You are on page 1of 8

Stereoscopically Observed

Deformations of a Compliant
Abdominal Aortic Aneurysm
Model
Clark A. Meyer
Eric Bertrand
Olivier Boiron1
Valerie Deplano2
e-mail: deplano@irphe.univ-mrs.fr
Equipe de Biomecanique,
Institut de Recherche sur les Phenome`nes Hors
Equilibre (IRPHE) UMR 6594,
Centre National de la Recherche
Scientifique (CNRS),
13384 Marseille, France

A new experimental setup has been implemented to precisely measure the deformations
of an entire model abdominal aortic aneurysm (AAA). This setup addresses a gap
between the computational and experimental models of AAA that have aimed at improving the limited understanding of aneurysm development and rupture. The experimental
validation of the deformations from computational approaches has been limited by a lack
of consideration of the large and varied deformations that AAAs undergo in response to
physiologic flow and pressure. To address the issue of experimentally validating these
calculated deformations, a stereoscopic imaging system utilizing two cameras was constructed to measure model aneurysm displacement in response to pressurization. The
three model shapes, consisting of a healthy aorta, an AAA with bifurcation, and an AAA
without bifurcation, were also evaluated with computational solid mechanical modeling
using finite elements to assess the impact of differences between material properties and
for comparison against the experimental inflations. The device demonstrated adequate
accuracy (surface points were located to within 0.07 mm) for capturing local variation
while allowing the full length of the aneurysm sac to be observed at once. The experimental model AAA demonstrated realistic aneurysm behavior by having cyclic strains consistent with reported clinical observations between pressures 80 and 120 mm Hg. These
strains are 12%, and the local spatial variations in experimental strain were less than
predicted by the computational models. The three different models demonstrated that the
asymmetric bifurcation creates displacement differences but not cyclic strain differences
within the aneurysm sac. The technique and device captured regional variations of strain
that are unobservable with diameter measures alone. It also allowed the calculation of
local strain and removed rigid body motion effects on the strain calculation. The results
of the computations show that an asymmetric aortic bifurcation created displacement
differences but not cyclic strain differences within the aneurysm sac.
[DOI: 10.1115/1.4005416]
Keywords: abdominal aortic aneurysm, stereo correlation, compliant models, experimental deformations

Introduction

Abdominal aortic aneurysm is present when the diameter of the


aorta is more than 1.5 times nominal in the region between the
renal arteries and the aortic bifurcation. Aneurysm rupture leads
to potentially catastrophic blood loss and death if not treated;
however, the treatments carry significant risk. There is uncertainty
in predicting rupture likelihood and a general lack of understanding of the factors that contribute to AAA rupture. To address this
lack of understanding, many computational models have been
built that consider the solid mechanics (finite element analysis
(FEA)), the fluid mechanics (computational fluid dynamics
(CFD)), or both through fluid-solid interaction (FSI).
Computational models have shown a predictive capability, better than diameter measurement alone in forecasting aneurysm rupture, by identifying aneurysms at greater risk of rupture than
others of larger size, despite the general correlation between size
1

Also at Ecole Centrale, Marseille, France.


Address for correspondence: Valerie Deplano, Technopole de Chateau Gombert,
49 rue F. Joliot Curie, B.P. 146, 13384 Marseille Cedex 13, France.
Contributed by the Bioengineering Division of ASME for publication in the JOURNAL OF BIOMECHANICAL ENGINEERING. Manuscript received March 4, 2011; final manuscript received October 25, 2011; published online November 28, 2011. Assoc.
Editor: Dalin Tang.
2

Journal of Biomechanical Engineering

and rupture risk [1]. However, due to the complexities of the


in vivo AAA and its relationship to surrounding tissue, it is difficult to estimate strength and stress precisely, and a lack of a universal approach persists as noted in Refs. [1,2]. Studies with
computational models of AAA utilizing FSI have also shown various estimates (0.1% to 1030%) of blood flows impact on the
AAA on maximum wall stress, von Mises or maximum principle, as
summarized in Borghi et al. [3]. Their summary included references
to several key papers [49]. Maximum wall stress has been associated with rupture risk [1]. Yet wall shear stress, which is of a lower
magnitude than von Mises stress, has been implicated in controlling
vascular remodeling and, thus, rupture [10]. Wall shear stress cannot
be estimated for AAA having compliant walls without FSI.
Experimental validations of computational approaches to AAA
are essential to understanding and fidelity. Specifically, experimental validation of the FSI models has been limited. Previous
validations have not considered the large displacements present
with AAA [11]. Large displacements and related biomechanical
factors (e.g., strain) have been observed clinically. These clinical
observations as well as isolation studies have shown a role for
these factors in changing AAA development and rupture risk.
Recently, an experimental model of AAA was evaluated by diameter measurements [12] but did not consider local strains. The
problem of AAA rupture prediction is further complicated by the

C 2011 by ASME
Copyright V

NOVEMBER 2011, Vol. 133 / 111004-1

Downloaded From: http://biomechanical.asmedigitalcollection.asme.org/ on 11/11/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

fact that some of the controlling factors of size, pressure, material


strength, mechanical properties, loads, and thickness of AAA can
be variable over time even within the same patient. Reports indicate that biomechanical factors at the tissue level and mechanical
factors at the cellular and sub-cellular levels are involved in AAA
failure [13] and in the growth of AAA [14].
The deformations in response to pressure loads in the computational and experimental models can be verified with clinical measures of wall motion in AAA from cine MRI/CT, which have been
summarized by van Keulen et al. [15]. Clinical measurements of
deformation such as Ep (also called Petersons or pressure-strain
modulus) also have shown promise in rupture prediction [16].
These clinical measurements only considered the change in diameter (or the perimeter) of axial slices of AAA at only the AAAs
maximum diameter. An unweighted average of the measurements
in that summary (converted to diameter when necessary) shows
that the maximum diameter changed by 1.1 mm between diastole
and systole, corresponding to a circumferential stretch ratio of
1.021. Due to the lack of surface markers in their techniques, these
researchers were only able to classify motions relative to the two
dimensional slice and could not verify against the locally varying
strain/stress patterns observed in computational analyses. The calculated peaks of local strain/stress are not typically located at the
maximum diameter position but instead at regions of maximally
changing curvature [17].
Stereoscopic technique provides the ability to measure locations and displacements in three dimensions (3D) over a large
area with high precision [18] by utilizing imaging from different
perspectives as a basis for the triangulation of 3D positions.
Strains can be calculated from these 3D locations and their displacements. Stereoscopic technique, using digital image correlation (DIC), has been previously utilized for, among other things,
an analysis of the inflation of mouse arteries [18]. However, DIC
typically relies on statistical correlation instead of an exact match
of markers and requires denser markings.
We used point-tracking, as has Genovese [19], instead of DIC,
to measure wall motion throughout the sac region of experimental
aorta shapes made with EstaneV, a polyurethane. The AAA shape
has been used previously in a flow study [20]. The measurements
taken in this study provide the first local measurements of whole
experimental aneurysm deformation. The approach was validated
by confirming the ability of the system to measure a rigid plate
and by comparison with a model of a healthy, bifurcated aorta
section. This validation and further results of inflated mock aortas
incorporating an AAA are presented in conjunction with computational models that used a common hyperelastic isotropic material
model [21] for aneurysmal tissue.
R

Methods

2.1 Geometry. Three representative shapes: AAA with bifurcation (AB), AAA without bifurcation (AU), and healthy bifurcated aorta (HB), are shown in Fig. 1. The geometry is the same
as that used by Deplano et al. [20]. Briefly, the AAAs bounding
curves in each plane are based upon exponentials of the form of
Eq. (1) with coefficients in Table 1, making the AAA sac asymmetric in the anterior-posterior (A-P) plane but otherwise symmetric. The geometry of the bifurcation is asymmetric in its
branching angle (take off angles of 29 and 18 for the right and
left common iliacs, respectively) and based upon patient M11
from Shah et al. [22]. Surfaces constructed from this geometric information provided the foundation of a computer aided design
(CAD) model.
  !P1


2
C1 P2  xc2
e
(1)
coordinate Di p
pC3
At the maximum diameter location of the AAA sac in the CAD
model, the lateral width is 64.8 mm and the anterior-posterior
111004-2 / Vol. 133, NOVEMBER 2011

Fig. 1 Rendered views of aorta models (top to bottom): AAA


without bifurcation, AAA with bifurcation, and healthy aorta
with bifurcation

Table 1 Coefficients to define curves of AAA model shapes,


where x is a variable for the axial coordinate and is varied
between 0 and 150
Curve
Lateral near ends
Lateral middle
Anterior near ends
Anterior middle
Posterior

Di

C1

C2

C3

P1

P2

10.2
10
9.7
10.5
10.4

100
100
100
100
100

7
10
8
9
8

0.8
0.8
0.8
0.8
0.8

0.009
0.007
0.0095
0.006
0.008

0.85
0.75
0.95
0.81
0.39

distance is 54.6 mm. The proximal neck and healthy aorta are
21.6 mm in diameter. The overall length is 540 mm for each
model, and the AAA sac dimensions are the same in both models
with aneurysm (AB and AU).
2.2 Manufacture of Experimental Models. CAD models
were utilized for the machining of metal plates that were used by
a glassblower to make seamless glass molds. The AAA models
were constructed by repeatedly coating the glass molds interior
with polyurethane, as has been done previously for compliant
AAA models used for studying fluid flow patterns [20,23]. Briefly,
a mixture of EstaneV 5714 (1:10 by weight, of polymer to tetrahydrafurane) was made. The mixture was poured into a glass mold.
The mixture was drained out until dripping, and the mold was
loaded onto a dual-rotating platform inside a dryer, leaving a thin
layer adherent on the interior. The mold was dried for 30 min at
35  C. The coating process was repeated until there were six
layers, with the rotation directions alternated each time. The
model then finished curing overnight. A pilot study of polyurethane model shrinkage conducted on a uniform tube 25 mm in diameter indicated that, for the thickness of interest in these models
R

Transactions of the ASME

Downloaded From: http://biomechanical.asmedigitalcollection.asme.org/ on 11/11/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 2 Optical imaging setup with AB model and representative images from the device of the AU model with temporary
point markings

(0.2 mm), the expected shrinkage of the model due to curing


would be 1.4%. Thickness was measured by caliper and graduated
cone.
2.3 Mechanical Characterization. To quantify the properties of the molded material for our application and strain range,
uniaxial deformation-controlled traction tests were conducted.
The samples were cut from the straight inlet tubular section of a
typical molded model and pulled uniaxially in the formerly circumferential direction. These unloaded samples were approximately 20 mm wide  0.2 mm thick  40 mm long (between
grips)the exact dimensions of each sample were used in the calculation of stress and strain. The thickness is the same as that used
in the inflated models and is thinner than typical AAA because the
material is stiffersee Sec. 3.1. The tests were conducted at
strains up to 0.1 (length change/original length) with a 10 -N
capacity load cell using a custom designed traction bench at a rate
of 0.2 strain/sec. Relaxation tests, which allowed for 600 s of
relaxation at increments of 2% strain, up to 10%, were also conducted to determine the time required to allow the material deformation to reach steady state.
2.4 Stereoscopic Device for Observing Model AAA
Inflation. The optical measurement system that we constructed,
shown in Fig. 2, provided a non-contacting means to quantify the
detailed AAA deformations under pressurization. The system consists of a rigid frame, AAA model mounts, an angle measurement
plate, an inflation system, two cameras connected to a synchronizer, and a computer with an image acquisition card. The rigid
frame provides a base upon which the components are mounted in
fixed positions. The AAA model mounts can be moved axially in
tracks to modify stretch but also contain bearings to allow rotation, allowing modification of which portion of the model is presented to the cameras and to prevent unintended twisting of the
compliant AAA model. Rotational alignment of a sample is
achieved through the angle measurement plate and a removable
plumb line. A hand pump and pressure gauge (Boso, Germany)
are connected through the top mount for pneumatic inflation. The
two cameras were mounted at 620 from perpendicular to the
model AAAs long axis. This orientation allows for accurate triangulation of the tracking marker positions from the captured images.
The two grayscale cameras (2 Megapixel PowerViewTM Plus, TSI
Inc., Shoreview, MN) capture images of 1600  1192 pixels with
28 mm lenses. The cameras allow simultaneous capture of the
images, which are recorded for subsequent analysis. The camera
mounts allow adjustment of the lenses to satisfy the Scheimpflug
condition and get the plane of the long axis in full focus.
2.5 Stereoscopic Device Calibration. To incorporate distortion correction and transform between pixel coordinates and
Journal of Biomechanical Engineering

spatial coordinates, equations with a fourth order dependence


were used. The coefficients for the factors in the equations were
found by solving, with a least squares method, the overdetermined system of equations that related the pixel coordinates
from the images to idealized 3D positions. The system of equations allowed interactions between terms, resulting in 35 coefficients for each dimension. This method is an extension to three
dimensions of the two-dimensional method described in Wolberg
[24].
The calibration positions came from moving a flat calibration
plate, with 37 rows of 37 dots (uniformly spaced 2.5 mm apart on
centers), in 1 mm increments over a 50 mm height using two micrometer positioning stages aligned in series. The cameras were
focused prior to calibration at the halfway point (25 mm) of the
full height incremented over. At each height increment, image
pairs were recorded. Each image was processed using the same
process as was applied to the AAA model images to detect the
centroid (in pixel coordinates) of each dot seen on the calibration
plate.
Specifically, within a custom MATLAB program, a median filter
was applied to reduce noise before converting to a binary image
with a threshold. The binary image was filtered to consolidate the
points representing the markers and to fill any gaps created by
specular reflections. The centers of each markers pixel groups
were then calculated for each image by averaging the positions of
the pixels comprising the group. These centers were then associated with the corresponding pixel group center from the paired
image to allow the triangulation of the position in pixel coordinates. These triangulated pixel coordinates were paired with their
matching idealized spatial coordinates as defined by calibration
plate parameters and position to create the over-determined system of equations that was solved for the transformation
coefficients.
These transformation coefficients were then useful for transforming any 3D triangulated pixel coordinate to a physical location. Rotations of the coordinate system of the physical location
could then be used for full reconstructions of objects rotated about
the axis.
2.6 Inflation Experiments. The AAA models were loaded
in an upright position to minimize any impact of gravity induced
motion on the 3D reconstruction of the model shape. Multiple
views were used because only a portion (typically a quadrant) of
the aneurysm was directly visible by the cameras at a given rotation. The sides and backside of the AAA were obscured by the
surface closest to the camera.
The tracking markers were dots manually drawn on the surface
in erasable ink one quadrant at a time. The dots were spaced an
average of 4.0 mm apart on centers (calculated 3D distance, with
an average standard deviation of 1.15 mm). Each dots position
was determined with equivalent accuracy. However, the strain calculation relies on the position of the three points making up the
triangular element to be considered planar. This presumption is
more accurate when the points are less spaced and/or the surface
is less curved.
The models were each inflated using the hand pump to 120 mm
Hg. This pressure was maintained for 10 min, and images were
taken. The pressure was then reduced to 80 mm Hg and maintained for another 10 min before taking images. The markings,
except for one edge row, were then removed. This row provided a
boundary that was used to ensure rotational completeness between
image sets. The model was rotated 90 , markings applied, and the
inflation and imaging process repeated. This rotation process was
repeated until the full surface of the aneurysmal portion had been
observed. Each model type was analyzed once.
2.7 Numerical Simulation. Computational models were
built to compare the expected behavior of the experimental aneurysm with the behavior expected from models utilizing established
NOVEMBER 2011, Vol. 133 / 111004-3

Downloaded From: http://biomechanical.asmedigitalcollection.asme.org/ on 11/11/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

techniques and material properties. The models geometries were


created in ABAQUS (v. 6.7, ABAQUS, Inc.) using uniform
reductions (by 1.4%) of the dimensions of the CAD models used
for the glass molds. This reduction compensated for the shrinkage
expected from the molding process. Consistent sizing allowed for
comparison of the effects of the different material models proposed for AAA and the polyurethane material for our experimental AAA. The computational techniques were similar to those
Fillinger et al. [25] and Vande Geest et al. [26] utilized. The models were analyzed with internal pressures of 10.667 kPa and
16 kPa (80 and 120 mm Hg) corresponding to systolic and
diastolic pressures and fixed ends. The models consisted of shell
elements (three node triangular Abaqus type S3), and mesh independence was confirmed when peak stresses between subsequent
meshes varied by less than 1%, with the final mesh containing at
least 11,000 elements.
Material properties were either a commonly used, hyperelastic,
isotropic material model with constants derived by Raghavan and
Vorp [21] to fit aneurysmal tissue or a simplistic linear elastic
material model for the molded material. The isotropic model
(Eq. (2)) is from Ref. [21], and it is hereafter referred to as R&V.
W aI1  3 bI1  32

(2)

with W as the strain energy function, C as the right Cauchy-Green


deformation tensor, I1 tr(C), a 174 kPa, and b 1,881 kPa.
Wall thicknesses of 1.5 mm and 2 mm were used with this material to allow for consideration of different thicknesses that have
been reported and utilized in computational models (examples
include Refs. [9,27] as well as Refs. [2831]). For the linear elastic material based model, subsequently referred to as LE, a uniform, thinner wall (0.2 mm) was used, with the Youngs Modulus
of E 17.5 MPasee Sec. 3.1 for a full explanation. Linear elastic models for AAA have been used previously within some FSI
papers [32,33].
The R&V material model has been frequently used for AAA
computational modeling, including in models that demonstrated
clinical significance [1]. Other material models, including a more
recently proposed anisotropic model [26] and multilayered models
may prove more useful due to their increased complexity, but they
have yet to demonstrate clinical significance [34].
2.8 Evaluation and Visualization. In order to evaluate the
experimental setup, the results of inflation to diastolic and systolic
pressures were compared in terms of diastolic shape, displacement
ranges, principal strain, and Ep. The Ep is defined by Eq. (3),
using pressure and diameter values where systolic and diastolic
are considered equivalent to and interchangeable with maximum
and minimum, respectively.
Ep 

Psys  Pdia

dsys  ddia =ddia

(3)

The surfaces of the geometrical models are represented by nodal


locations, either of the computational mesh or of experimental
points. Two positions are associated with each node, diastolic and
systolic (the unloaded position is also available for the computational models). Triangulation of the experimental points into elements produces a mesh covering the whole of the AAA.
Comparisons between AAA model types include only the central,
aneurysmal region, which was observable in the experiments.
Cyclic strain was calculated for each triangular element using
the method for strain described by Genovese [19]. However, we
utilized the diastolic position for reference instead of the unloaded
position, as is possible with clinical measurements. In clinical
measurements, the unloaded reference positions cannot be directly
observed. Our cyclic strain calculation routine was shown to
exactly match the strains calculated by ABAQUS, when using the
initial unloaded positions as reference, in order to prove the
111004-4 / Vol. 133, NOVEMBER 2011

accuracy of the routine. To allow comparison of cyclic strains


across model types, nodal values of cyclic strain were determined
from the element-wise calculation. This calculation generated a
smoothing effect by relying on the average of element values at
their connection points (i.e., nodes).

Results

3.1 Material Testing. Evaluation of the stress-strain relationship for the molded polyurethane used in the experiment was
based upon uniaxial extension tests. Cyclic tests (10% strain at
1 Hz) showed an essentially linear behavior with a modulus of
17.5 MPa (1.75  107 Pa). This test is comparable to the loading
frequency when simulating flow, though the amplitude is exaggerated, as the minimum strain is not representative of that at diastolic pressure. A relaxation test was also conducted to assess
viscoelastic properties [35]. The relaxation test showed an 8%
decrease in the linearly estimated modulus when utilizing the
relaxed stress values instead of the peak values at each strain. The
relaxation period of 600 s was more than adequate for the stabilization of stress (not shown).
3.2 Calibration and Validation. The calibration of the measurement device provided a system capable of a sub-pixel level of
accuracy as demonstrated by a mean absolute 3D distance error of
0.0118 mm across all points, with a maximum error 0.108 mm.
The device accuracy was verified by quantifying the idealness of
positions resulting from a rotation ( 45 deg) of the calibration
plate. The mean position error was 0.07 mm, with a maximum
error of 0.18 mm. These levels of precision are adequate for
observing the expected motions of the sample while maintaining
the ability to view the full length of the AAA sac at once.
Further assessment of the system was done using HB shape for
a comparison between the LE computational model and the experimentally observed. The experimental HB case showed the uniform size reduction consistent with the pilot study and provides a
test case for the reconstruction technique. They differed by
< 0.05% in lateral and A-P peak distances at diastolic pressure in
the central region. The difference in peak cyclic strain was
< 0.35%, indicating the close match between LE and experiment
for this simple case.
3.3 Comparisons. Comparisons with differing materials
highlight the differences attributable to material models (between
columns of Figs. 35), and the consistency among the experiments
and computations expresses some of the limits in the validity in
the replacement of one type of model with another. Comparisons
between AU and AB shapes highlight the role of the bifurcation
(first two rows of Figs. 35), and comparisons between AB and
HB shapes highlight the role of the AAA sac (second and third
rows). The axial positions, shown in Fig. 3(b), illustrate that the
same region was compared in all cases. Within Table 2, the Ep
values of the different models are compared.
3.3.1 Effects of Material Model. The LE computational
model did not capture the full experimentally observed material
behavior of molded polyurethane when aneurysm was present.
Although a similarity in strain behavior and dimensions were noted
for the HB case, the AB and AU model types at diastolic pressure
are smaller in the A-P direction than the shapes observed with stereoscopic imaging. At diastolic pressure in the AU model, the A-P
maximum distances were 60.85 mm and 55.71 mm in the LE computational model and experiment, respectively. The lateral widths
are closer together, at 63.69 mm and 63.58 mm, respectively.
The differences between our LE computational model and
experiment are further manifest in the A-P component of motion,
as shown in Fig. 4(a), where the patterns as well as the magnitudes of motion are different. Specifically, the experiment showed
greater upward motion at the lateral portion above the midline. It
Transactions of the ASME

Downloaded From: http://biomechanical.asmedigitalcollection.asme.org/ on 11/11/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 3 Positions of the surface at diastolic pressure are color


coded with contour lines, showing shape variations between
models with (a) anterior-posterior direction, (b) axial direction,
and (c) lateral direction

is unconfirmed as to whether or not the difference is attributable


to uneven shrinkage during the molding process. Also, as noted in
the methods, the density of markings affects how well the experimental deformation can be quantified. In this case, the regional
quantification of strain shows limited variation, but the uniformity
might also be attributable to creep of the material.
By comparison of the LE computational models to the R&V
1.5 mm thick models, we show other differences that are attributable to the material model. The comparisons between LE and
R&V with 1.5 mm are based upon the same original meshes. The
diastolic position ranges in models with AAA are quite similar,
with the lateral position increased by 0.81 to 0.83% and the A-P
position ranges increased by 0.88 to 0.90% in the LE model. The
cyclic displacements ranges, however, are much larger in the LE
models as compared to the R&V, at 47 to 42% more displacement
laterally, and 61 to 62% A-P. Consequently, the maximum principal cyclic strains are 6162% higher in the LE AAA models, at
0.0360.037 for the LE as compared to 0.023 for the R&V. The
Journal of Biomechanical Engineering

Fig. 4 Cyclic displacements in the (a) anterior-posterior direction, (b) axial direction, and (c) lateral direction. These cyclic
displacements are defined as the positions of nodes at 120 mm
Hg minus their position at 80 mm Hg. They are plotted on position observed at 80 mm Hg.

minimum cyclic strains are more similar with 0.014 (mm/mm) for
the LE and 0.015 for the R&V. The locations of these peak values
for maximum principal strain are consistent between LE and
R&V. The peak values occur in the lateral shoulders of the aneurysm shapes, where the diameter is tapering. The lowest strains
are observed after taper is complete in the LE models but at peak
sac diameter in the R&V model.
R&V with a thicker wall (2.0 mm versus 1.5 mm) is also considered because some computational models have used it, and
AAA wall thickness has been shown to vary over this range and
beyond [36]. Wall thickness has been shown to vary regionally
over an individual aneurysm, in addition to varying merely
between patients, and this is not accounted for in these models.
The effect of thickness is essentially uniform within each of the
measured parameters. The extreme positions at diastole in the AU
model with the 1.5 mm thickness are 64.2, 61.4, and 115.7 mm
apart in the lateral, A-P, and axial dimensions, respectively, and
63.6, 60.5, and 115.7 mm in the corresponding dimensions in the
NOVEMBER 2011, Vol. 133 / 111004-5

Downloaded From: http://biomechanical.asmedigitalcollection.asme.org/ on 11/11/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Fig. 5

Strain patterns where strain is relative to diastolic positions

Table 2 Ep values at the level of maximum diameter from computation and experiment, from average and planar dimensions
(Units of kPa)
Ep of AB models
Model type
LE
R&V 1.5 mm
R&V 2.0 mm
Experiment

Ep of AU models

Average

A-P

Lateral

Average

A-P

Lateral

274
365
421
328

178
280
294
227

369
450
548
460

276
366
427
341

177
277
293
323

354
432
522
360

2.0 mm thick model (Figs. 3(a), 3(b), and 3(c)). These reductions
indicate the 1.5 mm thick model is larger at diastolic pressure by
1.04% in the lateral direction and 1.44% in the A-P direction. The
axial dimension is consistent between models as the region is limited to the region visible with the stereoscopic system. Within the
AB models, the dimensions are almost exactly the same as the AU
except the lateral extremes are 0.03 mm closer together. Notably,
the thickness effect causes unequal changes in the A-P and lateral
directions as the aneurysm is asymmetric and not centered in the
A-P plane.
The effect of thickness difference with the R&V models is also
manifest in the response to systolic pressure. Using diastole as the
reference, the thinner model has greater motion in the lateral
direction in the AU and AB models by 10.08% (0.98 mm) AU and
12.09% (1.02 mm) AB. The A-P direction shows a more moderate
influence than the lateral, with motion difference values of 6.24%
AU and 5.41% AB on extreme ranges of 1.21 and 1.20 mm in the
1.5 mm thick models. These displacement differences are also
apparent in the results of Ep calculations, where using the average
of lateral and A-P diameters, the values are 3.65, 3.66, 4.21, and
4.27  105 Pa for the AB 1.5, AU 1.5, AB 2.0, and AU 2.0, respectively. Simplistically, a 25% reduction in thickness caused a 1314% reduction in Ep for R&V models with our idealized shapes.
The thickness effect on cyclic strain is nearly uniform (0.2%
strain). In absolute terms, thinner models had increases of cyclic
strains of 0.16% in both AAA outlet types. In relative terms, differences in cyclic strains of 8.96% in AB and 9.18% in AU were
found. For the same level of thickness decrease (25%), Venkatasubramaniam et al. [37] observed a 20% increase in stress. However, this value is not directly comparable, because that study
looked at the peak value and not the value relative to diastolic.
3.3.2 Geometric Effects. The bifurcation has often been
excluded in computational models of the AAA [6,38]. Consideration
111004-6 / Vol. 133, NOVEMBER 2011

of this geometrical parameter is given in this study because of the


known effects a bifurcation can have on fluid flow. The role of the
bifurcation in this solid mechanics study is demonstrated by comparisons between the AB and AU series of models with similar
material conditions. The bifurcation contributes to an asymmetric
displacement pattern in the AB computational models, and this is
observable most clearly in Fig. 4(c) but also in Fig. 4(b). The
contour lines in the HB models also show this same asymmetric
pattern. The variation in displacements is consistent with a small
rotation about the z-axis (A-P). There is essentially no impact on
the strains within the sac (Fig. 5), regardless which of the material
models was employed. These results would not have been observable in other idealized models that did not include a physiological
lateral asymmetry in their bifurcations [39].
The influence of a sac is abundantly clear in the comparison
between AB and HB models (Figs. 35). The sac provides for
greater cyclic deformations than would be present in an aorta
without aneurysm. These differences are observable in all material
models considered computationally and in the experiment. In
these models, that incorporate no residual stress and no remodeling, the presence of the AAA sac also creates a greater variation
in cyclic strain values. The strain distributions in the same portion
of the aorta in the HB models have standard deviations of
1.43 E-04, 8.74 E-05, and 6.49 E-05 depending on the material
model used (linear elastic, R&V 1.5 mm and R&V 2.0 mm,
respectively) whereas the comparable standard deviations in the
AB models are 5.14 E-03, 2.03 E-03, and 1.94 E-03. The greater
degree of variation in the LE models versus R&V, as quantified
by standard deviation, is associated with greater motions (Fig. 4),
despite the similar diastolic size (Fig. 3).

Discussion

The rupture risk of an individual aneurysm is specific to a combination of factors (thickness, smoking, genetics, shape, etc.)
within each patient, though some factors, such as maximum diameter, typically dominate. To best treat a patient, an understanding
of their AAA as it currently is and as it is likely to become is necessary. Before a greater understanding of a patient specific case is
possible, there must be an understanding of the idealized case, and
this study provides insight into one of the most important aspects
(mechanical stress and motion) through both experimental and
computational methods.
The aneurysm shape used in this study attempted to strike a balance between some of the essential features of AAA and simplification. The shape is less symmetric than others [40,41],
particularly at the bifurcation, which allowed isolation of that features role. Increasing the complexity of AAA models provides
Transactions of the ASME

Downloaded From: http://biomechanical.asmedigitalcollection.asme.org/ on 11/11/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

transitional steps to aid in understanding patient specific shapes


and resulting emergent behaviors.
By utilizing a cyclic strain measure instead of only displacement, the effect of the asymmetric bifurcation on AAA wall
motion can be shown to be nearly entirely a rigid body type
motion that does not affect strains or stress throughout the AAA
sac. Calculating principal cyclic strain based upon large deformation theory is a coordinate system free approach that is independent of patient alignment or changes thereto between scans.
However, cyclic strain observations alone do not allow for stress
calculations in a nonlinear material because the undeformed, natural configuration (zero-stress state) is not known. There have been
methods published for estimating a zero-stress state from deformed
configurations for AAA [42,43], but they require knowledge of the
material properties and thickness a priori. In instances where the
properties are not uniform or unknown, the system is indeterminate.
A system such as described in this paper, where the full AAA shape
is reconstructed and observable, could prove useful for assessing
the accuracy of whole AAA models and their material models.
The device described in this paper showed a method for measuring the motions of AAA of an entire side at once, and through
repetition and rotation, measuring of an entire aneurysm sac. The
accuracy is excellent (resolving locations to within 0.07 mm of
true 3D position on average), thus, allowing for the observation of
even small motions. The technique, as implemented, is limited to
shapes that do not obstruct views of themselves as some extremely
tortuous patient aneurysm might. However, this is not an issue for
idealized AAA shapes. The method is applied to clear models that
can also be used for particle image velocimetry (PIV) because of
their good optical clarity and density in a glycerin-water solution
approximating blood. However, the device/technique as applied to
wall motion alone could be utilized even more easily with opaque
surfaces (including tissue) as the image processing issues are simplified. The method used here is unique in its scale and application
to AAA.
The material is partially suitable for capturing AAA and normal
aortic behaviors. The HB models demonstrated the similarity
between the experimentally observed mock aortic wall and the
computational models. However, the AB and AU models illustrated some of the limitations of the material choice. These limitations included the uneven shrinkage during the glass molding and
the materials relaxed response to loading. The materials testing
showed that while similar to an AAA tissue model (R&V) at systolic to diastolic strain levels, the molded polyurethane material
has a time-dependent response. It is more linear than R&V over
the full range of strains examined, making it not well suited for
modeling tissue response over a large range of pressures. The ability of polyurethane models to have their compliance tuned by
adjusting their thicknesses was not directly shown but is a method
for creating more or less compliant AAA. The difference in material stiffness was the reason our mock AAA was thinner than clinical AAA. Models with compliance tuned by adjusting thickness
alone also suffer in comparison to true AAA tissue because of the
linearity of their stress-strain curves. Additionally, the isotropic
nature of our material prevents it from modeling anisotropic material behavior in isolation.
The peak cyclic motions we observed experimentally, and in
our computational models, are larger than those reported in Doyle
et al. [12]. Their values were 0.08 mm with Sylgard 160 and
0.34 mm with Sylgard 170, when considering changes of maximum diameter between 80 and 120 mm Hg. The cyclic motions
observed in this study are in better agreement with expected
motions at the maximum diameter and to motions overall of most
untreated AAA studies (motions of 1.0 mm) as is summarized
by van Keulen et al. [15], though they are less than those reported
in Faries et al. [44], which showed a mean motion of 3.51 mm
prior to endograft treatment. The motions observed (summarized
in Table 2) are also in line with what is expected from the observations of Long et al. [16] of Ep with the interquartile range of
2.6  105 Pa to 4.3  105 Pa, with a median of 3.4  105 Pa.
Journal of Biomechanical Engineering

The experimental accuracy in the determination of strain by the


stereoscopic system could be improved by increasing marker (dot)
density and/or increasing camera resolution. The marker density
and camera resolution used in this study were adequate for demonstrating the technique and its effectiveness in assessing AAA
wall motion. Although the technique did not provide thickness information, this could be acquired in another way (e.g., MRI, micro
CT). The limited marker density made direct comparison between
cyclic strains infeasible.
The device was built with future additions in mind. These
include the application of the device to AAA in a flow loop for
the validation of wall motion and FSI, which would allow for the
combinational assessment of wall stresses (including maximum
principal/von Mises and shear). The cameras used are already
synchronized and capable of high frame rates. A combinational
assessment using FSI is probably necessary for understanding and
predicting the growth and remodeling of an AAA, since there is a
known dependency of growth and remodeling on both types of
stress.

Conclusion

In summary, a new device has been presented and shown to


measure realistic levels of cyclic wall deformation in mock AAA.
The measurements taken in this study provide the first local measurements of whole experimental aneurysm deformation. The
results demonstrate the physiologic nature of the motions
observed and the capability of the experimental setup. The limitation of a linear elastic model for the polyurethane utilized is also
shown. The experimental techniques developed are useful for a
subsequent validation of FSI calculations based upon future compliant wall flow studies of AAA.
The usefulness of the optical measurement system is its capability of providing high precision measurement of motion and the
capability of observing the entire sac length. The experimental
models appeared smaller in their A-P dimension than the computational models, likely due to additional shrinkage of the material
at this increased diameter since only the AAA models were
reduced. Despite this difference, the displacements and strain
ranges were well in line with most clinical observations of Ep and
cyclic strains. The technique provided additional information on
regional variations in displacement as compared to maximum
diameter changes. Unfortunately, the lack of fine detail prevents a
fine-grain comparison of cyclic strains against the computational
models. The viscoelastic nature of the molded polyurethane may
have also contributed to the lack of variation seen in the
experiments.
The computational models demonstrated that the linear elastic
material model created both higher average stress and higher
stress variability than the R&V models. They also showed a
nearly uniform response to globally reducing thickness and that
the impact of the asymmetric bifurcation was on sac displacement
but not sac strain.

Acknowledgment
The authors would like to acknowledge the support of the Whitaker Scholars program for C. Meyer.

References
[1] Fillinger, M. F., Marra, S. P., Raghavan, M. L., and Kennedy, F. E., 2003,
Prediction of Rupture Risk in Abdominal Aortic Aneurysm During Observation: Wall Stress Versus Diameter, J. Vasc. Surg., 37(4), pp. 724732.
[2] Vande Geest, J. P., Di Martino, E. S., Bohra, A., Makaroun, M. S., and Vorp,
D. A., 2006, A Biomechanics-Based Rupture Potential Index for Abdominal
Aortic Aneurysm Risk Assessment: Demonstrative Application, Ann. N.Y.
Acad. Sci., 1085, pp. 1121.
[3] Borghi, A., Wood, N. B., Mohiaddin, R. H., and Xu, X. Y., 2008, Fluid-Solid
Interaction Simulation of Flow and Stress Pattern in Thoracoabdominal Aneurysms: A Patient-Specific Study, J. Fluids Struct., 24(2), pp. 270280.

NOVEMBER 2011, Vol. 133 / 111004-7

Downloaded From: http://biomechanical.asmedigitalcollection.asme.org/ on 11/11/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

[4] Bluestein, D., Niu, L., Schoephoerster, R. T., and Dewanjee, M. K., 1996,
Steady Flow in an Aneurysm Model: Correlation Between Fluid Dynamics
and Blood Platelet Deposition, J. Biomech. Eng., 118(3), pp. 280286.
[5] Di Martino, E. S., Guadagni, G., Fumero, A., Ballerini, G., Spirito, R., Biglioli,
P., and Redaelli, A., 2001, Fluid-Structure Interaction Within Realistic ThreeDimensional Models of the Aneurysmatic Aorta as a Guidance to Assess the
Risk of Rupture of the Aneurysm, Med. Eng. Phys., 23(9), pp. 647655.
[6] Leung, J., Wright, A., Cheshire, N., Crane, J., Thom, S., Hughes, A., and Xu,
Y., 2006, Fluid Structure Interaction of Patient Specific Abdominal Aortic
Aneurysms: A Comparison With Solid Stress Models, Biomed. Eng. Online,
5(1), p. 33.
[7] Papaharilaou, Y., Ekaterinaris, J. A., Manousaki, E., and Katsamouris, A. N.,
2007, A Decoupled Fluid Structure Approach for Estimating Wall Stress in
Abdominal Aortic Aneurysms, J. Biomech., 40(2), pp. 367377.
[8] Scotti, C. M., Shkolnik, A. D., Muluk, S. C., and Finol, E. A., 2005, FluidStructure Interaction in Abdominal Aortic Aneurysms: Effects of Asymmetry
and Wall Thickness, Biomed. Eng. Online., 4, p. 64.
[9] Wolters, B. J., Rutten, M. C., Schurink, G. W., Kose, U., de Hart, J., and van de
Vosse, F. N., 2005, A Patient-Specific Computational Model of FluidStructure Interaction in Abdominal Aortic Aneurysms, Med. Eng. Phys.,
27(10), pp. 871883.
[10] Wayman, B. H., Taylor, W. R., Rachev, A., and Vito, R. P., 2008, Arteries
Respond to Independent Control of Circumferential and Shear Stress in Organ
Culture, Ann. Biomed. Eng., 36(5), pp. 673684.
[11] Kanyanta, V., Ivankovic, A., and Karac, A., 2009, Validation of a FluidStructure Interaction Numerical Model for Predicting Flow Transients in
Arteries, J. Biomech., 42(11), pp. 17051712.
[12] Doyle, B. J., Corbett, T. J., Cloonan, A. J., ODonnell, M. R., Walsh, M. T.,
Vorp, D. A., and McGloughlin, T. M., 2009, Experimental Modelling of Aortic Aneurysms: Novel Applications of Silicone Rubbers, Med. Eng. Phys.,
31(8), pp. 10021012.
[13] Vorp, D. A., 2007, Biomechanics of Abdominal Aortic Aneurysm, J. Biomech., 40(9), pp. 18871902.
[14] Li, Z.-Y., Sadat, U., U-King-Im, J., Tang, T. Y., Bowden, D. J., Hayes, P. D.,
and Gillard, J. H., 2010, Association Between Aneurysm Shoulder Stress and
Abdominal Aortic Aneurysm Expansion: A Longitudinal Follow-Up Study,
Circulation, 122(18), pp. 18151822.
[15] van Keulen, J. W., van Prehn, J., Prokop, M., Moll, F. L., and van Herwaarden,
J. A., 2009, Dynamics of the Aorta Before and After Endovascular Aneurysm
Repair: A Systematic Review, Eur. J. Vasc. Endovasc. Surg., 38(5), pp.
586596.
[16] Long, A., Rouet, L., Bissery, A., Rossignol, P., Mouradian, D., and Sapoval,
M., 2005, Compliance of Abdominal Aortic Aneurysms Evaluated by Tissue
Doppler Imaging: Correlation With Aneurysm Size, J. Vasc. Surg., 42(1), pp.
1826.
[17] Sacks, M. S., Vorp, D. A., Raghavan, M. L., Federle, M. P., and Webster, M.
W., 1999, In Vivo Three-Dimensional Surface Geometry of Abdominal Aortic
Aneurysms, Ann. Biomed. Eng., 27(4), pp. 469479.
[18] Sutton, M. A., Ke, X., Lessner, S. M., Goldbach, M., Yost, M., Zhao, F., and
Schreier, H. W., 2008, Strain Field Measurements on Mouse Carotid Arteries
Using Microscopic Three-Dimensional Digital Image Correlation, J. Biomed.
Mater. Res. Part A, 84A(1), pp. 178190.
[19] Genovese, K., 2009, A Video-Optical System for Time-Resolved Whole-Body
Measurement on Vascular Segments, Opt. Lasers Eng., 47(9), pp. 9951008.
[20] Deplano, V., Knapp, Y., Bertrand, E., and Gaillard, E., 2007, Flow Behaviour
in an Asymmetric Compliant Experimental Model for Abdominal Aortic
Aneurysm, J. Biomech., 40(11), pp. 24062413.
[21] Raghavan, M. L., and Vorp, D. A., 2000, Toward a Biomechanical Tool to
Evaluate Rupture Potential of Abdominal Aortic Aneurysm: Identification of a
Finite Strain Constitutive Model and Evaluation of its Applicability, J. Biomech., 33(4), pp. 475482.
[22] Shah, P. M., Scarton, H. A., and Tsapogas, M. J., 1978, Geometric Anatomy
of the AorticCommon Iliac Bifurcation, J. Anat., 126(Pt 3), pp. 451458.
[23] Gaillard, E., Bergeron, P., and Deplano, V., 2007, Influence of Wall Compliance on Hemodynamics in Models of Abdominal Aortic Aneurysm, J. Endovasc. Ther., 14(4), pp. 593599.
[24] Wolberg, G., 1994, Digital Image Warping, IEEE Computer Society Press,
New York.

111004-8 / Vol. 133, NOVEMBER 2011

[25] Fillinger, M. F., Raghavan, M. L., Marra, S. P., Cronenwett, J. L., and Kennedy,
F. E., 2002, In Vivo Analysis of Mechanical Wall Stress and Abdominal Aortic Aneurysm Rupture Risk, J. Vasc. Surg., 36(3), pp. 589597.
[26] Vande Geest, J. P., Schmidt, D. E., Sacks, M. S., and Vorp, D. A., 2008, The
Effects of Anisotropy on the Stress Analyses of Patient-Specific Abdominal
Aortic Aneurysms, Ann. Biomed. Eng., 36(6), pp. 921932.
[27] Scotti, C. M., and Finol, E. A., 2007, Compliant Biomechanics of Abdominal
Aortic Aneurysms: A Fluid-Structure Interaction Study, Comput. Struct.,
85(1114), pp. 10971113.
[28] Martufi, G., Di Martino, E. S., Amon, C. H., Muluk, S. C., and Finol, E. A.,
2009, Three-Dimensional Geometrical Characterization of Abdominal Aortic
Aneurysms: Image-Based Wall Thickness Distribution, J. Biomech. Eng.,
131(6), p. 061015.
[29] Shum, J., DiMartino, E. S., Goldhamme, A., Goldman, D. H., Acker, L. C., Patel,
G., Ng, J. H., Martufi, G., and Finol, E. A., 2010, Semiautomatic Vessel Wall
Detection and Quantification of Wall Thickness in Computed Tomography Images
of Human Abdominal Aortic Aneurysms, Med. Phys., 37(2), pp. 638648.
[30] Shum, J., Martufi, G., Di Martino, E., Washington, C. B., Grisafi, J., Muluk, S.
C., and Finol, E. A., 2011, Quantitative Assessment of Abdominal Aortic Aneurysm Geometry, Ann. Biomed. Eng., 39(1), pp. 277286.
[31] Shum, J., Xu, A., Chatnuntawech, I., and Finol, E. A., 2011, A Framework for
the Automatic Generation of Surface Topologies for Abdominal Aortic Aneurysm Models, Ann. Biomed. Eng., 39(1), pp. 249259.
[32] Khanafer, K., and Berguer, R., 2009, Fluid-Structure Interaction Analysis of
Turbulent Pulsatile Flow Within a Layered Aortic Wall as Related to Aortic
Dissection, J. Biomech., 42(16), pp. 26422648.
[33] Li, Z., and Kleinstreuer, C., 2005, Fluid-Structure Interaction Effects on SacBlood Pressure and Wall Stress in a Stented Aneurysm, J. Biomech. Eng.,
127(4), pp. 662671.
[34] Rissland, P., Alemu, Y., Einav, S., Ricotta, J., and Bluestein, D., 2009,
Abdominal Aortic Aneurysm Risk of Rupture: Patient-Specific FSI Simulations Using Anisotropic Model, J. Biomech. Eng., 131(3), p. 031001.
[35] Polansky, J., Boiron, O., and Novacek, V., 2005, Identification of Viscoelastic
Properties of Artificial Materials Simulating Vascular Wall, Comput. Methods
Biomech. Biomed. Eng., 8(4 supp 1), pp. 219220.
[36] Raghavan, M. L., Kratzberg, J., de Tolosa, E. M. C., Hanaoka, M. M., Walker,
P., and da Silva, E. S., 2006, Regional Distribution of Wall Thickness and
Failure Properties of Human Abdominal Aortic Aneurysm, J. Biomech.,
39(16), pp. 30103016.
[37] Venkatasubramaniam, A. K., Fagan, M. J., Mehta, T., Mylankal, K. J., Ray, B.,
Kuhan, G., Chetter, I. C., and McCollum, P. T., 2004, A Comparative Study of
Aortic Wall Stress Using Finite Element Analysis for Ruptured and Non-Ruptured
Abdominal Aortic Aneurysms, Eur. J. Vasc. Edovasc. Surg., 28(2), pp. 168176.
[38] Wang, D. H., Makaroun, M. S., Webster, M. W., and Vorp, D. A., 2002, Effect
of Intraluminal Thrombus on Wall Stress in Patient-Specific Models of Abdominal Aortic Aneurysm, J. Vasc. Surg., 36(3), pp. 598604.
[39] Doyle, B. J., Corbett, T. J., Cloonan, A. J., ODonnell, M. R., Walsh, M. T.,
Vorp, D. A., and McGloughlin, T. M., 2009, Experimental Modelling of Aortic Aneurysms: Novel Applications of Silicone Rubbers, Med. Eng. Phys.,
31(8), pp. 10021012.
[40] Fraser, K. H., Li, M. X., Lee, W. T., Easson, W. J., and Hoskins, P. R., 2009,
Fluid-Structure Interaction in Axially Symmetric Models of Abdominal Aortic
Aneurysms, Proc. Inst. Mech. Eng., Part H: J. Eng. Med., 223(2), pp. 195209.
[41] Doyle, B. J., Cloonan, A. J., Walsh, M. T., Vorp, D. A., and McGloughlin, T.
M., 2010, Identification of Rupture Locations in Patient-Specific Abdominal
Aortic Aneurysms Using Experimental and Computational Techniques, J. Biomech., 43(7), pp. 14081416.
[42] Lu, J., Zhou, X., and Raghavan, M. L., 2007, Inverse Elastostatic Stress Analysis in Pre-Deformed Biological Structures: Demonstration Using Abdominal
Aortic Aneurysms, J. Biomech., 40(3), pp. 693696.
[43] Speelman, L., Bosboom, E. M., Schurink, G. W., Buth, J., Breeuwer, M.,
Jacobs, M. J., and van de Vosse, F. N., 2009, Initial Stress and Nonlinear Material Behavior in Patient-Specific AAA Wall Stress Analysis, J. Biomech.,
42(11), pp. 17131719.
[44] Faries, P. L., Agarwal, G., Lookstein, R., Bernheim, J. W., Cayne, N. S., Cadot,
H., Goldman, J., Kent, K. C., Hollier, L. H., and Marin, M. L., 2003, Use of
Cine Magnetic Resonance Angiography in Quantifying Aneurysm Pulsatility
Associated With Endoleak, J. Vasc. Surg., 38(4), pp. 652656.

Transactions of the ASME

Downloaded From: http://biomechanical.asmedigitalcollection.asme.org/ on 11/11/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like