Atherosclerotic plaque characterization using plaque area variation in IVUS images during compression: A computational investigation

Abstract

INTRODUCTION: The rupture of atherosclerotic plaques causes millions of death yearly. It is known that the kind of predominant tissue is associated with its dangerousness. In addition, the mechanical properties of plaques have been proved to be a good parameter to characterize the type of tissue, important information for therapeutic decisions. METHODS: Therefore, we present an alternative and simple way to discriminate tissues. The procedure relies on computing an index, the ratio of the plaque area variation of a suspecting plaque, using images acquired with vessel and plaques, pre and post-deformation, under different intraluminal pressure. Numerical phantoms of coronary cross-sections with different morphological aspects, and simulated with a range of properties, were used for evaluation. RESULTS: The outcomes provided by this index and a widely used one were compared, so as to measure their correspondence. As a result, correlations up to 99%, a strong agreement with Bland-Altman and very similar histograms between the two indices, have shown a good level of equivalence between the methods. CONCLUSION: The results demonstrated that the proposed index discriminates highly lipidic from fibro-lipidic and calcified tissues in many situations, as good as the widely used index, yet the proposed method is much simpler to be computed.

Full text

Original Article
http://dx.doi.org/10.1590/rbeb.2014.013
*e-mail: matheuscardosomg@hotmail.com
Received: 6 August 2013 / Accepted: 4 December 2013
Atherosclerotic plaque characterization using plaque area variation
in IVUS images during compression: a computational investigation
Matheus Cardoso Moraes*, Fernando Mitsuyama Cardoso, Sérgio Shiguemi Furuie
Abstract Introduction: The rupture of atherosclerotic plaques causes millions of death yearly. It is known that the
kind of predominant tissue is associated with its dangerousness. In addition, the mechanical properties of
plaques have been proved to be a good parameter to characterize the type of tissue, important information for
therapeutic decisions. Methods: Therefore, we present an alternative and simple way to discriminate tissues.
The procedure relies on computing an index, the ratio of the plaque area variation of a suspecting plaque, using
images acquired with vessel and plaques, pre and post-deformation, under different intraluminal pressure.
Numerical phantoms of coronary cross-sections with different morphological aspects, and simulated with a
range of properties, were used for evaluation. Results: The outcomes provided by this index and a widely
used one were compared, so as to measure their correspondence. As a result, correlations up to 99%, a strong
agreement with Bland-Altman and very similar histograms between the two indices, have shown a good level of
equivalence between the methods. Conclusion: The results demonstrated that the proposed index discriminates
highly lipidic from bro-lipidic and calcied tissues in many situations, as good as the widely used index, yet
the proposed method is much simpler to be computed.
Keywords Intravascular Ultrasound (IVUS), Atherosclerosis, Intravascular elastography, Phantoms,
Strain map.
Introduction
Atherosclerotic plaque rupture is the main reason of
acute coronary events. In United States of America
(USA) 785,000 new coronary attacks occurs annually,
with 470,000 recurrences. A coronary event happens
every 25 seconds with one death per minute. Moreover,
in 2007 and 2008 the total cost of cardiovascular
disease and stroke of USA were U$286 and U$297.7
billions, respectively, greater than costs of other
diseases (Roger et al., 2012; Rosamond et al., 2007).
Atherosclerosis is the accumulation of lipidic,
brous, and calcied tissues in the arterial wall. An
atherosclerotic plaque rupture may lead to myocardial
infarction, unstable angina, or sudden cardiac death
(Davies, 2000; Falk et al., 1995; Maurice et al., 2005;
Viermani et al., 2000). The lesion severity is directly
related to plaque composition and morphological
features. A subset of vulnerable plaque is recognized
by having large lipid pool with macrophages, covered
by a thin cap (Baldewsing et al., 2004; Davies, 1996;
Viermani et al., 2000). Accordingly, by identifying
the preponderant tissue in a suspecting plaque, very
important lesion information is revealed, as the kind of
tissue is related to the dangerousness. Consequently,
this information provides extra aspects for more
accurate therapeutic decision or interventional
procedure (Fisher et al., 2000; Ohayon et al., 2001).
Consequently, tools and methods dedicated to
intravascular ultrasound (IVUS) have been growing
rapidly (De Korte and Van der Steen, 2002; Le
Floc’h et al., 2009; Maurice et al., 2005).
IVUS is a popular imaging modality for cardiac
interventional procedures. Its images supply the
cardiologists with anatomical, morphological and
pathological coronary and plaque information,
crucial for diagnostics, evaluations, and treatment
planning (Baldewsing et al., 2004; Liang et al., 2009).
Nonetheless, to accurately infer about atherosclerotic
plaque composition and dangerousness, additional
procedures, such as segmentation and intravascular
Elastography, are important (Baldewsing et al., 2004;
Cardenas et al., 2013; De Korte and Van der Steen,
2002; Le Floc’h et al., 2009; Maurice et al., 2005;
Moraes and Furuie, 2010, 2011).
Intravascular Elastography is a method in which
the mechanical properties of tissues from the arterial
wall are assessed through strain or elasticity (i.e. shear
or Young’s modulus) map (Céspedes et al., 1993;
Loree et al., 1994; Maurice et al., 2005; Ophir et al.,
1991). Therefore, extra information about tissue
composition and vulnerability is provided, as a result
Volume 30, Número 2, p. 159-172, 2014

Moraes MC, Cardoso FM, Furuie SS
the best therapeutic procedure can be carried out
(Céspedes et al., 1993; Ophir et al., 1991). Basically,
the Elastography procedure is divided into three main
blocks: First, the images are acquired with a variation
of intraluminal pressure - usually arterial pressure
and inated balloon are used as source of pressure
(De Korte et al., 1999; Shapo et al., 1996). Second,
the displacements of microstructures are computed,
followed by the strain map computation. Finally,
by knowing the intraluminal pressure and resultant
strain map, Finite Element Method (FEM) is usually
applied to calculate the corresponding force vector and
consequent elasticity map, the Young’s Modulus of
structures (Baldewsing et al., 2004; De Korte and Van
der Steen, 2002; Fisher et al., 2000; Le Floc’h et al.,
2009; Ohayon et al., 2001).
Prior approaches have dedicated efforts to
build well-documented and widely recognized
materials in coronary and plaque functional behavior
(Baldewsing et al., 2004; Céspedes et al., 1993; De
Korte and Van der Steen, 2002; Fisher et al., 2000; Le
Floc’h et al., 2009; Loree et al., 1994, Maurice et al.,
2005; Ohayon et al., 2001; Ophir et al., 1991). As a
result, anatomical and mechanical aspects such as Cap
thickness, lesion disposition and parameter values are
known. Hence, new tools and methods in Elastography
have been created, providing alternative tools for
therapeutic decisions, evaluation, and interventional
procedures (Fisher et al., 2000; Ohayon et al., 2001).
Although precise and reliable, IVUS Elastography is
still a costly and complex method to be implemented.
Therefore, simple and reliable alternatives for
extracting mechanical properties of tissue are needed.
The proposed approach presents a simple and more
practical procedure for extracting mechanical properties
of suspecting atherosclerotic tissues. The method
can be an alternative choice to IVUS Elastography.
In our procedure the mechanical properties and
corresponding composition of tissues of a suspecting
lesion, can be estimated by computing the area ratio
of the plaque under inspection in IVUS images
(Figure 1). Therefore, during imaging acquisition
(Figures 1a and b), different pressure is applied
by an expandable balloon to acquire the deformed
and non-deformed images. Thus, a distinguishable
deformation ratio is assured, and problems related
to uncertainty of intraluminal pressure, catheter
eccentricity and inclination are well overcome. After,
the Plaque Area Computation (Figures 1c and d), the
plaque areas under inspection in the IVUS images
are segmented and the area of a suspecting plaque
quantied in different pressures. Next, Ratio of the
Plaque Area Variation, the overall percentage of plaque
deformation is calculated (Cardoso et al., 2012). As
a result, the type of the predominant tissue of the
suspecting area is estimated based on the regional
deformation values. Therefore, the goal of this paper
is to introduce the procedure, and using deformable
numerical phantoms, demonstrating the capability of
tissue discrimination of the proposed index, Plaque
Area Variation (AR), in a range of situations.
Methods
The materials used for this work are comprised
by a personal computer with an Intel Core 2 Duo,
and microprocessor of 2.53 GHz, 4 GB of RAM,
Windows 7 64 bits, MATLAB® (2009a) (MathWorks,
Inc., Natick, MA, USA) with Imaging Processing
and Partial Differential Equation Toolbox. Due to
computational models advancements (Hoskins,
2008), numerical phantoms were used. The phantoms,
deformation outcomes and gold standard strain
maps were obtained by the framework described
by Cardoso et al. (2012). The complete method is
embedded in the Toolbox called IVUSSim. It is a
citationware software, available online at http://www.
leb.usp.br/IVUSSim, free of charge for research and
educational purposes (Cardoso et al., 2012). A brief
explanation of the framework is given below.
Overview of IVUSSim
Realistic IVUS image generation in Different
Intraluminal Pressures (IVUSSim) - The framework
is a tool, in which from different anatomical and
morphological coronary cross-section models and input
parameters, IVUS phantoms are created (Figure 2e). In
addition, some related results, such as the corresponding
non-deformed and deformed gray level images
(Figure 2c), strain maps (Figure 2d) are also generated
to serve as gold standard to many investigations
(Cardoso et al., 2012). In summary, the framework
works as follows: First the user selects an available,
or designs a new coronary cross section model,
representing a coronary cross section in a diastole
cardiac phase (Figure 2a). Second, the investigator
sets some parameters, such as ultrasound frequency,
number of transducers, and incremental intraluminal
pressures. Third, the algorithm automatically identies
the different regions of the selected model, such as
media, adventitia, and plaques, and asks the user to
insert the corresponding mechanical properties, such
as Young’s Modulus, for each region. Next, a 2D mesh
is constructed, and FEM is carried out, so that the
deformed mesh is generated (Figure 2b). Finally, a
morphism procedure is performed; hence, innitesimal
or very small structures, such as scatterers and pixels,
are repositioned according to FEM deformation. As
a result, the correspondent mesh and deformed mesh
Rev. Bras. Eng. Bioméd., v. 30, n.2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014
160

Atherosclerotic plaque characterization
led to the gray level images and realistic phantoms,
respectively (Figures 2c and e). In addition, as the
displacement of small structure is known, the strain
map is also generated (Figure 2d). Therefore, the
gray level images, and strain map (Figures 2c and
d) can be used as gold standards for investigation of
structural deformations and shifting of small particles,
such as speckle or pixel. For more information about
the phantom creation framework, please refer to
Cardoso et al. (2012).
Deformation measurements
Two different methods are used to compute the level of
plaque deformation, the Average of the Plaque Radial
Strain Values (AS) (Figure 2d), and the proposed
method, the Ratio of the Plaque Area Variation (AR)
(Figure 2c). The AR outcomes are obtained by:
( ) ( )
( )
High Pressure Low Pressure
Low Pressure
Area Plaque Area Plaque
AR Area Plaque
−
= (1)
where, Area(Plaque
Low Pressure
) and Area(Plaque
High Pressure
)
are the plaque areas with low, and high intraluminal
pressure, respectively. The plaque areas were obtained
from the gray level images (Figure 2c), since they are
the images representation of the mesh and deformed
mesh, gold standards, the plaque under investigation
in low and high pressure are isolated and their area
computed. The AS results are extracted from the
corresponding radial strain map by:
plaque plaque
AS = ε ±σ (2)
where, e
plaque
, and s
plaque
are the mean and standard
deviation of the plaque radial strain map values,
respectively (Figure 2d).
Figure 1. Block diagram of the proposed procedure. An illustration of an IVUS acquisition with a compliant balloon in: (a) low pressure; (b) high pressure. IVUS
phantoms corresponding to the previous acquisition at: (c) low pressure; (d) high pressure.
Rev. Bras. Eng. Bioméd., v. 30, n. 2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014 161

Moraes MC, Cardoso FM, Furuie SS
Figure 2. Block diagram of the overall methodology of this approach: (a) A coronary cross section model under evaluation (M); (b) The
mesh, c, and deformed mesh, c
deformed
, of M, after FEM; (c) The corresponding gray level images of the mesh, c, and deformed mesh, c
deformed
,
Gray(c) and Gray(cdeformed), respectively; (d) Strain Map corresponding to the resultant deformation; (e) The resultant non-deformed and
deformed IVUS phantoms.
Rev. Bras. Eng. Bioméd., v. 30, n.2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014
162

Atherosclerotic plaque characterization
The investigation
Using IVUSSim, the present approach aims to
demonstrate computationally the potential of the
proposed index, AR, for atherosclerotic tissue
characterization. In addition, the index reliability
is reinforced, by comparing its results with the
corresponding outcomes obtained by strain map
computations, a widely used index, which can be
found in many studies (Baldewsing et al., 2004; Le
Floc’h et al., 2009; Maurice et al., 2005; Shapo et al.,
1996). Specically, the phantoms were generated with
ultrasound frequency at 20 MHz and transducer with
256 RF-lines and the tissue acoustic parameters of
Table 1. The FEM was carried out in a set of coronary
cross section models under a variation of conditions,
such as different cap-thickness values, and Young’s
modulus. The deformation results of each kind of
plaque, highly-lipidic, bro-lipidic, and calcied,
were computed by the two indexes, so that they
could be directly compared. The overall procedure
of this approach is summarized as follows (Figure 2).
First, a Coronary Cross Section Model (Figure 2a)
is chosen from a set of models to be investigated
(Figure 3). Second, FEM was performed to simulate the
different physiological coronary and plaque behaviors
(Cardoso et al., 2012). Third, the two deformation
measurements, AR (Figure 2c) and AS (Figure 2d)
were used to obtain the plaque deformation values.
Finally, the correspondence between the two methods
and the reliability of the proposed method were directly
related, and statistically corroborated during the
Evaluation (Figure 2). The evaluation was performed
by computing and comparing the deformation values
obtained by the proposed method, AR, and a well
known method, strain map computations. Pearson
correlation, Bland-Altman plot, and the Histogram
of the deformation values computed by the two
indices were carried out so that the reliability of
the proposed method could be analyzed. Finally,
the results obtained by the gold standards, the gray
level images, were compared to the corresponding
outcome obtained by the segmented phantoms; thus,
the impact of segmentation for highly-lipidic plaque
could be measured.
Coronary cross section models
A set of coronary cross section models, with a variation
of morphological features, was created to have the
coronary plaque deformities investigated (Figure 3).
The set of models represents the anatomical shape of
artery coronary cross sections in a diastole cardiac
phase. The adventitia is represented by the lighter
gray region, the media is the dark gray, and the
blue, brown and white represent the highly-lipidic,
bro-lipidic, and calcied plaques, respectively
(Figure 3). Each model was simulated in different
morphological situations. The morphological situations
correspond to the different kinds of plaques and their
disposition in the coronary with the cap thickness
values, Cap = 100, 200, and 300 µm, common values
employed by apposite studies (Baldewsing et al.,
2004; Le Floc’h et al., 2009; Maurice et al., 2005).
Therefore, the plaque behavior when the plaque is
isolated, and when it is neighboring another plaque,
with different cap thickness, could be investigated.
Specically for models M
f
and M
i
(Figures 3f and i),
the calcied plaque is in front of the highly-lipidic
and bro-lipidic plaques, respectively. In a real
IVUS image, the calcied tissue would produce a
shadow covering the others plaques. For this reason,
the analysis of the highly-lipidic and bro-lipidic
plaques in this situation would be unfeasible, since
investigators and methods wouldn’t be able to see them.
However, the goal here, by using these two models,
is to investigate the calcied plaque behavior with
other neighbor’s tissues. Since the calcied can be
identied in real IVUS images and in the phantoms,
the investigation could be performed.
Coronary and plaques parameters for
performing FEM
The physiological coronary and plaque properties and
behaviors were numerically represented. In order to
do that, the coronary and plaque parameters used in
Table 1. Acoustic and mechanical tissue parameters used in the Models.
Tissue
Parameters
Acustic Mechanical
Attenuation [dB/(cm.MHz)] Acoustic impedance [MRayl] Young’s modulus [kPa]
Adventicia 1.57 1.81 80kPa, 150kPa, 300kPa, 450kPa, 600kPa
Media 1.09 1.62 600kPa
Higly lipidic 0.48 1.40 25kPa
Fibro lipidic 0.54 1.63 296kPa
Calcied 6.90 7.38 1500kPa
Rev. Bras. Eng. Bioméd., v. 30, n. 2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014 163

Moraes MC, Cardoso FM, Furuie SS
IVUSSim, were extracted by a number of previous
investigations (Baldewsing et al., 2004; Cardoso et al.,
2012; Le Floc’h et al., 2009; Zienkiewicz et al., 2010a,
2010b). The mesh was generated with triangular
element, using Delaunay triangulation. To maximize
precision, a higher mesh density was disposed at
the borders (Figure 2b). The intraluminal forces
were applied perpendicularly to the lumen border,
corresponding to a compliant balloon producing
from 1 to 0.05 atm of pressures. The minimum 0.05
corresponds approximately to 40 mmHg, which is
the normal difference between systolic and diastolic
cardiac pressure. The maximum of 1 atm was chosen,
because level of pressure provides a good tradeoff
between high deformation ratio and risk of plaque
rupture during acquisition procedures. Moreover, by
using the balloon, problems with catheter eccentricity
and inclination are overcome, ensuring the reliability of
the deformed values (Choi et al., 2002; De Korte et al.,
1999; Shapo et al., 1996). The xed nodes, at the
external border, after the adventitia tissue, were selected
to be the boundary condition. The arterial and plaque
mechanical properties used were: Poisson’s ratio,
ν = 0.49, and Young’s modulus, E = 600 kPa, 1500
kPa, 296 kPa, and 25 kPa, for the media, calcied,
bro-lipidic, and highly-lipidic with macrophages,
respectively. As the adventitia may have a variation
of elasticity, 5 values were considered for simulation
E
adventitia
= 80, 150, 300, 450, 600 kPa. All parameters
values are correspondent to mechanical properties of
in vivo coronaries, and were obtained from previous
related studies (Table 1) (Baldewsing et al., 2004;
Cardoso et al., 2012; De Korte and Van der Steen,
2002; Le Floc’h et al., 2009; Shapo et al., 1996;
Zienkiewicz et al., 2010a, 2010b).
Results
We evaluated the reliability of the proposed index
by analyzing two aspects, the correlation between,
AR and AS, as well as the equivalence between
deformation ratios and corresponding tissues. The
deformation values of AR and AS for the three plaques
using nine models (Figure 3) in different situations
Figure 3. Coronary cross section models (Mn) to investigate: (a) The highly-lipidic plaque; (b) The bro-lipidic plaque; (c) The calcied plaque; (d) Highly-lipidic
with a bro-lipidic plaque neighbor; (e) The bro-lipidic with a highly lipidic plaque neighbor; (f) The calcied with a highly-lipidic neighbor; (g) The highly-lipidic
with a calcied plaque neighbor; (h) The bro-lipidic tissue with a calcied plaque neighbor; (i) The calcied with a bro-lipidic neighbor.
Rev. Bras. Eng. Bioméd., v. 30, n.2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014
164

Atherosclerotic plaque characterization
were acquired at 6 different pressures, 1, 0.75, 0.5,
0.25, 0.1, 0.05 atm. The Models, M
a
, M
d
, and M
g
(Figures 3a, d, and g), were devoted to deformation
values of the highly-lipidic plaque, whereas Mb, Me,
M
h
, and M
c
, M
f
, M
i
were dedicated to the bro-lipidic,
and calcied plaque ones, respectively (Figures 3b,
e, h, and c, f, i). In addition, different morphological
situations were also considered for computing the
deformation outcomes, for instance, Cap = 100,
200, 300 µm, and adventitia elasticity Eadventitia = 80,
150, 300, 450, 600 kPa. The entire simulation for the
different models and morphological situations led
to 135 values for each pressure, for instance Table 2
shows the values for 1 atm of pressure, which were
computed by Equations 1 and 2. As it can be observed
in Table 2, the greatest majority of AR values are very
close and inside the mean and standard deviation
of the AS index, showing a strong correspondence
between them.
Comparison
The evaluation was performed by: First computing
the Pearson Correlation between each AR result with
the corresponding AS (reference method); Second,
analyzing residuals using Bland Altman plot; Third
verifying the separability among classes (Histogram)
for the deformation outcomes of each of the 6 applied
pressures (Figures 4, 5 and 6). In addition, the impact
of segmentation, for highly-lipidic plaque classication,
was measured.
Pearson Correlation is an index which measures
how similar two sets of data are; the level of correlation
is denoted by (ρ), and the closer it is to 1, the higher is
the correlation. Specically, the Pearson Correlation
was computed, and the correspondences between
AR and AS, for the results of the 6 pressures were
obtained (Figures 4). As can be observed in Figure 4,
there are two predominant clusters, as highlighted
in Figure 4a: bottom-left - cluster of highly-lipidic
values; top-right: the other tissues, for instance bro-
lipidic and calcied. The two clusters can be noticed
for the six applied pressures. The linear aspect of the
clusters in all cases (Figure 4) demonstrates the strong
correspondence between the two indexes, AR and AS.
It proved the strong relationship between the AR and
AS values, for highly-lipidic, bro-lipidic calcied
tissue results. In addition, a correlation, ρ, very close
to 1 in practically all cases (Figure 4), with a linear
correlation (AR ≈
1.2AS), strongly corroborate the
two indexes proportionality.
The Bland Altman plot provides information
about the level of agreement between two sets of data,
which are devoted to measure a common property;
the more points inside the limits, the greater the
agreement between two indices. In our evaluation, the
Bland Altman plot was performed between AR and
AS also for the 6 pressures (Figures 5). As expected,
it could be identied two major clusters of data, the
data provided by the highly-lipidic and other tissues
deformations, highlighted in Figure 5a. As can be
seen in Figure 5, the plots, for all pressures, show
the immense majority of points inside the limit of
agreement, indicating the strong agreement between
AS and AR outcomes.
The Histogram is a representation of data
distribution. By carrying out the histogram of a
dataset in different situations, besides the frequency
of occurrence, we can visually demonstrate the
tendency in shape and direction of each cluster.
Additionally, it permits the computation of some
related parameters, such as separability (η), and the
difference in distance between clusters (∆(%)). Again,
in this evaluation, the Histogram for AR and AS data,
for the 6 applied pressures was computed (Figure 6).
As can be observed in Figure 6, the highly-lipidic
and other tissue clusters of deformation values can
be well identied in both, the AR and AS histogram,
and for all applied pressure. In addition, both indexes,
AR and AS, provide similar distributions (Figure 6).
Therefore, the high correlation between the proposed
and usual method are reinforced.
Highly-lipidic discrimination
It can be assumed that the AR distributions discriminate
better the tissues than the values computed by AS
(Figure 6). As can be observed in Figure 6, the higher
average of highly-lipidic distribution of AR led to a
greater difference between the highly-lipidic and the
other cluster. We also analyzed the discrimination
capability of AR by computing two indices, the
histogram class separability, η, (Otsu, 1979), and
the difference in distance between classes of tissue
(∆(%)). The η quanties how two clusters of data can
be well discriminated; it varies from 0 to 1, the greater
the separability, the better the method is for classes
discrimination. The ∆(%) provides the minimum
distance between edges of two classes, illustrated
in Figures 6a and b. Both indices, η and ∆(%) were
computed between the two clusters for the two indices,
AR and AS, and for all pressures (Figure 6). As can
be observed in Figure 6, η is almost the same, while
∆(%) for the AR values is higher than for the AS in
all cases. Consequently, highly-lipidic plaque can be
reliably identied by the proposed method.
Impact of segmentation for highly-lipidic
discrimination
Phantoms from the models with lipidic plaques were
created for every pressure, using the parameters
described in Table 1. The highly-lipidic plaques
Rev. Bras. Eng. Bioméd., v. 30, n. 2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014 165

Moraes MC, Cardoso FM, Furuie SS
Table 2. Deformation values of each plaque for 1 atm of pressure, measured in the 9 models with two indexes, AR and AS, three cap-thickness values, 100, 200, and 300um, and ve different adventitia elasticity, 80,
150, 300, 450, 600kPa.
Ma ( highly-lipidic ) Mb (bro-lipidic) Mc (calcied)
80kPa 150kPa 300kPa 450kPa 600kPa 80kPa 150kPa 300kPa 450kPa 600kPa 80kPa 150kPa 300kPa 450kPa 600kPa
100um AR(%) –61.50 –64.03 –66.01 –67.20 –67.95 –1.88 –2.32 –4.03 –3.38 –4.91 1.71 1.13 0.72 0.48 0.44
AS(%) –47.64 –50.59 –53.72 –55.26 –56.17 –1.45 –2.48 –3.65 –4.26 –4.63 0.92 0.51 0.00 –0.29 –0.47
±21.97 ±22.71 ±23.52 ±23.92 ±24.16 ±1.40 ±1.40 ±1.42 ±1.44 ±1.45 ±0.62 ±0.57 ±0.55 ±0.56 ±0.57
200um AR(%) –45.19 –48.10 –51.09 –51.90 –53.37 –0.97 –1.74 –2.40 –3.33 –3.49 1.82 1.28 0.85 0.97 1.16
AS(%) –35.32 –38.21 –41.39 –42.97 –43.92 –1.03 –1.96 –3.06 –3.63 –3.99 0.89 0.50 0.01 –0.27 –0.44
±17.73 ±18.64 ±19.65 ±20.16 ±20.46 ±1.18 ±1.20 ±1.23 ±1.26 ±1.28 ±0.58 ±0.56 ±0.56 ±0.58 ±0.59
300um AR(%) –39.77 –42.78 –44.98 –46.96 –47.78 –0.34 –1.59 –2.50 –2.89 –3.49 1.55 1.12 0.69 0.78 1.12
AS(%) –30.95 –33.56 –36.54 –38.07 –39.00 –1.15 –1.97 –2.95 –3.48 –3.82 0.75 0.39 –0.06 –0.32 –0.48
±15.61 ±16.48 ±17.49 ±18.00 ±18.32 ±1.10 ±1.13 ±1.18 ±1.22 ±1.24 ±0.53 ±0.51 ±0.53 ±0.55 ±0.56
Md ( highly-lipidic ) Me (bro-lipidic) Mf (calcied)
80kPa 150kPa 300kPa 450kPa 600kPa 80kPa 150kPa 300kPa 450kPa 600kPa 80kPa 150kPa 300kPa 450kPa 600kPa
100um AR(%) –71.74 –74.52 –76.64 –77.26 –77.45 –0.36 –1.29 –1.78 –2.02 –2.23 2.06 1.49 1.42 1.13 1.25
AS(%) –54.01 –58.89 –62.72 –64.23 –65.03 0.51 –0.84 –1.92 –2.34 –2.57 1.51 1.03 0.63 0.47 0.38
±23.41 ±24.76 ±25.82 ±26.23 ±26.46 ±1.84 ±1.86 ±1.92 ±1.95 ±1.98 ±1.12 ±1.12 ±1.19 ±1.23 ±1.26
200um AR(%) –53.19 –57.68 –61.20 –62.08 –63.06 0.21 –0.71 –1.46 –1.67 –1.79 1.94 1.73 1.21 1.38 1.42
AS(%) –41.94 –46.82 –50.75 –52.31 –53.15 0.44 –0.74 –1.70 –2.08 –2.29 1.36 0.93 0.56 0.41 0.33
±18.71 ±20.17 ±21.36 ±21.84 ±22.09 ±1.57 ±1.55 ±1.58 ±1.61 ±1.62 ±1.05 ±1.05 ±1.11 ±1.16 ±1.19
300um AR(%) –45.63 –50.30 –53.27 –55.09 –55.97 0.26 –0.74 –1.34 –1.55 –1.62 1.71 1.71 1.29 1.09 1.16
AS(%) –36.44 –41.30 –45.24 –46.82 –47.67 0.43 –0.66 –1.56 –1.92 –2.11 1.26 0.86 0.52 0.38 0.31
±15.26 ±16.72 ±17.92 ±18.40 ±18.66 ±1.26 ±1.27 ±1.33 ±1.37 ±1.40 ±0.85 ±0.87 ±0.95 ±1.00 ±1.04
Mg ( highly-lipidic ) Mh (bro-lipidic) Mi (calcied)
80kPa 150kPa 300kPa 450kPa 600kPa 80kPa 150kPa 300kPa 450kPa 600kPa 80kPa 150kPa 300kPa 450kPa 600kPa
100um AR(%) –73.93 –76.20 –77.37 –77.79 –78.00 –2.93 –4.82 –5.71 –6.39 –6.71 1.55 1.08 0.40 –0.13 0.06
AS(%) –56.56 –60.60 –63.90 –65.25 –65.98 –2.59 –4.11 –5.45 –6.03 –6.35 1.00 0.32 –0.32 –0.60 –0.76
±24.28 ±25.36 ±26.24 ±26.60 ±26.80 ±1.63 ±1.65 ±1.68 ±1.69 ±1.70 ±0.78 ±0.73 ±0.74 ±0.75 ±0.76
200um AR(%) –55.99 –58.91 –62.10 –63.06 –64.00 –2.54 –3.71 –5.05 –5.51 –5.99 1.48 1.13 0.27 0.25 0.17
AS(%) –44.23 –48.37 –51.82 –53.25 –54.02 –2.38 –3.76 –5.01 –5.56 –5.86 0.87 0.25 –0.34 –0.60 –0.75
±19.60 ±20.80 ±21.81 ±22.23 ±22.46 ±1.54 ±1.56 ±1.59 ±1.60 ±1.62 ±0.75 ±0.71 ±0.71 ±0.73 ±0.74
300um AR(%) –47.65 –51.63 –54.48 –55.97 –56.69 –2.32 –3.66 –4.55 –5.30 –5.15 1.38 0.66 0.24 0.20 0.20
AS(%) –38.59 –42.74 –46.25 –47.70 –48.49 –2.11 –3.43 –4.63 –5.16 –5.46 0.83 0.25 –0.31 –0.56 –0.70
±16.02 ±17.25 ±18.31 ±18.74 ±18.99 ±1.18 ±1.23 ±1.29 ±1.33 ±1.34 ±0.59 ±0.59 ±0.62 ±0.65 ±0.66
Rev. Bras. Eng. Bioméd., v. 30, n.2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014
166

Atherosclerotic plaque characterization
were manually segmented by specialists (Figure 7b).
Once the phantoms were ready two investigators,
with knowledge in IVUS image features and
segmentation, segmented manually in agreement
to each other, the corresponding lipidic regions of
the phantoms. Once the corresponding areas were
obtained, the Ratios of the Plaque Area Variation
were computed for the segmented plaques. The
impact of segmentation (IS) was obtained by
computing the error, using:
Segmented
IS AR AR=− (3)
where, AR and AR
Segmented
, are the area ratios computed
by the gold standards (Figure 7a), and corresponding
segmented images (Figure 7b), respectively, and IS is
the impact or error of segmentation. The obtained IS
result, was 4.3 ± 3.3%. This value may only inuence
the discrimination between highly-lipidic and other
tissues, for pressures equal and below 0.25 atm, where
the pressure does not cause enough deformation to
overcome the discrimination between highly lipidic
and other tissues. However, better segmentation result
can be obtained using other segmentation approaches,
or with a good preprocessing.
Figure 4. Pearson Correlation between AR and AS values for pressures of: (a) 1 atm; (b) 0.75 atm; (c) 0.5 atm; (d) 0.25 atm; (e) 0.1 atm;
(f) 0.05 atm.
Rev. Bras. Eng. Bioméd., v. 30, n. 2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014 167

Moraes MC, Cardoso FM, Furuie SS
Discussion
Previous works have demonstrated the importance,
and contributed to Elastography, and coronary disease
investigations (Baldewsing et al., 2004; De Korte
and Van der Steen, 2002; Le Floc’h et al., 2009;
Liang et al., 2009; Loree et al., 1994; Maurice et al.,
2005; Ophir et al., 1991). Consequently, information
of Elastography, coronary and atherosclerosis, has
been provided. This knowledge is important to help
cardiologists and investigators to improve diagnostic,
therapy, evaluation, as well as for creating new
tools and methods. However, despite efforts, and
advancements, atherosclerosis is still a dramatic
problem. In addition, for the majority of cardiologic
centers in the world, implementing, and managing the
methods presented by the literature is difcult. The
reason is that most of clinics and hospitals around
the world lack nancial support, and specialists with
expertise in computing and mathematical skills. As
Figure 5. Bland Altman Analyses between AR and AS values for pressures of: (a) 1 atm; (b) 0.75 atm; (c) 0.5 atm; (d) 0.25 atm; (e) 0.1 atm;
(f) 0.05 atm.
Rev. Bras. Eng. Bioméd., v. 30, n.2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014
168

Atherosclerotic plaque characterization
a result, buying new equipment, or implementing
complex methods, is not always feasible. Therefore,
new approaches with simpler implementation, more
practical, yet as reliable as previous methods, are
welcome.
We presented a simple and practical method, in
which by the area ratio of a plaque the predominant
tissue is classied. The evaluation has shown that
the proposed index reliably distinguishes the highly-
lipidic plaque from others tissues in many situation.
The methodology is based on three simple steps,
Imaging Acquisition, Plaque Area Computation, and
Ratio of the Plaque Area Variation. The reliability
of the method was demonstrated by showing the
equivalence between tissue and deformed values
(Table 2), and the high correlation and agreement
between deformed values computed by AR and
corresponding AS (Figures 4 and 5). As can be seen
Figure 6. Histogram of the AR and AS values for pressures of: (a) 1 atm; (b) 0.75 atm; (c) 0.5 atm; (d) 0.25 atm; (e) 0.1 atm; (f) 0.05 atm.
Rev. Bras. Eng. Bioméd., v. 30, n. 2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014 169

Moraes MC, Cardoso FM, Furuie SS
in Figure 4, for all plaques, and considering all
situations, the AR deformation values are proportional
to the tissue stiffness, following the AS outcomes; as
a result, a correlation close to 100% was obtained.
The strong agreement between the two indices is
reinforced by the Bland Altman analyses, in which
the great majority of points are inside the limits of
agreement (Figure 5). In addition, the AR distributions
provide good tissue discrimination (Figure 6). The
discrimination capability was quantied by computing
the histogram class separability (Otsu, 1979), and
the difference in distance between classes in which,
the strain values computed by AR provided a higher
difference (Figure 6). Besides providing better
discrimination among tissues, the AR estimation
requests a very simple area computing operation after
any segmentation method. The segmentation method,
for the specied acoustic parameters, provided an
accuracy of 4.3±3.3%, this permits reliability for all
results above 0.25 atm of pressure, since they provide
difference above 7.2%. Moreover, the image processing
operations can be entirely performed using basic
operations of free license software, such as ImageJ.
In summary, the resultant high correlation between
the proposed and well-known method, alongside the
simplicity, proved the method feasibility, and low-cost.
As a result, the contribution and implications are: (a) A
combination of simple and usual procedures for plaque
stiffness estimation; (b) The proposed method can be
applied directly to similar modalities, for instance,
Intravascular Optical Coherence Tomography (IOCT)
and Intravascular Magnetic Resonance (IMR); (c)
The proposed method may be an alternative in many
places, by allowing clinics and centers to have an
extra tool to support their exams, therapy planning,
and evaluation.
The inclination and eccentricity of the catheter
during image acquisition, not enough tissue contrast
for segmentation process, may be seen as limitation
for this approach. Consequently, they are also a
restriction for any 2D algorithm, such as strain or
elastic map, as they also rely only on 2D IVUS images.
The Poisson’s ratio, ν = 0.49, does not varies in this
study, the value were obtained and used in other
related studies, which considered the coronary a quasi-
incompressible material. Indeed, its variation would
result in different deformation values, which would not
represent the coronary strains, yet as this parameter
is applied in the entire model and not locally these
changes would be correspondent in the two indexes
AR and AS. Moreover, 3D methods would be the best
choice to provide more complete vessel and coronary
information, since they provide deformations of all
directions, and not only the transversal one. However,
2D approaches resultant of well acquired IVUS images,
with transducer carefully placed perpendicular to
vessel wall, provide good information to localize and
estimate predominant tissues of suspect areas. Indeed,
depending on the segmentation method accuracy,
higher pressure by the balloon is required, so that the
separation between highly-lipidic and other tissues
is increased. Therefore, the possession and correct
use of a compliance balloon is an important part of
the presented procedure. As a result, a simpler and
alternative method to estimate mechanical properties
of plaques can be used and evolved.
Finally, since we do not currently possess clinical
data with desirable features and pressure information,
clinical data for validation was not in the scope of
this paper. In addition, computational phantoms have
become a very solid tool, which are exible tool,
providing numerous investigations (Culjat et al.,
2010). Accordingly, numerical phantoms were the most
Figure 7. Computation of the impact of segmentation. (a) AR deformation computed using the gray level images, gold standards. (b) AR deformation
computed using the corresponding segmented phantoms.
Rev. Bras. Eng. Bioméd., v. 30, n.2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014
170

Atherosclerotic plaque characterization
advantageous and feasible choice for the methodology
evaluation. In addition, transducers with different
properties may change the segmentation accuracy,
since the main goal herein was to verify the two
indexes equivalence, a wider analysis about the
impact of segmentation depending on the transducer
properties was not the scope of this paper. However,
investigations using images acquired from physical
phantom and clinical data, with desired coronary
features and controlled parameters, will be designed
and carried out with a cardiac center, and results
presented in future works.
Acknowledgements
São Paulo Research Foundation – Brazil (FAPESP):
Brazilian National Council of Scientific and
Technological Development, (CNPq), Biomedical
Engineering Laboratory of the University of São
Paulo, Brazil (LEB-USP). Institute of Science and
Technology of the Federal University of São Paulo,
Brazil (ICT-UNIFESP).
References
Baldewsing RA, De Korte CL, Schaar JA, Mastik F, Van
der Steen AFW. A nite element model for performing
intravascular ultrasound Elastography of human
atherosclerotic coronary arteries. Ultrasound in Medicine
& Biology. 2004; 30(6):803-13. http://dx.doi.org/10.1016/j.
ultrasmedbio.2004.04.005
Cardenas DAC, Moraes MC, Furuie SS. Segmentação do
lúmen em imagens de IOCT usando Fuzzy Connectedness
e Reconstrução Binária Morfológica. Revista Brasileira de
Engenharia Biomédica. 2013; 29(1):32-44. http://dx.doi.
org/10.4322/rbeb.2013.004
Cardoso FM, Moraes MC, Furuie SS. Realistic IVUS image
generation in Different Intraluminal Pressures. Ultrasound
in Medicine & Biology. 2012; 38(12):2104-19. http://dx.doi.
org/10.1016/j.ultrasmedbio.2012.08.005
Céspedes EI, Ophir J, Ponnekanti H, Maklad N. Elastography:
Elasticity imaging using ultrasound with application to muscle
and breast in vivo. Ultrasonic Imaging. 1993; 15(2):73-88.
Choi CD, Skovoroda AR, Emelianov SY, O’Donnell M .
An integrated compliant balloon ultrasound catheter for
intravascular strain imaging. IEEE Transaction on Ultrasonics
Ferroelectric and Frequency Control. 2002; 49(11):1552-60.
http://dx.doi.org/10.1109/TUFFC.2002.1049737
Culjat MO, Goldenberg D, Tewari P, Singh RS. A review
of tissue substitutes for ultrasound imaging. Ultrasound in
Medicine & Biology. 2010; 36(6):861-73. http://dx.doi.
org/10.1016/j.ultrasmedbio.2010.02.012
Davies MJ. The pathophysiology of acute coronary
syndromes. Heart and Education in Heart. 2000; 83(3):361-
6. http://dx.doi.org/10.1136/heart.83.3.361
Davies MJ. Stability and instability: two faces of coronary
atherosclerosis. The Paul Dudley White lecture 1995.
Circulation. 1996; 94(8):2013-20. http://dx.doi.
org/10.1161/01.CIR.94.8.2013
De Korte CL, Van der Steen AFW. Intravascular ultrasound
Elastography: an overview. Ultrasonics. 2002; 40:859-65.
http://dx.doi.org/10.1016/S0041-624X(02)00227-5
De Korte CL, Céspedes EI, Van der Steen AFW. Inuence
of catheter position on estimated strain in intravascular
Elastography. IEEE Transaction on Ultrasonics, Ferroelectrics,
and Frequency Control. 1999; 46(3):616-25. http://dx.doi.
org/10.1109/58.764848
Falk E, Shah PK, Fuster V. Coronary plaque disruption.
Circulation. 1995; 92:657-71. http://dx.doi.org/10.1161/01.
CIR.92.3.657
Fisher A, Gustein DE, Fayad ZA, Fuster V. Predicting plaque
rupture: enhancing diagnosis and clinical decision-making in
coronary artery disease. Vascular Medicine. 2000; 5:163-72.
Hoskins PR. Simulation and validation of arterial ultrasound
imaging and blood ow. Ultrasound in Medicine &
Biology. 2008; 34(5):693-17. http://dx.doi.org/10.1016/j.
ultrasmedbio.2007.10.017
Le Floc’h S, Ohayon J, Tracqui P, Finet G, Gharib
AM, Maurice RL, Cloutier G, Pettigrew RI. Vulnerable
atherosclerotic plaque elasticity reconstruction based on a
segmentation-driven optimization procedure using strain
measurements: Theoretical framework. IEEE Transaction
on Medical Imaging. 2009; 28(7):1126-37. http://dx.doi.
org/10.1109/TMI.2009.2012852
Liang Y, Zhu H, Friedman MH. The correspondence
between coronary arterial wall strain and histology in a
porcine model of atherosclerosis. Physics in Medicine
and Biology. 2009; 54(18):5625-41. http://dx.doi.
org/10.1088/0031-9155/54/18/018
Loree HM, Tobias BJ, Gibson LJ, Kamm RD, Small DM,
Lee RT. Mechanical properties of model atherosclerotic
lesion lipid pools. Arteriosclerosis, Thrombosis, and Vascular
Biology. 1994; 14(2):230-4. http://dx.doi.org/10.1161/01.
ATV.14.2.230
Maurice RL, Brusseau E, Finet G, Cloutuer G. On the potential
of the Lagrangian speckle model estimator to characterize
atherosclerotic plaques in endovascular Elastography: in
vitro experiments using an excised human carotid artery.
Ultrasound in Medicine & Biology. 2005; 31(1):85-91. http://
dx.doi.org/10.1016/j.ultrasmedbio.2004.07.009
Moraes MC, Furuie SS. An approach to automatically
segment the media-adventitia borders in IVUS. Revista
Brasileira de Engenharia Biomédica. 2010; 26(3):219-33.
Moraes MC, Furuie SS. Automatic coronary wall
segmentation in intravascular ultrasound images using binary
morphological reconstruction. Ultrasound in Medicine &
Biology. 2011; 37(9):1486-99. http://dx.doi.org/10.1016/j.
ultrasmedbio.2011.05.018
Ohayon J, Teppaz P, Finet G, Rioufol G. In-vivo prediction
of human coronary plaque rupture location using
intravascular ultrasound and the nite element method.
Rev. Bras. Eng. Bioméd., v. 30, n. 2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014 171

Moraes MC, Cardoso FM, Furuie SS
Coronary Artery Disease. 2001; 12:655-63. http://dx.doi.
org/10.1097/00019501-200112000-00009
Ophir J, Céspedes EI, Ponnekanti H, Yazdi Y, Li X.
Elastography. A quantitative method for imaging the elasticity
of biological tissues. Ultrasonic Imaging. 1991; 13(2):111-34.
Otsu N. A threshold selection method from gray-level
histograms. IEEE Transaction on Systems, Man, and
Cybernetics - Part C Applications and Reviews. 1979; 9(1):62-
6. http://dx.doi.org/10.1109/TSMC.1979.4310076
Roger VL, Go AS, Lloyd-Jones DM, Benjamin EJ, Berry
JD, Borden WB, Bravata Dawn M, Dai S, Ford ES, Fox CS,
Fullerton HJ, Gillespie C, Hailpern SM, Heit JA, Howard VJ,
Kissela BM, Kittner SJ, Lackland DT, Lichtman JH, Lisabeth
LD, Makuc DM, Marcus GM, Marelli A, Matchar DB, Moy
CS, Mozaffarian D, Mussolino ME, Nichol G, Paynter NP,
Soliman EZ, Sorlie PD, Sotoodehnia N, Turan TN, Virani SS,
Wong ND, Woo D, Turner MB. Executive summary: heart
disease and stroke statistics - 2012 update: a report from
the American Heart Association. http://dx.doi.org/10.1161/
CIR.0b013e3182456d46irculation. 2012; 125(1):188-97.
Rosamond W, Flegal K, Friday G, Furie K, Go A, Greenlund
K, Haase N, Ho M, Howard V, Kissela B, Kittner S, Lloyd-
Jones D, McDer-mott M, Meigs J, Moy C, Nichol G,
O’Donnell CJ, Roger V, Rumsfeld J, Sorlie P, Steinberger J,
Thom T, Wasserthiel-Smoller S, Hong Y. Heart disease and
stroke statistics - 2007 update: a report from the American
Heart Association Statistics Committee and Stroke Statistics
Subcommittee. Circulation. 2007; 115(5):e69-171. http://
dx.doi.org/10.1161/CIRCULATIONAHA.106.179918
Shapo BM, Crowe JR, Skovoroda AR, Eberle MJ, Cohn
NA, O’Donnell M. Displacement and strain imaging
of coronary arteries with intraluminal ultrasound.
IEEE Transaction on Ultrasonics, Ferroelectrics, and
Frequency Control. 1996; 43(2):234-46. http://dx.doi.
org/10.1109/58.485949
Viermani R, Kolodgie FD, Burke AP, Farb A, Schwartz SM.
Lessons from sudden coronary death: A comprehensive
morphological classication scheme for atherosclerotic
lesions. Arteriosclerosis, Thrombosis, and Vascular
Biology. 2000; 20:1262-75. http://dx.doi.org/10.1161/01.
ATV.20.5.1262
Zienkiewicz OC, Taylor RL, Zhu JZ. The nite element
method: its basis & fundamentals. 6
th
ed. Great Britain:
Elsevier; 2010a.
Zienkiewicz OC, Taylor RL, Zhu JZ. The nite element
method: for solid and structural mechanics. 6th ed. Great
Britain: Elsevier; 2010b.
Authors
Matheus Cardoso Moraes*
Departamento de Ciência e Tecnologia, Instituto de Ciência e Tecnologia – ICT,
Universidade Federal de São Paulo – UNIFESP, Rua Talim, 330, CEP 12231-280, São José dos Campos, SP, Brasil.
Fernando Mitsuyama Cardoso, Sérgio Shiguemi Furuie
Laboratório de Engenharia Biomédica, Departamento de Engenharia de Telecomunicações e Controle,
Escola Politécnica – Poli, Universidade de São Paulo – USP, São Paulo, SP, Brasil.
Rev. Bras. Eng. Bioméd., v. 30, n.2, p. 159-172, jun. 2014
Braz. J. Biom. Eng., 30(2), 159-172, June 2014
172