Unified model based classification with FCM for brain tumour segmentation

Abstract

This paper presents a novel method for brain tumour segmentation which unifies two different models. The segmentation procedure goes through three stages. First stage employs an Extended Hyperbolic Tangent (EHT) model to separate natural cells from deceased ones. Second stage makes use of fuzzy clustering technique to sharpen the separation done in the previous stage. The final stage isolates tumour effected cells from edema part using a mixture model of edema and tumour. The computationally efficient algorithm is applied to multichannel Magnetic Resonance Image slices and works good for both high grade and low grade tumours. The only interaction needed from the user is the slice number selection. The algorithm yields a comparatively better result by making use of T1 weighted, T2 weighted, T1 contrast and FLAIR MRI channels.