Brain Tissue Segmentation using Patch-wise M-net Convolutional Neural Network

Abstract

The anatomical structures of the brain can be visualized and changes in the brain and delineating pathological regions are analyzed by automatic segmentation of brain tissues into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) from magnetic resonance images (MRI). In this paper, we propose a patch-wise M-net architecture for the automatic segmentation of brain structures in MRI. In the proposed method, the limitations of conventional approaches can be overcome by the patch-wise M-net with more retention of local information. In our method, the slices from an MRI scan are divided into non-overlapping patches and then the nonoverlapping patches with their corresponding patches of ground truth are fed to the M-net model to train the network. It is shown that the segmentation performance has been improved with the proposed patch-wise M-net architecture as compared to state-of-the-art methods.