Novel Level Set Approach For Medical Image Segmentation Based On Local Region Information

Medical image segmentation is one of the key steps of medical image analysis, efficient and accurate medical image segmentation take a crucial position in reconstruction of human anatomy, treatment planning, identification and diagnosis of lesions and graphics to guide the operation. It has important research value and practical significance for biomedical field. This dissertation focuses on the study of accurate and efficient segmentation method for medical image with intensity inhomogeneity.This dissertation describes the traditional image segmentation methods, study the medical images'imaging principles and image characteristics, and further discussion on level set image segmentation methods. In traditional medical images, there are significant changes in the gray of some human tissues; and some regions of different human tissues have the same gray scale. It is named intensity inhomogeneity. Chan-Vese proposed a level set segmentation method based on simplified Mumford-Shah model. It is able to handle the low contrast, blurred edges and have a good anti-noise performance, but it does not properly segment the medical images with intensity inhomogeneity. So this dissertation proposed a novel level set segment method with local region information. Firstly, we apply the image's local region information to construct a novel energy function. Secondly, the penalty-function is introduced to our method, therefore completely eliminates the need of the costly re-initialization procedure. The novel level set method utilize the local region information, it can segment the medical images with intensity inhomogeneity. The experimental results of MR images, blood vessel angiography images and X ray images show that the method is practical.