Classification of brain tumours Using Genetic Algorithms as a Feature Selection Method (GAFS)

Abstract

A brain tumour image classification technique based on Genetic Algorithm (GA) for feature selection is proposed in this paper. The searching capability of genetic algorithms is explored for appropriate selection of features from input data and to obtain an optimal classification. The objective of this paper is to present a novel method for feature selection. The method is implemented to classify and label brain MR Images into 5 tumour types viz. Glioma, Intra Ventricular Malignant Mass, Central Neuro Cytoma, Glioblastoma and Metastasis. A number of spatial features (texture, Gray Level Co-occurrence Matrix (GLCM), shape etc.) can be extracted from an image, so choosing the best features to avoid poor generalization and over specialization is of paramount importance. The proposed method was tested by varying two factors, the population size and the desired number of features. A set of 56 features distributed in 5 domains (Texture, GLCM, Shape, Haralicks, and color moments) were extracted and optimal features were selected using the genetic algorithm and fed to a K Nearest Neighbor (KNN) classifier for classification. The results were compared with the standard principal component analysis feature selection method. The same methodology was also tested on a publicly available external machine learning dataset to classify 16 types of Arrhythmia.