A NOVEL APPROACH FOR THE EXTRACTION AND CLASSIFICATION OF TUMOR IN MR IMAGES OF THE BRAIN VIA PRINCIPLE COMPONENT ANALYSIS AND KERNEL SUPPORT VECTOR MACHINE

Meenakshi Pareek

doi:10.26483/ijarcs.v8i7.4585

Abstract

Medical imaging plays an important role in the medical field. In present time there are various computerized methods for medical imaging to diagnose the inner portion of human body. Brain tumor detection is an important application in recent days. In this paper, various methods have been used for brain tumor detection and classification taking magnetic resonance images as input. For tumor classification, we had done experimentation with 50 MRI images taken from “figshare brain data set”. We propose an efficient method for brain tumor classification to classify cancerous and noncancerous tumor. The proposed method has three major steps 1.) Feature extraction 2) Feature reduction and 3) Classification. The present approach extracts the statistical texture features using 2D Discrete Wavelet transformation (Daubechies) and GLCM. PCA (principle Component Analysis) is used for feature reduction. We create a training data set that was carried out with 50 MRI images and applied SVM classifier to classify the tumor whether it is Benign and Malignant and evaluate the performance by Kernel based SVM.

Highlights

Nowadays computer technology used in medical field to focus on vast scale of medical field, such as tumor, brain research, cardiac research etc
According to the “Central Brain Tumor Registry of the United States (CBTRUS), there will be 64,530 new cases of primary brain and central nervous system tumors diagnosed by the end of 2011”
The proposed system incorporate with different types of techniques, for feature extraction and selection they used DWT(Discrete wavelet transformation) and principle components analysis (PCA)(Principle component analysis) and classified tumor with SVM

Summary

INTRODUCTION

Nowadays computer technology used in medical field to focus on vast scale of medical field, such as tumor, brain research , cardiac research etc. When large no of MRI images are analyzed and visualized by radiologist manually or with conventional method, leads to inaccurate classification. [6]SVM optimized better result than other classifier; it is binary classification, based on supervised learning. Training set comprises one “target value” and second “attribute” (feature) [6] This manuscript is structured as follows:- Section II provides a interrelated work of different “feature extraction” techniques, “feature selection” techniques and different “classification” techniques to classify the tumor whether it is cancerous and non-cancerous.

RELATED WORK

PROPOSED WORK

EXPERIMENT AND RESULTS

CONCLUSION AND FUTURE WORK