Prostate Segmentation On CT Image Based On Level Set

Medical image segmentation is an important part of medical image processing. The effect of segmentation directly affects the availability of computer-aided diagnosis system. Although the new medical image segmentation algorithms have been proposed, but due to the special nature of medical images and the complexity of the algorithm often leads to universality has been limited. Therefore, CT images for prostate segmentation, we must conduct a separate study.Difficulties for prostate segmentation, this paper studies regional information-based level set active contour model, using the model method can outline the completion of segmentation, but require human jobs to locate the center of the prostate area on initial pelvic CT images, to assist the curve in evolution, does not implement automatic image segmentation.To achieve automatic segmentation of the prostate on CT images, the paper proposes a combination of shape and texture information using genetic algorithm and level set. The zero level set represents the level set segmentation contour. The traditional level set contour evolution model has extremely demanding for the selection and organization of the initial contour edge sharpness. Splitting away from the initial outline of objectives often to drive the evolution not split objectives because of lack of ability; medical image-specific local body effect also makes the deviation from the prostate boundary curve evolution. In contrast, use genetic algorithms to optimize the level set function without calculating the energy function and parameters and reduce the complexity of the algorithm. Make using of the manual segmentation by experts and PCA shape analysis model to obtain the average deviation of the shape and the average shape. Utilize texture information in different regions to establish Fisher criterion, and use genetic algorithms to implement the level set function evolution of segmentation with texture classification results. Experiments show that the proposed method is an accurate automatic image segmentation method.