The Research Of Medical Image Segmentation Algorithm Based On Multi-Agent System

In order to distinguish normal tissue and abnormal pathological changes, We need to segment the medical image. Lower contrast of medical image, the variety of tissue character, the illegibility between different tissues or between tissue and focus as well the distributing complexity of imperceptibility structure, such as vas and nerve lead to the difficulty of medical image segmentation. The purpose of medical image segmentation is to separate the different regions with special signification and make the results approximate to the anatomic structure as much as possible, so that it can provide the credibility gist for clinic diagnose and pathology research. How to extract the information of medical image exactly and automatically by computer to meet the demand of medical image disposal is the key problem of image analysis experts.Based on the analysis of domestic and overseas related literatures about medical image segmentation, the paper brings forward a medical image segmentation method based on MAS with Reinforcement Learning (RL) and Agent. Its main contents and innovation are as follows:(1) After analyzing home and overseas main medical image segmentation methods in depth, I summarize roundly the concepts, theories and methods of medical image segmentation, and also expound the feasibility and necessity of medical image segmentation method based on MAS aimed at the characteristic of medical image.(2) In Edouard Duchesnay's irregular pyramid method, we need to input the parameters [Dn, Ds], and the segmentation result related to [Dn, Ds] immediately, this can't achieve automatic segmentation fully. We improve the Edouard Duchesnay method using the dynamic and cooperative pyramid model. Based on the improved method we advance the medical image segmentation method based on MAS.(3) The key skill of medical image segmentation method based on MAS is the dynamic and cooperative pyramid model (DCPM). The model permits the cooperation behavior to local operation (region/edge). The dynamic and cooperative pyramid organization controls and constrains agent society so global constraints can be guaranteed.(4) We operate on the left ventricle MRI using the method of medical image segmentation based on MAS. The result indicates the method can achieve automatic segmentation without initial figure. The segmentation effect is basically consistent with the expert of medical image segmentation field.This paper obtains the undergraduate creativity sustentation fund of Jiangsu University.