Research On Pancreatic Image Segmentation Based On Level Set Method

Pancreas is an important organ in human body.With the popularization of digital medicine,computer-aided diagnosis has become possible.The segmentation of pancreas with the help of computer can greatly improve the efficiency of doctors.However,the pancreas in medical CT images shows the characteristics of small size and deep location,and the pancreas and the surrounding liver,kidney and other abdominal organs do not have a good gray contrast,making the edge of the pancreas in CT images is not obvious.So it is difficult for computer to segment pancreas accurately with the help of traditional image segmentation methods.This paper studies the level set method for pancreatic medical image segmentation.The main work is as follows:(1)The traditional level set segmentation model(CV model)based on global image information needs to be re-initialized to ensure the stability of curve motion.In order to improve the efficiency of curve operation in CV model,an improved CV model is proposed,and a curve area term is fitted in the traditional CV model.Experiments on gray level uniform image segmentation by improved CV model show that the improved CV model proposed in this paper can indeed improve the speed of curve motion in CV model.(2)For the traditional level set segmentation model of global and local information(LGIF model),it is necessary to adjust the weight of global and local information items in order to obtain better segmentation results.In order to realize adaptive segmentation of LGIF model,an adaptive function is proposed to measure the weights of global and local information items in real time.The validity and rationality of the ALGIF model fitting adaptive function are verified by the segmentation experiments of pancreatic medical images and synthetic images.(3)To avoid reinitialization of LGIF model level set functions.In this paper,an improved RD-LGIF model is proposed by combining LGIF model based on local search strategy with reaction diffusion method.Experiments show that the improved RD-LGIF model can effectively segment pancreatic medical images.