Digital Image Processing Assessment in Multi Slice CT Angiogram using Liner, Non-Liner and Both Classifiers

Sarvadeva BhatlaMurali Krishna. Int. Journal of Engineering Research and Application www.ijera.com
ISSN: 2248-9622, Vol. 7, Issue 2, (Part -2) February 2017, pp.29-35
www.ijera.com DOI: 10.9790/9622- 0702022935 29 | P a g e
Digital Image Processing Assessment in Multi Slice CT
Angiogram using Liner, Non-Liner and Both Classifiers
Sarvadeva BhatlaMurali Krishna*, Dr. M Chandra Shekar**
*(PhD Scholar in ECE, at Rayalaseema University with Regd No. PP-ECE-0044, AP, India)
**(Senior Manager Bharath Dynamics Limited, Hyderabad, Telangana, India)
ABSTRACT
Nowadays with heart diseases most of peoples are dying lock process to find problem agile fashion in remote
areas. In this research help to peoples who are staying remote also to find problem in heart on which location
and how much problem to near and finally gives best analysis methodology for automated analysis for Multi-
Slice CT Angiogram images. Multi slice CT scanner is used to identify heart disease. Multi-detector CT is
considered convenient and reliable non-invasive imaging modality for assessment of human angiogram 3D
images. Automatic hart segmentation from Computed Tomography (CT) is highly demanded. Accurate hart
segmentation is a crucial for computer-aided heart disease diagnosis and treatment planning. After segmentation
and future extraction then identify whether the patient angiogram waveform has disease or not. For that, Support
Vector Machine method is used to confirm the presence of disease. Neural Networks can solve different types of
nonlinear problems in image classification and retrieval process. After that majorly focus on research learning
methodologies for improve performance of the Multi Slice CT Angiogram images.
So, in this research is summarizing the problem with liner of RBF neural network, Non-Liner SVM and RBF
NN with liner and non-liner. Final, RBF NN with liner and non-liner provided more value and proved as best
compare with other existing methodologies. This methodology consumes less time for both learning as well as
testing comparatively than any other methods like back propagation. This issue drastically improves the
estimation efficiency and accuracy for real time 128, 256 slices CT scan angiogram images.
I. INTRODUCTION
Nowadays medical systems are providing lot
of images for diagnosis problem in right way and
treatment process as well. It also links 'classical'
diagnosis to modern forms of treatment such as
image guided surgery. Major number of human
beings is suffering with heart artery disease. Most of
medications processes are not able to identify
blocked areas of a blood vessel in heart soon [1] [2].
Earlier and nowadays most of the heart Patients
undergoing a coronary CTA scan receive an iodine-
containing contrast material (dye) as an intravenous
injection to ensure the best possible images and allow
reproducible and accurate non-invasive evaluation of
coronary artery diseases. CT scans of internal organs,
bones, soft tissue and blood vessels provide greater
clarity and reveal more details than regular x-ray
exam. Basic CTA scan 3D image is shown below
Fig.1 [3].
Figure 1: Basic heart CTA scan 3D image
Preoperative coronary computed tomography (CT)
angiography is shown below figure 2. Figure 2 shows
the four images. Image A represents the location of
the leak in the left ventricle is the sub-mitral valve
prosthesis area. Image B represents the three large
lobulated aneurysms are seen bulging (an excessive
localized swelling of the wall of an artery) from the
base of the left ventricle with calcified walls. Images
C and D represents the Three-dimensional
reconstruction of preoperative coronary CT
angiography [4] [5]. PsA - Pseudo Aneurysm; LV -
RESEARCH ARTICLE OPEN ACCESS

Left Ventricle; AV- Aortic Valve; MV - Mitral
Valve; Dao - Descending Aorta. Based on this, we
diagnosed the patient with a pseudo aneurysm of the
left ventricle, and surgical treatment was planned [6]
[7].
Figure 2: Basic multi slice CT scan angiogram
image
Figure 3: Basic segmental reconstruction of CT scan
3D image
In process of medical analysis CT images
used normal first step as segmentation. Segmentation
is going on agile fashion the research for medical
image analysis. Now, CT scanner most advanced and
using 128, 256 slices CT scanners are providing the
most valuable information for processing. Several
general-purpose algorithms and techniques have been
developed for image segmentation. Multi slice CT
angiogram segmentation is performed with the help
of k-means++ algorithm. Segmental reconstructions
of angiogram scan shows high-resolution images of
(clockwise from top) the right coronary artery, left
anterior descending artery, and left circumflex artery
acquired on a multi slice CT scanner[8] [9].
Features are an important measurement for
image understanding. A complete feature
representation of the segmented region is an
important issue for object classification and analysis.
After the segmentation and feature extraction next
step is to decide whether the patient angiogram
waveform has disease or not. For that, Support
Vector Machine method is used to confirm the
presence of disease. Here SVM is non-liner method
when use any type of application will miss most liner
ways of blood vessels and lack of speed in process
[10]. Basic SVM method is most supported for binary
classification. Recently SVM implemented for multi
class classification problems. As per details from the
WIKI, SVMs are also useful in many applications to
classify with up to 90% of the compounds classified
correctly. Many recent methods are used to extended
SVM from binary classification to multi class
classification problem, in that main popular methods
are four: One Against One (OAO), One Against All
(OAA), Fuzzy Decision Function (FDF) and
Decision Directed Acyclic Graph (DDAG).So, way
of solution is with which method can solve the
problem of linear and nonlinear classifiers. The best
algorithm of liner and non-liner classifier is single
layer Neural Network, i.e., Radial Based Function
Neural Network. Final analysis of the research gives
more improve of the accuracy in terms of the time
and classifications images [11].
II. BACK GROUND
Now way days most of the humans suffer
the hart problem. Multi slice CT scanner is used to
identify heart disease. Multi-detector CT is
convenient and reliable non-invasive imaging
modality for assessment of human angiogram 3D
images. Accurate heart segmentation is a decisive
requirement for computer-aided hepatic disease
diagnosis and treatment planning. If the segmentation
can be performed rapidly, the results can also be used
in intraoperative guidance. Manual annotation is error
prone and time-consuming. Automatic hart
segmentation from Computed Tomography (CT)
volumes is therefore highly demanded. Computed
tomography (CT) datasets facilitates evaluation of 3D
CT angiogram images [12]. Multi-detector CT is
considered convenient and reliable non-invasive
imaging modality for assessment of human
angiogram 3D images. This paper provides
information on expanding your imaging services with
multislice CT. First Multislice CT invention by
Kalender in the 1980s; Kalender argues that the terms
spiral and helical are synonymous and equally
acceptable. Spiral computed tomography is
a computed tomography (CT) technology in which
the source and detector travel along a helical path
relative to the object. Typical implementations
involve moving the patient couch through the bore of
the scanner whilst the gantry rotates. Spiral CT can
achieve improved image resolution for a given
radiation dose, compared to individual slice
acquisition [13].
A scanner with more slices allows faster
acquisition and makes it easier to examine
uncooperative children and frailer older patients. If
no of slices or increased the speed of the scan also
increased. And it gives the perfect visual of the heart,
since 20th century first introduce the 16 slice CT

scanner, after that 64 slice CT scanner .The multi
slice CT angiogram with 16, 32, 64 are facing
problem in the issue of quantitative evaluation of
segmentation of the san result. The research goes
with segmentation which is one of the difficult tasks
in image processing and determines the outcome of
analysis and evaluation of pathological regions. The
main objectives of this research are to improve the
accuracy of the multi slice CT scan with 128,256
slices for advanced image processing [14].
SVM is a Soft Computing system developed
by Dr. Vapnik (1995) in Bell Lab. Its basic concept is
to construct the best super plane in sample space so
that the margin between super plane and the sample
set of different types will be a maximum.
III. PROPOSED METHODOLOGY
This research proposed methodology
focuses on identification of different existing and
new ways of extracting features of Multi Slice CT
Images. The main aim is to provide a method for
recognizing and classifying CT Angiogram images
using liner, non-liner and both classifier. Proposed
methodology is shown Fig. 4.
The main objectives of this work is
 To study and review of Multi Slice CT
Angiogram Images
 Segmentation of Angiogram images with high
rated level.
 Feature extraction with Colour, Texture and
relational of analysis.
 Analysing the Machine Learning Algorithms
with current research trends and developed the
following algorithms to find the disease from
Multi Slice CT Angiogram Images:
 First Step is to analysis use existing method
Support Vector Machine (SVM)
 In second step, used Liner and Non-Liner RBF
Neural Network for analysis problem.
 To Test on different data sets of multi-slice CT
Angiogram Images and summarize the analysis
of accuracy for the above proposed algorithms.
In this paper, exhibit best process for Multi
Slice CT Angiogram image classification using liner,
non-liner and both methodologies.
Major process divided into four steps:
 First step is the Data Processing is collecting
data from different image datasets. This digital
data is pre-processed before data selection.
 After data processing the entire data is grouped
into different datasets. Each data set consists of
Train Data sets and Test Data sets. Train data set
is used to build the system and Test data set is
used to measure the system.
Figure 4: Block Architecture Proposed Methodology
 There are different methodologies proposed in
this research. In the medical imaging analysis,
features and measurements can also be used for
region segmentation to extract meaningful
structures, subsequently, interpret the result
using classification methods.
 Following methods are used in current research
areas: Support vector machines and neural
networks.
 The final one is accuracy assessment, after
finalizing the method build the system first and
then test with different queries for evaluating the
performance in terms of accuracy.
3.1 Test Data
In this Research, data is collected from one
source, which is osirix-viewer. Osirix-viewer
organization is help to do research for medical image
open data. Data sets are provided in form of DICOM
format and JPEG 2000. Our research source data is
converted to binary and worked as with jpeg images.
Collected three types of data sets are from
data source. About each data set, here explained
below:
Agecanonix Dataset:122 records from 13 subjects
this dataset contains, each voxel size is 0.47 x 0.47 x
0.50(mm). Each signal is digitized at 1106 samples
per second, with 16 bit resolution over a range of ±
16.384 mV. Each subject is specified in below Table
1 with number of records.

Table 1: Tested data from Agecanonix and #records
FIVIX Dataset:FIVIX dataset 1754 records from 3
subjects. Below Table 2 subjects are specified with
number of records.
Subject Name No of Records
CorCTA 0.6 B20f 65% dias– 11 414
CorCTA 3.0 B20f 0-90% 10ph – 13 530
pet cardiac 10 gates Volume (Corrected) – 404 810
Table 2: Tested data from FIVIX and #records
All subjects are 1754 simultaneously measure Multi
slice CT angiogram heart images. Each voxel size is
0.35 x 0.40 x 0.40(mm).
MAGIX dataset:MAGIX contains 760
simultaneously measure Multi slice CT angiogram
heart images. Each voxel size is 0.40 x 0.40 x
2.0(mm). All from 10 different subjects specified in
below Table 3.
Table 3: Tested data from MAGIX and #records
Here total 3 types of data set with 1106, 1754
and 760 Multi Slice CT images are used in this
research. Next step of research is extract feature set
from collection data set.
Technique Name Feature Name
Statistical Pixel-Level Mean
Statistical Pixel-Level Variance
Statistical Pixel-Level Gray Value
Statistical Pixel-Level Contrast of Pixel
Statistical Pixel-Level Edge Gradient
Shape Circularity
Shape Compactness
Shape Moments
Shape Chain Codes
Relational Relational Structure
Relational Hierarchical Structure
Table 4: Extract Features Descriptions
3.2 Non-Liner: Support Vector Machine– RBF
The original SVM algorithm was invented
by Vladimir N. Vapnik in 1995. Support vector
machines are supervised learning models with
associated learning algorithms . A support vector
machine constructs a set of hyper planes in
an infinite-dimensional space, which can be used for
classification, regression, or other tasks. In the case
of support vector machines, a data point is viewed as
a display style dimensional vector, and we want to
know whether we can separate such points with
a {(p-1)}dimensional hyper plane. This is called
a linear classifier [15] [16].
Basic SVM hyper plane image is shown below
Figure 5. Figure 5 shows the three hyper planes.
Subject Name No of Records
CorCTA w-c 1.0 B20f – 6 355
CorCTA w-c 3.0 B20f 0% - 9 51
CorCTA w-c 3.0 B20f 10% - 10 51
CorCTA w-c 3.0 B20f 20% - 11 51
CorCTA w-c 3.0 B20f 30% - 12 51
CorCTA w-c 3.0 B20f 40% - 13 51
CorCTA w-c 3.0 B20f 50% - 14 51
CorCTA w-c 3.0 B20f 60% - 15 51
CorCTA w-c 3.0 B20f 70% - 16 51
CorCTA w-c 3.0 B20f 80% - 17 51
CorCTA w-c 3.0 B20f DIASTOLE – 8 119
CorCTA w-c 3.0 B20f SYSTOLE – 7 119
CorCTA w-o 3.0 B20f – 4 58
Subject Name
# of
Records
Cir CardiacCirc 3.0 B20f 0-90%
RETARD_DECLECHEMENT 0 % - 10 76

1. H1 hyper plane does not support the classes. 2.
Hyper plane H2 does, but only small margin. 3.
Hyper plane H3 separates them with the maximum
margin.
Figure 5: SVM Hyper Plane
In a finite dimensional space set of hyper planes are
not linearly separable, so it is called as Non-linear
classifier. Non-linear classifier working is based on
kernel method.
Figure 6: Traditional Radial Basis Function Network
outline
Mainly in non-liner SVM using different kernel
methods to train data sets. Following kernel methods
using major popular:
 Homogeneous -Polynomial
 Inhomogeneous - Polynomial
 RBF - Gaussian Radial Basis Function
 HT - Hyperbolic Tangent
A common disadvantage of non-parametric
techniques such as SVMs is the lack of transparency
of results. SVMs cannot represent the score of all
companies as a simple parametric function of the
financial ratios, since its dimension may be very
high. It is neither a linear combination of single
financial ratios nor has it another simple functional
form. The weights of the financial ratios are not
constant. Thus the marginal contribution of each
financial ratio to the score is variable. Using a
Gaussian kernel each company has its own weights
according to the difference between the value of their
own financial ratios and those of the support vectors
of the training data sample [17].
3.3 Liner and Non-Liner: RBF Neural Network
Figure 2.4, shows the block diagram of RBF
Neural network. x1, x2,…. Xd are input vector and
wi are the weights among hidden and output layer.
The output function f compute by using the below
approximation function like below equation (1).
y x = wi φ x − xi
n
i=1
-- (1)
Figure 7: Block diagram of RBF neural network
Different optimization methods are used for
find weight values in Radial Basis Function Neural
Network. Most popular method is Gradient Descent.
GD is an optimization algorithm used for calculating
a local minimum of a function. The algorithm takes
steps proportional to the negative of the gradient of
the function at the current point. RBF output written
as below equation (2):
y=w.h -- (2)
Where w is defined as weight matrix, h is
hidden layer output values. Here supervised leaning
we already know output values y based on below
function way, used to find the w value with below
equation (3):
w=h.y^(-1) (3)
Here in this paper of research, used weight value
used matrix inversion showed above equation (3).
Above two methods are discussed following next
section, and its proved RBF NN accuracy is more
compare than SVM –RBF [18].
IV. ACCURACY ASSESSMENTS
In this paper of research is used one source
with three different data sets. About data sets are
specified in section 3.1. Total data sets are three
types, done research with final output is find the
place of blood vessels in images. Each data set,
defined some set of data has training data set and left
our part is test data set. Instance of data set is already
defined how much blood vessels part of data sets for
train data. Has first training data set is defined 90%
data set from original data and 10% consider has test
data set. Like that, decrease training data set % and
increase test data set % with different 7 cases are

divided as data sets and showed detailed each case in
below Table 5.
Table 5: Training and Test Data Sets % with
Different Test Cases
Used with above training and test data sets
for each analysis division of % shown in above Table
5. Assessment of accuracy done for following
methods with above train and test data set cases:
 Support Vector Machines With RBF
 Neural Network with RBF
Cases
Accuracy of
SVM-RBF Accuracy of RBF NN
Case #1 94% 96%
Case #2 91% 95%
Case #3 85% 95%
Case #4 87% 95%
Case #5 82% 90%
Case #6 75% 86%
Case #7 70% 85%
Table 6: Accuracy Assessment for proposed
Methodologies
Accuracy is exhibited above Table 6 for
research methodologies. Clear proved RBF – NN is
more accurate compare than SVM-RBF and another
thing is here clear specified even decrease train data
set also accuracy is stabilized in all aspect not much
decrease as like SVM-RBF [18].
Additional analyzed SVM RBF stabilized
different with Sigma and Cost values. Final best
Sigma and Cost value: 1.16 and 0.05. Additional
analyzed RBF NN stabilized different with Weights
and Bias values. Total weight values are 132 values
and 2 bias values.
V. CONCLUSIONS
This paper, proposed a method to find the
Heart diseases. In this research using Computed
Tomography (CT) scanner for identified blood vessel
in heart area. The problem of identification blood
vessels in heart is like most binary classification.
Majorly binary classification algorithms are classified
in two forms, linear and nonlinear. Max of the times
non-liner classification algorithms are best comparing
with linear algorithms for large data sets like Images,
Application Big Data. And most popular in binary
and non-liner classification algorithm is Support
Vector Machine (SVM), and it is most successfully
industry algorithm. Here SVM is non-liner method
when use any type of application will miss most liner
ways of blood vessels and lack of speed in process.
So, way of solution is with which method can solve
the problem of linear and nonlinear classifiers. The
best algorithm of liner and non-liner classifier is
single layer Neural Network, i.e., Radial Based
Function Neural Network.
Future extended with advanced
methodology for image classification, feature
extractions and best accuracy assessment.
REFERENCES
[1]. Lee H, Lee KS, Sim SB, Jeong HS, Ahn
HM, Chee HK, Korean J Thorac
Cardiovascular Surg. "Left ventricular
pseudo aneurysm after valve replacement”
Published in 2015 Feb;
[2]. Wang M, Qi HT, Wang XM, Eur J Radiol.
“Dual source CT versus single source CT in
cardiac CT angiography’’ published in
2009; 72:396-400.
[3]. Baumüller S,Leschka S, Desbiolles
L.“Angiography at lower heart rates
comparison of accuracy and radiation”.
Radiology 2009; 253:56-64.
[4]. Guo SL, Guo YM, Zhai YN. “Diagnostic
accuracy of first generation dual-source
computed tomography in the assessment of
Image Segmentation and Feature Extraction
ofcoronary artery disease”.Int J
Cardiovascular Imaging 2011; 27:755-71.
[5]. Lei Z, Han P, Xu H, Yu J. “Application
study of 512-slice computed tomography low
dose coronary angiography”. Digit Med
2015; 1:28-33.
[6]. A. P. Dhawan. “Coronary artery diseaseCT
Angiography”. World Scientific Publishing
Co. Pte. Ltd, 2008, pp 197-228.
[7]. Andrej Krenker, J Bester, and Andrej Kos.
“Introduction to the artificial neural
networks”. ISBN, pages 978–953, 2011.
[8]. J. Kennedy, R.C. Eberhart, Y. Shi.“Swarm
Intelligence”.Morgan Kaufmann Publishers
in 2001.
[9]. JinGu, He-shui Shi, Ping Han, Jie Yu, Gui-
na Ma & Sheng Wu. “Image Quality and
Radiation Dose for Prospectively Triggered
Coronary CT Angiography”. Scientific
Seq. No
Case
Training Data Set
% from Original
Data Set
Test Data Set %
from Original
Data Set
Case #1 90 10
Case #2 80 20
Case #3 70 30
Case #4 60 40
Case #5 50 50
Case #6 40 60
Case #7 30 70

reports 6, article number: 34795(18 October
2016).
[10]. Dr. Kazuhiro Katada, chairman of the
department of radiology at Fujita Health
University School of Medicine in Japan.
“256-slice CT at Stanford University's
International Symposium on Multi detector
Row CT technology”.Autominnie.com
during the sanfranciscomeeting. August 21
2006.
[11]. NarendraKohli, Nishchal, K.Verma
“Arrhythmia classification using SVM with
selected features”.International Journal of
Engineering, Science and TechnologyVol. 3,
No. 8, 2011, PP 122-131.
[12]. Nico R Mollet, Filippo Cademartiri, Pim J
de Feyter Heart.“Non-invasive multislice CT
coronary imaging”. 2005 Mar; 91(3): 401–
407. Doi: 10.1136/hrt.2004.034009
PMCID: PMC1768758
[13]. Knez A, Becker CR, Leber A, et
al. Usefulness of multislice spiral computed
tomography angiography for determination
of coronary artery stenoses. Am J
Cardiol 2001; 88:1191–4. [PubMed]
[14]. Vogl TJ, Aboimaali ND, T Diebold.“Multi-
detector row CT coronary
angiography”. Radiology 2002; 223:212–2,
PubMed.
[15]. A. P. Dhawan and S. Dai.“Clustering and
Pattern Classification”, in Principles and
Advanced Methods in Medical Imaging and
Image Analysis. Singapore”. World
Scientific Publishing Co. Pte. Ltd, 2008, pp
229-265.
[16]. Flohr TG, Schoepf UJ, Kuettner
A. Advances in cardiac imaging with 16-
section CT systems. AcadRadiol 2003;
10:386–401, PubMed.
[17]. Budoff, M.J., Achenbach, S., Blumenthal,
R.S., Carr, J.J., Goldin, J.G., Greenland, P.,
Guerci, A.D., Lima, J.A., Rader, D.J.,
Rubin, G.D., Shaw, L.J., Wiegers, S.E.
“Assessment of coronary artery disease by
cardiac computed tomography”.
Circulation 114, 1761–1791
(2006)CrossRef.
[18]. Syeda-Mahmood, T., Wang, F., Beymer, D.,
Amir, A., Richmond, M., Hashmi, S.
Aalim.“Multimodal mining for cardiac
decision support”. Comput. Cardiol. 34,
209–212 (2007).

Digital Image Processing Assessment in Multi Slice CT Angiogram using Liner, Non-Liner and Both Classifiers

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Digital Image Processing Assessment in Multi Slice CT Angiogram using Liner, Non-Liner and Both Classifiers (20)

Recently uploaded (20)

Digital Image Processing Assessment in Multi Slice CT Angiogram using Liner, Non-Liner and Both Classifiers