Study On Segmentation Method For Medical Images Based On Level Set

Medical image segmentation is a traditional and challenging project, and it is the basis of medical image visualization, computer-aided analysis and disease diagnosis. The medical image segmentation method based on level set has many advantages on medical image segmentation. Therefore the medical image segmentation method based on level set is focused in this thesis.The level set method translates the problem of evolution of 2-D(3-D)close curve(surface) into the evolution of level set function in the space with higher dimension. As an accurate and steady algorithm, the level set method has a wide application. A deep study is made in this thesis and it covers four parts which is organized as follows:Firstly, the basic theory on level set method is systematically researched, and its application in image segmentation field is discussed and proved through experiments, which lays a solid foundation for the following research.Secondly, the boundary-based level set methods and the region-based level set methods are researched. After their advantages and disadvantages are analyzed through experiments, a novel hybrid segmentation method integrating boundary and region information of images is proposed in this thesis, and it is applied to segment brain MRI in the IBSR (Internet Brain Segmentation Repository) and real clinical medical image. The effectiveness and accuracy of the hybrid segmentation method are proved through the experimental comparison and differences evaluation.Thirdly, a hybrid segmentation method integrating improved prior shape is proposed after summarizing the drawbacks of the hybrid segmentation method integrating boundary and region information of images. Experimental results for real clinical images shows that the hybrid segmentation method integrating improved prior shape is superior to the hybrid segmentation method integrating boundary and region information of images.Finally, the summary and the prospect on this study t are given.