Research On Anklebone Images Segmentation Based On Active Shape Models

The development and popularization of the physical equipment provides more evident for the clinic diagnosis. However, what the equipment can provide is just images, not particular information dug from the data. With the help of image segmentation technique, we can distinguish the organ from the image, based on which, the physiological characteristics and functional information of the whole organization can be gained. Furthermore, the possible positions of pathological focus may be analyzed to assist doctor to diagnosis. Therefore, it is also the hotspot of research.Deformable model is one of the most popular images segmentation methods. The past thirty years witnessed the grows of the model technology, however, most of them are not suitable for the medical images segmentation. Because the active shape models (ASM) can make full advantage of prior knowledge of medical images training set, it can be applied to the segmentation of the medical images based on the anatomic signification. The ASM was applied to the segmentation of the anklebone images in this dissertation. Firstly, we set up statistical shape models of anklebones using the training sets of medical anklebone images. The experimental results showed that the ASM method presented in this dissertation has higher accuracy compared with former methods. Furthermore, it can achieve excellent segmentation effect in noisy position, and validate the validity and feasibility of the anklebone images segmentation based on ASM. Secondly, we found some problems arose in the process of statistical modeling including inaccurate mean-shape and premature partly iteration and disorder of shape. With the help of global searching characteristic of genetic algorithm introduced in this dissertation improved the original shape, analyzed every fitness function which was used to reform the original shape. Meanwhile, it added the sub-shape segmentation to shape global segmentation. Finally, we adopted adjusting the global shape transformation to optimize the whole shape segmentation by controlling transformation of sub-shape.The research of the dissertation indicates that ASM can effectively realize the 2-D medical images segmentation. It is also the powerful basis of theory and practice for the other organ image segmentation and 3-D modeling.