Wall Shear Stress:
Theoretical Considerations and
Methods of Measurement
Demosthenes Katritsis, Lambros Kaiktsis, Andreas Chaniotis, John Pantos,
Efstathios P. Efstathopoulos, and Vasilios Marmarelis
In arterial blood flow, the wall shear stress
expresses the force per unit area exerted by the
wall on the fluid in a direction on the local tangent
plane. There is substantial evidence that the wall
shear stress induced by the pulsatile blood flow in
the arterial system affects the atherogenic process.
It is now widely accepted that the vessel segments
that appear to be at the highest risk for develop-
ment of atherosclerosis are those with low wall
shear stress or oscillating wall shear stress. The
purpose of this article is to define wall shear stress,
to introduce relevant concepts of fluid mechanics
to nonexperts, and to critically review the various
methods that have been used for the assessment
of wall shear stress in animal and human blood
circulation, paying special attention to the case of
coronary arteries.
n 2007 Elsevier Inc. All rights reserved.
W all shear stress (WSS) expresses the force
per unit area exerted by a solid boundary
on a fluid in motion (and vice-versa) in a direction
on the local tangent plane. Currently, issues
related to WSS distribution in arterial flow are
receiving attention because of emerging evidence
that it is associated with vascular disease and,
specifically, atherosclerosis. It is now well estab-
lished that WSS plays a significant role in
endothelial homeostasis and the focal distribution
of atherosclerotic lesions.1,2 Several studies on the
distribution of atherosclerotic plaques in human
arterial systems have shown that atherosclerosis
occurs predominantly at certain locations of the
vascular tree where the arteries have relative
complex geometry that results in bdisturbedQ
blood flow behavior.3,4 These observations have
led to the notion that local hemodynamic factors,
such as WSS, may play a role in the initiation and,
Progress in Cardiovascular Diseases, Vol. 49, No. 5 (March/April), 2007: pp 307-329
perhaps, progression of the disease.3 WSS values,
therefore, at various sections of the vascular tree,
may be of clinical significance and the methods
employed for their measurement are critical for
the validity of derived results.
A complete description of hemodynamics
within a particular vessel or lesion requires
knowledge of the pattern of blood velocities
within the flow.5,6 The latter depends on the
geometry and mechanical properties of the
vascular wall, an overall pressure difference,
and the rheological characteristics of blood
(viscosity, density). Knowledge of the pattern
of velocities near the vessel wall allows evalua-
tion of WSS values and thus may provide
indications for estimating the risk for develop-
ment and progression of atherosclerosis, rupture
of atherosclerotic plaques, and hemorrhage
associated with cerebral aneurysms and arterio-
venous malformations. At the earlier stages of
research in this area, contradictory data were
reported that had associated both high and low
WSS values with atherosclerosis.7,8 It is now
widely accepted that the vessel segments that
appear to be at the highest risk for development
From the Department of Cardiology, Athens Euroclinic,
Athens, Greece, Department of Naval Architecture and
Marine Engineering, National Technical University of
Athens, Greece, Medical and Radiation Physics, 2nd
Department of Radiology, Athens University Medical
School, Greece and Department of Biomedical Engineer-
ing, University of Southern California, LA.
Address reprint requests to Demosthenes Katritsis, MD,
PhD, FRCP, Department of Cardiology, Athens Euroclinic,
9 Athanassiadou Str., 115 21 Athens, Greece.
E-mail: dkatritsis@euroclinic.gr
0033-0620/$ - see front matter
n 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.pcad.2006.11.001
307
308
of atherosclerosis are those with low or oscillat-
ing WSS values.9
The purpose of this article is to introduce
relevant concepts of fluid mechanics to nonex-
perts, to define WSS, and to critically review the
various methods that have been used for the
assessment of WSS in animal and human blood
circulation, giving special attention to the case of
coronary arteries.
Definition of Wall Shear Stress
In this section, we introduce the concept of WSS.
For simplicity, we consider the case of flow in a
cylindrical tube. The reader is referred to
Appendix A for a more general presentation of
relevant concepts of fluid mechanics. When a
fluid is flowing inside a straight tube, the velocity
of flow is not equal at all points in the tube.10
The velocity is highest at the center of the tube
and drops at points toward the tube wall. This
velocity gradient (ie, the pattern of velocity
changes between adjacent points in space) that
exists inside the tube is due to frictional forces
that are exerted between the adjacent layers of
the flowing fluid and between the fluid and the
walls of the tube.11 The frictional forces arise
from the viscous properties of the fluid, as
defined by the viscosity of the fluid. The
viscosity of the fluid expresses its resistance to
flow, and it is low if a trivial force on a fluid layer
produces a high velocity of that layer relative to
an adjacent layer and vice versa11,12 (for further
details, please see Appendix A).
Wall shear stress (WSS) is defined as the
tangential force per unit area that is exerted by
the flowing fluid on the surface of the conduit
tube. The magnitude of WSS is proportional to the
velocity gradient near the tube wall, that is, how
fast the flow velocity increases when moving from
a point on the tube wall to an immediately
adjacent point in a direction normal (ie, perpen-
dicular) to the tube wall toward the center of the
tube. Evidently, low WSS values are associated
with low local velocities and, thus, long residence
times of fluid in the near-wall region. This
velocity gradient near the wall is termed wall
shear rate. The WSS value s w is defined as
proportional to the wall shear rate11,13,14:
sw ¼ ld ċc w ð1Þ
KATRITSIS ET AL
where l is the dynamic viscosity of the fluid and ġw
is the wall shear rate. Shear rate is measured in
reciprocal seconds (s1), shear stress is measured in
pressure units, dyne/cm2 or Pa, ie N/m2 (1 N/m2 =
10 dyne/cm2), and dynamic viscosity is measured
in poise (1 poise = 1 dyne s/cm2) or in Pad s. In the
case of blood circulation, the flow of blood induces
a bdragQ between the outermost layer of blood and
the inner layer of the vascular wall in contact with
blood, that is, the vascular endothelium.9 It has
been reported that this vascular WSS can affect both
the functional and the structural integrity of the
endothelial cells.13
Eq. 1 implies that dynamic viscosity is an
inherent physical property of the fluid, showing
how easily the fluid is sliding.13 It also implies
that the velocity gradient at the vessel wall is a
linear function of the applied shear stress and
vice versa. Water and many other liquids exhibit
such behavior, and they are called Newtonian or
ideal fluids.13,14 A Newtonian fluid has, by
definition, constant dynamic viscosity at all rates
of shear.14 When the dynamic viscosity depends
on the shear rate, the fluid is termed non-
Newtonian. The latter often occurs when the
fluid contains sizable particles, as in the case of
blood, which is essentially a suspension of
erythrocytes, leukocytes, and platelets in blood
plasma. Blood plasma itself is a suspension of
molecular organic/inorganic substances and
macromolecules (fibrinogen, immunoglobulin,
albumins, and lipoproteins) in an electrolyte
solution and has been considered to behave
approximately as a Newtonian fluid.14 Fibrino-
gen is probably the greatest contributor to
plasma viscosity because of its distinctly asym-
metric molecular configuration.13,15,16 Fibrino-
gen plasma level has a concentration-dependent
effect on red blood cell (RBC) aggregation that is
an important component of whole blood viscos-
ity, especially at low shear rates.17
However, whole blood should be viewed as a
non-Newtonian fluid because it shows shear-
dependent viscous properties of 2 kinds: (a) the
apparent viscosity increases at low shear rates—
blow shearQ effect, and (b) in small vessels, the
apparent viscosity is smaller (at higher shear
rates) than it is in larger vessels—bhigh shearQ or
bsmall tubeQ effect.14 This shear-dependent vis-
cosity is due to the presence of sizable particles
in the plasma (see below). The single most
WALL SHEAR STRESS 309
important determinant of whole blood shear– more complicated and may require the eventual
dependent viscosity is the RBC content in the use of non-Newtonian models.
blood, which is described by the Hct. Nonethe-
less, the far greater complexity of analyzing the Human Blood as a Non-Newtonian Fluid
behavior of a non-Newtonian fluid has forced
As with all fluids, the (dynamic) viscosity of
many computational studies of blood flow to
consider the whole blood as an approximate blood depends on its composition. Blood is
composed of plasma and several kinds of formed
Newtonian fluid under specific conditions.
elements (corpuscles).19,20 The plasma consists
The flow under steady (nonpulsatile) condi-
of approximately 91% water by weight, 7%
tions of a Newtonian fluid in a cylindrical tube
plasma proteins, and 2% inorganic and organic
with rigid walls is fully developed into a
substances.14,20 The formed elements consist of
parabolic velocity profile (ie, the velocity distri-
approximately 96% RBCs (erythrocytes), 3%
bution over a cross-section normal to the flow
white blood cells (leukocytes), and 1% platelets.
direction) where the velocity is highest at the
center of the tube and zero at the tube walls. This The formed elements are suspended normally in
the fluid medium of the blood (plasma). The
type of flow is usually referred to as Poiseuille
percentage of blood volume occupied by red
flow. In Poiseuille flow, the WSS can be
cells is the Hct, its normal range for a human
calculated analytically by the Hagen-Poiseuille
male being Hct=40-52%.
formula13,14 (please see Appendix B):
Although pure blood plasma can be consid-
4lQ ered as a Newtonian fluid with a dynamic
sw ¼ ð2Þ
pR3 viscosity of 0.0015 Pad s (at 358C), whole blood
viscosity demonstrates a rather complicated
where Q is the volume flow-rate, l is the dynamic behavior: it is a function of the Hct, the velocity
viscosity of the fluid and R is the inner radius of the profile (shear rate), the size (diameter) of the
conduit cylindrical tube. Another useful formula blood vessel, and blood temperature.12-14,19-23
for WSS that is derived for the parabolic velocity Because the viscosity of blood is dependent on
profile of Poiseuille flow is the following18: plasma viscosity and the Hct, a relative apparent
2lumax viscosity of blood can by defined as the ratio of
sw ¼ ð3Þ
R
where u max is the highest velocity of flow (velocity
at the center of the tube). If we assume that blood
is a Newtonian fluid and that the blood flow is
steady through rigid and cylindrical vessel walls,
we can use Eqs. 2 and 3 to estimate WSS. This is
an attractive method due to its simplicity and is
utilized by several techniques in clinical practice.
However, the consideration of blood flow as
Poiseuille flow requires assumptions regarding
blood flow and properties of the conduit vessels
that are not valid in the context of actual blood
circulation. Clearly, the major simplification is that
of steady (non-pulsatile) flow. Nonetheless, one
may accept these assumptions to obtain possibly
useful approximations of WSS based on measure-
ments in a clinical context (for elaboration on this,
please see Limitations of Shear Stress Assessment
Methods).
Because the Newtonian model is easier to Fig 1. Relative apparent blood viscosity (dynamic
viscosity of blood normalized by dynamic viscosity
study, it has been adopted in many flow of plasma) as a function of Hct. Drawn based on data
circulation studies. Clinical reality, however, is from reference 24.
310 KATRITSIS ET AL

RBC aggregates become larger. As the shear rate

increases, the RBC aggregates tend to break up,
and individual RBCs also deform slightly (elon-
gate) while simultaneously lining-up with the
streamlines of the flow. As a consequence, the
blood viscosity decreases with increasing shear
rate. Practically, the viscosity can be viewed as
constant in large- and medium-size arteries for
shear rates higher than 100 s1.13,14,22
Because of frictional forces, a certain mini-
mum shear stress, termed the yield stress,30 is
required in order to initiate blood motion. As
expected, the yield stress is a rapidly increasing
function of the Hct, as shown in Fig 2B.
Because of the presence of the sizable formed
elements (especially RBC) in whole blood, blood
viscosity also depends on the size (diameter) of
the blood vessel.12,24,29 This effect is known as
the Fahraeus-Lindqvist effect and plays a signif-
icant role for vessels with a diameter smaller
than 1 mm, as illustrated in Fig 3, for the case of
a Hct of 45%. For vessel diameters higher than
1 mm, the relative apparent viscosity has a value
close to 3, in accordance with the data of Fig 1.
The effect of temperature on blood viscosity31
is illustrated in Fig 4, where a 2% increase in
relative apparent viscosity is observed for a 18C
temperature increase.31 Note that viscosity is a
decreasing function of temperature in liquids;
thus, both the viscosity of plasma and that of

Fig 2. A, Relative apparent blood viscosity as a

function of shear rate for a Hct of 49%. B, Yield stress
as a function of Hct. Drawn based on data from
references 28 and 30.

the apparent viscosity of blood to the viscosity of

plasma, under the same conditions. At increasing
levels of the Hct, the blood viscosity increases
substantially,24 an effect illustrated in Fig 1 by
the rapid increase of the relative apparent
viscosity with Hct. When the shear rate (velocity
gradient) increases, blood viscosity decreases, as
shown in Fig 2A.12,21,25-29 This effect of shear
rate on viscosity is explained from the fact that
human RBCs can form aggregates (rouleaux), in
which the RBCs stick-on loosely together at
Fig 3. Relative apparent blood viscosity as a function
shear rates lower than 100 s1.19 At decreasing of vessel diameter for a Hct of 45%. Drawn based on
shear rates, the size and number-density of the data from reference 24.
WALL SHEAR STRESS
Fig 4. Relative apparent blood viscosity as a function
of temperature for a Hct of 45%. Drawn based on data
from reference 31.
blood decrease with increasing temperature.
Because of the absence of formed elements in
plasma, its viscosity decreases faster with tem-
perature than in the case of whole blood. Thus,
the ratio of blood-to-plasma viscosity (the
relative apparent viscosity) is an increasing
function of temperature.
Other parameters also exist that have a
secondary effect on blood rheology in a manner
dependent on the Hct and the plasma proper-
ties.19,21,22 It must be also noted that, due to the
pulsatile character of blood flow, the shear rate
during the cardiac cycle is not always greater than
100 s1 because it varies from zero to approxi-
mately 1000 s1 in several large arterial vessels.32
Therefore, it is clear that blood behaves as a non-
Newtonian fluid, and several models have been
proposed to mimic the rheological behavior and
properties of blood.33-35 Most of these non-
Newtonian models have been developed by
parameter fittings to experimental data for blood
viscosity measured at certain shear rates.32
In summary, blood is expected to behave as a
non-Newtonian fluid in actual circulation, and
proper attention should be paid in computational
studies of blood circulation to the modeling of
blood viscosity and its dependence on shear rate.
A shear stress threshold (the yield stress) is
required to initiate blood flow, and it is an
increasing function of the Hct. The apparent
311
viscosity of blood is a strongly increasing func-
tion of the Hct, while it is a decreasing function of
the local shear rate. Finally, the apparent viscos-
ity of blood can only be considered as indepen-
dent on vessel size for vessel diameters larger
than approximately 1 mm.
Assessment of WSS
The estimation of wall shear rate and shear
stress and their effect on the structure and
function of the vascular wall has been the
subject of many studies. Measurements and
calculations have been obtained through in vitro
and in vivo studies, both in animals and
humans. This section provides an overview of
the various methods and techniques concerning
the estimation of wall shear rate and shear
stress. Special attention is given to the particu-
larly challenging studies concerning the coro-
nary arteries of humans.
In Vitro Measurements
Many in vitro observations and measurements of
wall shear rate and shear stress have been
performed on transparent glass tubes with side
branches and different degrees of curvature, or
casts of real vessels. The most common methods
used include:
! flow visualization by means of dye injec-
tion and high-speed cinemicrographic
techniques36
! laser Doppler anemometry37,38
! particle image velocimetry39-41
In Vivo Measurements
The most straightforward methods for in vivo
estimation of WSS that do not require sophis-
ticated calculations are based on the Hagen-
Poiseuille formula (see Eq. 2), which requires
simply the measurement of the blood volume
flow-rate and the radius of the vessel lumen at
the site of flow measurement. Equivalently, the
shear rate can be expressed in terms of
the average velocity, (cd = 4ū/R) or in terms
of the maximum velocity (cd = 2u max/R). Thus,
the determination of WSS requires the mea-
surement of the vessel radius and either the
average or the maximum flow velocity. The
312 KATRITSIS ET AL

values of viscosity l that are commonly used

in these formulas are 3.0 cP,42-44 3.5 cP,45,46 or
4.0 cP47 (1 cP = 103 Pad s), which apply to
arterial blood with a normal Hct. The above
expressions are based on the parabolic profile,
of Poisseuille flow, and consider blood
as a Newtonian fluid of constant dynamic
viscosity. Because these conditions are not
generally met in the arterial circulation, the
estimation of WSS using these expressions is
only approximate.
A more precise method requires the determi-
nation of wall shear rate, which is the blood
flow velocity gradient near the vessel wall,
because the magnitude of WSS depends on
how fast the blood velocity increases when
moving from the vessel wall to an immediately
adjacent point toward the center of the vessel in
a direction normal (perpendicular) to the vessel
wall. According to Eq. 1, WSS is proportional to
the wall shear rate under the Newtonian
assumption of constant dynamic viscosity.
Next, brief review of the most common
approaches to the measurement or calculation
of WSS in animals and humans is presented. Fig 5. The velocity distribution (A) and the shear rate
(B) calculated by jdu/drj as function of depth and time
at the common carotid artery of a healthy volunteer.
Reprinted with permission from Ultrasound Med Biol
Animal Studies 1995;21:171-185.

Animal studies have been reported that aimed at

obtaining estimates rather than exact measure-
ments of shear rate and shear stress because of
the complex rheological conditions found in
living circulation such as pulsatile flow and
moving/elastic vessels. Some examples are the
following:
! Hot film coronary artery velocity measure-
ments in horses48
! Wall shear stress estimation by upstream-to-
downstream pressure drop measurements in
unbranched arterioles, capillaries, and ven-
ules, coupled with length and diameter
measurements in the cat mesentery49
! Doppler ultrasound velocity measurements
and shear stress estimation at the aortas of
Yucatan micropigs50
Human Studies
There are 3 ways to obtain blood velocity data in
order to calculate WSS in a blood vessel: (a)
noninvasively, (b) invasively, and (c) through
simulations using computational fluid dynamics.
Noninvasive Methods
Ultrasound
Velocity profiles in vascular flow can be derived
by ultrasound through offline processing of
radiofrequency signals.18,51,52 Accurate measure-
ment of low blood flow velocities close to the
vessel wall can be achieved when the high-
amplitude, low-frequency signals reflected by the
artery wall are adequately suppressed without
losing the low blood flow velocity information
near the wall.51 The radial derivative of the
velocity profile at each site and time instant gives
the shear rate distribution, from which the WSS
value can be calculated by multiplying with the
blood viscosity. Based on the shear rate distri-
bution, the following parameters can be deter-
WALL SHEAR STRESS
mined51: (1) the mean wall shear rate at the
anterior and the posterior vessel wall, (2) the
time-averaged and peak wall shear rate over one
cardiac cycle, (3) the value at peak systole, and
(4) the maximal cyclic change in wall shear rate
within the cardiac cycle (Fig 5). The maximum
shear rate has been reported to be assessed at
250 to 300 lm from the vessel wall,51,52 which
indicates that blood flow shear rates cannot be
determined accurately at the wall. Another
important limitation of the ultrasound technique
is that wall shear rates can only be determined
reliably in relatively straight vessels.51 This
method is not applicable to the coronary arteries
due to overlying bone and tissue.
Pulsed Doppler ultrasound
Doppler ultrasound measures average velocities
in the targeted volume of the focused acoustic
wave,9 which can be reduced to a smallest
possible size of about 1-mm cube.53 Thus, the
velocity profiles determined by this technique
313
suffer from reduced accuracy due to the inade-
quate spatial resolution18 and the inability of the
technique to discriminate unambiguously be-
tween signals induced by slowly moving struc-
tures, like vessel walls, and those induced by
slowly moving blood near the vessel wall.18
However, the wall shear rate can be alternatively
calculated by using measurements of the central
velocity and the vessel diameter.53-55 In this
method, the mean WSS is calculated according
to the formula s mean = lcd M, where l is the
apparent viscosity of blood and c M is the mean
wall shear rate. Wall shear rate is not directly
measured but can be calculated assuming Pois-
euille flow and, thus, parabolic velocity distri-
bution across the arterial lumen. Under these
assumptions, as indicated previously, wall shear
rate can be calculated by dc M = 2u max/R M , where
u max is the time-averaged centerline velocity and
R M is the corresponding lumen radius.54,55 This
method suffers from the limitations imposed by
the assumption of Poiseuille flow, and it is not

Fig 6. Left, Phase velocity images from peak systole and diastole. The white color represents flow toward the
head, and the black color in the opposite direction. Right, Typical examples of mean wall shear rate waveforms
from the 3 arteries. Reprinted with permission from J Magn Reson Imaging 2004;19:188-193.
314
applicable to the coronary arteries due to over-
lying bone and tissue.
Phase-contrast magnetic resonance imaging
Blood flow velocity can be determined non-
invasively by magnetic resonance imaging
(MRI) phase contrast velocity mapping tech-
niques.47,51,56,57 From the derived velocity pro-
files, the shear rate close to the vessel wall and,
thus, the WSS can be calculated (Fig 6). The
velocity data can be precisely matched with
anatomic pictures, thus providing an anatomic
and functional examination that is superior to
ultrasound.58 Magnetic resonance imaging can
be used to examine almost any vessel in the
body without regard to overlying bone or bowel
gas.9 One important limitation in the use of
MRI velocity measurements for the calculation
of WSS is the precise identification of the blood
vessel wall boundary within the image pixel
because that pixel may be partially covered with
moving blood and partially covered by the
stationary vessel wall.9 This problem arises
from the low spatial resolution of the technique
(0.5-1 mm),51,59 which limits the true spatial
resolution of velocity assessment to about 1 to
2 mm51 (which is comparable to the diameter
of the coronary arteries). Blood flow velocities
cannot be measured closer than 1000 to
1200 lm51 to the wall, and as a consequence,
the wall shear rate values are underestimated.
Another limitation of this method is its low
temporal resolution because MRI measurements
are obtained over 25 to 30 s51,59; hence, special
precautions are required to be able to study
dynamic processes involving movement and
temporal changes. As with ultrasound, accurate
velocity profiles can only be obtained in
relatively straight vessels. Because the MRI
techniques have a spatial resolution of velocity
assessment comparable to the size of the
coronary vessels, they are not suitable for
obtaining intracoronary velocity profiles.
Invasive Methods
Intravascular Doppler ultrasound
This technique measures flow velocities by
means of intravascular Doppler wires.42,45 It
can only provide the peak value of the velocity
profile and cannot provide information regard-
KATRITSIS ET AL
ing the vessel wall.60 Doppler ultrasound meth-
ods have been used in various studies for the
measurement of WSS in the coronary arteries
where the coronary flow velocities can be
selectively measured during cardiac catheteriza-
tion, and the corresponding cross-sectional areas
of vessels can be determined with computer-
assisted angiographic methods. It is thus possible
to calculate absolute flow-rates and, in turn,
determine average WSS values at various sites
along the coronary arteries by means of the
Hagen-Poiseuille formula (Eq. 2). This method
suffers from the limitations of the assumption of
Poiseuille flow, on which the Hagen-Poiseuille
formula is based, and also by the estimation of
volumetric blood flow by means of the formula
Q = 0.5U maxA (where U max is the time-averaged
centerline velocity of the flow and A is the cross-
sectional area of the vessel), which gives only
approximate results especially for very low flow
rates. Finally, the invasive nature of the method
results in the disturbance of the original flow
conditions in the vessel, and thus, the measured
values of WSS may vary from the actual
(undisturbed) values.
Intravascular ultrasound
This technique measures velocity profiles based
on the decorrelation of the ultrasonic radio-
frequency signal.61 It uses miniaturized trans-
ducer assemblies inserted into catheters as small
as 2 mm in external diameter. It obtains velocity
profiles, from which the flow shear rate estima-
tion is possible, and subsequent calculation of
the WSS by multiplying with the blood viscosity
(under the Newtonian assumption). Simulta-
neously, this method provides cross-sectional
information of the morphology of the vessel
wall.61 This method cannot be used for the
estimation of WSS at the coronary arteries
because there are no available probes with the
required spatial resolution that are capable of
measuring intracoronary velocity fields.42,62 In
addition, the invasive nature of this technique
alters the original intravascular flow conditions.
Computational fluid dynamics
Computational fluid dynamics (CFD) is a general
term used to describe the body of knowledge and
WALL SHEAR STRESS
methods that seek the numerical solution of the
governing (Navier-Stokes) equations of fluid
flow. Typically, a CFD problem consists of
simulating the flow patterns in a given geometry
(ie, velocity and pressure distributions), subject
to certain boundary conditions for the flow
variables. This simulation involves the numerical
solution of the governing equations by decom-
posing the given geometry into a large number of
small geometric segments (finite volumes or finite
elements) forming a computational mesh on
which the variables of interest are calculated.
The discretization (in space and time) of the
governing differential equations results in a
system of algebraic equations, whose numerical
solution yields the problem unknowns at the
mesh points (ie, the distributions of the velocity
components and the pressure in incompressible
flow). To perform a CFD simulation of flow in a
blood vessel, a 3-dimensional description of the
vessel lumen is required. Several methods have
been developed for this purpose, the most widely
used being the reconstruction based on biplanar
angiography data.35,63 Other approaches are
based on the combination of simultaneous intra-
vascular ultrasound and biplanar angiogra-
phy43,44,64-68 and on vessel imaging by computed
tomographic scanning.46 In cases involving cor-
onary arteries, simulations have been based on
averaged experimental coronary data rather than
patient-specific geometries. An example of CFD
flow simulation in a prototype vessel with bifur-
cation is provided in Appendix C.
The field of CFD has made great progress in the
past 2 decades, taking advantage of the availabil-
ity of fast computing capabilities. In terms of the
application of CFD to the study of arterial blood
flows, the main limitations arise from the uncer-
tainties in defining the problem with the requisite
precision rather than the inherent limitations of
the numerical techniques or hardware capabili-
ties. For example, in the simulation of blood flow
in coronary arteries, the ambiguities associated
with the precision of the 3-D reconstruction of
the coronary arterial tree and the definition of the
inflow/outflow boundary conditions, as well as
the motion of the coronary vessels during the
cardiac cycle, give rise to considerable uncertain-
ty regarding the validity of computational results.
Further progress can be achieved through proper
315
definition of the problem and synergies between
simulations and experiments.
Limitations of Shear Stress
Assessment Methods
All the methods that have been described for the
determination of WSS are subject to several
limitations and inherent methodological prob-
lems, which are discussed below.
Methods based on the Hagen-Poiseuille formula
The Poiseuille law and the resulting Hagen-
Poiseuille formula are based on specific assump-
tions regarding the blood flow and the properties
of the conduit vessels. The accuracy of the
associated methods of WSS estimation depends
on the validity of these assumptions in the case of
blood circulation. We summarize these assump-
tions below and comment on their applicability:
Assumption 1. Blood behaves as a Newtonian
fluid. As discussed earlier, this assumption of
constant viscosity, independent of the velocity
gradients, holds for water and some other
liquids but not always for blood. Blood
behaves approximately like a Newtonian fluid
only for shear rates above 100 s1 but
becomes non-Newtonian for lower shear rates
because the viscosity increases exponentially
with decreasing shear rate.13,14 Note that the
instantaneous shear rate varies from zero to
approximately 1000 s1 in several large
arteries32 during the cardiac cycle.
Assumption 2. The flow is laminar and fully
developed. This assumption holds in those
parts of vascular circulation where low veloc-
ities of blood flow and small diameters of
blood vessels lead to small Reynolds numbers
(b2000) that correspond to laminar flow (ie,
the streamlines of blood flow are parallel to
the vessel wall). The assumption of laminar
flow is valid in most parts of normal circula-
tion, although the flow can become turbulent
in the largest blood vessels or at sites of
pathologic deformations of arterial structure
(eg, stenoses) where the streamlines of blood
flow may not be parallel to the vessel wall, and
eddies of different sizes may be formed.13,14
Assumption 3. Blood does not slip at the vessel
wall. This assumption implies that the lamina
in contact with the vessel wall is stationary.
316
Hydrodynamic experiments confirm this as-
sumption because bslippageQ has not been
observed at any liquid-solid interface.13,14
Assumption 4. The flow is steady. This
condition is not met in blood circulation
where flow is pulsatile. The flow velocity
profiles vary with time during the cardiac
cycle and may not be parabolic in shape.13,14
Assumption 5. The vessel is cylindrical. This
assumption implies that the vessel cross-
section is circular and the vessel diameter
remains constant. Most arteries have approx-
imately circular cross-section, but many veins
and the pulmonary arteries tend to have
elliptical rather than circular cross-section.
The requirement of constant diameter is never
met in blood vessels because individual
arteries tend to become narrower as they
progress toward the periphery, whereas in
veins, the process is reverse.13,14
Assumption 6. The vessel wall is rigid. This
condition is not met in reality because the
blood vessels are distensible and their diam-
eter changes with transmural pressure (intra-
vascular minus extravascular pressure). Given
a fixed pressure gradient and external pres-
sure, a distensible tube will assume a slightly
tapered shape because transmural pressure
will fall along the length of the vessel.13,14
The above considerations point to the fact that
the pulsatile nature of blood flow and occasional
turbulence, the tapered cross-section and disten-
sibility of blood vessels, and the non-Newtonian
behavior of blood all seriously limit the applica-
tion of Poiseuille’s law and, thus, limit accord-
ingly the accuracy of WSS estimation by means
of the Hagen-Poiseuille formula.
Methods Based on Measurements of Wall
Shear Rate
The accurate determination of wall shear rate
from velocity profiles requires a velocity mea-
surement method that is able to assess unambig-
uously the low blood flow velocities near the
vessel wall. In practice, it is feasible to obtain
velocity measurements at a discrete number of
points in the blood vessel and then use the
measured values to bbuildQ the velocity profile
using an interpolation algorithm.69 The wall shear
rate calculation requires an estimate of the slope
of the velocity profile at the wall from at least 2,
KATRITSIS ET AL
preferably 3, velocity values measured near the
vessel wall.70 The accuracy of the calculated value
of WSS by this method depends on the spatial
resolution of the velocity measurements (which is
the ability of the measurement method to define
accurately the varying velocities at adjacent
points71) and on the interpolation algorithm used
to bbuildQ the velocity profile.9
The limited spatial resolution of the methods
used for the measurement of intravascular
velocity profiles affects the accuracy of WSS
estimation, especially near the vessel wall, where
the velocity of flow changes rapidly. Extrapola-
tion of the data on shear rate assessed in venules
as a function of vessel radius72 indicates that
shear rate determined 250 to 300 lm from the
wall (as in ultrasound systems) underestimates
WSS by about 10%.51 Clearly, methods that
determine shear rates even further from the
vessel wall (as in MRI systems) will result in
additional underestimation of WSS.
The interpolation algorithm used to generate
shear rate estimates also affects the resulting
WSS assessment. Various algorithms have been
used for the estimation of shear rate, such as
linear extrapolation, linear extrapolation with
correction for wall position, and quadratic
extrapolation,9,73 with varying levels of accuracy
when compared to the results from the solution
of the equations of fluid flow.9,59,73
Methods Based on Computational Fluid
Dynamics Simulations
Some key aspects and simplifications of the CFD
simulations that affect the accuracy of the
obtained results and their potential use in
clinical practice are discussed below.
1. The model of blood viscosity used in the
simulations affects the resulting estimates
of WSS because the latter is calculated
as the product of viscosity and shear
rate. For reasons of simplicity, a lot of
studies consider blood as a Newtonian
fluid44,46,62,74 and, thus, use a constant
value of viscosity. Because blood behaves
like a Newtonian fluid only for shear rates
above 100 s1, whereas for lower shear rates
the viscosity is increasing exponentially
with decreasing shear rate,13,14 the accuracy
WALL SHEAR STRESS
of the WSS estimates is limited, as previ-
ously discussed. In the case of the coronar-
ies, it is observed that the distribution
pattern of WSS is similar for the Newtonian
and the various non-Newtonian models in
steady flow conditions, but the magnitude
of the WSS estimates depends on the model
used.35 The differences are significant only
for low inlet velocities because, as the inlet
velocity becomes larger, the magnitude of
WSS produced by the various viscosity
models increases and becomes similar
among the different models.35 In the case
of simulation of pulsatile flow, it is observed
that the Newtonian model yields consistent-
ly lower WSS values throughout the entire
cardiac cycle, particularly during systole,75
but as in steady flow, the distribution
pattern of WSS is similar for the Newtonian
and non-Newtonian models.63,75 Thus, it
seems that the assumption of Newto-
nian fluid is an adequate approximation
when the distribution pattern of WSS is
examined, but not its actual values.63,75
2. The CFD simulations generally assume
that the arterial wall is rigid,43,63,76 al-
though it is known that the vessel diameter
changes during the cardiac cycle.
3. The vessel 3-D reconstruction used in the
CFD simulations is usually performed at
end diastole in order to eliminate the
problem of heart motion; thus, the motion
of vessels is not incorporated in the calcu-
lations. A comparative simulation between
a steady coronary vessel, and one with
varying curvature mimicking the movement
due to heart motion, revealed an increase of
the mean wall shear rate of 6.7% for the
dynamic model.77 Thus, the use of a static
geometry to predict the mean wall shear
rates may yield acceptable results.
4. To simplify computations, many simula-
tions consider steady flow. However, flow
throughout the vascular tree is pulsatile by
nature, and the resulting temporal varia-
tion of WSS has been proposed as a
decisive atherogenic factor.75 Pulsatile
flow simulations in coronaries have used
an average physiologic flow velocity wave-
form for the human left coronary. These
studies have the additional limitation that
317
the non-Newtonian models, which are
obtained by parameter fitting to experi-
mental viscosity data obtained at certain
shear rates under steady flow (rather than
pulsatile flow) conditions.63
5. Most simulations consider blood as a
homogenous liquid and use either a New-
tonian or a non-Newtonian model to
describe its properties. Recently, the
non-homogenous nature of blood has
been also considered in simulations.68
6. In many studies, the velocity at the
entrance of the vessel under investigation
is considered to be uniform over the vessel
cross-section.63,76 However, velocity pro-
files are non uniform, with the velocity at
the center of the vessel being higher than
the velocity toward the vessel walls. This
assumption introduces errors only within
a few diameters of the inlet because after
that distance, the flow is fully developed.62
In some simulations, a straight entry tube
is attached to the reconstructed vessel to
remove inlet effects on the computed flow
field and ensure a fully developed velocity
profile in the entering region.68
In summary, the most important limitation
that applies to all techniques currently used for
WSS measurement/estimation is that, no gold
standard for measuring blood flow exists. Thus,
obtained data on shear rate cannot be verified by
actual in vivo measurements, and the resulting
WSS estimates inevitably represent only approx-
imations of the actual values. This is currently
the most important methodological problem of
WSS measurement techniques, which has to be
addressed in order to assure the reliability of
scientific inference and the clinical relevance of
these important investigations.
Conclusions
Wall shear stress is a potentially important
factor of atherogenesis and rupture of athero-
sclerotic plaques. Therefore, methods of esti-
mation of the distribution of WSS in the arterial
system are of clinical relevance. However, all
techniques that have been used so far for the
measurement of shear rate and the estimation of
WSS are subject to several limitations and
318
exhibit inherent methodological problems. Fur-
thermore, measurements and accurate calcula-
tions suffer from the lack of reliable means for
validation against a standard reference. There-
fore, none of the existing methods provides
unequivocally reliable results. Improvements of
existing methods of measurement or estimation,
as well as reliable means for validation of the
obtained results, are clearly needed in order to
advance the state of the art in clinical research
on this important topic.
Appendix A. Basic Concepts of
Incompressible Flow Fluid Mechanics
and Blood Rheology
In this Appendix we introduce some fundamen-
tal concepts of fluid mechanics of viscous
incompressible flow. The reader is referred to
fluid mechanics textbooks (eg, Panton78, Kundu
and Cohen79, and Schlichting and Gersten80) for
a more extensive presentation of the concepts
introduced here.
The forces acting on a fluid volume can be
classified as follows:
1. Body forces: they arise from the presence of
a force field (gravitational, magnetic, etc).
Body forces are proportional to the fluid
mass (equivalently: fluid volume) and are
therefore expressed per unit mass or per
unit volume (in SI units: N/kg or N/m3).
2. Surface forces: they are exerted by the envi-
ronment through direct contact. Surface
forces are proportional to the contact area;
therefore, they are expressed per unit area as
surface stresses (in SI units: N/m2). A sur-
Fig 7. Sketch of the differential forces exerted by the
differential surface dA on a fluid.
KATRITSIS ET AL
Fig 8. Infinitesimal fluid control volume and stress
components (in the limit of zero-size control volume)
acting on its surfaces. The stress components shown
are assumed to be in a positive direction.
face stress can be decomposed into a compo-
nent which is normal to the contact area
(normal stress), and one which is tangential
to the contact area (shear stress) (Fig 7):
dFn
normal stress : sn ¼
dA
dFs
shear stress : ss ¼
dA
3. Line forces: they appear at the interface
between 2 fluids and originate from at-
tractive forces between molecules. Line
forces are proportional to the length of a
line along the interface and are therefore
expressed per unit length of the interface
line as surface tension (in SI units: N/m).
The stress at a point is, in general, charac-
terized by nine components. Consider, for
example, the infinitesimal fluid control volume
shown in Fig 8, the edges of which are parallel
to the coordinate axes. A stress acting on a
surface is denoted by s ij , where the subscript i
denotes the direction of the unit vector normal
to the surface, whereas the subscript j denotes
the direction of the stress component. For
example, s 12 is the stress component acting
on the surface perpendicular to the x1 axis in
the x2 direction. A stress component is a
normal stress for i = j, whereas it is a shear
stress for i p j. Corresponding stress compo-
WALL SHEAR STRESS 319
nents acting on opposing faces have a direction In Newtonian fluids, a linear relationship
opposite to the one of their counterparts, exists between a viscous stress and the corre-
becoming also equal in magnitude when the sponding strain rate:
infinitesimal volume shrinks to zero. Thus, the
rij ¼ 2lSij ðA  4Þ
stress at a point is indeed specified by a 9-
component tensor, denoted by s ij (where i, j = where l is the fluid’s dynamic viscosity, a thermo-
1, 2, 3). Furthermore, in the limit of zero-size physical property depending on temperature and
control volume, the requirement of zero torque pressure (SI units: Pad s where 1 Pa = 1 N/m2). The
with respect to the 3 axes gives s ij = s ji , that is, behavior of common fluids as water and air can be
the stress tensor is symmetric, consisting of considered Newtonian, that is, the viscosity coeffi-
only six independent components. cient l is dependent only on the pressure and
For a fluid at rest, the only nonzero compo- temperature and not on the flow field.
nents of the stress tensor are the diagonal ones By applying the principle of mass conserva-
(normal stresses), their magnitude being equal to tion and Newton’s law to the control volume of
the local fluid pressure: Fig 8, the following equations are obtained for
incompressible flow:
sij ¼  dij p ðA  1Þ
Bui
where d ij is the Kronecker d (d ij = 1 for i = j, d ij = 0 ¼ 0 ðmass conservation equationÞ ðA  5Þ
Bxi
for i p j). For a fluid in motion, the intermolecular
forces between fluid elements must be taken into  
account, and thus, the stress tensor becomes: Bui Bui Bsij
q þ uj ¼ þ q fi ðmomentum equationsÞ
Bt Bxj Bxj
sij ¼  dij p þ rij ðA  2Þ
ðA  6Þ
where r ij is the contribution due to the intermo-
where u i, u j are the velocity components in the i and
lecular forces, becoming zero for a stationary fluid.
j directions, respectively; q the fluid density; s ij , the
The r ij components depend on a thermophysical
total stress, as previously defined; and f i, a body
property of the fluid, its viscosity, which expresses
force component (per fluid unit mass) in the i
the fluid’s resistance to deformation; thus, they are
direction. Note that, in the above 2 equations, the
referred to as viscous stresses. It is straightforward
Einstein summation convention is used, that is, the
to express the stress tensor components in a rotated
appearance of an index twice in a term implies
coordinate system and, consequently, to calculate
summation over the index. For example, the term
the force components acting on a surface of known
Bu i /Bx i in the mass conservation equation is equiva-
orientation (eg, see Schlichting and Gersten80).
lent to Bu 1/Bx 1 + Bu 2/Bx 2+ Bu 3/Bx 3. By expressing
Next, we relate the stress tensor to the rate of
the total stress s ij in terms of both the pressure and
deformation (or strain rate) of fluid elements.
viscous components and assuming Newtonian
The strain rate is expressed in terms of a tensor,
fluid, the momentum equations take the form:
S ij (where i, j = 1, 2, 3). A diagonal component
(i = j) of the S ij tensor corresponds to a normal 
Bui Bui

Bp

B 2lSij

strain rate (rate of change of a fluid element q þ uj ¼ þ þ q fi ðA  7Þ
Bt Bxj Bxi Bxj
length per unit length), whereas an off-diagonal
component (i p j) expresses a shear strain rate or, for constant l:
(rate of change of a corresponding 908 angle). It
can be proved (see Kundu and Cohen79 and Bui Bui 1 Bp B2 ui
þ uj ¼  þm þ fi ðA  8Þ
Panton78) that all elements of the strain rate Bt Bxj q Bxi Bxj Bxj
tensor can be expressed as:
  where m = l/q is the fluid’s kinematic viscosity (SI
1 Bui Buj
Sij ¼ þ : ðA  3Þ units: m2/s). In many applications, the body forces
2 Bxj Bxi
are only due to earth’s gravitational field, thus f 1 =
Evidently, the strain rate tensor is also symmetric, f 2 = 0; f 3 = g, where g is the acceleration of gravity.
that is, it consists of only 6 independent The conservation equations for mass and momen-
components. tum are referred to as the Navier-Stokes equations.
320 KATRITSIS ET AL

Note that, because only the gradient of pressure
appears in the momentum equations, the level of
absolute pressure is not a relevant quantity in
incompressible flow: identical pressure differences
produce identical flow patterns, irrespective of the
absolute pressure values.
The Navier-Stokes equations form a system of
4 partial differential equations with 4 unknowns
(the 3 velocity components and the pressure),
which should be solved with proper initial and
boundary conditions. In the presence of walls,
the no-slip condition is implemented for the
velocity components at the corresponding
boundaries, that is, it is assumed that the fluid
velocity at the interface coincides to that of the
solid, an assumption justified by experiments
and molecular dynamics simulations.
In the following, we consider flows in which
the only body forces present are due to earth’s
gravitation. We introduce now appropriate
scales for length (L 0), velocity (U 0), time (t 0 =
L 0/U 0), and pressure ( p 0 = qU 02), where L 0 is a
representative geometrical length, and U 0 a
representative velocity of the fluid. New nondi-
mensional variables can now be defined:
xi ui tU0 p principles, we consider the class of simple shear
x4i ¼ ; u4i ¼ ; t4 ¼ ; p4 ¼
qU02
:
L0 U0 L0
Expressing the (dimensional) quantities in the
Navier-Stokes equations in terms of the new
nondimensional variables, the governing equa-
tions for Newtonian incompressible fluids can be
written in nondimensional form:
Bu4i
¼0 ðA  9Þ
Bx4i
expresses the ratio of inertia to viscous forces,
whereas the Froude number expresses the square
root of the ratio of inertia to gravitational forces.
In general, both parameters have to be equal for
flows in geometrically similar setups to also be
kinematically similar (ie, to be characterized
by the same nondimensional flow patterns).
Flows which are both geometrically and kine-
matically similar are referred to as being dynam-
ically similar.
In many applications, the presence of gravity
simply adds a hydrostatic component to the
pressure field, which can be absorbed into the
pressure. Thus, the flow dynamics can depend
only on the Reynolds number, Re. At low values
of Re, flows are in the laminar state, character-
ized by smooth variations in space and in time.
As Re increases, flows undergo a sequence of
hydrodynamic instabilities, becoming turbulent
at high Re. Turbulent flows are characterized by
irregular flow patterns and by increased rates of
momentum transfer and mixing.
Next, we discuss cases in which the viscosity
coefficient is not merely a thermophysical property
but is also dependent on the flow field, that is the
fluid is non-Newtonian. To illustrate the basic
flows u = f( y). A representative example of such
flows is the flow between a stationary and a moving
(at a constant velocity) plate, which is known as
the Quette flow. If the fluid is Newtonian, the shear

Bu4i Bu4i Bp4 1 B2 u4i

þ u4j ¼  þ
Bt4 Bx4j Bx4i Re Bx4j Bx4j
1
 di3 ðA  10Þ
Fr2
where Re = U 0L 0/m is the Reynolds number, and
Fr ¼ pUffiffiffiffi
0
gL
is the Froude number.
0
Evidently, the dynamics of incompressible
flow depends on 2 nondimensional parameters,
the Reynolds number, Re, and the Froude
number, Fr. (In the case of external excitation
or imposed unsteadinesss, eg pulsatile flow, the
dynamics depends, additionally, on the non-
Fig 9. Steady-state velocity distribution for flow of a
dimensional amplitude and frequency of the Newtonian fluid between a moving and a stationary
excitation pattern). The Reynolds number plate.
WALL SHEAR STRESS 321
Table 1. Characteristic Parameters of Non-
Newtonian Fluids
Shear thinning fluid so = 0 0bnb1
Shear thickening fluid so = 0 1bn
Bingham fluid so p 0 n=1

stress s is evidently proportional to the velocity

gradient Bu/By (shear rate), the proportionality
constant l being the fluid’s dynamic viscosity.
du
s¼l ðA  11Þ
dy
Note that, according to its general definition, the
shear rate is equal to twice the strain rate, as defined in
Eq. A-3. For laminar flow and Newtonian fluid, the
problem of Quette flow has an analytical solution,
which consists of a linear velocity profile (Fig 9).
As indicated above, fluids that exhibit a more
complicated relationship between shear stress
and shear rate are known as non-Newtonian
fluids. For a Newtonian fluid, the viscosity
depends only on temperature and pressure (for
given chemical composition) and not on the
shear forces acting on the fluid. In a non-
Newtonian fluid, the viscosity changes with the
applied shear force, and as a result, the determi-
nation of the viscosity coefficient is more
complicated in such fluids.
Depending on the transient behavior of viscosity
under the same flow conditions, non-Newtonian
fluids can be characterized by time-independent or
time-dependent viscosity. For time-dependent
Fig 10. Relationship between shear stress and shear
rate in the case of a simple flow, u = f( y), for different
types of fluids.
Fig 11. Sketch of shear stress versus shear rate for
blood using the Casson model and for a Newtonian
fluid.
viscosity, assuming constant strain rate, a fluid is
called brheopecticQ if the shear stress increases with
time and bthixotropicQ if it decreases with time. For
many non-Newtonian fluids of time-independent
viscosity, the shear rate–shear stress dependency
can be expressed by the general equation34:
 n
du
s ¼ so þ k ðA  12Þ
dy
known as the Herschel-Bulkley equation, where k
and n are the consistency and flow behavior indexes,
respectively. The stress s o known as the yield stress,
corresponds to the minimum stress required to
initiate flow of a non-Newtonian fluid (see below).
The classification of fluids that obey the above
equation can be summarized as presented in Table 1.
The rheological curves of representative non-
Newtonian fluids with time-independent viscos-
ity are shown in Fig 10. For all time-independent
non-Newtonian fluids, we can define an appar-
ent viscosity l app as the slope of the rheological
curve at a specific shear rate:
 n1
du
lapp ¼ k
dy
This apparent viscosity l app is a constant function of
time for time-independent non-Newtonian fluids,
while it is time-varying for time-dependent non-
Newtonian fluids (rheopectic, thixotropic). In mul-
tiphase flows, a relative apparent viscosity l*app can
be defined as the ratio of the apparent viscosity of the
solution and the apparent viscosity of the solvent.
322 KATRITSIS ET AL

One commonly used non-Newtonian blood ity profile does not vary in the streamwise flow
viscosity model is the model of Casson, which direction). Fully developed steady flow in a
assumes a stress-shear rate relation of the form35: straight tube is referred to as the Pouiseuille
pffiffiffi pffiffiffiffiffi pffiffiffiffiffi flow. The governing Navier-Stokes equations can
s ¼ s0 þ kc ðA  13Þ be written in cylindrical coordinates (see, eg,
Panton78), and by applying the above assump-
where s 0 is the yield stress, and dc , the shear rate. tions and the no-slip condition at the wall, the
The Casson model evidently implies that blood following analytical solution is obtained for the
requires a finite stress before it begins to flow. For velocity profile:
high stress values s H s 0 and K = l app, the
modeled viscosity of blood demonstrates a New- dp 1  2
R  r2

tonian like linear behavior (see Fig 11). uðrÞ ¼ ðA  14Þ
dx 4l
Other non-Newtonian models of blood are
the Carreau model,32 the Walburn-Schneck where r is the local radius; x, the streamwise
model,81 the Power Law model,32 and the coordinate; and dp/dx, the pressure gradient in
Generalized Power Law model.82 Because of the streamwise direction. As the flow is fully
the high complexity of blood rheology and the
fact that none of the non-Newtonian models is
generally accepted, blood viscosity is often
considered to be constant, and the relation
between shear stress and shear rate Newtonian
(linear). Evidently, the standard Newtonian
model is a good approximation in regions of
medium and high shear rate but not in regions
of low shear rate. Most of the non-Newtonian
models illustrate a superior performance in low
shear rate regions, and exceptionally, only the
Generalized Power Law model is a good
approximation at medium and high shear
rates.35 In arterial flows, the strain rates are
maximum in the wall regions, and thus, the
differences between the Newtonian and a non-
Newtonian model are depicted in the details of
bulk flow, whereas the wall shear stress values
are not substantially affected.35,63

Appendix B. Fully Developed Laminar

Flow in a Straight Tube
For hemodynamics applications, laminar flow in
a rigid cylindrical tube forms a reference
problem. Studies of blood circulation under
more realistic conditions, such as pulsatile flow
and elastic and branching vessels, can be
interpreted with reference to the simpler prob-
lem discussed here.
We consider Newtonian incompressible flow
in a straight rigid tube of radius R. To arrive to
the problem solution, we make the additional
Fig 12. Nondimensional velocity profile (A) and
assumptions that the flow is steady (ie, it is time- corresponding normalized shear stress (B) in fully
independent) and fully developed (ie, the veloc- developed flow in a straight tube.
WALL SHEAR STRESS 323
developed, the pressure gradient is also indepen-
dent of x; thus, the velocity profile can be
equivalently expressed as:

p1  p2  2
R  r2

uðrÞ ¼ ðA  15Þ
4lL

where p 1 and p 2 denote the pressure values at the

beginning and at the end of a segment of length L.
Clearly, the velocity distribution is a parabolic
function of the local radius, becoming maximum
on the centerline and zero on the wall (Fig 12).
The maximum velocity is thus:
ðp1  p2 ÞR2
umax ¼ uðr ¼ 0Þ ¼ ðA  16Þ
4lL
The nondimensional velocity distribution is read-
ily obtained (see Fig 12):
u r2
¼1 ðA  17Þ
umax R2
Clearly, fully developed laminar flow in a straight
tube takes place in infinitely thin concentric
layers of fluid that flow parallel to the tube wall,
with each layer moving at a different velocity. The
velocity gradient in the tube radius arises from
viscous friction between the adjacent fluid layers.
The differential volume flow rate in a circular
layer of width dr is dQ = u(r)2prdr. By integrating
over an entire tube cross-section, the following
Fig 13. Geometry and detail of mesh for flow in a pipe
bifurcation.
Fig 14. Local velocity distributions (m/s) at different
cross-sections in steady flow in a pipe bifurcation.
The Reynolds number at inflow is Re = 80.45.
formula is obtained for the total volume flow rate,
known as Poiseuille’s Law13,78,80:
pR4 ðp1  p2 Þ
Q¼ : ðA  18Þ
8lL
Poiseuille’s Law describes the relation between
volume flow-rate pressure gradient along the tube,
tube size, and fluid viscosity.
The average velocity ū of flow across the tube is
obtained by dividing the volume flow rate Q by
the cross sectional area pR 2:
R2 ðp1  p2 Þ
ū ¼ ðA  19Þ
8lL
It is apparent from the comparison of Eqs. A-16
and A-19 that the average velocity ū is half of the
maximum value u max.
Based on the concepts presented in Appendix A,
the shear stress can now be computed:
du
s¼l ðA  20Þ
dr
The shear rate du/dr can be obtained by differen-
tiating the u(r) expression in Eq. A-15. Thus, the
local shear stress magnitude is given by:
du rðp1  p2 Þ
s¼l ¼ ðA  21Þ
dr 2L
Note that the shear stress is directly proportional
to the streamwise pressure gradient. It is clear
that the distribution of shear stress is maximum
324 KATRITSIS ET AL

Fig 15. Color-coded contours of wall shear stress magnitude (in Pa) in steady flow in a pipe bifurcation. The
Reynolds numbers at inflow are (a) Re = 80.45 and (b) Re = 148.92.

at the tube wall (r = R), whereas it is zero at the wall friction. The human heart is the pump of
tube axis where r = 0. The maximum value of the entire vascular network.
shear stress Thus, the flow in blood circulation induces a
bdragQ between the outermost layer of blood and
Rðp1  p2 Þ
sw ¼ ðA  22Þ the vessel wall. This shearing force imposed by
2L
blood is applied mainly to the inner layer of the
is defined as the bwall shear stress,Q s w, and is vascular wall in contact with blood, the vascular
used as the characteristic shearing stress of endothelium, and can affect the functional and
incompressible flow in a tube. We underline structural integrity of the endothelial cells13.
that Eq. A-22 only holds for the case of fully Wall shear stress thus expresses the force per
developed laminar flow. In the entrance region of unit area in a plane tangential to the vessel wall,
a vessel, the laminar flow gradually develops from imposed by blood flow due to the viscous
a rather uniform velocity profile to the parabolic properties of blood. According to Eq. A-20,
profile of Eq. A-14. The development length of WSS is proportional to the viscosity of blood
laminar flow (nondimensionalized with the vessel
diameter) is a linear function of the flow’s
Reynolds number, defined in terms of the average
flow velocity and the tube diameter.80,83 The more
uniform velocity profiles in the development
region are characterized by higher velocity gra-
dients (shear rates) in the wall region and, thus,
higher levels of WSS.
The integral of WSS over the entire vessel area
in contact with the fluid gives the overall friction
force exerted by the vessel on the fluid. This
force is balanced by a force equal to the total
pressure difference times the vessel’s cross-
section. The power dissipated into heat is given
by the product of the total pressure difference
times the volume flow rate. Thus, even in the
absence of other areas of dissipation of mechan-
ical power (eg, regions of local flow separation),
an input of mechanical power (a pump) is Fig 16. Inlet mean velocity versus time for a cardiac
required to supply the power dissipated due to cycle. Drawn based on data from reference 91.
WALL SHEAR STRESS
and the radially directed gradient of the velocity
at the vessel wall. Maintaining the assumption of
Newtonian fluid, we can combine the Poiseuille
Law (Eq. A-18) with Eq. A-21, to express the
WSS in terms of the volume flow rate in fully
developed laminar pipe flow:
4lQ
sw ¼ ðA  23Þ
pR3
Eq. A-23, known as the Hagen-Poiseuille formu-
la,13 is often used in in vitro and animal experi-
ments for the estimation of shear stress from the
measured flow rate Q lumen radius R, and the
blood’s dynamic viscosity l.42,45,51 We underline
that the above equation has been derived for the
Fig 17. Local velocity distributions (m/s) at different cross-sections in pulsatile flow in a pipe bifurcation. The
average Reynolds number at inflow is Re = 80.45. The 8 distributions shown (A through H) correspond to the time
instants (a through h) of Fig 16.
325
case of parabolic flow profile. Deviations have
been observed in arterial flows, in particular,
velocity profiles in the form of flattened parab-
olas.51 These deviations are related to the fact
that viscosity is higher at the center of the vessel
because the RBCs tend to stream in the center of
the vessel,84,85 thereby reducing the strain rates
they are exposed to. In the common case of
curved vessels, the velocity profile is not only
flattened but also skewed.51 As a consequence,
the wall shear rate differs over the circumference
of a curved vessel cross-section. Considering
the above facts, it is clear that, in arterial flows,
the estimation of WSS using Eq. A-23 is
only approximate.
326 KATRITSIS ET AL

Appendix C. Laminar Flow in mately 400 000 finite volumes. A parabolic

Pipe Bifurcations
In this Appendix we consider the problem of flow
in a pipe bifurcation, which is of relevance to
arterial flows. In general, no analytical solutions
exist for flow problems in complex geometries.
Thus, we will obtain a numerical solution to the
Navier-Stokes equations for incompressible flow
of a Newtonian fluid. To this end, the Adapco
STAR-CD code is used,86-89 based on a finite
volume approach (eg, see Ferziger and Peric90).
The problem geometry is shown in Fig 13
and is decomposed into a total of approxi-
velocity profile is prescribed at the inflow
boundary; at the outflow Neumann boundary,
conditions are implemented for the velocity
(velocity gradient equal to zero), whereas a
specified volume flow rate division ratio is
imposed at the outflow boundaries.
We present results for both steady and pulsatile
flow. In the case of pulsatile flow, the Reynolds
number value refers to the time-averaged flow.
We have calculated both cases for a Reynolds
number equal to Re = 80.45, a value represen-
tative of hemodynamic applications in small

Fig 18. Wall shear stress magnitude distributions (in Pa) in pulsatile flow in a pipe bifurcation. The average
Reynolds number at inflow is Re = 80.45. The 8 distributions shown (A through H) correspond to the time instants
(a through h) of Fig 16.
WALL SHEAR STRESS
arteries.14 To demonstrate the effect of a higher
Reynolds number, we have also performed a
steady flow simulation at Re = 148.92. Fig 14
shows representations of local velocity profiles
at different cross-sections for steady flow. One
can identify the steeper boundary layers in the
inner regions of the flow divider, as opposed to
the thicker boundary layers in the outer
regions. In both branches, the flow gradually
develops to the fully developed parabolic
profile of laminar flow farther downstream.
The local WSS consists, in general, of 2
components, both perpendicular to the normal
to the wall unit vector. The distribution of
WSS magnitude (square root of the sum of
squares of the two local shear stress compo-
nents) is shown in Fig 15 for both Reynolds
number values investigated. Clearly, the WSS is
minimal in the outer regions of the bifurcation,
consistent with the thicker boundary layers
(corresponding to smaller velocity gradients).
These results also demonstrate a direct scaling
of WSS values with Reynolds number (Fig 15).
To simulate pulsatile flow, we have imple-
mented an inflow boundary condition in terms
of a parabolic velocity profile, whose variation
in time is periodic, characterized by a mean
value history shown in Fig 16. Eight character-
istic time instant are depicted in the time history
of Fig 16. Figs 17 and 18 show the computed
velocity and WSS distributions at those charac-
teristic time instant, respectively. Clearly, the
computed values always scale with the instan-
taneous velocity at inflow, whereas the distri-
butions are qualitatively similar to those of
steady flow.
References
1. Chien S, Li S, Shyy YJ: Effects of mechanical forces
on signal transduction and gene expression in
endothelial cells. Hypertension 31:162 - 1698 1998
2. Davies PF: Flow-mediated endothelial mechano-
transduction. Physiol Rev 75:519 - 5608 1995
3. Giddens DP, Zarins CK, Glagov S, et al: Flow and
atherogenesis in the human carotid bifurcation. In
Schettler G, (ed). Fluid dynamics as a localizing
factor for atherosclerosis. New York: Spinger-Verlag
Press8 1993
4. Fung YC: Biomechanics. Circulation. New York:
Springer-Verlag8 1997
5. Song S, Leahy R, Boyd D, et al: Determining cardiac
velocity fields and intraventricular pressure distribu-
327
tion from a sequence of ultrafast CT cardiac images.
IEEE Trans Med Imaging 13:386 - 3978 1994
6. Adler R, Chenevert T, Fowlkes B, et al: Calculation of
pressure gradients from MR velocity data in a laminar
flow model. J Comput Assist Tomogr 15:483 - 4888
1991
7. Fry DL: Certain histological and chemical responses
of the vascular interface to acutely induced me-
chanical stress in the aorta of the dog. Circ Res
24:93 - 1088 1969
8. Caro CG, Fitz-Gerald JM, Schroter RC: Atheroma
and arterial wall shear. Observation, correlation and
proposal of a shear dependent mass transfer mech-
anism for atherogenesis. Proc R Soc Lond B Biol Sci
177:109 - 1598 1971
9. Shaaban AM, Duerinckx AJ: Wall shear stress and
early atherosclerosis: A review. AJR Am J Roentgenol
174:1657 - 16658 2000
10. Strackee J, Westerhof N: The physics of heart and
circulation. Bristol: Institute of Physics Publishing8
1993
11. Rowan JO: Physics and the circulation. Bristol: Adam
Hilger Ltd8 1981
12. Pearson TC: Hemorheology in the erythrocytoses.
Mt Sinai J Med 68:182 - 1918 2001
13. Milnor W: Hemodynamics. Williams & Wilkins8 1982
14. Nichols W, O’Rourke M: McDonald’s blood flow in
arteries: Theoretical, experimental and clinical princi-
ples. Hodder Arnold8 2005
15. Spring S, van der Loo B, Krieger E, et al: Decreased
wall shear stress in the common carotid artery of
patients with peripheral arterial disease or abdominal
aortic aneurysm: Relation to blood rheology, vascular
risk factors, and intimamedia thickness. J Vasc Surg
43:56 - 638 2006 (discussion 63)
16. Kannel WB: Overview of hemostatic factors involved
in atherosclerotic cardiovascular disease. Lipids
40:1215 - 12208 2005
17. Marton Z, Kesmarky G, Vekasi J, et al: Red blood cell
aggregation measurements in whole blood and in
fibrinogen solutions by different methods. Clin Hem-
orheol Microcirc 24:75 - 838 2001
18. Brands PJ, Hoeks AP, Hofstra L, et al: A noninvasive
method to estimate wall shear rate using ultrasound.
Ultrasound Med Biol 21:171 - 1858 1995
19. Goldsmith HL, Skalak R: Hemodynamics. Annu rev
fluid mechanics 7:213 - 2478 1975
20. Guyton AC, Hall JE: Textbook of medical physiology
(ed 11). Saunders8 2005
21. Merrill EW: Rheology of blood. Physiol Rev 49:
863 - 8888 1969
22. Stuart J, Nash GB: Technological advances in blood
rheology. Crit Rev Clin Lab Sci 28:61 - 938 1990
23. Chien S: Red-cell deformability and its relevance to
blood-flow. Annu Rev Physiol 49:177 - 1928 1987
24. Pries AR, Neuhaus D, Gaehtgens P: Blood-viscosity
in tube flow—Dependence on diameter and hemat-
ocrit. Am J Physiol 263:H1770 - H17788 1992
25. Merrill EW, Pelletie GA: Viscosity of human blood—
transition from Newtonian to non-Newtonian. J Appl
Physiol 23:178 - 1828 1967
328
26. Chien S: Shear dependence of effective cell volume
as a determinant of blood viscosity. Science 168:
977 - 9798 1970
27. Reinke W, Johnson PC, Gaehtgens P: Blood-viscos-
ity during tube flow—effect of shear rate. Int J
Microcirculation-Clin Exp 4:908 1985
28. Zydney AL, Oliver JD, Colton CK: A constitutive
equation for the viscosity of stored red-cell suspen-
sions—effect of hematocrit, shear rate, and suspend-
ing phase. J Rheol 35:1639 - 16808 1991
29. Box FM, van der Geest RJ, Rutten MC, et al: The
influence of flow, vessel diameter, and non-New-
tonian blood viscosity on the wall shear stress in a
carotid bifurcation model for unsteady flow. Invest
Radiol 40:277 - 2948 2005
30. Picart C, Piau JM, Galliard H, et al: Human blood
shear yield stress and its hematocrit dependence.
J Rheol 42:1 - 128 1998
31. Cinar Y, Senyol AM, Duman K: Blood viscosity and
blood pressure: Role of temperature and hypergly-
cemia. Am J Hypertens 14:433 - 4388 2001
32. Lee RT, Grodzinsky AJ, Frank EH, et al: Structure-
dependent dynamic mechanical behavior of fibrous
caps from human atherosclerotic plaques. Circulation
83:1764 - 17708 1991
33. Fung YC: Biomechanics: properties of living tissues.
Berlin: Springer8 1993
34. Barnes HA: The yield stress—a review or dsa pamsa
qeiT—everything flows? Journal of Non-Newtonian
Fluid Mechanics 81:133 - 1788 1999
35. Johnston BM, Johnston PR, Corney S, et al: Non-
Newtonian blood flow in human right coronary arteries:
Steady state simulations. J Biomech 37:709 - 7208 2004
36. Kerber CW, Hecht ST, Knox K, et al: Flow dynamics
in a fatal aneurysm of the basilar artery. AJNR Am
J Neuroradiol 17:1417 - 14218 1996
37. Deters OJ, Mark FF, Bargeron CB, et al: Comparison
of steady and pulsatile flow near the ventral and
dorsal walls of casts of human aortic bifurcations.
J Biomech Eng 106:79 - 828 1984
38. Bharadvaj BK, Mabon RF, Giddens DP: Steady flow
in a model of the human carotid bifurcation. Part II—
Laser-Doppler anemometer measurements. J Bio-
mech 15:363 - 3788 1982
39. Hochareon P, Manning KB, Fontaine AA, et al: Wall
shear-rate estimation within the 50cc Penn State
artificial heart using particle image velocimetry.
J Biomech Eng 126:430 - 4378 2004
40. Bale-Glickman J, Selby K, Saloner D, et al: Exper-
imental flow studies in exact-replica phantoms of
atherosclerotic carotid bifurcations under steady
input conditions. J Biomech Eng 125:38 - 488 2003
41. Natarajan S, Mokhtarzadeh-Dehghan MR: A numer-
ical and experimental study of periodic flow in a
model of a corrugated vessel with application to
stented arteries. Med Eng Phys 22:555 - 5668 2000
42. Doriot PA, Dorsaz PA: Residual stenosis poststent-
ing and subsequent decrease in the proximal
reference diameter are correlated: excessive axial
wall stress is a plausible explanation. J Endovasc
Ther 11:310 - 3188 2004
KATRITSIS ET AL
43. Wentzel JJ, Corti R, Fayad ZA, et al: Does shear
stress modulate both plaque progression and regres-
sion in the thoracic aorta? Human study using serial
magnetic resonance imaging. J Am Coll Cardiol
45:846 - 8548 2005
44. Wentzel JJ, Krams R, Schuurbiers JC, et al: Rela-
tionship between neointimal thickness and shear
stress after wallstent implantation in human coronary
arteries. Circulation 103:1740 - 17458 2001
45. Thury A, van Langenhove G, Carlier SG, et al: High
shear stress after successful balloon angioplasty is
associated with restenosis and target lesion revas-
cularization. Am Heart J 144:136 - 1438 2002
46. Boutsianis E, Dave H, Frauenfelder T, et al: Com-
putational simulation of intracoronary flow based on
real coronary geometry. Eur J Cardiothorac Surg
26:248 - 2568 2004
47. Stokholm R, Oyre S, Ringgaard S, et al: Determina-
tion of wall shear rate in the human carotid artery by
magnetic resonance techniques. Eur J Vasc Endo-
vasc Surg 20:427 - 4338 2000
48. Butcher JT, Tressel S, Johnson T, et al: Transcrip-
tional profiles of valvular and vascular endothelial
cells reveal phenotypic differences: Influence of shear
stress. Arterioscler Thromb Vasc Biol 26:69 - 778 2006
49. Lipowsky HH, Kovalcheck S, Zweifach BW: The
distribution of blood rheological parameters in the
microvasculature of cat mesentery. Circ Res
43:738 - 7498 1978
50. Sawchuk AP, Unthank JL, Davis TE, et al: A
prospective, in vivo study of the relationship between
blood flow hemodynamics and atherosclerosis in a
hyperlipidemic swine model. J Vasc Surg 19:58 - 638
1994 (discussion 63-4)
51. Reneman RS, Arts T, Hoeks AP: Wall shear stress—
An important determinant of endothelial cell function
and structure—In the arterial system in vivo. Discrep-
ancies with theory. J Vasc Res 43:251 - 2698 2006
52. Samijo SK, Willigers JM, Brands PJ, et al: Reproduc-
ibility of shear rate and shear stress assessment by
means of ultrasound in the common carotid artery of
young human males and females. Ultrasound Med
Biol 23:583 - 5908 1997
53. Irace C, Cortese C, Fiaschi E, et al: Wall shear stress
is associated with intima-media thickness and carotid
atherosclerosis in subjects at low coronary heart
disease risk. Stroke 35:464 - 4688 2004
54. Gnasso A, Carallo C, Irace C, et al: Association
between intima-media thickness and wall shear
stress in common carotid arteries in healthy male
subjects. Circulation 94:3257 - 32628 1996
55. Gnasso A, Irace C, Carallo C, et al: In vivo association
between low wall shear stress and plaque in subjects
with asymmetrical carotid atherosclerosis. Stroke
28:993 - 9988 1997
56. Oyre S, Pedersen EM, Ringgaard S, et al: In vivo wall
shear stress measured by magnetic resonance
velocity mapping in the normal human abdominal
aorta. Eur J Vasc Endovasc Surg 13:263 - 2718 1997
57. Oyre S, Paaske WP, Ringgaard S, et al: Automatic
accurate noninvasive quantitation of blood flow,
WALL SHEAR STRESS
cross-sectional vessel area, and wall shear stress by
modelling of magnetic resonance velocity data. Eur J
Vasc Endovasc Surg 16:517 - 5248 1998
58. Naderson C, Edelman R: Clinical magnetic resonance
angiography. Raven Pr8 1993
59. Yim P, Demarco K, Castro MA, et al: Characteriza-
tion of shear stress on the wall of the carotid artery
using magnetic resonance imaging and computa-
tional fluid dynamics. Stud Health Technol Inform
113:412 - 4428 2005
60. Doucette JW, Corl PD, Payne HM, et al: Validation
of a Doppler guide wire for intravascular measure-
ment of coronary artery flow velocity. Circulation
85:1899 - 19118 1992
61. Li W, van der Steen AF, Lancee CT, et al: Potentials of
volumetric blood-flow measurement. Semin Interv
Cardiol 2:49 - 548 1997
62. Gibson CM, Diaz L, Kandarpa K, et al: Relation of
vessel wall shear stress to atherosclerosis progres-
sion in human coronary arteries. Arterioscler Thromb
13:310 - 3158 1993
63. Johnston BM, Johnston PR, Corney S, et al: Non-
Newtonian blood flow in human right coronary arteries:
Transient simulations. J Biomech 39:1116 - 11288
2006
64. Bourantas CV, Kourtis IC, Plissiti ME, et al: A method
for 3D reconstruction of coronary arteries using biplane
angiography and intravascular ultrasound images.
Comput Med Imaging Graph 29:597 - 6068 2005
65. Wentzel JJ, Whelan DM, van der Giessen WJ, et al:
Coronary stent implantation changes 3-D vessel
geometry and 3-D shear stress distribution. J Bio-
mech 33:1287 - 12958 2000
66. Krams R, Wentzel JJ, Oomen JA, et al: Evaluation of
endothelial shear stress and 3D geometry as factors
determining the development of atherosclerosis and
remodeling in human coronary arteries in vivo.
Combining 3D reconstruction from angiography and
IVUS (ANGUS) with computational fluid dynamics.
Arterioscler Thromb Vasc Biol 17:2061 - 20658 1997
67. Garcia J, Crespo A, Goicolea J, et al: Study of the
evolution of the shear stress on the restenosis
after coronary angioplasty. J Biomech 39:799 - 8058
2006
68. Jung J, Hassanein A, Lyczkowski RW: Hemodynamic
computation using multiphase flow dynamics in a
right coronary artery. Ann Biomed Eng 34:393 - 4078
2006
69. Lou Z, Yang WJ, Stein PD: Errors in the estimation of
arterial wall shear rates that result from curve fitting of
velocity profiles. J Biomech 26:383 - 3908 1993
70. Oshinski JN, Ku DN, Mukundan Jr S, et al: Determi-
nation of wall shear stress in the aorta with the use of
MR phase velocity mapping. J Magn Reson Imaging
5:640 - 6478 1995
71. Guo Z, Moreau M, Rickey DW, et al: Quantitative
investigation of in vitro flow using three-dimensional
329
colour Doppler ultrasound. Ultrasound Med Biol
21:807 - 8168 1995
72. Long DS, Smith ML, Pries AR, et al: Microviscom-
etry reveals reduced blood viscosity and altered
shear rate and shear stress profiles in microvessels
after hemodilution. Proc Natl Acad Sci U S A
101:10060 - 100658 2004
73. Masaryk AM, Frayne R, Unal O, et al: In vitro and in
vivo comparison of three MR measurement methods
for calculating vascular shear stress in the internal
carotid artery. AJNR Am J Neuroradiol 20:237 - 2458
1999
74. Wentzel JJ, Gijsen FJ, Stergiopulos N, et al: Shear
stress, vascular remodeling and neointimal formation.
J Biomech 36:681 - 6888 2003
75. Soulis JV, Giannoglou GD, Chatzizisis YS, et al:
Spatial and phasic oscillation of non-Newtonian wall
shear stress in human left coronary artery bifurca-
tion: An insight to atherogenesis. Coron Artery Dis
17:351 - 3588 2006
76. Soulis JV, Farmakis TM, Giannoglou GD, et al: Wall
shear stress in normal left coronary artery tree.
J Biomech 39:742 - 7498 2006
77. Santamarina A, Weydahl E, Siegel Jr JM, et al:
Computational analysis of flow in a curved tube
model of the coronary arteries: Effects of time-varying
curvature. Ann Biomed Eng 26:944 - 9548 1998
78. Panton RL: Incompressible flow. New York: Wiley-
Interscience8 1984
79. Kundu PK, Cohen IM: Fluid mechanics. (ed 2). San
Diego: Academic Press8 2002
80. Schlichting H, Gersten K: Boundary-layer theory.
(ed 8). Berlin: Springer-Verlag8 2000
81. Walburn FJ, Schneck DJ: A constitutive equation for
whole human blood. Biorheology 13:201 - 2108 1976
82. Ballyk PD, Steinman DA, Ethier CR: Simulation of
non-Newtonian blood-flow in an end-to-side anasto-
mosis. Biorheology 31:565 - 5868 1994
83. Triton DJ: Physical fluid dynamics. (ed 2). Oxford:
Clarendon Press8 1988
84. Tangelder GJ, Teirlinck HC, Slaaf DW, et al: Distri-
bution of blood platelets flowing in arterioles. Am J
Physiol 248:H318 - H3238 1985
85. Eckstein EC: Rheophoresis—a broader concept of
platelet dispersivity. Biorheology 19:717 - 7248 1982
86. CD Adapco Group. Star-CD Version 3.20 Methodolo-
gy, 2004
87. CD Adapco Group. Star-CD Version 3.20 User Guide,
2004
88. CD Adapco Group. Star-CD Version 3.24 Methodolo-
gy, 2004
89. CD Adapco Group. Star-CD Version 3.24 User Guide.
2004
90. Ferziger JH, Peric M: Computational methods for
fluid dynamics (ed 3). Berlin: Springer8 2001
91. Berne RM, Levy MN: Cardiovascular physiology
(ed 8). Mosby8 2001