The Research On Interactive Segmentation In MR Brain Images Based On Snake Model

Medical image segmentation is the foundation of image 3D reconstruction, structure analysis, motion analysis and other after operation of natural tissues and pathological tissues. The veracity of segmentation is very important for the doctor to get actual instance and make right diagnosis. Due to the complexity of anatomical structure, the scrambling of organ and the otherness between individuals, together with that the quality of imaging is restricted by multiplicate factors, make medical image segmentation a difficulty. There are inhomogeneity, constructed defect, weak boundary, boundary rupture etc. in MR image, so the segmentation result of traditional methods are not very well. Currently, medical image automatic segmentation draws continuous attention. At the same time, interactive segmentation method has become the research emphasis of medical image segmentation.Snake model is a main trend used for object extraction in image process recent years. It provides a very flexible tool with various restrict mechanisms. By analyzing the rationale of Snake model and expounding the mathematic model and numerical solution method in detail, this paper develops and applies the technique of Snake for the research of interactive segmentation method in brain MR image. Aim at the disadvantages of traditional Snake model, this paper proposes an improved Snake model based on greedy algorithm. This model widens the capture range of exterior energy, so the result of segmentation doesn't relying on the initial contour. And the improved algorithm also improves the capability for concave regions. Based on this, this paper analyzes the application of gradient vector flow in the improving of Snake model and a new produce method of edge map based on Canny operator is presented in order to solve the problems in practical application in brain MR image segmentation. And an improved GVF with normalized the force vectors in the diffusion equations and added exterior sanction energy is introduced. The proposed algorithms have good performances.Aimed at the continuity of slices, this paper uses the segmented result of monolayer slice as initialized contour of neighbor slices, and achieves satisfying segmentation of multilayer serial images. This paper discusses the specific application of improved Greedy Snake algorithm, GVF Snake model and the improved algorithms in brain MR image separately, and in the end analyzes the experiment result and choosing of parameters.Based the research of above methods, by integrating characteristics of the Snake model, this paper designs and develops an interactive image segmentation system with C++ language. In order to realize the local interactive rectify to the result of segmentation friendly and ensure the veracity of segmentation, Bezier curve is adopted to draw the initial contour of Snake model and implement interactive control of the model under the GDI+ environment. This system not only can be applied to brain MR image, but also can be applied to segment other medical images (such as CT, PET etc.) interactively. Experiment results demonstrate that it is fast, effective, and has good repeatability and also it's easy to manipulate and has high precision and practical value in the medical image analysis.