Study On Medical Image Segmentation Method Based On Snake Model

Medical image segmentation is the base of medical image processing and analysis, it is the basement of other medical imaging post-treatment (such as three-dimensional visualization, surgical navigation, etc.) it is also the principal step in computer vision,therefor many researchers pay close attention to it in recent years. Parametric active contour models (or snake models) are very suitable to medical image segmentation and contour extracting due to its highly dynamic quality.First, a survey on purpose and significance of image segmentation, the background and status, and parameters of active contour model segmentation method are reviewed in this thesis. Second, the discourse introduces the basic principles of the traditional Snake model, the relationship between the model and variations, and the simulation results of the algorithm. Then, the discourse describes the balloon model, the potential model, expatiate on GVF Snake and their simulation experiments are given.In connection with the problem of snake model, such as capturing region limited, weak or leakage detected of the boundary. Starting from the minimum energy function, and a novel external force for snakes named Directional Preserving Gradient Vector Field is proposed in the thesis. The method is validated on both phantom and clinical images. The experimental results show that the method largely prevents interference from the initial contour to the border closely and weak boundary, leakage of the border clearance and other issues. Additionally, the DPGVF Snake has larger capture ranges and better robustness to initialization and it has better segmentation results for complex medical images.Due to the diversity of medical images which make them have gray uniform, noise and other issues, when the position of initial contour is not suitable, it may not get the accurate and stable segmentation results. The the concept of neural network is inducted so that it can locate the position of initial contour around the boundary directly.Experiments show that the method not only make the time of evolution shorter. but also has stable performance even in the case of strong noise, it is more applicable to the field of medical image segmentation.Finally. the research summary are conducted and the work in future prospect are given.