Mathematics and Computer Science, Volume 1

15Feature Extraction and Selection for Classification of Brain Tumors

Saswata Das

Department of Radiophysics and Electronics, University of Calcutta, Calcutta, India

Abstract

When malignant masses grow in an uncontrolled way, then it produces obscureness in the boundaries, shape and location. Sometimes it produces imprecise feature values. So selection of proper feature is very necessary to classify the brain tumors and diagonose it correctly. In this case, the tumor image is contrast enhanced and segmented. The segmented image is edged by using canny edge detection algorithm. From this segmented edged image, the proper feature is extracted. After that, the feature is selected using genetic algorithm and particle swarm optimization algorithm. Finally, classifiers like decision tree, support vector machine, K-nearest neighbor, etc., are used to classify the images. Experimental results show that this method reduces the feature redundancy to produce high efficiency in reasonably low time. The aim of our study is to check whether the tumor is benign or malignant.

Keywords: Brain tumor, feature extraction, feature selection, classification, SVM, KNN

15.1 Introduction

The main objective of this work is to analyze the brain MRI for proper detection of lesion characteristics in early stage. Different brain tumors like glioma, oligodendroglioma, astrocytoma, acoustic neuroma are very fertile to human being if not diagnosed properly. It causes different abnormalities life vision problem, ...