Download Title Characterization of the heart muscle aniosotropy using

Title
Author(s)
Citation
Issued Date
URL
Rights
Characterization of the heart muscle aniosotropy using
ultrasound Nakagami imaging
Yu, X; Lee, WN
The 2014 IEEE International Ultrasonics Symposium (IUS),
Chicago, IL., 3-6 September 2014. In IEEE Ultrasonics
Symposium Proceedings, 2014, p. 2367-2370
2014
http://hdl.handle.net/10722/208745
IEEE Ultrasonics Symposium Proceedings. Copyright © IEEE.;
©2014 IEEE. Personal use of this material is permitted.
Permission from IEEE must be obtained for all other uses, in any
current or future media, including reprinting/republishing this
material for advertising or promotional purposes, creating new
collective works, for resale or redistribution to servers or lists,
or reuse of any copyrighted component of this work in other
works.; This work is licensed under a Creative Commons
Attribution-NonCommercial-NoDerivatives 4.0 International
License.
10.1109/ULTSYM.2014.0590
Characterization of the Heart Muscle Aniosotropy
Using Ultrasound Nakagami Imaging
Xue Yu
Wei-Ning Lee
Electrical and Electronic Engineering
The University of Hong Kong
Hong Kong
[email protected]
Electrical and Electronic Engineering
Medical Engineering Programme
The University of Hong Kong
Hong Kong
[email protected]
Abstract—The myocardial architecture of the left ventricle is
important for diagnosis, as it is associated with the mechanical
and electrical functions. Nakagami imaging has been used in
modeling the statistics of backscattered ultrasound signals in soft
tissues, and the m value is found capable of differentiating
various scattering conditions. Therefore, this study aims to apply
Nakagami imaging to explore its feasibility in evaluation of the
structural anisotropy in the myocardium. Three excised porcine
left ventricles were embedded in gelatin phantoms and imaged in
standard long- and short-axis views using the coherent plane
wave compounding technique and a linear array probe operating
at 5.2 MHz. The Nakagami parametric map was generated using
a window-based method using an optimized window size of 12λ x
12λ. Four parameters, including mean, variance, skewness and
kurtosis, were used to characterize the distribution of m values.
Our results showed that the distributions of m values in the
middle layers were statistically different from the subepicardial
and subendocardial layers (p < 0.05). The mean m values were
significantly different both between anterior and posterior heart
walls and between long- and short-axis views (p < 0.05). Our
preliminary results showed a potential of Nakagami imaging in
myocardial anisotropy characterization.
Keywords—Nakagami distribution; myocardium; ansiotropy;
backscattering
I. INTRODUCTION
The myocardium is composed of myofibers, arranged into a
layered structure. These layers in the left ventricle (LV) exhibit
a transmural variation in orientation, from -60° near the
epicardium to +60° close to the endocardium [1]. The myofiber
architecture is closely related to the mechanical [2] and
electrical [3] properties of the heart. Heart dysfunction, caused
by aging or myocardial diseases, can be associated with
myofiber disorientation. Arrhythmia, for example, is majorly
caused by the abnormalities in conduction of cardiac impulse,
which is partially induced by the architectural changes in the
myocardium [4]. Therefore, evaluation of the myocardial
architectural features may aid the understanding of the cardiac
function and potentially benefit diagnosis.
Non-invasive investigation of myocardial fiber architecture
has been performed using various medical imaging techniques,
for example, magnetic resonance imaging (MRI) and
echocardiography-based imaging. Diffusion tensor imaging
(DTI), for instance, is currently the gold standard in mapping
978-1-4799-7049-0/14/$31.00 ©2014 IEEE
2367
the myocardial fiber orientation in a non-invasive fashion. DTI
is sensitive to the restricted diffusion of water molecules in
tissue, which is related to the local tissue structure, and enables
two- and three-dimensional tracking of the myofiber
architecture [5,6]. The major disadvantages of DTI are that it is
time consuming and cost inefficient. Meanwhile, the
echocardiography-based methods are routinely applied in the
clinic for functional and geometrical analysis in the heart due
to its wide availability and real-time imaging capability. The
anisotropy of the myocardium has been studied by ultrasound
imaging. The backscatter-based approach proposed in the
1980s can correlate the variations in backscattered signals with
the directions of insonification based on the phenomenon that
attenuation is greater when ultrasound propagates in a direction
parallel to the fibers than when it travels perpendicular to the
fibers [7]. This approach, however, depends on the ultrasound
system settings and requires image calibration before
subsequent analyses. Echocardiography-based shear wave
imaging [8,9] has been exploited in mapping the myocardial
fiber orientation in good agreement with histology and DTI.
Nonetheless, shear wave imaging has not been widely available
in clinical practice yet due to its special hardware requirements.
Nakagami imaging is a non-invasive technique for tissue
characterization through the analysis of the backscattered
ultrasonic signals and does not require special hardware for
data collection. The Nakagami distribution is a two-parameter,
generalized distribution that can model the backscattered
signals. It encompasses various scatterer statistics and exhibits
computational simplicity. According to simulation and
phantom study findings, the Nakagami shape parameter, m,
Fig. 1. Three scatterer statistics in the resolution cell that can be
modeled by Nakagami distribution, including (a) a small number of
scatterers, (b) a large number of randomly distributed scatterers, and (c)
randomly distributed scatterers with regularly spaced scatterers.
2014 IEEE International Ultrasonics Symposium Proceedings
Fig. 2. Illustration of ultrasound scanning of the anterior wall at the mid-left-ventricular level in (a) a long-axis view and (b) a short-axis view, and their
corresponding Nakagami m parametric maps were overlayed on the B-mode images .
was sensitive in differentiating the number density of
scatterers, the different scattering cross-sections, and the
presence and absence of periodic structures [10]. This study
aims to evaluate the feasibility of Nakagami imaging in
characterizing the anisotropy in the myocardium.
II. MATERIALS AND METHODS
A. Nakagami distribution
The probability density function of the envelope of the
backscattered ultrasonic signal, f(R), fitted by the Nakagami
distribution is as follows [10]:
f(R) = (2mmR2m-1)/(Γ(m)Ωm))exp((-m) R2/ Ω)U(R) (1)
where Γ( · ) and U( · ) are the gamma function and unit step
function, respectively; R is the envelope of the backscattered
signal. The shape parameter (m) and the power parameter (Ω)
are calculated by:
m = [E(R)2]2/E[R2 – E(R)2]2
Ω = E(R2),
(2)
(3)
where E( · ) denotes the statistical mean. The m value can
differentiate various scatterer statistics in soft tissues (Fig. 1):
when m is between 0.5 and 1, the envelope statistics is
Rayleigh distributed; when m is smaller than 0.5, the envelope
statistics conforms to a pre-Rayleigh distribution; when m is
larger than 1, the envelope statistics becomes a post-Rayleigh
distribution.
B. Porcine hearts and data acquisitoin
Three freshly excised porcine hearts were examined in this
study, and the LVs were carefully dissected. The apex and the
base of the LVs were marked before embedded in a 5% gelatin
phantom.
Ultrasound in-phase and quadrature (IQ) data were
acquired using Vantage 256 (Verasonics, Redmond, WA,
2368
USA) and a 128-element linear array probe operating at 5.2
MHz at a sampling rate of 20.8 MHz. The coherent plane wave
compounding sequence [11] was used in data acquisitions, with
a total of 71 steered plane waves emitted sequentially. The
overall angular aperture was 28.6°, from -14.3° to 14.3° at 4.1°
increments, and the images for each plane wave insonification
were coherently added.
The LV was divided into anterior and posterior regions, and
both the anterior and posterior walls were imaged in both
standard long- and short-axis views. For both the anterior and
posterior walls, the long-axis direction was first identified by
linking the two markers at the base and the apex, and imaging
planes were taken at the mid-ventricular level. Then the
phantom was rotated by 90° about the center of the transducer
before acquiring the images from the short-axis view. Image
acquisitions were repeated for three times to test
reproducibility.
C. Data analysis
While Nakagami imaging is a window-based method, a
previous study showed that Nakagami parametric map quality
was related to the size of the sliding window [12] and that
decreasing the size of the sliding window improved the
resolution of Nakagami imaging while sacrificing the stability
of estimation. Based on our pilot study, a window of 12λ x
12λ was chosen to balance the trade-off between estimation
stability and image resolution.
In order to explore the feasibility of Nakagami imaging in
differentiating wall regions in the heart wall, the myocardium
was manually segmented into 3 equally thick layers, i.e.,
subepicardial, middle and subendocardial layers, along the
radial direction (Fig. 2). Regions of interest (ROIs) were
contoured on B-mode images, and overlaid on the Nakagami
image in MatLab (The Mathworks, Natick, MA, USA).
D. Statistical analysis
2014 IEEE International Ultrasonics Symposium Proceedings
Fig. 3. Histograms of Nakagami m values in (a) subepicedial, (b) middle and (c) subendocardial layers of three porcine myocardium (each with 3 repeated
measurements). All layers were manually segmented and had equal thickness.
Histograms of the m value within the three layers were
plotted for the anterior and posterior walls in both the longand short-axis views. The shape of the histograms was
described by mean, skewness, kurtosis and variance of m
values. For comparison between the anterior and the posterior
walls, the afore-mentioned parameters were analyzed by
Student’s t-tests. Similarly, the comparison between long- and
short-axis views was also performed by Student’s t-test for
mean, skewness, kurtosis and variance. The m value
distributions for different layers in the myocardium were
evaluated using one-way ANOVA for the four parameters.
Significance was assumed when p-value was smaller than
0.05.
III. RESULTS AND DISCUSSION
Fig. 3 shows the differences of the m distribution in each
of the three layers across the wall, and the different m
distributions for the anterior and posterior walls in the longand short-axis views. The histograms of m values in the longaxis view resembled those in the short-axis view in both the
anterior (solid line) and posterior (dash line) walls. The
anterior and posterior walls in either the long- or the short-axis
view, however, exhibited distinct distributions of m values in
the subepicardial and middle layers of the myocardium. The
subepicardial layer of the posterior wall showed a larger
amount of low m values (below 0.3), while that of the anterior
wall had m values widely spread from 0 to 1.3. In the middle
layer, the anterior wall had a comparatively symmetrical m
value distribution, and was less affected by small m values.
The mid-wall of the anterior region, on the contrary, exhibited
a wide spectrum of m values ranging from 0 to 1.0. The
differences between the anterior and posterior walls in the
long- or short-axis view in the subendocardial layer were less
notable compared to those in the subepicardial and middle
layers.
The distributions of the m values were described by four
parameters as demonstrated in Fig. 4. In addition to mean and
variance, skewness measures the asymmetry of a distribution
deviating from the Normal distribution, and kurtosis measures
the height of the peak of one distribution compared to its tails.
The anterior and posterior walls had significant differences (p
< 0.05) in mean m values but similar skewness, kurtosis and
variances. Similarly, under different echocardiography views,
only significant differences were observed in the mean m
values (p < 0.05). For the layer-specific m value distributions,
the middle layer had significantly higher mean m values
compared to the subepicardial and subendocardial layers (p <
0.05). This echoed the findings in Fig. 3 that low m values
were more likely to be found in the subepicardial and
subendocardial layers. The kurtosis of the m distribution in the
Fig. 4. Comparison of mean, kurtosis, skewness and variance in (a) the anterior/posterior, (b) the long-/short-axis, and (c) the
subepicardial/middle/subendocardial layers. (a) The m values of the anterior wall were higher than that of the posterior wall. (b) The m values of the long-axis
views were higher than that of the short-axis views. (c) Among the 3 layers across the myocardiall wall, the middle layer had the highest m values, a lest
skewed distribution and smallest variance of m values. *Indicates statistical significance at p < 0.05.
2369
2014 IEEE International Ultrasonics Symposium Proceedings
middle layer was significantly larger than that in the
subepicardial layer (p < 0.05), indicating that the m values in
the middle layer were more concentrated at higher m values
(Fig. 3b). The middle layer also had the least skewed
distribution compared to the subepicardial and subendocardial
layers (p < 0.05) as observed in Fig. 3b. Among all layers, the
middle layer had the smallest variances although the deviation
from the subendocardial layer was not statistically significant.
To our best knowledge, this is the first study applying
Nakagami imaging in the heart. In this preliminary study, we
found that the middle layer under Nakagami imaging was
different from the subepicardial and subendocardial layers.
This may be explained by the widely known transmural
variance in the fiber orientation in the LV. Fibers run
circumferentially (0±22.5°) in the middle region of the
myocardium at mid-cavity level, while they rotate clockwise
and counterclockwise with respect to the circumferential
direction in the subepicardial and subendocardial layers,
respectively [1]. Since both long- and short-axis views imaged
the cross-section of the myocardial fibers in the subepicardial
and subendocardial layers, the clockwise and counterclockwise
rotation of fibers may have resulted in similar scattering
patterns. Therefore, Nakagami imaging might have limited
sensitivity in differentiating the subepicardial and
subendocardial layers. Nakagami imaging also revealed the
difference between the anterior and posterior walls in the
middle layer, and the anterior wall had significantly larger m
values.
In the future studies, anisotropy revealed by Nakagami
imaging should be correlated with a gold standard, such as
histology. Since the backscattered signals are probe-angle
dependent, the influence of insonification angles on Nakagami
imaging shall also be investigated.
IV. CONCLUSION
Our results showed that Nakagami imaging was able to
differentiate the middle layer of the myocardium from the
subepicardial and subendocardial layers. The anterior wall had
a larger m values compared to the posterior wall in the middle
layer. In addition, the long-axis view also had a larger mean m
value in comparison with the short-axis view. Therefore,
Nakagami imaging may have the potential in characterizing
the structural anisotropy in the myocardium.
ACKNOWLEDGMENT
This work is supported by the Research Grants Council of
Hong Kong (HKU 739413E).
2370
REFERENCES
1. Streeter DD, Jr., Spotnitz HM, Patel DP, Ross J, Jr.,
Sonnenblick EH (1969) Fiber orientation in the canine left
ventricle during diastole and systole. Circ Res 24 (3):339-347
2. Costa KD, Takayama Y, McCulloch AD, Covell JW (1999)
Laminar fiber architecture and three-dimensional systolic
mechanics in canine ventricular myocardium. The American
journal of physiology 276 (2 Pt 2):H595-607
3. Sano T, Takayama N, Shimamoto T (1959) Directional
difference of conduction velocity in the cardiac ventricular
syncytium studied by microelectrodes. Circ Res 7 (2):262-267
4. Peters NS, Wit AL (1998) Myocardial architecture and
ventricular arrhythmogenesis. Circulation 97 (17):1746-1754
5. Tseng WY, Dou J, Reese TG, Wedeen VJ (2006) Imaging
myocardial fiber disarray and intramural strain hypokinesis in
hypertrophic cardiomyopathy with MRI. Journal of magnetic
resonance imaging : JMRI 23 (1):1-8. doi:10.1002/jmri.20473
6. Reese TG, Weisskoff RM, Smith RN, Rosen BR, Dinsmore
RE, Wedeen VJ (1995) Imaging myocardial fiber architecture
in vivo with magnetic resonance. Magn Reson Med 34
(6):786-791
7. Mottley JG, Miller JG (1988) Anisotropy of the ultrasonic
backscatter of myocardial tissue: I. Theory and measurements
in vitro. J Acoust Soc Am 83 (2):755-761
8. Lee WN, Pernot M, Couade M, Messas E, Bruneval P, Bel
A, Hagege AA, Fink M, Tanter M (2012) Mapping
myocardial fiber orientation using echocardiography-based
shear wave imaging. IEEE Trans Med Imaging 31 (3):554562. doi:10.1109/TMI.2011.2172690
9. Lee WN, Larrat B, Pernot M, Tanter M (2012) Ultrasound
elastic tensor imaging: comparison with MR diffusion tensor
imaging in the myocardium. Phys Med Biol 57 (16):50755095. doi:10.1088/0031-9155/57/16/5075
10. Shankar PM (2000) A general statistical model for
ultrasonic backscattering from tissues. Ieee T Ultrason Ferr 47
(3):727-736
11. Montaldo G, Tanter M, Bercoff J, Benech N, Fink M
(2009) Coherent Plane-Wave Compounding for Very High
Frame Rate Ultrasonography and Transient Elastography. Ieee
T
Ultrason
Ferr
56
(3):489-506.
doi:Doi
10.1109/Tuffc.2009.1067
12. Tsui PH, Chang CC (2007) Imaging local scatterer
concentrations by the Nakagami statistical model. Ultrasound
Med Biol 33 (4):608-619. doi:S0301-5629(06)01887-4 [pii]
10.1016/j.ultrasmedbio.2006.10.005
2014 IEEE International Ultrasonics Symposium Proceedings