Brain Tissue Segmentation in MR Images with FGM

Abstract

We propose an automatic integrated novel approach to segmenting the human brain tissue in magnetic resonance (MR) images into White Matter (WM), Grey Matter (GM) and Cerebrospinal fluid (CSF) via fusing the Fuzzy C-Means (FCM) and Gaussian Mixture Model (GMM). The proposed approach, Fuzzy Gaussian Method (FGM) allows pixels with higher intensities (≥ 0.8) to retain their GMM clusters while others obtain the fuzzy membership in different clusters. Dice similarity coefficient (DSC) indexes as high as 0.9306 and 0.8684 for the WM and GM respectively with corresponding accuracy ratios of 0.9916 and 0.9783 are realized with the proposed FGM model. For the CSF, FGM also have a higher accuracy value of 0.9781 compared to FCM’s 0.9507 and GMM’s 0.9434 as well as DSC Index of 0.8049 to 0.7273 and 0.6467 of FCM and GMM respectively.