Sie sind auf Seite 1von 15

Evangelos Boutsianis

Laboratory of Thermodynamics in Emerging


Technologies,
Department of Mechanical and Process
Engineering,
ETH Zurich,
CFD and PTV Steady Flow
8092 Zurich, Switzerland
Investigation in an Anatomically
Michele Guala
Institute of Environmental Engineering,
ETH Zurich,
Accurate Abdominal Aortic
8092 Zurich, Switzerland
Aneurysm
Ufuk Olgac
Laboratory of Thermodynamics in Emerging There is considerable interest in computational and experimental flow investigations
Technologies, within abdominal aortic aneurysms (AAAs). This task stipulates advanced grid genera-
Department of Mechanical and Process tion techniques and cross-validation because of the anatomical complexity. The purpose
Engineering, of this study is to examine the feasibility of velocity measurements by particle tracking
ETH Zurich, velocimetry (PTV) in realistic AAA models. Computed tomography and rapid prototyping
8092 Zurich, Switzerland were combined to digitize and construct a silicone replica of a patient-specific AAA.
Three-dimensional velocity measurements were acquired using PTV under steady aver-
Simon Wildermuth aged resting boundary conditions. Computational fluid dynamics (CFD) simulations were
Institute of Diagnostic Radiology, subsequently carried out with identical boundary conditions. The computational grid was
University Hospital of Zurich, created by splitting the luminal volume into manifold and nonmanifold subsections. They
Raemistrasse 100, were filled with tetrahedral and hexahedral elements, respectively. Grid independency
8091 Zurich, Switzerland was tested on three successively refined meshes. Velocity differences of about 1% in all
three directions existed mainly within the AAA sack. Pressure revealed similar variations,
Klaus Hoyer with the sparser mesh predicting larger values. PTV velocity measurements were taken
Institute of Environmental Engineering, along the abdominal aorta and showed good agreement with the numerical data. The
ETH Zurich, results within the aneurysm neck and sack showed average velocity variations of about
8092 Zurich, Switzerland 5% of the mean inlet velocity. The corresponding average differences increased for all
velocity components downstream the iliac bifurcation to as much as 15%. The two do-
Yiannis Ventikos mains differed slightly due to flow-induced forces acting on the silicone model. Velocity
Department of Engineering Science, quantification through narrow branches was problematic due to decreased signal to noise
University of Oxford, ratio at the larger local velocities. Computational wall pressure and shear fields are also
Parks Road, presented. The agreement between CFD simulations and the PTV experimental data was
Oxford OX1 3PJ, UK confirmed by three-dimensional velocity comparisons at several locations within the in-
vestigated AAA anatomy indicating the feasibility of this approach.
Dimos Poulikakos1 关DOI: 10.1115/1.3002886兴
Laboratory of Thermodynamics in Emerging
Technologies, Keywords: abdominal aortic aneurysm, particle tracking velocimetry, computational
Department of Mechanical and Process fluid dynamics, grid generation, hemodynamics
Engineering,
ETH Zurich,
8092 Zurich, Switzerland
e-mail: dimos.poulikakos@ethz.ch

Introduction pathological boundary conditions. The plausible connection of


atherosclerosis with regions of low and oscillating WSS 关4,5兴 sus-
Abdominal aortic aneurysms 共AAAs兲 form secondary to athero-
tains the wide interest in arterial fluid mechanics.
sclerosis due to atrophy or necrosis of the arterial media. Several
biomechanical parameters can be taken into account to formulate Hemodynamics in the Physiologic Abdominal Aorta. Studies
case specific indicators of the risk of rupture 关1兴. Computational in the physiological abdominal aorta preceded those of AAA
fluid dynamics 共CFD兲 is used to quantify relevant hemodynamics cases. Recent CFD investigations include Lee and Chen 关6兴 who
indices based on velocity, wall shear stress 共WSS兲, and pressure produced steady three-dimensional simulations in a rigid model of
within the aneurismal sack 关2,3兴. This is a task of increased diffi- the abdominal aorta that included tapering, the celiac trunk, the
culty necessitating advanced grid generation techniques and ex- renal and iliac bifurcations, and the superior and inferior mesen-
perimental cross-validation because of the complexity of the un- teric arteries. Steady flow simulations revealed basic branching
derlying anatomy and the need for in vivo physiological and/or flow characteristics, i.e., WSS maxima located around bifurcation
apexes and low WSS regions with possible recirculation along the
lateral walls depending on the applied mass discharge ratios. Fur-
1
Corresponding author. ther regions of low WSS with or without recirculation are pre-
Contributed by the Bioengineering Division of ASME for publication in the JOUR-
NAL OF BIOMECHANICAL ENGINEERING. Manuscript received July 18, 2007; final manu-
dicted along the posterior aortic wall at the level of the diaphragm
script received June 27, 2008; published online November 21, 2008. Review con- due to the visceral branches. Taylor et al. 关7,8兴 conducted pulsatile
ducted by B. Barry Lieber. simulations under resting and exercise conditions in a similar rigid

Journal of Biomechanical Engineering Copyright © 2009 by ASME JANUARY 2009, Vol. 131 / 011008-1

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
model, which included the physiological lumbar curvature. Under Scope of the Present Investigation. Modern medical imaging
resting conditions, a recirculation zone formed along the posterior modalities, i.e., computed tomography 共CT兲 and magnetic reso-
wall of the infrarenal aorta in late systole lasting through diastole nance imaging 共MRI兲, enable the reconstruction of patient-specific
as well. This zone, in addition to the region between the superior arterial geometry. Anatomically accurate AAA reconstructions
and inferior mesenteric arteries upon the anterior wall, correlated were used to simulate blood flow dynamics 关27兴, or pure mechani-
well with low time-averaged WSS and with high WSS temporal cal wall stress distributions 关28兴 and more recently fluid-solid in-
oscillations. teractions 关29,30兴. Unstructured tetrahedral grids are the intuitive
Multibranched stiff aortic models have been investigated ex- option but more efficient techniques suited to patient-specific ana-
perimentally with a variety of techniques. Ku et al. 关9兴 and Moore tomical characteristics are now appearing. Wolters et al. 关30兴 put
et al. 关10兴 used a blown glass model of the physiological abdomi- forward a hexahedral meshing technique that adapts a standard-
nal aorta for flow visualization by dye injections under various ized bifurcated mesh according to the centerlines and the bound-
steady and pulsatile flow regimes. Pedersen et al. 关11兴 produced ary surface of the segmented AAA.
Although much has been done on the experimental front, there
transient flow visualizations by photographing the path lines of
is still considerable room for improvement. MR velocimetry is
neutrally buoyant particles and subsequently 关12兴 used the same
usually hampered by long acquisition times, especially when all
model and flow conditions to quantify instantaneous orthogonal three velocity components are interrogated, and suffers from in-
velocity components along the aortic centerline with laser Doppler sufficient resolution both in in vivo and in vitro settings. PIV
anemometry 共LDA兲. Noninvasive in vivo velocity quantification methods achieve high spatial resolution but are inherently limited
in the abdominal aorta was benefited by development of special- to an essentially two-dimensional observation plane. LDV is a
ized magnetic resonance 共MR兲 procedures 关13兴. Moore and Ku very accurate three-dimensional velocimetry technique with high
关14,15兴 successfully acquired axial pulsatile velocity measure- temporal resolution. However, being a single point measurement,
ments by MR in the same glass aortic model and under the bound- it requires complex positioning hardware to fully resolve the ve-
ary conditions that were used for flow visualization in Refs. 关9兴 locity field within a realistic AAA model.
and 关10兴. A comparison of in vitro and in vivo axial velocity An alternative is particle tracking velocimetry 共PTV兲, which is
measurements by MR 关16兴 revealed good qualitative agreement a quantitative optical technique comprising of 共1兲 individual par-
but at the same time underlined the importance of actual in vivo ticle identification via stereoscopic matching, 共2兲 coordinate as-
boundary conditions. signment in physical space via a ray tracing mathematical model
based on external and internal camera calibration parameters, and
Hemodynamics in Abdominal Aortic Aneurysms. CFD in- 共3兲 the repeated tracking of individual particles with time. PTV
vestigations in AAA cases commenced with simplified branchless relies on recording multiple synchronous images of the illumi-
three and/or two-dimensional axis-symmetrical models under nated particles within the flow field. Maas et al. 关31兴 enhanced
steady and unsteady boundary conditions for a physiologically classical PTV exploiting digital image processing to determine
realistic range of Reynolds numbers 关17,18兴. The importance of through photogrammetry the three-dimensional coordinates,
asymmetry was pointed out by three-dimensional investigations in whereas Malik et al. 关32兴 implemented automated particle track-
single and double aneurysm geometries by Finol et al. 关19兴 and ing schemes. This system was capable of determining coordinate
Kumar 关20兴, respectively. Asymmetry develops mainly because sets of some 1000 particles in a flow field at a time resolution of
the posterior wall of the infrarenal aorta is constrained by the 25 datasets per second. This method was successfully used by
presence of the vertebral column. These studies have indicated Virant and Dracos 关33兴 to obtain Lagrangian velocity fields and by
that recirculation zones with annular vortex growth and shedding Luthi et al. 关34兴 for vorticity dynamics in turbulent flows. Kieft et
are present within the aneurismal sack during the systolic decel- al. 关35兴 made measurements in the wake flow behind a heated
eration and diastole. Low and oscillatory WSS regions reside cylinder and validated its accuracy with the exact solution. In
along most of the aneurismal wall depending on the degree of these PTV configurations, the entire volumetric region of interest
asymmetry. The highest WSS and WSS gradients are reached at is simultaneously illuminated and the particle trajectories are re-
the distal end. On the other hand, wall pressure remains nearly corded. Recently, Hoyer et al. 关36兴 developed scanning particle
uniform along the expansion. Significant pressure oscillation has tracking velocimetry 共SPTV兲. In SPTV, the flow field is recorded
been reported at or near the distal aneurysm exit as well, indicat- by sequential tomographic high-speed imaging of the region of
ing susceptibility to rupture. interest accomplishing tracking of an average of 3500 particles
Several researchers have undertaken experimental investiga- per time step.
tions in simplified transparent AAA models. Fukushima et al. 关21兴 The purpose of this study is to examine the feasibility of veloc-
visualized pulsatile flow in glass aneurysm models by releasing ity measurements by PTV in realistic AAA models. CFD simula-
disk shaped aluminum particles and photographing their trajecto- tions are utilized by developing appropriate grid generation tech-
ries. Asbury et al. 关22兴 utilized color Doppler imaging for visual- niques and are cross-validated by PTV experiments in an
ization and laser Doppler velocimetry 共LDV兲 to quantify axial anatomically accurate AAA replica under averaged steady resting
velocities within symmetric elliptic and rigid silicone AAA mod- boundary conditions.
els under steady inflow conditions. LDV was also utilized to mea-
sure centerline velocity in pulsatile experiments at realistic resting Methods
and exercise inflows by Egelhoff et al. 关23兴 in a wide range of
symmetric AAA sizes in addition to an asymmetric one. In vivo Anatomical Data Acquisition and Model Reconstruction. To
pulsatile resting flow conditions were successfully simulated in produce a realistic three-dimensional model of a patient’s
fusiform AAAs by Peattie et al. 关24兴. This investigation presented anatomy, CT-datasets were acquired by a four-row-detector
two-dimensional transient velocity data across the longitudinal CT-scanner.2 The selected patient had a fusiform aneurysm in the
center-plane by a dual-beam LDV system. Digital particle image infrarenal aorta with a maximum diameter of approximately
velocimetry 共PIV兲 was used by Bluestein et al. 关25兴 under steady 5.5 cm. An event triggering protocol3 was used to start data ac-
conditions in another axis-symmetric AAA model at a 10:1 scale quisition at a predefined level of opacification in a monitored re-
to enhance resolution. Nevertheless, refractive index mismatching gion. In the present case, this region was placed inside the distal
produced significant optical distortion near the wall. Unsteady thoracic aorta. Helical type data were captured at a rotation time
PIV velocity quantifications were reported by Yu et al. 关26兴 under
sinusoidal inflow conditions along the longitudinal symmetry 2
Somatom Volume Zoom, Siemens Medical Solution, Erlangen, Germany.
3
plane of axis-symmetric AAA Pyrex glass models. CARE Bolus System, Siemens Medical Solution, Forchheim, Germany.

011008-2 / Vol. 131, JANUARY 2009 Transactions of the ASME

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Table 1 Distribution of outlet volumetric flow rates

Average weight Distribution Volumetric flux


Outlet 共g兲 共%兲 共m3 / s兲

Sup. 654.5 10.817 1.821⫻ 10−5


mesenteric
Left renal 744.5 12.305 2.071⫻ 10−5
Right renal 738.5 12.206 2.055⫻ 10−5
Celiac 785.5 12.982 2.185⫻ 10−5
(a) Left ext. iliac 785.0 12.974 2.184⫻ 10−5
Left int. iliac 713.0 11.784 1.984⫻ 10−5
Right int. iliac 844.5 13.958 2.349⫻ 10−5
Right ext. iliac 785.0 12.974 2.184⫻ 10−5
Total 1.683⫻ 10−4

quired CT scans were segmented and smoothed as before to gen-


erate the luminal surface reconstruction for the computational do-
main. To allow for a direct comparison between experiments and
simulations, the computational domain was retained in the in vitro
scale.
Experimental Arrangement
Flow Circuit. To establish flow through the silicone phantom, a
closed circuit system was assembled, shown in Fig. 1共b兲. A mix-
(b) ture of 40% water and 60% glycerol at 23° C with a density of
1156 kg/ m3 served as the working fluid through the model as
Fig. 1 Overview of the experimental setup. Panel „a…: Photo- well as for filling the glass tank around it. This was necessary for
graph of the AAA model tank. The terminal branches are la-
beled as „1… supraceliac aorta, „2… celiac trunk, „3… superior me-
matching the index of refraction between the running fluid and the
senteric, „4… left renal, „5… right renal, „6… left external iliac, „7… phantom’s walls to minimize optical distortion. The refraction in-
right external iliac, „8… left internal iliac, and „9… right internal dex of the silicone model was measured by a handheld
iliac. Panel „b…: Schematic of the experimental setup with indi- refractometer7 at 1.413. To prevent any permanent deformation of
cated parts: „1… reservoir, „2… volumetric pump, „3… bypass-able the flexible AAA walls, the silicone model and its tank were filled
filter, „4… electromagnetic flow meter, „5… inflow conditioner, „6… simultaneously at very slow rate. The enclosing tank was filled to
AAA model tank, „7… outflow lines from the lower branches, „8… the maximum possible and an airtight plate sealed the topside.
outflow lines from the upper branches, and „9… movable optical This configuration restricted any overall volumetric expansion
module. during the measurements even after the establishment of flow.
A variable speed volumetric gear pump drove the fluid at
10.1 l / min. Its speed was regulated by measuring the total inflow
of 0.5 s. The patient had received 150 ml of nonionic contrast in the AAA model with an electromagnetic flow meter8 having a
material4 at an injection rate of 3 ml/ s. nominal accuracy of 0.5%. The flow rate was determined through
The lumen of the aortic section beginning 6 cm above the ce- dynamic similarity to achieve a Reynolds number of 560, corre-
liac trunk to the external iliac arteries including the celiac trunk, sponding to averaged inflow resting boundary conditions 关15兴.
the renal arteries, the superior mesenteric artery, and the common The inlet of the model was placed in the supraceliac section of the
and internal iliac arteries was segmented semi-automatically by aorta and had a diameter of 4.45 cm. The kinematic viscosity9 of
utilizing the specialized commercial software AMIRA 3.1.1.5 The the water-glycerin mixture is 8.3⫻ 10−6 m2 / s at the temperature
arteries protruding from the abdominal aorta were cut off distally of 23° C. An inflow conditioner was placed just upstream of the
to their respective origins after a distance equal to five local di- phantom’s inlet to control the spatial velocity profile at the inlet.
ameter lengths. A thrombus completely blocking the inferior me- The device consisted of two parts. The upstream part was a
senteric artery was present and was excluded. A triangular un- smooth contraction containing a coarse filter mat to achieve a
structured surface mesh was built over the marked volume using uniform velocity distribution. The flow strainer was followed by
the marching cube algorithm. Constrained smoothing was applied an interchangeable straight pipe. Its length permitted control of
by means of a Laplacian filter. The resulting geometrical informa- the development of the spatial velocity profile. A 60 l tank served
tion on the anatomy of the inner aortic walls was used to construct as the main reservoir of the working fluid as well as the collection
a thick, flexible, and transparent silicone elastomer phantom to point for all outlets, suitably positioned to maintain a constant
facilitate the experimental flow investigation. External arterial outflow pressure. To simplify things, no effort was made to adjust
walls were built to ensure structural stability, targeting the mini- the flow discharge ratios through the various aortic branches. Nev-
mum possible wall thickness to minimize optical distortion. The ertheless, the acquired volumetric flux through each outlet was
thinnest walls were approximately 1.5 mm thick. This model was measured by the stopwatch and receptacle method to provide
manufactured by a specialist,6 at a scale of 1.5:1 to increase spa- boundary conditions to the computational simulations. For 30 s
tial resolution and was mounted within a custom built transparent the fluid flowing from each outlet was collected and afterward
glass tank, shown in Fig. 1共a兲. To minimize deviations between weighed. This procedure was repeated two times for each outlet,
the manufactured model and the CFD geometry, the tank with the and the averaged results are shown in Table 1.
phantom was scanned filled with a mixture of calcium chloride
and water that matched the density of the working fluid. The ac-
7
Eclipse, Bellingham & Stanley Ltd.
8
KROHNE Altometer IFC 080, KROHNE Messtechnik GmbH & Co. KG, Duis-
4
Visipaque 320, Amersham Health, Buckinghamshire, United Kingdom. burg, Germany.
5 9
Mercury Computer Systems Inc., Duesseldorf, Germany. The Dow Chemical Company,
6
Elastrat Sàrl, Geneva, Switzerland. http://www.dow.com/glycerine/resources/table18.htm.

Journal of Biomechanical Engineering JANUARY 2009, Vol. 131 / 011008-3

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
d2 Re ␯
Vs = · 共␳ p − ␳ f 兲 · 兩g兩 and Vm = 共1兲
18␮ D
where d = 300 ␮m is the particle diameter, ␳ p = 1026 kg/ m3 is the
particle’s density, ␳ f = 1156 kg/ m3 is the fluid’s density, ␮ = 9.6
⫻ 10−3 Pa s is the fluid’s dynamic viscosity at the operating tem-
perature, 兩g兩 = 9.81 m / s2 is the gravitational acceleration, ␯ = 8.3
⫻ 10−6 m2 / s is the fluid’s kinematic viscosity, Re= 560 is the Rey-
nolds number, and D = 4.45 cm is the AAA inlet’s diameter. The
resulting velocities are Vs = 6.6⫻ 10−4 m / s and Vm ⬇ 0.1 m / s, re-
spectively. The velocity ratio, Vs / Vm ⬇ 6.4⫻ 10−3, shows that the
buoyancy effects can be neglected and that these particles are
suitable for this experiment.
The actual experiments were comprised of three sequential
steps: 共i兲 Calibration of the camera orientation with respect to the
observation volume, 共ii兲 recording of particle images, and 共iii兲
postprocessing, involving stereoscopic matching, tracking over
time, and further analysis of the resulting velocity field. Camera
calibration was performed prior to each measurement by photo-
graphing a dedicated target having markers with known coordi-
nates placed at three different depth levels around the current
region of interest of the AAA model. The utilized setup is auto-
mated and is capable of tracking about 1000 particles simulta-
neously. The maximum resolvable velocity was estimated at ap-
proximately 2 m / s for a camera frame rate of 250 Hz according
to the so-called tracking criterion, which requires that the maxi-
mum individual displacement per time step be of the order of the
mean particle separation. The absolute position error is estimated
at 50 ␮m and results from the error of the epipolar line intersec-
tions with respect to the particle image. The circulating fluid was
seeded at a density of 1.4 particles/ cm3. These numbers imply a
mean particle distance of about 9 mm. The obtained results con-
sisted of three-dimensional coordinates and corresponding veloc-
ity vectors within each region of interest. An extensive description
Fig. 2 Overview of the PTV setup. Panel „a…: Photograph of the
of the entire experimental setup and procedure may be found in
image acquisition system. Its main components are labeled as Ref. 关37兴.
„1… camera, „2… four way splitter prism, „3… mirrors and „4… opti-
cal rail. Panel „b…: Photographic output showing the four differ- Computational Fluid Dynamics Setup
ent perspectives.
Grid Generation. The computational domain is complex and
consists of the main trunk of the abdominal aorta and eight
branches including seven bifurcations. The various anatomical re-
gions as well as the names of the included branches are labeled in
Particle Tracking Velocimetry Setup. The PTV hardware setup Fig. 3共a兲. The elements of a volumetric mesh should satisfy cer-
consists of three main components: An image acquisition system, tain criteria of geometrical quality based on aspect ratio and skew-
an illumination facility, and the tracer particles. A high-speed ness definitions. Modern commercial grid generators ensure that
video camera system10 was used, with a lens11 of focal length, f the vast majority of the generated elements satisfy these criteria.
= 50 mm. The camera, Fig. 2共a兲, was positioned in front of a In addition, we decided to impose the following rules. First, the
four-way splitter prism, and a set of four secondary mirrors was elements’ size or volume should be smaller in the near wall re-
used to capture four different perspectives of the region of inter- gions because of the higher velocity gradients there. Second, both
est. The mirrors, the prism, and the camera were mounted onto an surface and volumetric elements should be allowed to rescale
optical rail. The raw image depicted in Fig. 2共b兲 shows the entire from a branch with a large diameter to a branch with a smaller
1024⫻ 1024 imager chip. The four different perspectives, as seen one. Any two cross sections should approximately have the same
through the image splitter, were recorded simultaneously at a rate number of elements regardless of their diameter. Third, hexahe-
of 250 frames per second. Splitting the image along the vertical dral elements are better suited to regions with streamlined flow.
and horizontal symmetry lines resulted in a 512⫻ 512 pixels im- This allows us to increase their aspect ratio along the main flow
age for each viewpoint. A mercury vapor lamp12 was used as a direction and thus decrease the required number of elements. Tet-
light source to illuminate the tank through the bottom window. To rahedral elements perform better in regions with intense
limit illumination only to the region of interest a light guide and recirculation.
an aperture were used underneath the tank. Particles with diam- We propose a hybrid technique based on splitting the luminal
eters spanning from 200 ␮m to 315 ␮m were used as fluid tracers volume into manifold and nonmanifold subsections and filling
and were added directly to the reservoir. To check whether sig- them with tetrahedral and hexahedral elements, respectively, Fig.
nificant discrepancies might arise between the fluid and particle 3共a兲. The procedure starts by specifying the number of nodes upon
motion we estimated their settling velocity Vs by Stokes’ law, all boundary and dividing edges. The length of the created inter-
assuming that they were spherical, and compared it with the ex- vals is given as a constant fraction of each edge’s perimeter, R1.
pected average inlet velocity Vm by Eq. 共1兲. The luminal walls that belong to nonmanifold volumes are
meshed with quadrilateral elements. The size of the generated
10
Fastcam-ultima APX.
quadrilaterals is controlled by R1 in the circumferential sense and
11
Nikkor. by R2 in the longitudinal direction. R2 is defined as a weighted
12
HTI 400, Osram. fraction of the end edges’ perimeters and is in general larger than

011008-4 / Vol. 131, JANUARY 2009 Transactions of the ASME

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 3 Panel „a…: Overview of the generated luminal surface mesh showing
the division into manifold and nonmanifold domains. Panel „b…: Definition of
the main gridding parameters Ri on the inlet. Panel „c…: Close-up of the
surface grid in the vicinity of the visceral branches.

Journal of Biomechanical Engineering JANUARY 2009, Vol. 131 / 011008-5

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 4 Perspective, panel „a…, and planar, panel „b…, views of the cross section and the polyline utilized
in the grid independency study

Table 2 Generated mesh particulars continuity equation and the incompressibility of the running fluid,
the flux through this specific outlet matched the indicated value of
Mesh No. R1 R2 R3 Total elements Table 1. The walls of the AAA model did not move during the
performed steady flow experiment. A no-slip velocity boundary
1 0.075 0.150 0.025 597,550 condition was imposed at the inner arterial walls.
2 0.060 0.120 0.020 963,299
3 0.045 0.090 0.015 1,813,568
To allow for an easier comparison with the experimental results
all simulations were performed at the in vitro scale. The compu-
tational fluid was assumed to have a Newtonian behavior with the
density and viscosity that were specified above for the water-
R1. The wall surfaces that bound manifold volumes are meshed glycerin mixture. The flow was assumed to be incompressible,
with triangular elements. The gradual resizing of the surface ele- laminar, and steady. A multiphysics solver, CFD-ACE⫹,14 was used
ments is controlled by linear size limitation functions with a to solve the Navier–Stokes equations. This code is based on a
growth rate of 1.05. The cross-sectional faces bounded by both finite volume discretization method and employs a pressure-
boundary and dividing edges are gridded either with quadrilateral velocity coupling variant of the SIMPLEC algorithm. The numeri-
or with triangular elements depending on whether they belong to a cal solution was acquired with second order spatial accuracy,
nonmanifold or manifold volume, respectively. Those faces form- while using an algebraic multigrid convergence acceleration tech-
ing the intersections between manifold and nonmanifold domains nique.
possess different grids on each side forming arbitrary interfaces.
The nonmanifold volumes are gridded with hexahedral ele- Results
ments by sweeping the mesh of the cross-sectional end faces.
Manifold volumes are filled with tetrahedral elements. To manipu- Grid Independency Study. To check the grid independency of
late the size of the generated elements we used a combination of the computational results, we performed simulations with the
prismatic boundary layers and additional linear sizing functions meshes of Table 2. All calculations were carried out in parallel on
attached upon each surrounding wall surface. The size of each a 32 bit dual xeon 3.0 GHz workstation with 4 Gbytes of RAM. It
prismatic layer grows according to a geometric series, and the can be seen from Table 2 that the number of elements nearly
total height is limited to 7.5% of the average of the corresponding doubles between successive refinements. A decrease of six orders
edges’ perimeters divided by ␲. A new parameter, R3, again de- of magnitude in the equation residuals was achieved in all cases.
fined as a weighted fraction of the corresponding end edges’ pe- To establish a common ground for comparing the simulation
rimeters was used to control the size of the first layer of the results we studied several planar cross sections within the AAA
prismatic elements as well as the element minimum allowable size model, all of which revealed similar behavior. One of them, at
in the sizing function definition. The growth rates of both the Y = 0.14 m, running through most of the computational domain
prismatic boundary layers and the sizing functions were set at 1.2. and intersecting both tetrahedral and hexahedral domains is shown
This procedure was programed in GAMBIT 2.2.3013 for the para- in Fig. 4共a兲. Velocity and pressure comparisons are shown in Figs.
metric generation of grids based on the a priori specification of R1, 5 and 6 for each mesh. Four panels contain line contour plots for
R2, and R3. A depiction of the definition of these parameters in the each velocity component and pressure. Corresponding line plots
inlet region is given in Fig. 3共b兲. We produced three successively for each variable and mesh along the indicated line path in Fig.
finer grids whose particulars are detailed in Table 2. A view of the 4共b兲 are plotted in Fig. 6. A plot of the cell volumes along this line
surface grid in the vicinity of the visceral branches is included in path is superimposed to the panels of this figure for each mesh. An
Fig. 3共c兲. inspection of the subplots of Figs. 5 and 6 reveals that there is
excellent convergence between the investigated grids for all
Main Assumptions and Boundary Conditions. In order to cross- solved variables. The velocity line contours overlap throughout
validate the computational and the experimental results the bound- the computational domain including the near wall regions. Visible
ary conditions were matched to those of the experimental setup. contour differences exist only within the central AAA region but
An experimentally derived three-dimensional velocity profile was in all directions remain below 1% along the indicated line path.
specified at the inlet. The volumetric outflow rates of Table 1 were By observing the cell volume lines we notice that the regions with
imposed at all outlets by means of fully developed paraboloid the most significant change in cell volume within the AAA
spatial profiles, except from the outlet of the right external iliac coincide with the regions of greater velocity differences. It is no-
artery where the static pressure was set to zero. By virtue of the ticed that this region is gridded with tetrahedral elements and

13 14
Fluent Inc., Darmstadt, Germany. ESI Group CFD, Paris, France.

011008-6 / Vol. 131, JANUARY 2009 Transactions of the ASME

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 6 Line plot comparisons of velocity components, panels
„a…–„c…, and pressure, panel „d…, along the indicated polyline in
Fig. 4„b…
Fig. 5 Contour plot comparisons of velocity components,
panels „a…–„c…, and pressure, panel „d…, along the planar cross
section indicated in Fig. 4„a…
parison with the cell volume distribution along the prescribed line
path reveals that the numerical differences intensify at the regions
experiences significant lateral recirculation. From a CFD point of of large resolution changes. In contrast to the velocity compari-
view this is expected and acceptable at the observed scale. sons, pressure variations are visible in both the upstream hexahe-
Pressure reveals similar variations of less than 1%, with the dral domain and within the AAA tetrahedral domain. Taking into
sparser mesh predicting larger absolute values. Once more, a com- account the small levels of discrepancy observed it is safe to as-

Journal of Biomechanical Engineering JANUARY 2009, Vol. 131 / 011008-7

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 7 Overview of the measurement volumes and the planar cross sections used for
the comparison between the PTV and CFD velocity quantifications

sume that we have reached at a converged solution. It is also measured by the electromagnetic flow meter and the integration of
worth mentioning that the intended mesh resizing of the proposed the experimental velocity field showed a difference, of 2.8%.
gridding technique is shown evidently in the cell volume distribu- In Figs. 8–10, we show representative plots between experi-
tion curve. The cell volume ratios between successive meshes mental and numerical velocities at two planar cross sections in the
along the investigated line path ranged between 40% and 80% X-direction for each region, i.e., the aneurysm sack, the aneurysm
with the exception of the level of the celiac trunk when passing neck, and the iliac bifurcation. The U-velocity component is pre-
from mesh 1 to mesh 2. The localization of the line path was sented with contour plots for each cross section in panels 共a兲, 共b兲,
chosen at the vessel’s centerline where larger cells reside. The 共e兲, and 共f兲. The V- and W-velocity components are included with
refinement of this region is the most difficult to achieve in un- embedded vector plots. Additional rows of line subplots accom-
structured meshes. Finally, it is also clear that the utilization of modate the quantitative comparison of the respective velocity
arbitrary interfaces, indicated by the abrupt cell volume changes components along the overlapping part of the indicated lines for
in Fig. 6, did not produce any spurious variations in the solved each cross section in panels 共c兲, 共d兲, 共g兲, and 共h兲.
variables. All subsequent numerical results presented in this study The large lateral slow recirculation zone within the aortic bulge
were acquired with the third mesh. is clearly defined by both approaches. There is only a thin forward
flowing stream that is confined in the anterior and right side of the
Experimental Measurements and Comparison With Simu- aorta, Fig. 8 panels 共a兲, 共b兲, 共e兲, and 共f兲. The posterior wall and
lation Results. Measurements were performed at four different most of the central region experience very low velocities. This
locations along the AAA model, as indicated in Fig. 7. These sites fact has been documented by both numerical and experimental
were 共1兲 the inlet of the silicone model in the supraceliac section investigations in models of the physiological aorta. In this AAA
of the aorta, 共2兲 the neck of the aneurysm, 共3兲 the anterior aneu- case the effect of the lumbar curvature is exaggerated by the di-
rysm sack, and 共4兲 the iliac bifurcation. Another observation vol- latation of the infrarenal aorta. The velocity comparisons, Fig. 8
ume was located at the level of the visceral branches; however, the panels 共c兲, 共d兲, 共g兲, and 共h兲, reveal small discrepancies in the
measurements could not be used due to unfavorable lighting con- alignment between the experimental and computational domains.
ditions caused by the increased thickness of the model wall in this The overlap between the PTV and CFD results is complete in the
region. At each sector, the experimental results were interpolated Y-direction. The shift in this direction increases up to 2 mm or
on to a fixed Cartesian grid of 30⫻ 25⫻ 20 points along the X, Y, 2.5% of the maximum aneurysm diameter in the experimental
and Z directions, respectively, due to memory constraints. The scale, Fig. 8共d兲. The predicted velocities show an average devia-
spacing between the grid points was varied to cover each region tion of less than 5% of the mean inlet velocity 共Vm = 0.1 m / s兲 in
of interest as follows: 共1兲 Supraceliac aorta 共3.3, 2.0, 2.5兲 mm, all directions with the exception of Fig. 8共d兲 where the differences
共2兲 aneurysm neck 共3.5, 2.9, 3.2兲 mm, 共3兲 aneurysm sack in the U-component rise to 11.2%.
共3.3, 3.5, 2.4兲 mm, and 共4兲 iliac bifurcation 共3.2, 3.9, 2.5兲 mm. The formation of a fast entrant jet moving from the center to the
The region of the supraceliac aorta was scanned to provide right wall is presented in agreement by both the experimental and
velocity boundary conditions at the inlet of the computational do- numerical results in Fig. 9. This is caused by the expansion of the
main, set at X = 0.545 m, Fig. 7. The inflow conditioner functioned aneurysm sack toward the left hand side as well as by the bending
properly, and a partially developed inflow profile was generated of the infrarenal aorta. Figures 9共a兲 and 9共e兲 also reveal the de-
with little or no secondary flow. We imposed this requirement velopment of the lateral recirculation zone that occupies the bulk
to the inflow velocity profile to approximate conditions of systol- of the AAA sack, shown in Fig. 8. The velocity plots of Fig. 9,
ic flow 关13兴. A comparison between the total volumetric flux panels 共c兲, 共d兲, 共g兲, and 共h兲, reveal negligible discrepancies in the

011008-8 / Vol. 131, JANUARY 2009 Transactions of the ASME

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 8 Comparison of experimental and computational velocity distributions along planar cross sec-
tions within the aneurysm sack as well as along indicated lines lying within these cross sections: Panels
„a…–„d… X = 0.19 m and panels „e…–„h… X = 0.22 m

alignment of the CFD and PTV domains. The velocity distribu- perimental velocity values in the cross sections X = 0.28 m and
tions show minor differences in all velocity components in the X = 0.31 m than those located in the aneurismal sack.
overlapping sections. The average deviation rises up to 5.3% in The iliac bifurcation is the focus of Fig. 10. The first cross
the case of the W-velocity component, shown in Fig. 9共h兲. Over- section, X = 0.16 m, was placed upstream of the bifurcation,
all, there is better agreement between the numerical and the ex- whereas the second one is set at X = 0.13 m and contains the left

Journal of Biomechanical Engineering JANUARY 2009, Vol. 131 / 011008-9

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 9 Comparison of experimental and computational velocity distributions along planar cross sec-
tions within the aneurysm neck as well as along indicated lines lying within these cross sections: Panels
„a…–„d… X = 0.28 m and panels „e…–„h… X = 0.31 m

and right common iliac arteries. A comparison with Fig. 8共a兲 re- cating branch due to the significant reduction in the cross-
veals a more uniform and slower distribution of the U-velocity sectional area. The predicted velocity values are in good agree-
component. The in-plane secondary rotation persists and extends ment with their experimentally measured counterparts in the
up to the aortic wall on all sides. The plots of the flow field at upstream cross section, Fig. 10 panels 共c兲 and 共d兲. The average
X = 0.13 m show an increase in the velocity values in each bifur- deviation goes up to 6.5% with a maximum of 12% registered in

011008-10 / Vol. 131, JANUARY 2009 Transactions of the ASME

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 10 Comparison of experimental and computational velocity distributions along planar cross sec-
tions within the iliac bifurcation as well as along indicated lines lying within these cross sections: Panels
„a…–„d… X = 0.16 m and panels „e…–„h… X = 0.13 m

the V-velocity component. A similar alignment discrepancy with and Z-directions, which affects the registered velocity differences
that of Fig. 8 is observed here too. The comparison deteriorates especially in the near wall regions. The mean velocity differences
significantly downstream the iliac bifurcation, Fig. 10 panels 共e兲– in the overlapping sections rise up to 15% in Fig. 10共h兲 with
共h兲. The corresponding cross sections are larger in the PTV than in maximum velocity deviations up to 40% of the average inlet
the CFD domain. Also there is a significant offset in both the Y- velocity.

Journal of Biomechanical Engineering JANUARY 2009, Vol. 131 / 011008-11

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 11 Normalized contour plot distributions upon the anterior and posterior sides of the aortic wall of pressure, panels
„a… and „b…, and wall shear stress magnitude, panels „c… and „d…

There are three presumable sources of error in the presented sections except in the near wall regions on the posterior side. All
experimental measurements under the assumption that the pre- three velocity components were predicted consistently within the
sented grid independent numerical solution is sufficiently accu- bulk of the AAA model.
rate. First, errors might creep in by the misalignment of the ex-
perimental versus the numerical coordinate systems. The second Wall Pressure and Wall Shear Stress Distributions. Conclud-
source is the intrinsic accuracy of the PTV method as affected by ing this study we present contour plots of wall pressure and WSS
magnitude for both the posterior and anterior sides in Fig. 11.
the local light refraction and reflection along with the presence of

冉 冊 冉冊
an adequate number of tracer particles. The accuracy of the
method is expected to deteriorate within narrow outflow vessels 1 Re ␯ 2
8 ␮ 2
PD = ·␳· and ␶ P = · Re 共2兲
and in the close proximity of the walls. Residual optical distortion 2 D ␳ D
remains a constraining factor when viewing tangentially to the
silicone wall. In such conditions, the scattered light follows an Wall pressure is normalized with the inlet’s dynamic head,
unfavorable optical path in terms of incident angle and perceived PDexpt = 5.78 Pa, and WSS with the equivalent Poiseuille inlet flow
wall thickness. Lastly, the phantom and the computational domain WSS, ␶ Pexpt = 0.18 Pa, according to Eq. 共2兲. The corresponding in
may differ slightly. Although the presence of the airtight plate seal vivo values at the same Reynolds number are PDblood = 1.99 Pa
prevented any net volumetric change, the hydrodynamic pressure and ␶ Pblood = 0.06 Pa.
and the momentum change in reference to the stagnant flow state Under steady flow conditions at Re= 560, the pressure drop
can still cause shape deformation, especially in the vicinity of the from the supraceliac aorta to the iliac bifurcation is very small at
iliac bifurcation where the main flow jet divides in the two out- ⌬P* = 0.8, which amounts to 4.6 Pa in the in vitro scale or equiva-
flow branches. Even though the computational domain was ac- lently 1.59 Pa in vivo. Wall pressure is nearly constant within the
quired by scanning a filled AAA model with its tank, it was not aneurismal bulge. The pressure drop intensifies through the vari-
possible to establish a flowing circuit within the confines of the ous branches and depends on the local morphology, size, and
CT scanner. branching angle as well as on the specified mass discharge ratios.
In general, the numerical results were in good qualitative and All outlets received approximately the same amount of flux but
quantitative agreement with the experimental results. The reso- apparently larger pressure drops are needed to drive the flow
lution of the experimental data was deemed satisfactory in all through the left and right renal arteries. Under realistic resting

011008-12 / Vol. 131, JANUARY 2009 Transactions of the ASME

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 12 Velocity streamlines within the AAA model colored with the velocity
magnitude

flow conditions the total outflow from the four visceral branches Discussion
should increase about 18% 关9兴. The maximum increase is required
The current study has used modern CT and rapid prototyping
at the celiac trunk by 7%, which shows a rather small pressure
technology to investigate a patient-specific AAA case. A com-
drop and receives too little flow in this experimental setup. Under
pulsatile flow conditions the amplitude of pressure difference os- bined PTV and CFD investigation of the flow within the entire
cillations is expected to reach approximately 300 Pa in the in vivo abdominal aorta was carried out under averaged steady physi-
scale from the supraceliac aorta to the iliac arteries 关38兴. This ological flow conditions. Our aim was to examine the feasibility
value corresponds to almost 871 Pa in the utilized experimental of performing reliable three-dimensional velocity measurements
setup. Significant pressure variations occur in the vicinity of the by PTV, which has not been attempted so far in an anatomically
iliac bifurcation even under steady flow conditions. Under pulsa- correct geometry.
tile conditions, values of the instantaneous pressure there are ex- The results confirmed the agreement between the selected nu-
pected to rise above the corresponding levels in the supraceliac merical and experimental approaches by successfully quantifying
aorta. This should be taken into account when planning for tran- the large recirculation zones that characterize aneurismal flow.
sient experimental investigations since it may affect the stability The significance of such predictions stems from the fact that the
of thin-walled silicone models. localization, morphology, and size of vortical structures have been
The results on WSS are in agreement with published data from shown to greatly influence blood clotting and the formation of
investigations in the physiological aorta 关7,38兴 as well as with intraluminal thrombus. The large velocity gradients that are ob-
more recent AAA patient-specific fluid-solid interaction simula- served at the interfaces between jet and recirculation zones as well
tions 关29,30兴. The posterior walls experience consistently lower as within the close proximity of the arterial walls are accompanied
WSS than the anterior ones. The anterior aneurismal wall reveals by strong shearing forces, which are known to stretch and deform
intense WSS spatial variation of up to ⌬␶* = 2.4, which corre- red blood cells. In addition, we presented computational quantifi-
sponds to 0.43 Pa in vitro and 0.14 Pa in vivo. This is caused by cations of the wall pressure and shear fields again revealing the
the coexistence of slow and fast lateral recirculating velocity re- characteristic patterns that have been linked to atherosclerosis and
gions. A broader overview of the flow field is shown in Fig. 12 by the formation of plaque. Besides CFD there are no available tech-
depicting velocity streamlines. Typically, maximum WSS values niques that can provide data on the distribution of shear stress
occur within the various branches and at the iliac bifurcation. The with comparable resolution.
wall regions around their orifices display both large and low WSS The grid independency of the presented computational results
areas in close proximity. This is a well documented observation in was accomplished in a systematic manner. Our meshing approach
the area of the visceral and the iliac arteries and is attributed to the has taken into account anticipated features of the investigated flow
complex branching flow patterns that develop in order to divert field, the local spatial scales of the surrounding geometry, and
sufficient flow from the main trunk of the aorta. The intensity of managed to retain the required resolution. It must be stressed here
the spatial WSS variations is expected to increase in accordance that the proposed gridding technique is static in character, and the
with the amount of diverting flow as well as under physiological grid does not deform according to the solution during the CFD
pulsatile flow conditions. It is also worth mentioning that the WSS calculations. Adaptive grid generation techniques, such as the one
magnitude is consistently lower than the pressure magnitude, proposed by Sahni et al. 关39兴, are possibly more powerful than the
which is an indication that WSS may only contribute to the weak- presented approach but at the same time increase the required
ening of the aortic wall in an indirect way through a malfunction- computational times.
ing endothelium. The direct mechanical loading of the wall must The proposed experimental methodology possesses advantages
be attributed to pressure. and some disadvantages that we need to overcome. The main

Journal of Biomechanical Engineering JANUARY 2009, Vol. 131 / 011008-13

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
advantage of PTV is its inherent three-dimensional character. determination of the relevant material properties. From a clinical
Tracer particles are known to avoid regions of large shear hinder- point of view, the investigation of the mechanics of stented AAAs
ing boundary layer measurements either with PTV or PIV. How- is of utmost importance. Li et al. 关41兴 moved toward computa-
ever, PTV does not require a minimum seeding density to get a tional fluid-solid interaction studies in patient-specific stented
good estimate of the velocity since it tracks individual particles. In AAA geometries, whereas LDA velocity quantifications have been
the presented experiment the number of the near wall resolved presented in an idealized stented AAA model by Chong and How
particle velocities was not sufficient to extract wall shear stress 关42兴.
estimations. Also, particle seeding must be increased for the im-
position of physiologically accurate resting boundary conditions. Acknowledgment
In this case the greatest part of the aortic inflow exits through the
The authors acknowledge the assistance of Rene Mathys and
visceral branches and we need to ensure that a sufficient number
Philipp Zurflueh, students of Mechanical and Process Engineering
of particles reach the AAA sack. At the same time, an increased
number of seeding particles requires an improved signal to noise in ETH Zurich, in the experimental work. Dr. Kevin Boomsma,15
ratio for particle detection, e.g., using fluorescent particles and Dr. Alex Liberzon,16 Dr. Beat Luthi,16 and Professor Wolfgang
spectral separation of the excitation wavelength scattering from Kinzelbach16 are kindly accredited for providing guidance and
the phantom wall. technical support in various subtasks of this work. We warmly
The main limitation of the presently available setup rests on the thank Dr. Med. Thomas Frauenfelder17 for his assistance in scan-
imposition of realistic boundary conditions. Obviously, this does ning and segmentation of the AAA model during the early stages
not concern the actual optical section of the experimental rig. It is of this work. This study was conducted within “Project 12” in the
first phase of the “Computer Aided and Image Guided Medical
connected with the active control of the flow circuit in order to
Interventions” 共CO-ME兲 project of the Swiss National Science
achieve the required mass flux distribution through the various
Foundation.
outlets. This can be achieved by mimicking the downstream im-
pedance characteristics of the truncated arterial branches. In the
case of steady flow the above requirement simplifies to the addi- References
tion of suitable linear resistances downstream the outflow 关1兴 Kleinstreuer, C., and Li, Z., 2006, “Analysis and Computer Program for
branches. We could also improve the construction of the inflow Rupture-Risk Prediction of Abdominal Aortic Aneurysms,” Biomed. Eng. On-
conditioner to induce the secondary motion that is present in the line, 5:19.
关2兴 Kleinstreuer, C., Hyun, S., Buchanan, J. R., Longest, P. W., Archie, J. P., and
descending aorta. CFD investigations are easier to perform under Truskey, G. A., 2001, “Hemodynamic Parameters and Early Intimal Thicken-
virtually any type of boundary conditions provided that rigorous ing in Branching Blood Vessels,” Crit. Rev. Biomed. Eng., 29共1兲, pp. 1–64.
grid generation methods are used and sufficient grid independency 关3兴 Thubrikar, M. J., Al-Soudi, J., and Robicsek, F., 2001, “Wall Stress Studies of
is established. The knowledge of the pertinent parameters is usu- Abdominal Aortic Aneurysm in a Clinical Model,” Ann. Vasc. Surg., 15共3兲,
pp. 355–366.
ally the limiting factor and not some numerical analysis or other 关4兴 Giddens, D. P., Zarins, C. K., and Glagov, S., 1993, “The Role of Fluid-
technical restriction. In vivo MR velocimetry and/or quantifica- Mechanics in the Localization and Detection of Atherosclerosis,” ASME J.
tions of the arterial impedances by ultrasound are the most reliable Biomech. Eng., 115共4兲, pp. 588–594.
关5兴 Bonert, M., Leask, R. L., Butany, J., Ethier, C. R., Myers, J. G., Johnston, K.
sources for physiologically accurate boundary conditions. Several W., and Ojha, M., 2003, “The Relationship Between Wall Shear Stress Distri-
approximations or simplifications based on entirely mathematical butions and Intimal Thickening in the Human Abdominal Aorta,” Biomed.
models are also possible 关38兴. Eng. Online, 2:18.
The utilization of a phantom with flexible walls can be per- 关6兴 Lee, D., and Chen, J. Y., 2002, “Numerical Simulation of Steady Flow Fields
in a Model of Abdominal Aorta With Its Peripheral Branches,” J. Biomech.,
ceived either as an advantage or as a disadvantage. For instance, it 35共8兲, pp. 1115–1122.
is much easier to handle a deformable model during the required 关7兴 Taylor, C. A., Hughes, T. J. R., and Zarins, C. K., 1998, “Finite Element
optical calibration. On the other hand, such a model can change its Modeling of Three-Dimensional Pulsatile Flow in the Abdominal Aorta: Rel-
shape under the action of hydrodynamic pressure. In steady flow evance to Atherosclerosis,” Ann. Biomed. Eng., 26共6兲, pp. 975–987.
关8兴 Taylor, C. A., Hughes, T. J. R., and Zarins, C. K., 1999, “Effect of Exercise on
investigations this can form a source of error. Conversely, this can Hemodynamic Conditions in the Abdominal Aorta,” J. Vasc. Surg., 29共6兲, pp.
turn into an advantage in pulsatile studies where the arterial walls 1077–1089.
are expected to deform. Another possible limitation is the adop- 关9兴 Ku, D. N., Glagov, S., Moore, J. E., and Zarins, C. K., 1989, “Flow Patterns in
tion of Newtonian working fluids. Blood displays a well docu- the Abdominal-Aorta Under Simulated Postprandial and Exercise
Conditions—An Experimental-Study,” J. Vasc. Surg., 9共2兲, pp. 309–316.
mented shear thinning behavior that becomes increasingly impor- 关10兴 Moore, J. E., Ku, D. N., Zarins, C. K., and Glagov, S., 1992, “Pulsatile Flow
tant in regions of low hydrodynamic shear. The extensive near Visualization in the Abdominal-Aorta Under Differing Physiological
stasis low strain regions that exist in AAAs might induce signifi- Conditions-Implications for Increased Susceptibility to Atherosclerosis,”
cant deviations from the hypothesized Newtonian behavior. ASME J. Biomech. Eng., 114共3兲, pp. 391–397.
关11兴 Pedersen, E. M., Yoganathan, A. P., and Lefebvre, X. P., 1992, “Pulsatile Flow
The presented results proved the feasibility of using PTV for Visualization in a Model of the Human Abdominal-Aorta and Aortic Bifurca-
three-dimensional velocity measurements within a phantom of an tion,” J. Biomech., 25共8兲, pp. 935–944.
anatomically accurate AAA, thus enriching already available mo- 关12兴 Pedersen, E. M., Sung, H. W., Burlson, A. C., and Yoganathan, A. P., 1993,
“2-Dimensional Velocity-Measurements in a Pulsatile Flow Model of the Nor-
dalities such as LDV, PIV, and MR velocimetry. It is important to mal Abdominal-Aorta Simulating Different Hemodynamic Conditions,” J.
complement the present experiment by acquiring velocity mea- Biomech., 26, pp. 1237–1247.
surements under purely pulsatile boundary conditions. Investiga- 关13兴 Boesiger, P., Maier, S. E., Liu, K. C., Scheidegger, M. B., and Meier, D., 1992,
tions on the transition to turbulence under simulated exercise con- “Visualization and Quantification of the Human Blood-Flow by Magnetic-
Resonance-Imaging,” J. Biomech., 25共1兲, pp. 55–67.
ditions are also of great clinical interest and PTV appears well 关14兴 Moore, J. E., and Ku, D. N., 1994, “Pulsatile Velocity-Measurements in a
suited for this task. The cross-validation of experimental with nu- Model of the Human Abdominal-Aorta Under Resting Conditions,” ASME J.
merical results has also shown that CFD can play a central role in Biomech. Eng., 116共3兲, pp. 337–346.
the investigation of the vascular hemodynamics of patients pro- 关15兴 Moore, J. E., and Ku, D. N., 1994, “Pulsatile Velocity-Measurements in a
Model of the Human Abdominal-Aorta Under Simulated Exercise and Post-
vided that the imposed boundary conditions are sufficiently accu- prandial Conditions,” ASME J. Biomech. Eng., 116共1兲, pp. 107–111.
rate and that advanced grid generation strategies are utilized to 关16兴 Moore, J. E., Maier, S. E., Ku, D. N., and Boesiger, P., 1994, “Hemodynamics
render such computations free of spurious grid dependency ef- in the Abdominal-Aorta—A Comparison of In-Vitro and In-Vivo Measure-
fects. The increase in the available computational resources has ments,” J. Appl. Physiol., 76共4兲, pp. 1520–1527.
also cleared the way toward anatomically realistic fluid-solid in-
teraction simulations 关30兴. Vorp 关40兴 presented an excellent re- 15
Laboratory of Thermodynamics in Emerging Technologies, ETH Zurich.
view of research efforts aiming both at the modeling of the aneur- 16
Institute of Environmental Engineering, ETH Zurich.
17
ismal wall mechanical behavior and at the experimental Institute of Diagnostic Radiology, University Hospital of Zurich.

011008-14 / Vol. 131, JANUARY 2009 Transactions of the ASME

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
关17兴 Taylor, T. W., and Yamaguchi, T., 1994, “3-Dimensional Simulation of Blood- Aortic Aneurysms: A Fluid-Structure Interaction Study,” Comput. Struct.,
Flow in an Abdominal Aortic-Aneurysm—Steady and Unsteady-Flow Cases,” 85共11–14兲, pp. 1097–1113.
ASME J. Biomech. Eng., 116共1兲, pp. 89–97. 关30兴 Wolters, B. J. B. M., Rutten, M. C. M., Schurink, G. W. H., Kose, U., De Hart,
关18兴 Finol, E. A., and Amon, C. H., 2001, “Blood Flow in Abdominal Aortic An- J., and Van De Vosse, F. N., 2005, “A Patient-Specific Computational Model of
eurysms: Pulsatile Flow Hemodynamics,” ASME J. Biomech. Eng., 123共5兲, Fluid-Structure Interaction in Abdominal Aortic Aneurysms,” Med. Eng.
pp. 474–484. Phys., 27共10兲, pp. 871–883.
关19兴 Finol, E. A., Keyhani, K., and Amon, C. H., 2003, “The Effect of Asymmetry 关31兴 Maas, H. G., Gruen, A., and Papantoniou, D., 1993, “Particle Tracking Veloci-
in Abdominal Aortic Aneurysms Under Physiologically Realistic Pulsatile metry in 3-Dimensional Flows.1. Photogrammetric Determination of Particle
Flow Conditions,” ASME J. Biomech. Eng., 125共2兲, pp. 207–217. Coordinates,” Exp. Fluids, 15共2兲, pp. 133–146.
关20兴 Kumar, B. V. R., 2003, “A Space-Time Analysis of Blood Flow in a 3D Vessel 关32兴 Malik, N. A., Dracos, T., and Papantoniou, D. A., 1993, “Particle Tracking
With Multiple Aneurysms,” Comput. Mech., 32共1–2兲, pp. 16–28. Velocimetry in 3-Dimensional Flows. 2. Particle Tracking,” Exp. Fluids,
关21兴 Fukushima, T., Matsuzawa, T., and Homma, T., 1989, “Visualization and 15共4–5兲, pp. 279–294.
Finite-Element Analysis of Pulsatile Flow in Models of the Abdominal Aortic- 关33兴 Virant, M., and Dracos, T., 1997, “3d Ptv and Its Application on Lagrangian
Aneurysm,” Biorheology, 26共2兲, pp. 109–130. Motion,” Meas. Sci. Technol., 8共12兲, pp. 1539–1552.
关22兴 Asbury, C. L., Ruberti, J. W., Bluth, E. I., and Peattie, R. A., 1995, “Experi- 关34兴 Luthi, B., Tsinober, A., and Kinzelbach, W., 2005, “Lagrangian Measurement
mental Investigation of Steady Flow in Rigid Models of Abdominal Aortic- of Vorticity Dynamics in Turbulent Flow,” J. Fluid Mech., 528, pp. 87–118.
Aneurysms,” Ann. Biomed. Eng., 23共1兲, pp. 29–39. 关35兴 Kieft, R. N., Schreel, K. R. A. M., Van Der Plas, G. A. J., and Rindt, C. C. M.,
关23兴 Egelhoff, C. J., Budwig, R. S., Elger, D. F., Khraishi, T. A., and Johansen, K. 2002, “The Application of a 3d Ptv Algorithm to a Mixed Convection Flow,”
H., 1999, “Model Studies of the Flow in Abdominal Aortic Aneurysms During Exp. Fluids, 33共4兲, pp. 603–611.
Resting and Exercise Conditions,” J. Biomech., 32共12兲, pp. 1319–1329. 关36兴 Hoyer, K., Holzner, M., Luthi, B., Guala, M., Liberzon, A., and Kinzelbach,
关24兴 Peattie, R. A., Riehle, T. J., and Bluth, E. I., 2004, “Pulsatile Flow in Fusiform W., 2005, “3d Scanning Particle Tracking Velocimetry,” Exp. Fluids, 39共5兲,
Models of Abdominal Aortic Aneurysms: Flow Fields, Velocity Patterns and pp. 923–934.
Flow-Induced Wall Stresses,” ASME J. Biomech. Eng., 126共4兲, pp. 438–446. 关37兴 Mathys, R., and Zurflueh, P., 2004, “Experimental Measurements of the Ve-
关25兴 Bluestein, D., Niu, L., Schoephoerster, R. T., and Dewanjee, M. K., 1996, locity Field Within an Anatomically Realistic Abdominal Aortic Aneurysm,”
“Steady Flow in an Aneurysm Model: Correlation Between Fluid Dynamics Ph.D. thesis, ETH Zurich, Zurich, Switzerland.
and Blood Platelet Deposition,” ASME J. Biomech. Eng., 118共3兲, pp. 280– 关38兴 Vignon-Clementel, I. E., Figueroa, C. A., Jansen, K. E., and Taylor, C. A.,
286. 2006, “Outflow Boundary Conditions for Three-Dimensional Finite Element
关26兴 Yu, S. C. M., Chan, W. K., Ng, B. T. H., and Chua, L. P., 1999, “A Numerical Modeling of Blood Flow and Pressure in Arteries,” Comput. Methods Appl.
Investigation on the Steady and Pulsatile Flow Characteristics in Axi- Mech. Eng., 195共29–32兲, pp. 3776–3796.
Symmetric Abdominal Aortic Aneurysm Models With Some Experimental 关39兴 Sahni, O., Muller, J., Jansen, K. E., Shephard, M. S., and Taylor, C. A., 2006,
Evaluation,” J. Med. Eng. Technol., 23共6兲, pp. 228–239. “Efficient Anisotropic Adaptive Discretization of the Cardiovascular System,”
关27兴 Boutsianis, E., Frauenfelder, T., Wildermuth, S., Poulikakos, D., and Ventikos, Comput. Methods Appl. Mech. Eng., 195共41–43兲, pp. 5634–5655.
Y., 2003, “Anatomically Accurate Haemodynamic Simulations of Abdominal 关40兴 Vorp, D. A., 2007, “Biomechanics of Abdominal Aortic Aneurysm,” J. Bio-
Aortic Aneurysms,” 2003 Advances in Bioengineering, Washington, DC, paper mech., 40共9兲, pp. 1887–1902.
No. IMECE2003-42766, pp. 61–62. 关41兴 Li, Z. H., and Kleinstreuer, C., 2005, “Blood Flow and Structure Interactions
关28兴 Fillinger, M. F., Raghavan, M. L., Marra, S. P., Cronenwett, J. L., and in a Stented Abdominal Aortic Aneurysm Model,” Med. Eng. Phys., 27共5兲, pp.
Kennedy, F. E., 2002, “In Vivo Analysis of Mechanical Wall Stress and Ab- 369–382.
dominal Aortic Aneurysm Rupture Risk,” J. Vasc. Surg., 36共3兲, pp. 589–597. 关42兴 Chong, C. K., and How, T. V., 2004, “Flow Patterns in an Endovascular Stent-
关29兴 Scotti, C. M., and Finol, E. A., 2007, “Compliant Biomechanics of Abdominal Graft for Abdominal Aortic Aneurysm Repair,” J. Biomech., 37共1兲, pp. 89–97.

Journal of Biomechanical Engineering JANUARY 2009, Vol. 131 / 011008-15

Downloaded 01 Dec 2008 to 136.142.183.22. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Das könnte Ihnen auch gefallen