Добавил:
Sekretar
kiopkiopkiop18@yandex.ru
t.me/Prokururor I Вовсе не секретарь, но почту проверяю
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз:
Предмет:
Файл:Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_3592_Библиотеки_им_академика_М_И_Перельмана
.pdf
Performance
https://t.me/medicina_free
evaluation
Validation
against GT
Cross
validation
Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
N/A N/A
ACC:
83.7%
SVM
PPV:
81.8%
Sn: 80%,
Sp: 86.4%
N/A N/A
ACC:
90.66%
Sn:
83.33%
Sp:
95.39%
N/A N/A
83.0%
(a) ACC:
GMM, DT,
(a) SVM,
Sn:
87.4%
KNN,
NBC,
Sp:
79.7%
RBPNN,
Fuzzy
89.5%
(b) ACC:
KNN,
(b) SVM,
Sn:
89.6%
Sp:
RBPNN
88.9%
Table B10. Comparison of various tissue characterization techniques from literature against our proposed work.
Wall plaque/wall
segmentation Data size Features Classifier
SN Author (year) Algorithm
346 Total of 3 features:
Manual
Risk assessment in
1 Acharya et al
DWT-based average (Dh1),
average (Dv1), energy (E)
segmentation
for plaque
carotid plaque
images
(2011)
region
(Atheromatic™)
160 Total of 32 features: texture SVM
segmentation
Manual
characterization
Carotid plaque tissue
(2012a)
2 Acharya et al
for plaque
region
and classification
(Atheromatic™)
Ent
Ene
18
18
LBP
LBP
(a) Total of 8 features:
plaque
(a) 346
(b) 342 wall
segmentation
for plaque
(a) Manual
using texture-
based features
Risk stratification
(2012b)
3 Acharya et al
Ent
Ene
216
216
LBP
LBP
region;
(b) Automated
(Atheromatic™)
Ene
Ene
5
324
LBP
LTE
segmentation
for wall using
Ene
6
LTE
AtheroEdge™
Ene
8
LTE
(b) Total of 4 features:
Ene
324
LBP
Ene
2
LTE
Ene
8
LTE
poly
IMTV
13-30

Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
± 0.223
and 6.67
± 0.186
N/A SACI*: 7.04
ACC:
91.7%
SVM
Sn: 97%
Sp: 80%
N/A N/A
85.3%
Sn:
84.4%
(a) ACC:
SVMDTFuzzy
Sp:
85.9%
(b) ACC:
93.1%
Sn:
99.0 %
Sp:
80%
N/A
against
Done
ACC:
77%
activity
Enhanced
GT
Sn: 70%
index based
Sp:
on Bayes
(Continued)
80.13%
factor
features (3), DWT (2) and
146 Total of 7 features: texture
segmentation
Manual
carotid plaque
Classification of
(2013a)
4 Acharya et al
HOS features (2)
for plaque
region
images by tissue
characterization
16 (UK dataset)
12 (Portugal dataset)
FGLCM: E, contrast,
entropy, correlation,
correlation,
homogeneity
FRLM: SRE,
GLNU, ASM, mean,
LRE, RLNU, RP
parameter (4th, 5th, 6th),
GLCM, wavelet, percentile
(10, 50), DoS, echogenic cap,
appearance, mean, skewness,
mixture components (5th, 6th,
No.), and plaque disruption
Total features:
346
146
segmentation
for plaque
risk assessment in
carotid plaque
(2013b)
region
images
(Atheromatic™)
Manual
(Atheromatic™)
Segmentation and
5 Acharya et al
146 Total of 16 features: Rayleigh
segmentation
Manual
characterization
Tissue
(2013)
6 Pedro et al
for plaque
region
for stroke risk
(AtheroRisk™)
13-31

Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
Performance
evaluation
Validation
against GT
N/A N/A
Cross
validation
ACC:
SVM,
91.8%
Sn: 83.3%
KNN,
RBPNN,
6 HOS features and
2 wall features (IMT and
Sp: 95%
DT
IMTVpoly)
95.86%
PoM:
against
Done
98%
SVM Far wall:
GLCM, GLRLM
1; Chaotic features: FDi
16 texture features:
GT
98%
Near wall:
with feature
wall:
99%
Combined
optimization
118 Total of 7 features:
Wall plaque/wall
segmentation Data size Features Classifier
Table B10. (Continued )
SN Author (year) Algorithm
wall
Automated far
characterization
Tissue
(2015)
7 Acharya et al
segmentation
for far wall
407
segmentation
for wall region
Automated
(Atheromatic™)
segmentation and
risk assessment in
Automated
method
8 Proposed
carotid far wall
and near wall
ACC = Accuracy; Sn: Sensitivity; Sp: Specificity.
13-32

Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
References
[1] Stroke statistics: Internet Stroke Center http://strokecenter.org/patients/about-stroke/stroke-
statistics/
[2] WHO CVD http://who.int/mediacentre/factsheets/fs317/en/
[3] Saba L, Gao H, Acharya U R, Sannia S, Ledda G and Suri J S 2012 Analysis of carotid
artery plaque and wall boundaries on CT images by using a semi-automatic method based on
level set model Neuroradiology
[4] Sanches J M, Laine A F and Suri J S 2012 Ultrasound Imaging: Advances and Applications
(London: Springer)
[5] Saba L, Tallapally N, Gao H, Molinari F, Anzidei M, Piga M, Sanfilippo R and Suri J S
2013 Semiautomated and automated algorithms for analysis of the carotid artery wall on
computed tomography and sonography a correlation study J. Ultrasound Med.
[6] Ross R 1995 Cell biology of atherosclerosis Annu. Rev. Physiol. 57 791–804
[7] Tracqui P, Broisat A, Toczek J, Mesnier N, O hayon J and Riou L 2011 Mapping elasticity
moduli of atherosclerotic plaque in situ via atomic force microscopy J. Struct. Biol.
115–23
[8] Teng Z, Zhang Y, Huang Y, Feng J, Yuan J, Lu Q, Sutcliffe M P F, Brown A J, Jing Z and
Gillard J H 2014 Material properties of components in human carotid atherosclerotic
plaques: a uniaxial extension study Acta Biomater.
[9] Amato M, Montorsi P, Ravani A, Oldani E, Galli S, Ravagnani P M and Baldassarre D
2007 Carotid intima–media thickness by B-mode ultrasound as surrogate of coronary
atherosclerosis: correlation with quantitative coronary angiography and coronary intra-
vascular ultrasound findings Eur. Heart J.
[10] Kao A H et al 2013 Relation of carotid intima–media thickness and plaque with incident
cardiovascular events in women with systemic lupus erythematosus Am. J. Cardiol.
1025–32
[11] Eigenbrodt M L, Bursac Z, Rose K M, Couper D J, Tracy R E, Ewans G W, Brancati F L
and Mehta J L 2006 Common carotid arterial interadventitial distance (diameter) as an
indicator of the damaging effects of age and atherosclerosis, a cross-sectional study of the
Atherosclerosis Risk in Community Cohort Limited Access Data (ARICLAD) Cardiovasc.
Ultrasound
[12] Eigenbrodt M L, Sukhija R, Rose K M, Tra cy R E, Couper D J, Ewans G W, Bursac Z
and Mehta J L 2007 Common carotid artery wall thickness and external diameter as
predictors of prevalent and i ncident cardiac events i n a large popu lation study Cardiovasc.
Ultrasound
[13] Ikeda N, Kogame N, Iijima R, Nakamura M and Sugi K 2013 Impact of carotid artery
ultrasound and ankle–brachial index on prediction of severity of SYNTAX score Circulation
77 712–6
[14] Araki T, Ikeda N, Dey N, Acharjee S, Molinari F, Saba L, Godia E C, Nicolaides A and
Suri J S 2015 Shape-based approach for coronary calcium lesion volume measurement on
intravascular ultrasound imaging and its association with carotid intima–media thickness
J. Ultrasound Med.
[15] Araki T et al 2016 PCA-based polling strategy in machine learning framework for coronary
artery disease risk assessment in intravascular ultrasound: a link between carotid and
coronary grayscale plaque morphology Comput. Methods Programs Biomed.
4 1–10
5 1–11
34 469–82
54 1207–14
32 665–74
174
10 5055–63
28 2094–101
112
128 137–58
13-33

Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
[16] Ogata T, Yasaka M, Yamagishi M, Seguchi O, Nagatsuka K and Minematsu K 2005
Atherosclerosis found on carotid ultrasonography is associated with atherosclerosis on
coronary intravascular ultrasonography J. Ultrasound Med.
[17] Ziembicka K A, Tracz W, Przewlocki T, Pieniazek P, Sokolowski A and Konieczynska M
2004 Association of increased carotid intima–media thickness with the extent of coronary
artery disease Heart 90 1280–90
[18] Elias-Smale S E et al 2012 Carotid intima–media thickness in cardiovascular risk strat-
ification of older people: the Rotterdam Study Eur. J. Prev. Cardiol.
[19] Polak J F, Pencina M J, Meisner A, Pencina K M, Brown L S, Wolf P A and D’Agostino R B
2010 Associations of Carotid Artery Intima–Media Thickness (IMT) With risk factors and
prevalent cardiovascular disease comparison of mean common carotid artery IMT with
maximum internal carotid artery IMT J. Ultrasound Med.
[20] Polak J F, Pencina M J, Herrington D and O’Leary D H 2011 Associations of edge-detected
and manual-traced common carotid intima–media thickness measurements with framingham
risk factors: the multi-ethnic study of atherosclerosis Stroke
[21] Cinthio M, Jansson T, Eriksson A, Ahlgren A R, Persson H W and Lindstrom K 2010
Evaluation of an algorithm for arterial lumen diameter measurements by means of ultra-
sound Med. Biol. Eng. Comput.
[22] Bots M L, Baldassarre D, Simon A, de Groot E, O’Leary D H, Riley W and Grobbee D E
2007 Carotid intima–media thickness and coronary atherosclerosis: weak or strong relations?
Eur. Heart J.
[23] Mirek A M and Wolińska-Welcz A 2012 Is the lumen diameter of peripheral arteries a good
marker of the extent of coronary atherosclerosis? Med. Biol. Eng. Comput. 71 810–7
[24] Delsanto S, Molinari F, Giustetto P, Liboni W, Badalamenti S and Suri J S 2007
Characterization of a completely user-independent algorithm for carotid artery segmentation
in 2-D ultrasound images IEEE Trans. Instrum. Meas.
[25] Molinari F, Zeng G and Suri J S 2010 Intima–media thickness: setting a standard for
completely automated method for ultrasound IEEE Trans. Ultrason. Ferroelectr. Freq.
Control
[26] Molinari F, Zeng G and Suri J S 2010 An integrated approach to computer-based automated
tracing and its validation for 200 common carotid arterial wall ultrasound images a new
technique J. Ultrasound Med.
[27] Molinari F, Zeng G and Suri J S 2010 A state of the art review on intima–media thickness
(IMT) measurement and wall segmentation techniques for carotid ultrasound Comput.
Methods Programs Biomed.
[28] Suri J S, Yuan C and Wilson D L (ed) 2005 Plaque Imaging: Pixel to Molecular Level
(Amsterdam: IOS)
[29] Gupta A et al 2015 Plaque echolucency and stroke risk in asymptomatic carotid stenosis: a
systematic review and meta-analysis Stroke
[30] Inzitari D, Eliasziw M, Gates P, Sharpe B L, Chan R K, Meldrum H E and Barnett H J 2000
The causes and risk of stroke in patients with asymptomatic internal-carotid-artery stenosis.
North American Symptomatic Carotid Endarterectomy Trial Collaborators N. Engl. J.
Med.
342 1693–700
[31] Acharya U R, Faust O, Alvin A P, Sree S V, Molinari F, Saba L, Nicolaides A and Suri J S
2012 Symptomatic vs. asymptomatic plaque classification in carotid ultrasound J. Med. Syst.
36 1861–71
28 398–406
57 1112–24
48 1133–40
29 399–418
100 201–21
46 91–7
24 469–74
19 698–705
29 1759–68
42 1912–6
56 1265–74
13-34

Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
[32] Araki T et al 2016 Reliable and accurate calcium volume measurement in coronary artery
using intravascular ultrasound videos J. Med. Syst.
[33] Pedro L M, Sanches J M, Seabra J, Suri J S, Fernandes E and Fernandes J 2014 Asymptomatic
carotid disease a new tool for assessing neurological risk Echocardiogrphy
[34] Acharya U R, Vinitha Sree S, Rama Krishnan M M, Molinari F, Saba L, Sin Yee S H,
Ahuja A T, Ho S C, Nicolaides A and Suri J S 2012 Atherosclerotic risk stratification
strategy for carotid arteries using texture-based features Ultrasound Med. Biol.
[35] Acharya U R, Faust O, Alvin A P, Vinitha Sree S, Molinari F, Saba L, Nicolaides A and
Suri J S 2012 Symptomatic vs. asymptomatic plaque classification in carotid ultrasound
J. Med. Syst.
[36] Acharya U R, Vinitha Sree S, Rama Krishnan M M, Saba L, Gao H, Mallarini G and Suri J S
2013 Computed tomography carotid wall plaque characterization using a combination of
discrete wavelet transform and texture features: a pilot study J. Eng. Med.
[37] Acharya U R, Rama Krishnan M M, Vinitha Sree S, Afonso D, Sanches J, Shafique S,
Nicolaides A, Pedro L M, Fernandes F J and Suri J S 2013 Atherosclerotic plaque tissue
characterization in 2D ultrasound longitudinal carotid scans for automated classifi cation: a
paradigm for stroke risk assessment Med. Biol. Eng. Comput.
[38] Acharya U R, Rama Krishnan M M, Vinitha Sree S, Sanches J, Shafique S, Nicolaides A, Pedro
L M and Suri J S 2013 Plaque tissue characterization and classification in ultrasound carotid
scans: a paradigm for vascular feature amalgamation IEEE Trans. Instrum. Meas.
[39] Sharma A M, Gupta A, Kumar P K, Rajan J, Saba L, Nobutaka I, Laird J R, Nicolades A
and Suri J S 2015 A review on carotid ultrasound atherosclerotic tissue characterization and
stroke risk stratification in machine learning framework Curr. Atheroscl. Rep.
[40] Saba L et al 2016 Carotid inter-adventitial diameter is more strongly related to plaque score
compared to lumen diameter: an automated and first ultrasound study in Japanese diabetic
cohort J. Clin. Ultrasound
[41] Araki T et al 2016 Two automated techniques for carotid lumen diameter measurement:
regional versus boundary approaches J. Med. Syst.
[42] Jain A K, Duin R P W and Mao J 2000 Statistical pattern recognition: a review IEEE Trans.
Pattern Anal. Mach. Intell.
[43] Acharya U R, Sree S V, Molinari F, Saba L, Nicolaides A and Suri J S 2015 An automated
technique for carotid far wall classification using grayscale features and wall thickness
variability J. Clin. Ultrasound
[44] Noor N M, Than J C, Rijal O M, Kassim R M, Yunus A, Zeki A A, Anzidei M, Saba L and
Suri J S 2015 Automatic lung segmentation using control feedback system: morphology and
texture paradigm J. Med. Syst.
[45] Molinari F, Constantinos P, Zeng G, Nicolaides A and Suri J S 2012 Completely automated
multi-resolution edge snapper (‘CAMES’)—a new technique for an accurate carotid ultra-
sound IMT measurement: clinical validation and benchmarking on a multi-institutional
database IEEE Trans. Image Process.
[46] Molinari F et al 2012 Ultrasound IMT measurement on a multi-ethnic and multi-institu-
tional database: our review and experience using four fully automated and one semi-
automated methods Comput. Methods Program. Biomed.
[47] Molinari F, Meiburger K M, Saba L, Zeng G, Acharya U R, Ledda M, Nicolaides A and
Suri J S 2012 Fully automated dual snake formulation for carotid intima–media thickness
measurement: a new approach J. Ultrasound Med.
36 1861–71
44 210–20
22 4–37
43 302–11
39 1–18
21 1211–22
40 1–20
31 353–61 .
38 899–915
227 643
51 513–23
62 392–400
17 955
40 182
108 946–60
31 1123–36
13-35

Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
[48] Molinari F, Meiburger K M, Saba L, Acharya U R, Ledda M, Nicolaides A and Suri J S
2012 Constrained snake vs. conventional snake for carotid ultrasound automated IMT
measurements on multi-center data sets Ultrasonics
[49] Araki T et al 2016 A new method for IVUS-based coronary artery disease risk stratification:
a link between coronary and carotid ultrasound plaque burdens Comput. Methods Programs
Biomed.
[50] Saba L, Lippo R S, Tallapally N, Molinari F, Montisci R, Mallarini G and Suri J S 2011
Evaluation of carotid wall thickness by using computed tomography and semiautomated
ultrasonographic software J. Vasc. Ultrasound
[51] Shrivastava V K, Londhe N D, Sonawane R S and Suri J S 2015 Reliable and accurate
psoriasis disease classification in dermatology images using comprehensive feature space in
machine learning paradigm Exp. Syst. Appl.
[52] Shrivastava V K, Londhe N D, Sonawane R S and Suri J S 2015 Exploring the color feature
power for psoriasis risk stratification and classification: a data mining paradigm Comput.
Biol. Med.
[53] Shrivastava V K, Londhe N D, Sonawane R S and Suri J S 2016 A novel approach to
multiclass psoriasis disease risk stratification: machine learning paradigm Biomed. Signal
Process. Control
[54] Soh L K and Tsatsoulis C 1999 Texture analysis of SAR sea ice imagery using gray level
co-occurrence matrices IEEE Trans. Geosci. Remote Sens.
[55] Kalyan K, Jakhia B, Lele R D, Joshi M and Chowdhary A 2014 Artificial neural network
application in the diagnosis of disease conditions with liver ultrasound images Adv.
Bioinform.
[56] Tang X 1998 Texture information in run-length matrices IEEE Trans. Image Process. 7
1602–9
[57] Mandelbrot B B 1983 The Fractal Geometry of Nature (New York: Freeman)
[58] Vapnik V 1998 Statistical Learning Theory (New York: Wiley)
[59] Muller K R, Mika S, Ratsch G, Tsuda K and Scholkopf B 2001 An introduction to kernel
based learning algorithms IEEE Trans. Neural Netw.
[60] Kohavi R 1995 A study of cross-validation and Bootstrap for accuracy estimation and model
selection Int. Joint Conf. Artif. Intell. 14 1137–43
[61] Molinari F, Meiburger K M, Zeng G, Nicolaides A and Suri J S 2012 CAUDLES-EF:
carotid automated ultrasound double line extraction system using edge flow J. Ultrasound
Imaging 24 129–62
[62] Saba L, Than J C, Noor N M, Rijal O M, Kassim R M, Yunus A, Ng C R and Suri J S 2016
Inter-observer variability analysis of automatic lung delineation in normal and disease
patients J. Med. Syst.
[63] Acharya U R, Faust O, Vinitha Sree S, Molinari F, Saba L, Nicolaides A and Suri J S 2012
An accurate and generalized approach to plaque characterization in 346 carotid ultrasound
scans IEEE Trans. Instrum. Meas.
[64] Acharya U R, Faust O, Alvin A P, Krishnamurthi G, Seabra J C, Sanches J and Suri J S
2013 Understanding symptomatology of atherosclerotic plaque by image-based tissue
characterization Comput. Methods Programs Biomed.
124 161–79
65 54–68
28 27–40
2014 708279
40 1–8
61 0018–9456
52 949–61
35 136–42
42 6148–95
37 780–95
12 181–201
110 66–75
13-36

IOP Publishing
https://t.me/medicina_free
Vascular and Intravascular Imaging Trends, Analysis, and
Challenges, Volume 1
Stent applications
Petia Radeva and Jasjit S Suri
Chapter 14
An improved framework for IVUS-based
coronary artery disease risk stratification by
fusing wall-based and texture-based features
during a machine learning paradigm
Sumit K Banchhor, Narendra D Londhe, Tadashi Araki, Luca Saba, Petia Radeva,
John R Laird and Jasjit S Suri
The planning of percutaneous interventional procedures involves pre-screening
andriskstratification of the coronary artery disease (CAD). Current screening
tools use stand-alone plaque texture-based features and therefore lack the ability to
stratify the risk. This institutional research board (IRB) approved study presents a
novel strategy for CAD risk stratification using an amalgamation of intravascular
ultrasound (IVUS) plaque texture-based and wall-based measurement features. As
it is a common genetic plaque make-up, the carotid plaque burden was chosen as a
gold standard for risk labels during the training phase of the machine learning
(ML) paradigm. A cross-validation protocol was adopted to compute the accuracy
of the ML framework. A set of 59 plaque texture-based features was padded with
six wall-based measurement features to show the improvement in stratificat ion
accuracy. The ML system was executed using a principle component analysis
(PCA)-based framework for dimensionality reduction and uses a support vector
machine (SVM) classifier for the training and testing phases. The ML system
produced a stratification accuracy of 91.28%, demonstrating an improvement of
5.69% when wall-based measurement features were combined with plaque texturebased features. The fused system showed an improvement in mean sensitivity,
specificity, positive predictive value and area under the curve (AUC) by 6.39%,
4.59%, 3.31% and 5.48%, respectively, wh en compared to the stand-alone system.
In addition to meeting the stability criterion of 5%, the ML system also showed a
high average feature retaining power and mean reliability of 89.32% and 98.24%,
doi:10.1088/2053-2563/ab01fach14 14-1 ª IOP Publishing Ltd 2019

Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
respectively. The ML system showed an improvement in risk stratification
accuracy when the wall-based measurement features were fused with the plaque
texture-based fea tures.
14.1 Introduction
Atherosclerotic cardiovascular disease accounts for the largest number of deaths in
the USA [1]. The disease of atherosclerosis over time causes calcium to build-up in
the coronary arteries [2]. During the advanced stage of the disease, the combined risk
of the patient includes higher plaque growth leading to plaque rupture. This also
includes the risk of developing different plaque components such as fibro-fatty,
macrophages, calcium and fatty tissue. When these components increase in size,
there is a risk of an increase in stenosis and stress on the fibrous cap thickness, which
can cause the risk of rupture leading to myocardial infarction (MI). All of the above
can be categorized as the ‘ risk of arterial wall rupture’ or ‘risk of MI’. Thus, rupture
of the arterial wall cap can cause calcium to dislodge, blocking the oxygen-rich
blood flow in the arteries, leading to myocardial infarction or stroke [3]. Current
screening methods such as computed tomography (CT) and magnetic resonance
imaging (MRI) [4, 5] suffer from excess radiation and magnetic interference,
respectively. Further, these devices take a long time to reconstruct the images,
thus they lack a real-time interface [4]. IVUS screening, on the other hand, has lowradiation exposure, is economic compared to MR/CT, is ergonomic and offers realtime diagnosis [6, 7].
Prior to stenting and percutaneous interventional procedures, cardiologists are
interested in performing pre-screening and risk stratification of CAD. Studies for
risk stratification of cardiovascular events are mainly categorized into two groups.
The first group attempts to predict the risk by quantifying the plaque characteristics
(i.e. texture-based features) while the second group predicts the risk by quantifying
wall-based measurement features [8–13]. Christodoulou et al [8], in 2003, proposed a
neural network for carotid plaque classification. Ten dominant texture feature sets
were selected from a total of 61 texture features showing a low accuracy of 73.10%
on a data size of 230 images. Two years later, Kyriacou et al [9] showed a carotid
classification system that used a neural network classifier with ten different textures
and carotid wall-based features and achieved even a slightly lower accuracy 71.2%
on a data size of 274 images. The same group in 2009 applied a SVM classifier on the
same database using only the texture features and showed an improvement in
accuracy of 2.5%. In a hybrid neural network approach on B-mode carotid
ultrasound images, Mongiakakou et al [10] in 2007 used 21 statistical and law
features on 108 images. The neural network was trained on the combined use of
genetic algorithms and back-propagation with momentum and adaptive learning
rate and showed an accuracy of 99.10%.
Our team, led by author JSS, has been working on the characterization of carotid
plaque. Acharya et al [12], in 2012, proposed an Atheromatic™ system for plaque
stratification into symptomatic and asymptomatic plaques showing an accuracy of
82.40% and 81.70%, using SVM and AdaBoost classifiers, respectively. The same
14-2

Vascular and Intravascular Imaging Trends, Analysis, and Challenges, Volume 1
https://t.me/medicina_free
group [13], in 2012, obtained an accuracy of 83% by fusing the plaque texture-based
and wall-based measurement features using an SVM classifier. A year later, the same
group [14] obtained a further high accuracy of 85.3% by fusing discrete wavelet
transform, higher order spectra and textural features on a large data size consisting
of 492 images. A year later, Pedro et al [15] fused the clinical and texture features for
the classification of carotid plaque. An enhanced activity index was proposed and
was correlated with the presence or absence of ipsilateral appropriate ischemic
symptoms. Leave-one-patient-out was applied to 146 carotid plaques obtained from
99 patients and a cross-validation accuracy of 77% was obtained. Araki et al [16]
showed a CADx system using the SVM that demonstrated a training and testingbased ML system using plaque texture features for coronary artery risk assessment.
Later, the same group [17] modified and improved their CADx system by introducing the PCA-based polling technique for selection of the grayscale dominant features
for improving the stratification accuracy. These prior studies had ignored how
plaque growth affects the walls of the arteries and lacked the prominent features
contributed by the wall-based parameters. This study is an extension of the above
studies using an amalgamation of IVUS plaque texture-based and wall-based
measurement features. This is motivated by the current strategy by JSS and his
team in stroke imaging where carotid intima–media thickness (cIMT) variability
was fused with carotid longitudinal grayscale features to improve the stroke risk
stratification [18, 19]. But in our current study, circular wall parameters along with
the plaque calcium are derived as measurement features from IVUS coronary walls.
Thus, the objective is to demonstrate the importance of wall-based measurement
features and their integration with plaque texture-based grayscale features for better
ML system design.
Calcium accumul ations always occur in the atheroma region which lies
betweenthelumen(innerwallorinternalelasticwall)regionandvessel(outer
wall or external elastic wall) region [20]. Therefore, an expansion of the walls is
purely a reflection of the growth of calcium in the arteries. Moreover, due to the
multi-focal nature of calcium [21], the wall thickness can vary along the circular
walls of the coronary artery. Figures 14.1 and 14.2 show typical examples of
images showing a g rayscale ring, along with calcium, lumen, vessel and athero ma
regional areas corresponding t o five different low-risk and high-risk patients,
respectively. Furthermore, our study is based on two hypotheses: (i) fusion of
plaque texture-based and wall-based measurement feat ures can offer an improvement in the coronary artery risk stratification; (ii) due to the genetic make-up of
the plaque, carotid plaque burden, which is considered as a biomarker for stroke
risk [22–28], can be used as a risk label for patients with coronary artery disease
[16, 17].
The novelty of this study is to demonstrate an improvement in the accuracy of the
CADx system built for the coronary artery risk assessment by fusing plaque texturebased features with wall-based measurement features compared to a stand-alone
system consisting of only plaque texture-based features. Our objective in this paper is
to predict the class label of the plaque type as high-risk or low-risk.
14-3
Соседние файлы в папке Библиотека им академика М.И. Перельмана
