Histopathological Image Segmentation Based on Probabilistic Graphical Model

The purpose of this paper is to design a method of pathological image segmentation, which can solve the problem of recognition and segmentation of conglutinated and partially overlapped objects on the background of complex texture in pathological image.In this paper a mixed probabilistic graphical model was constructed which is suitable for pathological image segmentation. By using undirected graph and directed graph, this paper built statistical models of the probability dependence of nuclear position, contour on region and edge information, and also researched the parameter estimation method in the graphical model, designed and implemented the pathological image segmentation algorithm based on the probabilistic graphical model. The experiment shows this method improves the accuracy of recognition of nuclei in pathological images.