Automatic Segmentation of Brain Tumor Magnetic Resonance Imaging Based on Multi-Constrains and Dynamic Prior

The most difficult and challenging problem in medical image analysis is image segmentation. Due to the limited imaging capability of magnetic resonance (MR), the sampled magnetic resonance images from clinic always suffer from noise, bias filed (also known as intensity non-uniformity), partial volume effects and motive artifacts. In additional, for the complex shape boundary and topology of brain tissues and structures, segmenting magnetic resonance image of brain tumor fast, accurately and robustly is very difficult. In this paper, we propose an image segmentation algorithm based on multiconstrains and dynamic prior. Through introducing a novel big scale constrain into Markov random filed model from magnetic resonance image we realize automatic segmentation under the principle of maximum a Posterior and a modified expectation-maximization algorithm according to the Bayesian frame. Finally, a set of human body detection and tracking experiments are designed to demonstrate the effectiveness of the proposed algorithms.