Automatic MR Brain Tumor Image Segmentation

Traditional Dirichlet process mixture (MDP) models has the characteristic that the image segmentation can be done without initialization of clustering numbers. For the computing speed of the classical MDP segmentation is jogging, a new kind of nonparametric segmentation (DMMDP algorithm) combined with anisotropic diffusion and Markov Random Fields (MRF) prior was inferred in this paper. The experiment results of menigioma MR images segmentation showed that the properties, such as accuracy and computing speed, of the DMMDP algorithm were significantly greater than the classical MDP model segmentation.