You are on page 1of 12

vii

TABLE OF CONTENTS

CHAPTER NO. TITLE PAGE NO.

ABSTRACT iii
LIST OF TABLES xii
LIST OF FIGURES xiv
LIST OF SYMBOLS AND ABBREVIATIONS xvii

1 INTRODUCTION 1
1.1 PROLOGUE 1
1.1.1 Human Brain 2
1.1.2 Brain Tumor 4
1.1.3 Categories of Brain Tumor 6
1.1.3.1 Primary brain tumor 6
1.1.3.2 Secondary brain tumor 6
1.1.3.3 Benign Tumor 6
1.1.3.4 Malignant Tumor 6
1.1.4 Types of Brain Tumor 7
1.1.5 Brain Tumor Statistics 8
1.1.6 Brain Tumor Symptoms 9
1.1.7 Stages of Brain Tumor 10
1.1.8 Magnetic Resonance Image 11
1.1.8.1 Advantages of MRI 12
1.1.8.2 Drawbacks of MRI 13
viii

CHAPTER NO. TITLE PAGE NO.

1.1.9 Computer-aided Diagnosis System 13


1.1.10 Database 14

1.2 MOTIVATION 15
1.3 PROBLEM DEFINITION 16
1.4 RESEARCH APPROACH 16
1.5 METHODOLOGY 17
1.5.1 Preprocessing and Enhancement 17
1.5.2 Image Segmentation 17
1.5.3 Feature Extraction and Selection 19
1.5.4 Classification 19
1.5.5 Performance Analysis 21
1.6 ORGANIZATION OF THE THESIS 22
1.7 SUMMARY 24

2 LITERATURE REVIEW 25
2.1 PROLOGUE 25
2.2 PREPROCESSING AND ENHANCEMENT 25
2.2.1 Preprocessing 26
2.2.2 Enhancement 30
2.3 REGISTRATION 37
2.4 SEGMENTATION 41
2.5 DETECTION 51
2.6 FEATURE EXTRACTION AD SELECTION 53
2.7 CLASSIFICATION 55
2.8 ROC ANALYSIS 58
2.9 MERITS OF PROPOSED SYSTEM 60
2.10 SUMMARY AND CONCLUSION 61
ix

CHAPTER NO. TITLE PAGE NO.

3 PREPROCESSING AND ENHANCEMENT 62


3.1 PROLOGUE 62
3.2 MRI BRAIN DATABASE 62

3.3 IMAGE ACQUISITION 64


3.4 PREPROCESSING 64
3.4.1 Removal of Film Artifacts 65
3.5 ENHANCEMENT 66
3.5.1 Median Filter 67
3.5.2 Center Weighted Median Filter 69
3.6 PERFORMANCE EVALUATION 73
3.6.1 Experiment and Result 74
3.7 SUMMARY AND CONCLUSION 78

4 SEGMENTATION 79
4.1 PROLOGUE 79
4.2 MARKOV RANDOM FILED 82
4.3 SEGMENTATION OF MRI IMAGE USING
FUZZY C MEANS (FCM) 83
4.4 SEGMENTATION OF MRI IMAGE WITH
MODIFIED PSO 87
4.4.1 Prelimaries on PSO 87
4.4.2 PSO Algorithm 92
4.5 SEGMENTATION OF MRI IMAGE WITH PSO
AND FUZZY C MEANS (FCM) 95
4.6 EXPERIMENT AND RESULTS 100
4.7 SUMMARY AND CONCLUSION 105
x

CHAPTER NO. TITLE PAGE NO.

5 CLASSIFICATION 107
5.1 PROLOGUE 107
5.2 FEATURE EXTRACTION 108
5.2.1 Surrounding Region Dependency Matrix 108
5.2.2 Spatial Gray Level Dependency Matrix 111
5.3 FEATURE SELECTION 115
5.3.1 Feature Selection using Genetic Algorithm 117
5.3.2 Feature Selection using Artificial Bee
Colony 118
5.3.3 Feature Selection using Bacterial Foraging
Optimization (BFO) 119
5.3.4 Feature Selection using Ant Colony
Optimization 120
5.3.5 Experiments and Results 121
5.4 BACK PROPAGATION ARTIFICIAL
NEURAL NETWORKS 121
5.5 BPN WITH ARTIFICIAL BEE COLONY
OPTIMIZATION ALGORITHM 127
5.6 SUMMARY 131

6 PERFORMANCE ANALYSIS 133


6.1 PROLOGUE 133
6.1.1 Sensitivity versus Specificity 134
6.2 PERFORMANCE ANALYSIS OF
CLASSIFIERS 135
xi

CHAPTER NO. TITLE PAGE NO.

6.3 JACK KNIFE METHOD 139


6.4 SUMMARY AND CONCLUSION 140

7 CONCLUSIONS AND RECOMMENDATIONS TO


FUTURE WORKS 141
7.1 CONCLUSION 141
7.2 RECOMMENDATIONS TO FUTURE WORKS 142
7.3 CONTRIBUTION OF THE THESIS 143

APPENDIX 1 144

APPENDIX 2 158

APPENDIX 3 162

APPENDIX 4 168

REFERENCES 173

LIST OF PUBLICATIONS 191


xii

LIST OF TABLES

TABLE NO. TITLE PAGE NO.

1.1 Types of brain tumor and the affected rate 7


1.2 Brain tumor symptoms and the percentage of patients 10
1.3 Brain tumor grade and survival period 11
2.1 Existing methods and techniques of preprocessing 27
2.2 Previous methods and techniques of enhancement 30
2.3 Previous approaches for registration 37
2.4 Previous approaches and methods for segmentation 42
2.5 Earlier methods and techniques for brain
tumor detection 52
2.6 Foregoing approaches for feature extraction
and selection 54
2.7 Previous approaches for classification 56
2.8 Previous approaches for ROC analysis 59
2.9 Merits and Methods for Proposed System 60

3.1 Properties of single MR brain image 63


3.2 Performance analysis of median filter 68
3.3 Performance analysis of center weighted median filter 71
3.4 Result of enhancement techniques 75
3.5 Performance analysis of enhancement techniques 76
3.6 Comparative study of ASNR and PSNR 77
xiii

TABLE NO. TITLE PAGE NO.

5.1 SRDM based feature extraction 114


5.2 Selected features from feature selection algorithms 121
6.1 Classification performance of SRDM 137
6.2 Classification performance of GLRLM 138
6.3 Classification performance based on ten fold
Validation method 139
6.4 Performance analysis of previous works and the
Proposed work 140
A4.1 Time Complexity when number of cluster varying 172
xiv

LIST OF FIGURES

FIGURE NO. TITLE PAGE NO.

1.1 The Structure of the brain 4


1.2 Types of primary brain tumor and the affected rates 8
1.3 Flow diagram of CAD system 14
1.4 Sample MRI brain tumor images 15
1.5 Preprocessing and segmentation steps 18
1.6 Feature extraction, selection and classification 20
1.7 A Typical curve 22
2.1 The sample ROC curve 60
3.1 Sample Brain MRI’s from patient database 63
3.2 Original MRI brain Image and preprocessed MRI 66
3.3 Median filtering for 3×3 sliding window on the MRI 67
3.4 Chart depicting performances analysis of
median filter 69
3.5 Algorithm for center weighted median filter 70
3.6 Chart depicting performances ANALYSIS of
center weighted median filter 72
3.7 Chart depicting the performance analysis
of enhancement techniques 76
3.8 Center weighted median filter applying of
(3x3, 5x5, 7x7, 9x9, 11x11, and 13 ×13 of
sliding window) for entire image 77
4.1 Algorithm of FCM 86
xv

FIGURE NO. TITLE PAGE NO.

4.2 Concept of modification of a searching point by PSO 89


4.3 Flow diagram for the PSO algorithm 91
4.4 Algorithm of modified PSO with FCM 99
4.5 Original brain MRI 100
4.6 Preprocessing stage in brain tumor detection 100
4.7 Enhanced MRI 101
4.8 Segmented MRI using FCM 101
4.9 MRI from database 102
4.10 Brain MRI in preprocessing 102
4.11 Enhanced MRI 103
4.12 Segmented MRI using MPSO 103
4.13 Segmentation using MPSO -FCM 104
4.14 User interaction screen 104
4.15 Segmentation page on automatic system 105
5.1 Algorithm of SRDM 109
5.2 Sub image for constructing SRDM 110
5.3 A typical SRDM matrix 111
5.4 Sub image of a segmented MRI image
for constructing SGLDM 113
5.5 A typical SGLDM matrix 113
5.6 Back propagation neural network 123
5.7 Weight updating using ABC algorithm 128
5.8 Flow chart for BPN classifier 129
5.9 BPN with ABC 130
6.1 Classification performances for the SRDM 136
6.2 Classification performances for the SGLDM 136
6.3 Chart for classification performance of SRDM 137
6.4 Chart for classification performance of SGLDM 138
xvi

FIGURE NO. TITLE PAGE NO.

6.5 ROC curves for texture analysis method


based on ten fold validation 139
6.6 Performance analysis 140
A2.1 Sample MR Images in DICOM Format 160
A3.1 Modern 3 Tesla clinical MRI scanner 165
A4.1 Registration process for automatic system 168
A4.2 Contrast difference betweeen
normal MRI and douted MRI 168
A4.3 Sum of square difference betweeen
normal MRI and douted MRI 169
A4.4 White cells difference between
Normal and abnormal MRI 169
A4.5 Segmented image from original MRI 170
A4.6 Classification report 170
A4.7 Segmentation using genetic algorithm 171
A4.8 Time complexity between segmentation methods 172
xvii

LIST OF SYMBOLS AND ABBREVIATION

Symbols

x - Center pixel of the label


CCp - Cluster Center Point
CII - Contrast Improvement Index
Cn - Current Neuron
Dij- - Distance Matrix
bi - Gray level value of Background region
T0 - Initial Pheromone
Y - Intensity value of pixels in the kernel
B - Mean Gray level value of Background
F - Mean Gray level value of foreground
- Mean value of the kernel
Tnew - New Pheromone
K - Number of Ants
n - Number of Iteration
Told - Old Pheromone
V - Potential function of the kernel
E - Rate of pheromone evaporation parameter In local
update
- Standard deviation of the kernel
N - Total number of pixel in the surrounding
Background region
Wv - Weight vector
Wn - Winning Neuron
xviii

Abbreviations

ABC - Artificial Bee Colony Algorithm


ACO - Ant Colony Optimization
ASNR - Average Signal-to-Noise Ratio
AT - Adaptive Threshold
BFO - Bacteria Foraging Optimization
CAD - Computer Aided Diagnosis
CAD - Computer Aided Diagnosis
CT - Computerized Tomography
FCM - Fuzzy C Means
FP - False Positive
FP - False Positive
GA - Genetic Algorithm
Gbest - Global best
Gmin - Global minimum
Lbest - Local best
Lmin - Local Minimum
MPSO - Modified Particle Swarm Optimization
MRI - Magnetic Imaging Resonance
PET - Position Emision Tomography
PSNR - Peak Signal-to-Noise Ratio
PSO - Particle Swarm Optimization
ROC - Receiver Operating Characteristic
SGLM - Spatial Grey Level Dependency Matrix
SRDM - Surrounding Region Dependency Method
TP - True Positive

You might also like