You are on page 1of 18

ABSTRACT

Oral cancer is a serious problem among people due to its aggressive nature, as-
sociated with mostly unfavorable prognosis. Clinical examination by experienced
medical experts followed by biopsy for diagnosis are time taking.Image process-
ing techniques with feature extraction enables clear visualization of cancer affected
areas with substantial resolution detecting different types of oral cancer lesions.
Here textural and run-length features of camera images are used. Although many
systems are developed for binary classification of images into cancerous and
normal, there is an apparent scarcity for systems performing within-class
classification of malignant cases. Such systems are quite useful in the prognosis
part.

Texture is one of the important characteristics used in identifying objects or re-


gion of interest in an image. Features derived from Grey Level Co-occurrence
Matrix (GLCM) matrix are widely used for an image characterization based on
texture analysis.It helps in suitably selecting texture discriminating features for
classification of
oral cancer lesions in digital camera images into different groups. Support Vector
Machine is used to compare and validate the performance of different feature sets.

In this proposed work, the color image is captured and de-noised to enhance the
quality of the image and oral lesion is detected effectively. Multi-class
classification of lesions in color images is performed using different feature sets
Using ANN Classifier.The efficiency of the proposed system offers a vindication of
the relevance of texture features in discriminating oral cancer lesions.
Contents

1 Introduction 2
1.1 Image Representation . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Noise Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Noise removal using Filters . . . . . . . . . . . . . . . . . 4
1.3 Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Texture Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Classifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Problem Statement 8

3 Implementation 9
3.1 Modules Implemented . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.1 Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.2 Thresholding . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.3 Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.4 Feature Extraction........................................................................10
3.1.5 Classifier 10

4 Applications 11

5 Conclusion and Future Scope 12


5.1 Conclusion................................................................................................ 12
5.2 Future Scope.............................................................................................12

References 14
List of Figures

1.1 IMAGE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 MATRIX REPRESENTATION OF IMAGE . . . . . . . . . . . . . . . . . . . 3
1.3 BEFORE NOISE FILTERING . . . . . . . . . . . . . . . . . . . 4
1.4 AFTER NOISE FILTERING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 EXAMPLE OF SEGMENTATION . . . . . . . . . . . . . . . . . 5
1.6 GLCM REPRESENTATION . . . . . . . . . . . . . . . . . . . . 6
1.7 SVM CLASSIFIER . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.1 FLOW CHART.......................................................................................11
Texture Map Based Tumor Classification

Chapter 1

Introduction

1.1 Image Representation


MATLAB uses image processing toolbox. A digital image is a numeric represen-
tation of a two-dimensional image. Digital image is how computer record the photos
and pictures. The smallest unit is called pixel, normally consists of a 0-255 value
for gray images(0:black, 255 white). For color images, each pixel has three 0-255
values, representing RBG. So the whole picture or photo in computer is actually a
data table, or matrix, with different size of pixels.

Figure 1.1: IMAGE

Department of Computer Science Engineering, RVCE, Banglore 1


Texture Map Based Tumor Classification

Figure 1.2: MATRIX REPRESENTATION OF IMAGE

1.2 Noise Filtering


Digital images are prone to various types of noise. Noise is the result of errors in
the image acquisition process that result in pixel values that do not reflect the true
intensities of the real scene. There are several ways that noise can be introduced
into an image, depending on how the image is created. For example: If the image is
scanned from a photograph made on film, the film grain is a source of noise. Noise
can also be the result of damage to the film, or be introduced by the scanner itself.

Department of Computer Science Engineering, RVCE, Banglore 2


Texture Map Based Tumor Classification

1.2.1 Noise removal using Filters

Figure 1.3: BEFORE NOISE FILTERING

Figure 1.4: AFTER NOISE FILTERING

Department of Computer Science Engineering, RVCE, Banglore 3


1.3 Segmentation Texture Map Based Tumor Classification

In computer vision,image segmentation is the process of partitioning a digital im-


age into multiple segments(sets of pixels, also known as image objects). The goal of
segmentation is to simplify and/or change the representation of an image into some-
thing that is more meaningful and easier to analyze. Image segmentation is typically
used to locate objects and boundaries (lines, curves, etc.) in images.

Figure 1.5: EXAMPLE OF SEGMENTATION

1.4 Texture Analysis


Gray Level Co-occurrence Matrix texture analysis refers to the characterization of re-
gions in an image by their texture content [2 ]. Texture analysis attempts to quantify
intuitive qualities described by terms such as rough, smooth, silky, or bumpy as a
function of the spatial variation in pixel intensities. In this sense, the roughness or
bumpiness refers to variations in the intensity values, or gray levels. Texture
analysis is used in various applications, including remote sensing, automated
inspection, and medical image processing.

Department of Computer Science Engineering, RVCE, Banglore 4


Texture Map Based Tumor Classification

Figure 1.6: GLCM REPRESENTATION

1.5 Classifier
Classification is the process of predicting the class of given data points. Classes
are sometimes called as targets/ labels or categories. Classification predictive mod-
eling is the task of approximating a mapping function (f) from input variables (X) to
discrete output variables (y).

In machine learning, support-vector machines are supervised learning models


with associated learning algorithms that analyze data used for classification and re-
gression analysis. Given a set of training examples, each marked as belonging to
one or the other of two categories, an SVM training algorithm builds a model that
assigns new examples to one category or the other.

Department of Computer Science Engineering, RVCE, Banglore 5


Texture Map Based Tumor Classification

Figure 1.7: SVM CLASSIFIER

Department of Computer Science Engineering, RVCE, Banglore 6


Texture Map Based Tumor Classification

Chapter 2

Problem Statement

To identify tumor by analyzing features of image and to classify them


into different types of lesions. The objective is to select a reduced set
of features which can clearly distinguish between different groups of
malignancy caused by carcinoma of different areas of oral cavity. The
popular texture analysis methods for classification of oral cancer
lesions based on the differences in superficial patterns and surface
deformities caused by the particular type of malignancy are used.

Department of Computer Science Engineering, RVCE, Banglore 7


Texture Map Based Tumor Classification

Chapter 3

Implementation

Image source : https://oralcancerfoundation.org/dental/oral-cancer-images/

3.1 Modules Implemented


3.1.1 Filtering
Non local Means Algorithm is an algorithm in image processing for image denoising.
Non Local means filtering takes a mean of all pixels in the image, weighted by how
similar these pixels are.This results in much greater post-filtering clarity, and less
loss of detail in the image compared with other algorithms.[ 1]

1 Noisy Img=Y1 ;
2 % Get Image I n f o
3 [ Height , Width ] = s i z e ( Noisy Img ) ;
4 % Initialize the denoised
5 image u = z e r o s ( Height , Width ) ;
6 PaddedImg = p a d a r r a y ( NoisyImg , [ P a t c h S i z e H a l f , P a t c h S i z e H a l f ] , ’ symmetric ’ , ’ bot h ’ ) ;
7 PaddedV = p a d a r r a y ( NoisyImg , [ Window Size Half , Window Size Half ] , ’ symmetric ’ , ’ b o t h ’ ) ;

3.1.2 Thresholding
Thresholding is the simplest method of segmenting image. From a grayscale im-
age, thresholding can be used to create binary images .
1 D1= i m r e s i z e ( D1 , [ 1 3 9 1 0 3 ] )
2 ; Thresh = g r a y t h r e s h ( D1 ) ;
3 S=im2bw ( D1 , Thresh ) ;
4 % f i g u r e , imshow ( S ) ;
5 T= imcomplement ( S ) ;

3.1.3 Segmentation
Active Contour it is used in delineating an object outline from a possibly noisy 2D
image.[ ]
Department of Computer Science Engineering, RVCE, Banglore 8
%g e t s u b s c r i p t s f o r l o c a l r e g i
Texture Map Based Tumor Classification

xneg ( xneg <1) = 1 ; yneg ( yneg <1) = 1 ;%check


2

xpos ( xpos>dimx ) = dimx ; ypos ( ypos>dimy )


3

Ain= z e r o s ( s i z e ( i d x ) ) ; Aout= z e r o s ( s i z e ( i d
u= z e r o s ( s i z e ( i d x ) ) ; v= z e r o s ( s i z e ( i d x ) ) ;
%−− g e t windows f o r l o c a l i z e d s t a t i s t i c
5

i d x = f i n d ( p h i <= 1 . 2 & p h i >= −1 . 2 )


6

’[ ;y x ] = i n d 2 s u b ( s i z e ( p h i ) , i d x ) ;
9

10

re − ;i n i t i a l i z e u , v , Ain ,
%−− g e t t h e c u r v e ’ s narrow band
11

yneg = y−r a d ; ypos = y+ r a d ;


sxneg = x−r a d ; xpos = x+ r a d ;
12

13

3.1.4 Feature Extraction

bounds

= dimy
Gray Level Co-occurance Matrix graycomatrix(I) creates a gray-level co-occurrence

x));
Aout
%−−
matrix (GLCM) from image I.It calculates how often a pixel with gray-level
(grayscale intensity) value i occurs horizontally adjacent to a pixel with the value j.
[1] [6 ]= g r a y c o m a t r i x (WW, ’ O f f s e t ’ ) ;
GLCM2
1

3.1.5 Classifier
are supervised learning models with associated
Support Vector Machine Classification
learning algorithms that analyze data used for classification and regression analysis.
Given a set of training examples, each marked as belonging to one or the other of
two categories, an SVM training algorithm builds a model that assigns new
examples to one category or the other.
1 [ Tra in ing Time , Testing Time , T r a i n i n g A c c u r a c y , T e s t i n g A c c u r a c y ]=SRC( ’ T r a i n
i n g S e t ’ , ’ T e s t S e t ’ , 1 , 1 , ’ RBF kernel ’ , 1 0 0 ) ;
2 feature=stats1.autoc(1,1);

Department of Computer Science Engineering, RVCE, Banglore 9


Texture Map Based Tumor Classification

Figure 3.1: FLOW CHART

Department of Computer Science Engineering, RVCE, Banglore 10


Texture Map Based Tumor Classification

Chapter 4

Applications

• Content-based image retrieval. It is the application of computer vision tech-


niques to the image retrieval problem, that is, the problem of searching for
digital images in large databases ”Content-based” means that the search ana-
lyzes the contents of the image rather than the metadata such as keywords,tags,
or descriptions associated with the image.
• Object detection is a computer technology related to computer vision and
image processing that deals with detecting instances of semantic objects of a
certain class (such as humans, buildings, or cars) in digital images and videos
• Feature selection is important in fault diagnosis in industrial applications,
where numerous redundant sensors monitor the performance of a machine.[6 ]
• In genomics data, individual features correspond to genes, so by selecting the
most relevant features, one gains important knowledge about the genes that are
the most discriminate for a particular problem.
• Applications of Classification are speech recognition, handwriting recognition,
bio-metric identification, document classification etc.

Department of Computer Science Engineering, RVCE, Banglore 11


Texture Map Based Tumor Classification

Chapter 5

Conclusion and Future Scope

5.1 Conclusion
In this proposed work, the color image is captured and de-noised to enhance the
quality of the image , lesion is detected manually and classification to the type of
cancer is made.

The objective is to select a reduced set of features which can clearly distinguish
between different groups of malignancy caused by carcinoma of different areas of
oral cavity[3].

5.2 Future Scope


• The detection of region of interest could be done automatically.
• Can be further extended to non-cancerous lesions
• A combination of GLCM and GLRL features to improve classification accu-
racy.[ 6]

Department of Computer Science Engineering, RVCE, Banglore 12


Snapshots:

Department of Computer Science Engineering, RVCE, Banglore 13


Texture Map Based Tumor Classification

References

[1] Texture-Map-Based Branch-Collaborative Network for Oral Cancer Detection


; Chih-Hung Chan, Tze-Ta Huang, Chih-Yang Chen, Chien-Cheng Lee, Man-
Yee Chan, and Pau-Choo Chung, 2019
[2] Advances in early detection and diagnostic adjuncts in oral cavity cancer; C.
S. Farah ,2013

[3] Texture Analysis Based Segmentation and Classification of Oral Cancer Le-
sions in Color Images Using ANN ; Dr.Vinod Kumar ,Belvin Thomas, 2012

[4] Active Contour Without Edges ; T.F. Chan, L.A. Vese, 2010

[5]
https://in.mathworks.com/help/vision/ref/objectdetectortrainingdata.ht
ml ( as accessed on 1/1/2020 )

[6] Image characterizations based on joint gray level -run length distributions ;
B.V. Dasarathy, E.B. Holder, 2008

Department of Computer Science Engineering, RVCE, Banglore 14

You might also like