An effective tree metrics graph cut algorithm for MR brain image segmentation and tumor Identification

Abstract

The proposed algorithm describes the problem of Magnetic Resonance (MR) brain image segmentation using the tree-metric graph cuts (TM) algorithm, a novel segmentation algorithm and introducing a “tree-cutting” method to interpret the labeling returned by the TM algorithm as tissue classification for the input brain MR brain image. The algorithm has three sequential steps: 1) pre-processing, which generates a tree of labels as key to the TM algorithm; 2) a sweep of the TM algorithm, which proceeds a globally optimal labeling with respect to the tree of labels; 3) post-processing, which involves running the “tree-cutting” method to generate a mapping from labels to brain tissues such as Grey Matter (GM), White Matter (WM) and Cerebrospinal Fluid (CSF) producing a meaningful MR brain image segmentation. On comparison with the current approaches, the result obtained shows that the tree metrics graph cut algorithm is faster and the overall segmentation accuracy is better for segmenting both T1 and T2 weighted MR axial brain slice images.