Research On Medical Image Segmentation Based On Active Shape Model

Medical image segmentation is the basis of medical image processing and analysis. In recent years, the active shape model with prior information has become an important method for medical image segmentation because of its stability and validity. In this paper, firstly we introduce several traditional methods of medical image segmentation, and design some experiments to analysis the advantages and disadvantages of each method. For the complexity of 3D medical image segmentation, we deeply study the method based on active shape model, using statistical principle to establish the average shape and varied models of the target. At last, we propose an improved segmentation method based on auxiliary localization of high contrast organs. It can be more efficient and accurate to complete the automatic segmentation of organs and lay the foundation for establishing organ automatic segmentation system in the future. The main work of this paper is as follows:(1) To solve the complexity and diversity of 3D medical images and uncertainty of the object shape, in this paper, we extract prior information of all training sets, use principle component analysis to establish the active shape model and search for the tested sample iteratively to accomplish the segmentation. In this paper, the interested organ of the mice was segmented with this method. The experimental results show that, dynamic active shape models can segment the organs efficiently and automatically. While ensuring the accuracy of the segmentation, it avoids the complex of 3D volume data processing and shows good validity.(2) In the segmentation method based on active shape model, there is a problem that the organs are not easy to locate, especially those with no obvious geometrical features. It can easily affect the accuracy of segmentation. So we propose an improved segmentation method based on optimized initialization localization. According to the correlation between the location transformation of high-contrast and low-contrast organs, the boundary contour of the latter is initialized quickly and accurately based on the former. Then use the active shape model to optimize the segmentation. In this study, the interested organ of the mice was segmented with this method. The experimental results show that compared with the traditional method, this method can automatically find the initial position of the organ and accomplish automatic segmentation more efficiently. At the same time, the accuracy can also be improved significantly. It has good universality and can be used for segmentation of multiple organs systematically.