Research Of Brain MRI Images Segmentation Based On Biomimetic Pattern Recognition

Medical image segmentation is a hot issues and tough in the medical image processing and analysis field. Magnetic Resonance Imaging (MRI) is a medical imaging technique with noninvasive, high-resolution and clear anatomical structure. In the diagnosis of brain tissue, Magnetic Resonance Imaging has unique advantages, which has become a major clinical diagnostic technique. This paper studies Support Vector Clustering (SVC) and Biomimetic Pattern Recognition (BPR) in brain magnetic resonance image segmentation application, the main work include:1. For traditional medical image smooth filtering methods, the details and edges of images would be lessened as well as noise is removed. In this paper we present a image denoising method based on SVC multi-window. Firstly, the pixel point was marked as signal, the points of possible positive noise and the possible points of negative noise using local statistical characteristics; Secondly, in the latter two categories, we use iterating SVC on the gray value of pixels and dealing it with noise filtering. Thus, the position of the noise will be located and the noise will be handling. Experiments show that the method presented in this paper can not only remove noise from MRI brain images but also keep the details and edges of images well.2. The neural network based on the biomimetic pattern recognition principles is built. The biomimteic pattern recognition makes recognition from the views of"matter cognition"instead of"matter classification", which analyzes and cognizes the high dimensional geometrical distribution that consists of the sample sets in the high dimensional feature space. First,ψ3-neurons'weights are determined according to training samples and then Multi-Weights Neuron Networks is established. Second, the neuron networks is used to completely cover samples'high dimensional feature space. Finally, medical images are recognized and segmented based on the results of the optimal coverage of the samples. The experimental results show that compared with traditional medical image segmentation methods, the proposed method has higher accuracy, reliability, and better generalization. Besides, this algorithm emphasizes on"cognition". It can combine transcendental knowledge, and obtain the desired object from medical images quickly and reliably with highly intelligent.