Research On Medical Image Segmentation Based On Level Set

Medical image is important for pathological diagnosis and scientific research. Accurate medical image segmentation can assist doctors quickly determine the patient’s condition, find lesions. However, because of the imperfection of imaging devices and the complexity of human tissue’s structure, often leading to seriously intensity inhomogeneity and fuzzy boundaries phenomenon in medical image, making image segmentation more complicated. Therefore, the segmentation algorithm based on medical image is the hotspot in the field of image segmentation currently.This paper based on level set method for medical image segmentation, has achieved the following results:(1) Focus on intensity inhomogeneity medical image, constructing energy function through image local area information, achieving intensity inhomogeneity image segmentation and bias correction. At present, the main medical image segmentation method is geometric active contour model based on level set method. The geometric active contour model can be categorized into two major classes:edge- based models and region-based models. Traditional region-based models such as CV (Chan-Vese) model can effectively segment weak boundary and discrete boundary problem, but can’t segment intensity inhomogeneity image accurately. The LBF (Local Binary Fitting) model proposed by Dr. Li effectively solve this problem, and further put forward a novel LIC (Local Intensity Clustering) model, this model not only has remarkable effect for intensity inhomogeneity image, but also can correct bias field, restore image to original real image. Based on this, focus on the phenomenon of intensity inhomogeneity of medical image, this paper proposed a novel segmentation and bias corrected method for medical image base on local information, to make the image intensity homogeneity. This method fitting original image by bias field and ideal image, using local region information of image, through fitting image and original image structuring energy function. It solved by variational level set method. The experimental results show that the method can effectively realize the medical image segmentation and bias correction, compared with other segmentation and bias correction method, this method has more precise and efficient in segmentation and bias correction.(2) In recent years, as the image segmentation problem more and more complicated, there are many targets in an image, so, this paper proposed a multiphase level set segmentation and bias correction method for medical image. This method using double level set structure four phase segmentation partition, through image region information fitting level set method, and Gaussian filtering the level set function at each iterative process. The experimental results show that the method can effectively segment the multi-objective image of intensity inhomogeneity, and can correction the bias field. Compared with other methods, this method is more accurate.