Automatic glioblastoma multiforme detection using hybrid-SVM with improved particle swarm optimisation

Abstract

Gliomas is one of the harmful life-frightening brain dementia which occurs with the unnecessary growth of cells in the brain. In medical field, detection of brain tumours is a challenging task. Medical resonance imaging is one of the best techniques to identify the tumours. Gaussian filters are used in pre-processing to increase the image quality. Grey-level matter co-occurrence matrix can be used to extract the features by using spatial relationship between the image pixels. In this paper, an advanced classification technique called hybrid-SVM is introduced to classify the brain image with reduced features. To improve the efficiency hybrid-SVM is used with radial basis function kernel. The proposed algorithm classifies the brain tumour classes intelligently. Hybrid-SVM with improved particle swarm optimisation is used to overcome the drawback of hyperplane selection. The experimental result analysis of the proposed work outstrips the other recent classifiers with 92.75% accuracy rate.