Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography Images Using Wavelet Based Statistical Texture Features

International Journal of Computer Science, Engineering and Information Technology (IJCSEIT), Vol.1, No.3, August 2011
DOI : 10.5121/ijcseit.2011.1303 22
Automatic Diagnosis of Abnormal Tumor Region
from Brain Computed Tomography Images Using
Wavelet Based Statistical Texture Features
A. Padma1
and Dr.R. Sukanesh2
1
Dept of Information Technology, Velammal College of Engg and Tech, Madurai,India
giri_padma2000@yahoo.com
2
Prof of Electronics and Communication Eng, Thiagarajar College of
Engg,Madurai,India
rshece@tce.edu
Abstract
The research work presented in this paper is to achieve the tissue classification and automatically
diagnosis the abnormal tumor region present in Computed Tomography (CT) images using the wavelet
based statistical texture analysis method. Comparative studies of texture analysis method are performed
for the proposed wavelet based texture analysis method and Spatial Gray Level Dependence Method
(SGLDM). Our proposed system consists of four phases i) Discrete Wavelet Decomposition (ii)
Feature extraction (iii) Feature selection (iv) Analysis of extracted texture features by classifier. A
wavelet based statistical texture feature set is derived from normal and tumor regions. Genetic Algorithm
(GA) is used to select the optimal texture features from the set of extracted texture features. We construct
the Support Vector Machine (SVM) based classifier and evaluate the performance of classifier by
comparing the classification results of the SVM based classifier with the Back Propagation Neural network
classifier(BPN). The results of Support Vector Machine (SVM), BPN classifiers for the texture analysis
methods are evaluated using Receiver Operating Characteristic (ROC) analysis. Experimental results
show that the classification accuracy of SVM is 96% for 10 fold cross validation method. The system
has been tested with a number of real Computed Tomography brain images and has achieved satisfactory
results.
Keywords
Discrete Wavelet Transform(DWT), Genetic Algorithm(GA), Receiver Operating Characteristic (ROC)
analysis , Spatial Gray Level Dependence Method (SGLDM), Support Vector Machine(SVM).
1.Introduction
In recent years, medical CT Images have been applied in clinical diagnosis widely. That can
assist physicians to detect and locate Pathological changes with more accuracy. Computed
Tomography images can be distinguished for different tissues according to their different gray
levels. The images, if processed appropriately can offer a wealth of information which is
significant to assist doctors in medical diagnosis. A lot of research efforts have been directed
towards the field of medical image analysis with the aim to assist in diagnosis and clinical studies
[1]. Pathologies are clearly identified using automated CAD system [2]. It also helps the
radiologist in analyzing the digital images to bring out the possible outcomes of the diseases. The
medical images are obtained from different imaging systems such as MRI scan, CT scan, Ultra

23
sound B scans. The computerized tomography has been found to be the most reliable method for
early detection of tumors because this modality is the mostly used in radio therapy planning for
two main reasons. The first reason is that scanner images contain anatomical information which
offers the possibility to plan the direction and the entry points of radio therapy rays which have to
target only the tumor region and to avoid other organs. The second reason is that CT scan images
are obtained using rays, which is same principle as radio therapy. This is very important because
the intensity of radio therapy rays have been computed from the scanned image. Advantages of
using CT include good detection of calcification, hemorrhage and bony detail plus lower cost,
short imaging times and widespread availability. The situations include patient who are too large
for MRI scanner, claustrophobic patients, patients with metallic or electrical implant and patients
unable to remain motionless for the duration of the examination due to age, pain or medical
condition. For these reasons, this study aims to explore methods for classifying and segmenting
brain CT images. Image segmentation is the process of partitioning a digital image into set of
pixels. Accurate, fast and reproducible image segmentation techniques are required in various
applications. The results of the segmentation are significant for classification and analysis
purposes. The limitations for CT scanning of head images are due to partial volume effects which
affect the edges produce low brain tissue contrast and yield different objects within the same
range of intensity. All these limitations have made the segmentation more difficult. Therefore, the
challenges for automatic segmentation of the CT brain images have many different approaches.
The segmentation techniques proposed by Nathali Richard et al and Zhang et al [3][4]
include statistical pattern recognition techniques. Kaiping et al [5] introduced the effective
Particle Swarm optimization algorithm to segment the brain images into Cerobro spinal fluid
(CSF) and suspicious abnormal regions but without the annotation of the abnormal regions.
Dubravko et al and Matesin et al [6] [7] proposed the rule based approach to label the abnormal
regions such as calcification, hemorrhage and stroke lesion. Ruthmann.et al [8] proposed to
segment Cerobro spinal fluid from computed tomography images using local thresholding
technique based on maximum entropy principle. Luncaric et al proposed [9] to segment CT
images into background, skull, brain, ICH, calcifications by using a combination of K means
clustering and neural networks. Tong et al proposed [10] to segment CT images into CSF,brain
matter and detection of abnormal regions using unsupervised clustering of two stages. Clark et al
[11] proposed to segment the brain tumor automatically using knowledge based techniques.
From the above literature survey shows that intensity based statistical features are the straightest
forward and have been widely used, but due to the complexity of the pathology in human brain
and the high quality required by clinical diagnosis, only intensity features cannot achieve
acceptable result. In such applications, segmentation based on textural feature methods gives
more reliable results. Therefore texture based analysis has been presented for tumor
segmentation such as SGLDM method and wavelet based texture features are used and achieve
promising results.
Based on the above literature, better classification accuracy can be achieved using wavelet based
statistical texture features. In this paper, the authors would like to propose a wavelet based
statistical texture analysis method to segment the soft tissues and automatically diagnosis
abnormal tumor region from brain CT images. The proposed method is illustrated in Figure 1.
This system uses the classifiers SVM [12], BPN [13] to classify and segment the abnormal
tumor region from brain CT images and gives relatively good segmentation results as
compared to the literature discussed above.
In our work, first by applying 2 level Discrete Wavelet Transform(DWT),the image is
represented by one approximation and three detail sub bands and the co-occurrence
matrix[14,15] is derived for detail sub bands. Then from these co-occurrence matrices, the
statistical texture features are extracted using the SGLDM method.. The extracted texture
features are optimized by Genetic Algorithm(GA)[16] for improving the classification accuracy
and reducing the overall complexity. The optimal texture features are fed to the SVM,BPN
classifiers to classify and segment the abnormal tumor region from brain CT images.

24
2. Materials and methods
Most classification techniques offer intensity based statistical features. However in our approach,
we adopt wavelet based statistical texture features to classify and segment the abnormal tumor
region. The proposed system is divided into 4 phases (a) Discrete Wavelet Decomposition (b)
Feature extraction (c) Feature selection (d) Classification and Evaluation. In the proposed
system for feature extraction, we discovered two methods which are wavelet based statistical
texture feature extraction method, SGLDM method without wavelet transform. Firstly the two
level wavelet decomposition is performed to decompose the image into one approximation and
three detail images and the co-occurrence matrix is derived for 2nd
level detail images. Then
from these co-occurrence matrices, the statistical texture features are extracted using the
SGLDM method. Once all the features are extracted, then for feature selection, we use Genetic
Algorithm(GA) to select the optimal texture features. The selected optimal texture features are
given as input to the SVM ,BPN classifiers to classify and segment the abnormal tumor region
from brain CT images.
2.1 Discrete Wavelet Decomposition
A two level wavelet decomposition of region of interest(ROI) is performed which results in four
sub bands. Daubechies wavelet filter of order two is used. In 2D wavelet decomposition [17]
the image is represented by one approximation and three detail images representing the low and
high frequency contents image respectively. The approximation can be further to produce one
approximation and three detail images at the next level of decomposition, wavelet decomposition
process is shown in Figure 1. A1 and A2 represent the wavelet approximations at 1st
and 2nd
level respectively, and are low frequency part of the images. H1,V1,D1,H2,V2,D2 represent the
details of horizontal, vertical and diagonal directions at 1st
and 2nd
level respectively, and are
high frequency part of the images.
Figure 1. Two level discrete wavelet decomposition.
Among the high frequency sub bands, the one whose histogram presents the maximum variance
is the sub band that represents the clearest appearance of the changes between the different
textures. The textures features are extracted from these high frequency sub bands are useful to
classify and segment the abnormal tumor region from brain CT images.

25
2.2 Feature extraction
Texture analysis is a quantitative method that can be used to quantify and detect structural
abnormalities in different tissues .As the tissues present in brain are difficult to classify using
shape or intensity level of information, the texture feature extraction is founded to be very
important for further classification. The purpose of feature extraction is to reduce original data set
by measuring certain features that distinguish one region of interest from another. The analysis
and characterization of textures present in the medical images can be done by using wavelet
based statistical feature extraction method. Each sub image is taken from top left corner of the
original image is decomposed using two level DWT and co-occurrence matrices are derived for
detail or high frequency sub bands(i.e.,H2,V2,D2 sub bands). Then from these co-occurrence
matrices ,the Wavelet Co-occurrence Texture features (WCT)are computed.
Algorithm for feature extraction is as follows
• Obtain the sub-image blocks, starting from the top left corner.
• Decompose sub-image blocks using 2-D DWT.
• Derive SGLDM or Co-occurrence matrices for detail sub-bands of DWT with 1 for
distance and 0,45,90 and 135 degrees for θ and averaged.
• From these co-occurrence matrices, the following nine Haralick texture features
[ 18] called wavelet Co-occurrence Texture features(WCT) are extracted.
Then the feature values are normalized by subtracting minimum value and dividing by maximum
value minus minimum value. Maximum and minimum values are calculated based on the training
data set. In the data set, if the feature value is less than the minimum value, it is set to minimum
value. If the feature value is greater than the maximum value, it is set to maximum value.
Normalized feature values are then optimized by feature selection algorithm. Table 1 Shows the
WCT features extracted using SGLDM method.
Table 1. WCT Features extracted using SGLDM method
SI.No Second order WCT features
1 Entropy-ENT (Measure the disorder of an image)
2 Energy- ENE ( Measure the textural uniformity )
3 Contrast-CON (Measure the local contrast in an image)
4 Sum Average-SA (Measure the average of the gray level within an image)
5 Variance –VAR (Measure the heterogeneity of an image)
6 Correlation-COR (Measure a correlation of pixel pairs on gray levels)
7 Max probability-MP (Determine the most prominent pixel pair in an image)
8 Inverse Difference Moment - IDM (Measure the homogeneity of an image)
9 Cluster tendency-CT (Measure the grouping of pixels that have similar
2.3 Feature selection
Feature selection is the process of choosing subset of features relevant to particular application
and improves classification by searching for the best feature subset, from the fixed set of original
features according to a given feature evaluation criterion(i.e., classification accuracy). Optimized
feature selection reduces data dimensionalities and computational time and increase the
classification accuracy. The feature selection problem involves the selection of a subset of `d'
features from a total of `D' features, based on a given optimization criterion. The D features are
denoted uniquely by distinct numbers from 1 to D, so that the total set of D features can be
written as S = { 1, 2, . . . , D }. X denotes the subset of selected features and Y denotes the set of

26
remaining features. So, S = X U Y at any time. J(X) denotes a function evaluating the
performance of X. J depends on the particular application. Here J(X) denotes the classification
performance of classifying and segmenting abnormal tumor region from brain CT images using
the set of features in X. In this work, Genetic Algorithm (GA technique is used.
Genetic algorithm:
We consider the standard GA to begin by randomly creating its initial population. Solutions are
combined via a crossover operator to produce offspring, thus expanding the current population of
solutions. The individuals in the population are then evaluated via a fitness function, and the less
fit individuals are eliminated to return the population to its original size. The process of
crossover, evaluation, and selection is repeated for a predetermined number of generations or
until a satisfactory solution has been found. A mutation operator is generally applied to each
generation in order to increase variation. In the feature selection formulation of the genetic
algorithm ,individuals are composed of chromosomes : a 1 in bit position indicates that feature
should be selected; 0 indicates this feature should not be selected. As an example chromosome
00101000 means the 3rd
and 5th
features are selected. That is the chromosome represents X={3,5}
and Y={1,2,4,6,7,8}. Fitness function for given bit string X is defined as
Fitness(X) = J(X) - penalty(X) [1]
Where X is the corresponding feature subset , and penalty(X) = w * (|X| -d) with a penalty
coefficient w. The size value d is taken as a constraint and a penalty is imposed on chromosomes
breaking this constraint. The chromosome selection for the next generation is done on the basis of
fitness. The fitness value decides whether the chromosome is good or bad in a population. The
selection mechanism should ensure that fitter chromosomes have a higher probability survival.
So, the design adopts the rank-based roulette-wheel selection scheme. If the mutated chromosome
is superior to both parents, it replaces the similar parent. If it is in between the two parents, it
replaces the inferior parent; otherwise, the most inferior chromosome in the population is
replaced. The selected optimal feature set based on the test data set is used to train the SVM,BPN
classifiers to classify and segment the abnormal tumor region from brain CT images. Table 2
shows the Best chromosomes selected (i.e., best features) using Genetic Algorithm(GA) during
the execution.
Table 2 . Best chromosomes selected by GA
SI-NO Feature set Classification accuracy
1 IDM,ENT, ENE, VAR,CON 95%
2 IDM,CON,ENE, MP, VAR 95%
3 ENT,IDM,VAR, IDM, CT 96%
4 IDM,ENT, CT, ENE,CON 95%
5 CON, IDM,VAR,ENT,ENE 96%
6 ENT,SA, IDM, ENE,VAR 96%
7 VAR,ENT, ENE, SA,IDM 95.5%
8 ENT,CON,ENE,VAR,IDM 96%
9 IDM,ENT, CT,CON, VAR 96%
10 ENE,ENT, MP,CON, COR 95%
The texture features Energy(ENE), Entropy(ENT), Variance(VAR), and Inverse Difference

27
Moment(IDM) are present in most of the feature vectors or feature set selected by GA. The
features such as CT,CON,SA,MP,COR which are least significant. The classification accuracy of
96% is obtained with four of the available 9 features using GA. Therefore it is possible to
classify and segment abnormal tumor region from brain CT images.
2.4 SVM classifier
Classification is the process where a given test sample is assigned a class on the basis of
knowledge gained by the classifier during training. Support Vector Machine(SVM) performs the
robust non-linear classification with kernel trick. SVM is independent of the dimensionality of the
feature space and that the results obtained are very accurate. It outperforms other classifiers even
with small numbers of available training samples. SVM is a supervised learning method and is
used for one class and n class classification problems. It combines linear algorithms with linear or
non-linear kernel functions that make it a powerful tool in the machine learning community with
applications such as data mining and medical imaging applications. To apply SVM into non
linear data distributions, the data can be implicitly transformed to a high dimensional feature
space where a separation might become possible. For a binary classification given a set of
separable data set with N samples X = {Xi}, i = 1, 2 …. N, labeled as Yi = ± 1. It may be difficult
to separate these 2 classes in the input space directly. Thus they are mapped into a higher
dimensional feature space by X’ = f(x).
The decision function can be expressed as
f(x) = W.x + ρ [2]
Where W.x + P = 0 is a set of hyper planes to separate the two classes in the new feature space.
Therefore for all the correctly classified data,
Yi f(x) = Yi (W.x + ρ) > 0, i = 1, 2 ….. N [3]
By scaling W and ρ properly, we can have f(x) = W.x + ρ = 1 for those data labeled as +1 closes
to the optimal hyper plane and f(x) = W.x + ρ = -1 for all the data labeled as -1 closes to the
optimal hype r plane. In order to maximize the margin the following problem needs to be solved.
Min (||W||2
/2)
Subject to Yi f(x) = Yi (W.x + ρ) ≥ 1, i = 1, 2 ….. N [4]
It is a quadratic programming problem to maximize the margins which can be solved by
sequential minimization optimization. After optimization, the optimal separating hyper plane can
be expressed as
f(x) = ∑=
+
N
i
iii xxKY
1
),( ρα [5]
Where K(.) is a kernel function, ρ is a bias, α is the solutions of the quadratic programming
problem to find maximum margin. When α is non zero, are called support vectors, which are
either on or near separating hyper plane. The decision boundary (i.e.) the separating hyper plane
whose decision values f(x) approach zero, compared with the support vectors, the decision values
of positive samples have larger positive values and those of negative samples have larger negative
values. Therefore the magnitude of the decision value can also be regarded as the confidence of
classifier. The larger the magnitude of f(x), the more confidence of the classification by choosing
the Gaussian kernel function
K(x,y) = e- γ
||x-y||2
[6]

28
Where the value of γ was chosen to be 1 and has good performance for the following two
reasons. First reason is the Gaussian model has only one parameter and it is easy to construct the
Gaussian SVM classifier compared to polynomial model which has multiple parameters. Second
reason is there is less limitation in using Gaussian kernel function due to nonlinear mapping in
higher dimensional space.
2.5 Detection of abnormal tumor region
Detection is important in selecting the sub band of the image to be decomposed. The process is
done by applying the 2 level 2D DWT, the image is decomposed into four sub bands. After
decomposition , SGLDM or Co-occurrence matrices is derived on detail sub bands. From these
co-occurrence matrices, WCT features are extracted as given in the feature extraction algorithm
and the optimal texture feature set is selected by GA based on the classification performance of
SVM,BPN classifiers. From the experiments conducted for feature selection, it is found that the
optimal feature set which gives good classification performance are the second order WCT
features like energy entropy, variance , and inverse difference moment. The four texture
features from detail sub bands form the feature vectors or feature set. These feature vectors are
given as input to the SVM,BPN classifiers to classify and segment the abnormal tumor region.
Efficiency or accuracy of the classifiers for each texture analysis methods are analyzed based on
the error rate (i.e.) All tests could have an error rate. This error rate can be described by the terms
true and false positive and true and false negative as follows:
True Positive (TP): The test result is positive in the presence of the clinical abnormality.
True Negative (TN): The test result is negative in the absence of the clinical abnormality.
False Positive (FP): The test result is positive in the absence of the clinical abnormality.
False Negative (FN): The test result is negative in the presence of the clinical abnormality.
Based on the above terms to construct the contingency table.
Table 3. Contingency table of classifier performance.
Actual group
AActual group
Predicted group
Normal Abnormal
Normal TN FP
Abnormal FN FP
Sensitivity = TP / (TP + FN) Specificity = TN / (FP + TN)
Accuracy = (TP+TN)/(TP + TN + FN + FP)
Sensitivity measures the ability of the method to identify abnormal cases. Specificity measures
the ability of the method to identify normal cases. Correct classification rate or accuracy is the
proportion of correct classifications to the total number of classification tests. The SVM,BPN
classifiers were tested by using Leave one out cross validation method . Leave one out cross
validation can be used as a method to estimate the classifier performance in unbiased manner.
Here each iteration, one data set is left out and the classifier is trained using the rest and the
testing applied to the left out data set. This procedure is repeated such that each data set is left
out once . Classification accuracy is calculated by taking the average number of all the iterations
. To evaluate the classification accuracy of classifiers , the 10 fold cross validation is done on the
data set collected from 100 images. In this method, the images are divided into 10 sets each
consisting of 5 normal images and 5 abnormal images. Then 9 sets are used for training and one
set is used for testing. Hence 10 iterations are done. s. For eg. In the first iteration 2 images and

29
another two iterations 1 image are wrongly classified. For the remaining 7 iterations all are
correctly classified. Hence the cross validation accuracy obtained as 96/100 which is equal to
96%. The results show that, if the number of representative samples increased, we get good
classification accuracy for 10 fold cross validation method. Other statistical method known as
Receiver Operating Characteristics (ROC) analysis [19] is also used to analyze the experimental
results of all the classifiers. The ROC curve is a graphical representation of sensitivity versus
specificity as a threshold parameter is varied. The Area under ROC Curve (AUC) has been used
to determine the overall classification accuracy. By calculating AUC, we can measure the class
discrimination capability of a specific classifier. An area of above 0.5 represents a perfect test
while an area of less than or equal to 0.5 represents worthless test. The larger the area (the higher
AUC value) means higher the classification performance. In this research, the ROC analysis and
accuracy are used to measure the performance of the classifiers and texture analysis methods.
3.Results
Our proposed method is implemented on real human brain CT dataset based on proposed flow
diagram as shown in Figure 1. Figure 2(a) represents the original CT brain image and Figure.2(b)
represents abnormal tumor (benign) brain CT image and Figure. 2(c) represents the abnormal
tumor region(malignant) brain CT image. The input data set consists of 100 images: 40 images
are normal, 60 images are abnormal. For each texture analysis method, input data set is
partitioned into training and test sets which are classified using SVM,BPN classifiers. This
section describes the wavelet based texture analysis method of classifying and segmenting
abnormal tumor region of CT images.
2a.Normal image 2b benign image 2c.Malignant image
Figure 2. Example of CT Normal and tumor images
BPN classifier performance was analyzed based on the experiments with the data set of 100
images. In BPN classifier ,more than one hidden layer may be beneficial for some applications,
but one hidden layer is sufficient if enough hidden neurons are used The number of hidden nodes
used in BPN is 25 (i.e.) 2*number of nodes in the input layer +1, number of nodes in the input
layer is 12, so 2*12+1=25, because 12 features are given as input to the classifier (Four features
such as energy, entropy, variance, inverse difference moment are extracted from detail sub
bands. So ( 3*4=12). For the best performance of BPN ,the proper number of nodes in the hidden
layer is selected through trial and error method based on number of epochs needed to train the
network .It is observed that the network performed well with 25 hidden nodes. The learning rate
for input and hidden layer is 0.4, moment=0.2 and the error allowed is 0.01. Binary sigmoid
function is used. After successful training with 450 epochs the satisfactory results are obtained.
In BPN, the initial weights are randomly selected from [-0.5, 0.5]. The selected optimal texture
features are given as input to the BPN classifier.

30
Feature selection is carried out using GA. There are 9 features are extracted from detail sub
bands. So totally 9*3= 27 features are extracted. The next step is to determine the
relevance of each selected feature to the process of classifying and segmenting abnormal tumor
region. During the evaluation process by using GA, some features may be selected many times as
the number of generation increases. If the feature was selected more times that feature was given
as more important in the feature selection. The number of times the features selected was energy,
entropy, variance and inverse difference moment. The parameter set for the GA algorithm is as
follows: Population size is 30; Cross Over probability is 1.0; Mutation rate is 0.1; Penalty
coefficient is 0.5 and stopping condition is 100 generations . Results show that, if the number of
sample images increased, we get good classification accuracy for the 10 fold cross validation
method .
A comparative study of the classification accuracy is performed for both wavelet based texture
analysis method and Spatial Gray Level Dependency Matrix method. Table 5 shows the
classification performances of the SVM classifier with different kernel functions.
N-Normal image, A-Abnormal image
The accuracy of SVM with Gaussian kernel function and SVM with polynomial kernel
function and SVM with linear function are 95%, 95.4%.96% respectively for same training and
testing data sets .The accuracy of SVM with Gaussian kernel function is high while compared to
SVM classifier with linear and polynomial functions.
Figure 3. ROC analysis curves of classifiers in wavelet domain

31
Figure 4. ROC analysis curve of classifiers in gray level domain
Figure 3 and Figure 4 shows the Roc analysis curve of classifiers in wavelet domain and gray
level domain. From the ROC analysis also ,the Area under the ROC Curve of SVM classifier in
wavelet domain , gray level domain are 0.96,0.92 . The Area under the ROC Curve of BPN
classifier in wavelet domain , gray level domain are 0.91,0.89. To justify the choice of wavelet
domain for the same data set without applying wavelet transform, in the gray level domain, the
four texture features are extracted and the performance of the SVM classifier is shown in Table 6
and the performance of BPN classifier is shown in Table 7. The accuracy of SVM, BPN
classifiers in wavelet domain are 96%,92% and in gray level domain are 91%, 89% respectively.
Table 5 .Classification performances of SVM classifier for 100 images
Parameter used Wavelet domain Gray level domain
TP 49 47
TN 47 45
FP 2 5
FN 2 3
Sensitivity in % 96.07% 94%
Specificity in % 95.91% 90%
Accuracy in % 96% 92%
Table 6 .Classification performances of BPN classifier for 100 images
Parameter used Wavelet domain Gray level domain
TP 46 45
TN 45 44
FP 5 7
FN 4 4
Sensitivity in % 92% 91.8%
Specificity in % 90% 86.2%
Accuracy in % 91% 89%

32
Table 7 shows classification performances of our proposed technique and the SGLDM method.
The classification accuracy of our proposed method is 96% which is high while compared with
SGLDM method.
Table 7 . Classification accuracy of proposed technique
SI-No Technique Classification Accuracy
1 WT+SGLDM+GA+SVM 96%
2 SGLDM+GA+SVM 92%
3 WT+SGLDM+GA+BPN 91%
4 SGLDM+GA+BPN 89%
4. Conclusions
As a conclusion, we have presented a method for wavelet based texture feature extraction and
selecting the optimal texture features using GA , and evaluated the SVM, BPN classifiers to
classify and segment the abnormal tumor region. The algorithm has been designed based on the
concept of different types of brain soft tissues have different textural features. This method
effectively works well for detection of abnormal tumor region with high sensitivity, specificity
and accuracy. The classification accuracy of the SVM,BPN classifiers using wavelet based
texture analysis method and SGLDM method without wavelet transform to classify and segment
the abnormal tumor region are 96%,92% ,91%,89% for 10 fold cross validation method. From
the ROC analysis ,the Area under the ROC curve(AUC) values for SVM, BPN classifiers in
wavelet based method, SGLDM method are 0.96,0.91.0.90,0.89. Results show that SVM
classifier with wavelet based statistical texture features were yielding better results compared to
the statistical texture features extracted directly from the image without applying wavelet
transform. This justifies the choice of using wavelet transform. Use of large data bases is
expected to improve the system performance and ensure the repeatability of the resulted
performance. This approach has potential for further development because of this simplicity that
will motivate to classify the types of tumors. The developed classification system is expected to
provide valuable diagnosis for the physicians.
References
[1] Duncan J.S.,Ayache N.,(2000)” Medical Image Analysis Progress Over two decade and challenges
ahead”, IEEE Trans on Pattern Analysis and Machine Intelligence , Vol 22, No 1,pp 85 – 106.
[2] Tourassi G D,(1999),” Journey towards computer aided Diagnosis – Role of Image Texture
Analysis”, Radiology, Vol 213, No 2, pp 317 – 320.
[3] Nathalie Richards, Michael Dujata,(2007),” Catherine Garbay , “ Distributed Markovian
segmentation : Application to MR brain scans”, Journal of Pattern Recognition ,Vol 40, No 6, pp
3467 – 3478.
[4] Zhang Y, Brady M, Smith S (2001) “Segmentation of Brain MR images through Hidden Markov
Random field Model and the expectation maximization algorithm”, IEEE Transactions on Medical
Imaging ,Vol 20, No 6, pp 45-57.
[5] Kaiping Wei, Bin He, Tao Zhang, Xianjun Zhen (2007),”A Novel Method for Segmentation of CT
Head Images”, 1st
International conference on Bioinformatics and Biomedical Engineering”, July 6-
8; Wuhan: IEEE Explore; pp. 717 – 720.

33
[6] Dubravko Cosic ,Sven Loncaric (1997),” Rule based labeling of CT head image”, 6th
conference on
Artificial Intelligence in Medicine”, March 23-26; Europe: Springer ; pp 453 – 456.
[7] Matesn Milan, Loncaric Sven, Petravic Damir (2001),” A Rule based approach to stroke lesion
analysis from CT brain Images”, 2nd
International Symposium on Image and Signal Processing and
Analysis, June 19-21 ; Pula, Crotia : IEEE Explore, pp 219 – 223.
[8] Ruthmann,V.E, Jayce E.M, Reo D.E, Eckardit M.J (1993), “ Fully Automated segmentation of
cerebrospinal fluid in computed tomography”, Psychiatry Research: Neuro Imaging ,Vol 50, No pp
101 – 119.
[9] Loncaric S and Kova Cevic D,(1997),” A Method for segmentation of CT head Images, Lecture
Notes on Computer Science”, Vol 1311, pp 1388 – 305.
[10] Tong Hau Lee, Mohammad Faizal , Ahmad Fauzi and Ryoichi Komiya (2009), “Segmentation of
CT Brain Images Using Unsupervised Clustering”, Journal of Visualization ,Vol 12, No 2, pp131-
138.
[11] Clark M C., Hall L O., Goldgof D B., Velthuzien R., Muztagh F R., and Silbiger M S.(1998),
”Automatic tumor segmentation using knowledge based techniques”, IEEE Transactions on Medical
Imaging , Vol 17, No 2, pp 187-192.
[12] Chang R F., Wu W J., Moon W K., Chou Y H., Chen D R. (2003),” Support vector machines for
diagnosis of breast tumors on US images”, Academic Radiology, Vol 10,No 2, pp 189-197.
[13] Fausett L (1994),” Fundamentals of Neural Networks: Architectures Algorithms and Applications”,
Englewood cliffs, NJ: Prentice Hall International, pp 289-293.
[14] Haddon J F, Boyce J F (1993), “Co-occurrence Matrices for Image analysis”, IEE Electronic and
Communications Engineering Journal, Vol 5, No 3, pp 71 – 83.
[15] Mohd Khuzi, Besar R, Wan Zaki W M D, Ahmad N N (2009), “ Identification of masses in digital
mammogram using gray level co-occurrence matrices”, Biomedical Imaging and Intervention journal
, Vol 5, Issue 3,e17.
[16] Van G., Wouver P.,Scheunders and Van Dyck.D (1998), “Statistical texture characterization from
discrete wavelet representation”, IEEE Trans. Image processing ,Vol 8, No 4, pp 592-598.
[17] Frank Z. Brill., Donald E. Brown., and Worthy N. Martin(1992),” Fast Genetic Selection of features
for Neural Network Classifiers”, IEEE Trans.,on Neural Networks, Vol 3, No 2, pp 324-328.
[18] Haralick R M, Shanmugam K and Dinstein I (1973),” Texture features for Image classification
“,IEEE Transaction on System, Man, Cybernetics ,Vol 3,No 6, pp 610 – 621.
[19] Tom Fawcett (2006) ,”An introduction to ROC analysis”, Pattern Recognition Letters , Vol 27, No 8,
pp 861-874.

34
Padma received her B.E in Computer science and Engineering from Madurai
Kama raj University ,Tamilnadu ,India in 1990. She received her M.E in
Computer science and Engineering from Madurai Kamaraj University
,Tamilnadu, India in 1997. She is currently pursuing P.hd under the guidance
of Dr.(Mrs.).R.Sukanesh in Medical Imaging at Anna university, Trichy. She is
working as Asst Professor in Department of information Technology, Velammal
college of Engineering and Technology,Tamilnadu, India. Her area of interest
includes Image processing, Neural Networks.
Dr.(Mrs.) R.Sukanesh, Professor in Bio medical Engineering has received her
B.E from Government college of Engineering and Technology, Coimbatore in
1982. She obtained her M.E from PSG college of Engineering and
Technology, Coimbatore in 1985 and Doctoral degree in Bio medical
Engineering from Madurai Kamaraj university , Madurai in 1999. Since
1985, she is working as a faculty in the Department of Electronics and
Communication Engineering , Madurai and presently she is the Professor
of ECE , and heads the Medical Electronics division in the same college.
Her main research areas include Neural Networks, Bio signal processing and
Mobile communication. . She has published several papers in National ,
International journals .

Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography Images Using Wavelet Based Statistical Texture Features

More Related Content

What's hot (20)

Similar to Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography Images Using Wavelet Based Statistical Texture Features (20)

Recently uploaded (20)

Automatic Diagnosis of Abnormal Tumor Region from Brain Computed Tomography Images Using Wavelet Based Statistical Texture Features