Research On Segmentation Method For Medical Images Based On Level Set

Image segmentation is the theory basis of medical image analysis, registration, fusion, classification, retrieval and the implementation of computer-aided diagnosis system in clinic. Medical image segmentation is an important research topic in the area of medical imaging, and it also plays a significant role in accurate diagnosis of disease.Aiming at increasing the accuracy of segmentation results of two types of medical imaging which are multi-gray scale and non-homogeneous or uniform image, a series of studies on medical image segmentation is accomplished based on the status quo of image segmentation and basic theory of level set method in this paper.Different parts of images present different gray-scale as the structural complexity of medical images, which can easily cause segmentation errors. In view of the errors caused by the multi-gray scale image segmentation, a method of fuzzy level set segmentation based on image edge information and regional information is proposed in this paper. This algorithm combines the advantages of fuzzy C means clustering model, segments the target according to gray level and improves the accuracy of segmentation. Meanwhile, the energy function of level set is improved in this algorithm to eliminate the re-initialization process, and to resolve the problems such as great amount of calculation and the complexity in process caused by re-initialization. Experimental results in MATLAB simulation environment are compared and evaluated, which shows that the segmentation effect of image borders of deep depression and the accuracy of the multi-gray scale medical image segmentation are improved by the proposed medical image segmentation of fuzzy level set with boundary and region information, which means that the algorithm is practical and effective.The errors of image segmentation could also be caused by the non-homogeneity of brightness of images which caused by the differences of experimental equipment and imaging methods. A hybrid kernel function method for medical images based on level set is put forward in this paper to solve segmentation error caused by non-homogeneity medical images.The algorithm improves the level set method by adding hybrid kernel which can be sustained on the non-homogeneity and makes the pixels classification more accurate. Meanwhile, the image segmentation process takes both image local features and global features into consideration, which improves the speed and accuracy of segmentation. The improvement of segmentation accuracy of non-homogeneity medical image is proved by experimental comparison and differences evaluation.Finally, the research is summarized to present the results, innovation points and the future trends of medical image segmentation.