady conditions. Pressure condi-
radius variations under pulsatile pressure conditions. tions imposed at the inlet and outlet of the tube in the
1662 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665

rates at the outlet. While there is a rough agreement, we

notice that there is a clear phase shift between the
computational and experimental data. There is also a
phase shift between the experimental flow rate and the
imposed pressure condition. A smaller phase shift of
tube radius between experimental and numerical data
can also be seen in Fig. 10. This may be an indication of
viscoelasticity of the tube. A viscoelastic model is cur-
rently being developed to improve the elastic model [42].
The wall mechanical properties will also need to be
measured under unsteady flow conditions.
Fig. 12 shows pressure–area relationship (tube law)
calculated numerically under ‘‘no flow’’ conditions and
compared with experimental data. The agreement is
good for p > 0, remains reasonable for p > 20 mmHg,
and becomes poor for p < 20 mmHg. This indicates
that the wall model provides reasonable approximation
when tube collapse is minor. The approximation be-
comes less accurate as tube collapse becomes more se-
vere and caution must be taken when interpreting the
Fig. 11. Comparison of flow rates between numerical and ex-
numerical results.
perimental results: flow rate (a) pressure drop curve; (b) flow
rate curves. S0 ¼ 80%, pin ¼ 70–130 mmHg, pout-avg ¼ 10 mmHg
(0–20 mmHg). 6. Discussion

6.1. Comparison with previous computational models

experiment are 70–130 mmHg and 0–20 mmHg (average
10 mmHg) respectively. The outlet pressure was chang- A review of the previous numerical results from 1D
ing slightly because it was not possible to keep it con- models can be found from Downing and Ku [9]. Since
stant in the experiment. Fig. 11(a) plots the pressure the 1D models used only average pressure and axial
drop which is more relevant to flow rate changes. Fig. velocity, the complex pressure distribution near stenosis,
11(b) gives the computational and experimental flow axial and radial wall deformation, flow separation and

Fig. 12. Comparison of numerical pressure–area relationship (tube law) with experimental data. Calculations were conducted under
no-flow condition with pin ¼ pout ¼ 50–100 mmHg, S0 ¼ 80%.
 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665 1663

Table 3
Comparison of critical flow characteristics and wall deformation from three stenosis models showing severe stenosis has considerable
effects on wall compression and flow behavior
3D 80% stenosis (current 2D 80% attached stenosis [3] 3D 78% stenosis [43]
model)
Tube radius (cm) 0.4 0.2 0.4
Wall thickness (cm) 0.1 0.016 0.1
Axial stretch 36.5% 50% 2%
Radial expansion (under 100 60% 22% 12%
mmHg, with stretch)
pin (mmHg) 90–150 80–120 100 (steady)
pout (mmHg) 20 20 (min) 20
umax (cm/s) 583.5 650 594
Re at inlet 325 197 315
pmin (mmHg) )12.5 (at wall) )39 (centerline) )52.5 (at wall)
smax (dyn/cm2 ) 2233–3270 (throat) 7740 3127
Radius reduction )0.22 (distal) )0.007 (throat) )0.002 (throat)
Cyclic tube collapse Yes Unable to simulate Unable to simulate

shear stress information could not be obtained. Existing of atherosclerotic plaque fracture and subsequent throm-
2D and 3D models [43,45,46] were limited by the way bosis or distal embolization. Negative flow pressure is
tube law was implemented: tube cross-section area re- found in the stenotic region which is closely related to
duction under collapsed conditions was incorrectly im- compressive stress in the tube wall. In fact, maximum
plemented as tube radius reduction in their derivation of compressive stress is found in the plaque near the throat
the stress–strain relationship for the tube wall. Since of the stenosis using a thick-wall model [43]. High shear
tube geometry is one of the most important factors af- stress in the order of 2000–3000 dyn/cm2 at the stenosis
fecting flow and wall behaviors, and the thin-shell theory may cause damage to the endothelial layer of the vessel
provides a better interpretation of the tube law under wall and platelet aggregation [28]. The flow recirculation
both expansion and collapsed conditions, this new 3D region provides an environment with small and alter-
model provides more accurate information about wall nating shear stresses and prolonged cell residence time
deformation and collapse, shear stresses, flow velocity favorable for cell adhesion and thrombus formation
and pressure fields, and gives more accurate predictions [6,36]. It was also noticed both experimentally and
about collapse conditions. Comparison of the main re- computationally that if the upstream pressure was high
sults of three models is given by Table 3. enough and downstream pressure low enough, the tube
Because of the thin-wall assumption in the model, it may collapse and remain collapsed even when upstream
was not possible to obtain detailed stress–strain distri- pressure drops again. This means the flow would remain
butions in the tube wall. Tube compressive stress can choked (actually, fluttering will occur) and the patient
only be inferred from wall compression and collapse may have noticeable clinical symptoms.
from this model, not direct stress calculation. A thick-
wall FSI model is being developed to improve on this 6.3. Effect of stenosis severity and pressure conditions
model. Results from the thin-wall model can be used as
initial approximations. Since pressure decreases considerably when passing a
Phase delays between imposed pressure and wall de- severe stenosis, the effect of the stenosis severity on flow
formation and flow rates were noticed in the experi- and pressure fields become much more noticeable when
mental data which may be caused by viscoelasticity of the comparison is made with comparable flow rates. An
tube wall. The extend of the viscoelastic effects needs to 80% stenosis and a 50% stenosis are compared with pin
be investigated by a viscoelastic model. Some prelimi- set to 120  30 mmHg for both stenoses, pout set to 20
nary results have been obtained in this regard [42] and we mmHg for the 80% stenosis and 117  30 mmHg for the
are currently working on a 3D viscoelastic model. 50% stenosis respectively. The average flow rate is
11.0331 ml/s for the 80% stenosis and 11.0354 ml/s for
6.2. Physiological significance of the findings the 50% stenosis. Comparison of the two cases is given
in Table 4. While minimum pressure ()12.5 mmHg) and
Arteries are made to sustain positive pressure and wall collapse were observed for the 80% stenosis, pres-
expansions. Compressive stress and cyclic wall bending sure decreased less than 2 mmHg when passing the 50%
and compression may be important in the development stenosis and no negative pressure and wall compression
1664 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665

Table 4
Comparison of wall stress and critical flow characteristics from two stenoses under two pressure conditions showing (a) severe stenosis
has considerable effects on wall compression and flow behaviors; (b) high pressure causes more critical flow conditions related arterial
diseases
Case 1 (high pressure) Case 2 (mild stenosis) Case 3 (normal pressure)
Stenosis severity 80% 50% 80%
pin (mmHg) 90–150 90–150 70–110
pout (mmHg) 20 87–147 20
umax (cm/s) 583.5 104.8 486.1
pmin (mmHg) )12.5 85.4–144.3 )0.41
smax (dyn/cm2 ) 3270 149.4 2522
Hcmin )0.22 0.000186 )0.0025

were observed. Maximum velocity and shear stress for fully simulate the collapse process and make more ac-
80% stenosis are 583.5 cm/s and 3270 dyn/cm2 , enough curate physiologically-relevant predictions.
to be of physiological significance, while they are only
104.8 cm/s and 149.4 dyn/cm2 respectively for the 50%
stenosis which do not cause any clinical symptoms. Acknowledgements
These results are consistent with clinical observations
[16]. This research was supported in part by a grant
To see the influence of imposed pulsatile pressure on from the Whitaker Foundation and NSF grant DMS-
the flow and wall behaviors, pin was set to 120  30 0072873.
mmHg and 90  20 mmHg, representing high and nor-
mal pressures. pout is set to 20 mmHg. Stenosis severity is
still 80% by diameter. While high velocity and high shear
References
stress are observed for both cases, cyclic wall compres-
sion and collapse are observed only for the high pressure
[1] Aoki T, Ku DN. Collapse of diseases arteries with
case. Comparison of the two cases is also included in eccentric cross-section. J Biomech Eng 1993;26(2):133–42.
Table 4. More in vivo experimental validations are [2] Bathe KJ. Finite element procedures. Prentice Hall 1996.
needed before these results can become applicable in [3] Bathe M, Kamm RD. A fluid–structure interaction finite
clinic applications. element analysis of pulsatile blood flow through a compli-
ant stenotic artery. J Biomech Eng 1999;121:361–9.
[4] Biz S. Flow in collapsible stenoses: an experimental study.
M.S. Thesis, Georgia Institute of Technology, 1993.
7. Conclusion [5] Burke AP, Farb A, Malcom GT, Liang YH, Smialek JE,
Virmani R. Plaque rupture and sudden death related to
An experiment-based 3D computational model with exertion in men with coronary artery disease. JAMA
FSI was introduced and solved using a GFD method to 1999;281(10):921–6.
[6] Cao J, Rittgers SE. Particle motion within in vitro models
simulate blood flow in stenotic collapsible carotid ar-
of stenosel internal carotid and left anterior descending
teries. Cyclic wall compression and collapse were ob-
coronary arteries. Ann Biomed Eng 1998;26(2):190–9.
served under physiological pressure conditions. Stenosis [7] Constantinides PA. Plaque hemorrhages their genesis and
severity and pressure conditions are found to be the their role in supra-plaque thrombosis and atherogenesis.
dominating factors affecting wall compression and col- In: Pathobiology of the human atherosclerotic plaque.
lapse and related flow conditions. The unstructured New York: Springer-Verlag; 1990. p. 393–411.
nature of the GFD method made 3D mesh generation [8] Davies MJ, Thomas AC. Plaque fissuring-the cause of
and code development possible without much compli- acute myocardial infarction, sudden ischemic death, and
cations. The incremental boundary iteration technique crecendo angina. Brit Heart J 1985;53:363–73.
used in this paper can be applied to a wide range of [9] Downing JM, Ku DN. Effects of frictional losses and
problems with FSI involving large deformations. Criti- pulsatile flow on the collapse of stenotic arteries. J Biomech
Eng 1997;119:317–24.
cal flow conditions such as negative transmural pressure,
[10] Flaherty JE, Keller JB, Rubinow SI. Post buckling behav-
high shear stress and large wall deformation caused by ior of elastic tubes and rings with opposite in contact.
severe stenosis under physiological pressure conditions SIAM J Appl Math 1972;23:446–55.
were quantified which may be directly related to artery [11] Friedman MH. Arteriosclerosis research using vascular
collapse and plaque cap rupture. Further investigations flow models: from 2D branches to compliant replicas.
using viscoelastic 3D thick-wall models are needed to J Biomech Eng 1993;115:595–601.
 D. Tang et al. / Computers and Structures 80 (2002) 1651–1665 1665

[12] Fung YC. Biodynamics, circulation. New York: Springer- [32] Kamm RD, Pedley TJ. Flow in collapsible tubes: a brief
Verlag; 1984. review. J Biomech Eng 1989;111:177–9.
[13] Fung YC. In: Biomechanics, mechanical properties of [33] Peskin CS. Mathematical aspects of heart physiology.
living tissues. 2nd ed. New York: Springer-Verlag; 1993. Lecture notes of Courant Institute of Mathematical
[14] Fung YC, Liu SQ. Determination of the mechanical Sciences, New York 1975.
properties of the different layers of blood vessels in vivo. [34] Powell BE. Experimental measurements of flow through
Proc Natl Acad Sci USA 1995;92:2169–73. stenotic collapsible tubes. M.S. Thesis, Georgia Institute of
[15] Fuster V, Stein B, Ambrose JA, Badimon L, Badimon JJ, Technology, 1991.
Chesebro JH. Atherosclerotic plaque rupture and throm- [35] Ramaswamy B, Kawahara M. Arbitrary Lagrangian–
bosis. Circulation, Supplement II 1990;82(3):II-47–59. Eulerian finite element method for unsteady, convective,
[16] Giddens DP, Zarins CK, Glagov S. The role of fluid incompressible viscous free surface fluid flow. Int J Numer
mechanics in the localization and detection of atheroscle- Meth Fluids 1997;7:1053–75.
rosis. J Biomech Eng 1993;115:588–94. [36] Rittgers SE, Yu YH, Strony JT. Thrombus formation in
[17] Hafner CD. Minimizing the risks of carotid endarterec- moderate coronary stenosis using 2D-LDA. In: Proceed-
tomy. J Vasc Surg 1984:392–7. ings of the Third World Congress of Biomechanics, 1998.
[18] Hayashi K. Experimental approaches on measuring the p. 201.
mechanical properties and constitutive laws of arterial [37] Rugonyi S, Bathe KJ. On finite element analysis of fluid
walls. J Biomech Eng 1993;115:481–8. flows fully coupled with structural interactions. CMES
[19] Hughes T, Liu WK. Zimmermann, Lagrangian–Eulerian 2001;2(2):195–212.
finite element formulation for incompressible viscous flows. [38] Shapiro AH. Steady flow in collapsible tubes. J Biomech
Comput Meth Appl Mech Eng 1981;29:329–49. Eng 1977;99:126–47.
[20] Judd RM, Mates ME. Flow through a stenosis in a [39] Siebes M, Campbell CS, D’Argenio DZ. Fluid dynamics
compliant tube. In: Liepsch D, editor. Proc Second Intl of a partially collapsible stenosis in a flow model of the
Symp on Biofluid Mechanics and Biorheology, Munich, coronary circulation. J Biomech Eng 1996;118:489–97.
Germany, 1989. p. 417–23. [40] Simon BR, Kaufmann MV, McAfee MA, Baldwin AL.
[21] Kamm RD, Shapiro AH. Unsteady flow in a collapsible Finite element models for arterial wall mechanics. J
tube subjected to external pressure or body force. J Fluid Biomech Eng 1993;115:489–96.
Mech 1979;95:1–78. [41] Stergiopulos N, Moore Jr JE, Strassle A, Ku DN, Meister
[22] Kleiber M. Handbook of computational solid mechanics. JJ. Steady flow test and demonstration of collapse on
New York: Springer-Verlag; 1998. models of compliant axisymmetric stenoses. Adv Bioeng
[23] Kobayashi S, Mijovic B, Tang D, Ku DN. Collapse in 1993;ASME BED-26:455–8.
high-grade stenosis during pulsatile flow experiments. In: [42] Tang D, Chen XK, Yang C, Kobayashi S, Ku DN. Effects
Proceedings of the Third World Congress of Biomechanics, of dynamic wall properties on blood flow in stenotic
1998. p. 11. arteries. J Med Biomech 2000;15(2):77.
[24] Kobayashi S, Tang D, Ku DN. Pulsatile flow through a [43] Tang D, Yang C, Huang Y, Ku DN. Wall stress and strain
stenotic collapsible tube. In: Proc of the 76th JSME Fall analysis using a 3D thick-wall model with fluid–structure
Annu Meeting, 1998. p. 265–6. interactions for blood flow in carotid arteries with stenoses.
[25] Ku DN. Blood flow in arteries. Annu Rev Fluid Mech Comput Struct 1999;72:341–56.
1997;29:399–434. [44] Tang D, Yang C, Kobayashi S, Ku DN. Steady flow and
[26] Ku DN, Kobayashi S, Wootton DM, Tang D. Compres- wall compression in stenotic arteries: a 3D thick-wall
sion from dynamic pressure conditions in models of model with fluid–wall interactions. J Biomech Eng 2001;
arterial disease. Ann Biomed Eng 1997;25(supp. 1):S-22. 123:548–57.
[27] Ku DN, Zeigler MN, Downing JM. One-dimensional [45] Tang D, Yang C, Ku DN. A 3D thin-wall model with
steady inviscid flow through a stenotic collapsible tube. J flow–structure interactions for blood flow in carotid arteries
Biomech Eng 1990;112:444–50. with symmetric and asymmetric stenoses. Comput Struct
[28] Lee RT, Richardson SG, Loree HM, et al. Prediction of 1999;72:357–77.
mechanical properties of human atherosclerotic tissue by [46] Tang D, Yang J, Yang C, Ku DN. A nonlinear axisym-
high-frequency intravascular ultrasound imaging. Arterio- metric model with fluid–wall interactions for viscous flows
sclerosis 1992:1–5. in stenotic elastic tubes. J Biomech Eng 1999;121:494–
[29] Liszka T, Orkisz J. The finite difference method at 501.
arbitrary irregular grids and its application in applied [47] Wootton DM, Ku DN. Fluid mechanics of vascular
mechanics. Comput Struct 1980;11:83–95. systems, diseases, and thrombosis. Annu Rev Biomed
[30] McCord BM, Ku DN. Mechanical rupture of the athero- Eng 1999;1:299–329.
sclerostic plaque fibrous cap. In: Proceedings of 1993 [48] Yamaguchi T, Nakayama T, Kobayashi T. Computa-
Bioengineering Conference, BED-vol. 24. 1993. p. 324–7. tions of the wall mechanical response under unsteady
[31] Patankar SV. Numerical heat transfer and fluid flow. New flows in arterial diseases. Adv Bioeng 1996;BED-33:
York: Taylor & Francis Publishers; 1980. 369–70.