An Automated Approach for Segmentation of Brain MR Images using Gaussian Mixture Model based Hidden Markov Random Field with Expectation Maximization

Abstract

Manual segmentation of brain tissues from MR images for diagnosis purpose is time consuming and requires much effort even by experts. This has motivated generation of automated segmentation techniques. Moreover, due to presence of noise in an image and its low contrast, it is difficult to correctly delineate tumour from brain MR images. In this paper, a novel hybrid method using Gaussian Mixture Model based Hidden Markov Random Field (HMRF) with Expectation Maximization (EM) has been proposed which segments tissues from MR brain images efficiently and helps to separate out tumour area easily. The proposed method minimizes energy function during each iteration of EM and gives comparable results with ground truth. The results obtained are also compared with the results of fuzzy c-means clustering algorithm for image segmentation.