Research On Pancreas Segmentation Method Based On Level Set

The pancreas is an important organ in the human body with endocrine and exocrine functions,but it is also susceptible to various diseases,especially pancreatitis and pancreatic cancer.Pancreatic segmentation plays a very important role in detecting pathology and providing quantitative analysis,and with the continuous advancement of digital medical equipment,pancreatic segmentation has become an important prerequisite for computer-aided diagnosis.However,the complex anatomical structure of the pancreas and the shape,size and location of the pancreas vary from person to person,and the pancreas is difficult to distinguish from the surrounding tissues and organs,resulting in borders touching.Therefore,how to accurately and effectively segment the pancreas is very important.This paper takes pancreatic CT images as the research object,uses the level set method to perform semi-automatic segmentation of the pancreas,and performs fully automatic segmentation based on the UCTrans Net model as the initialization.The research contents mainly include the following aspects:Data preprocessing.Understand the format of medical image storage.After reading the pancreatic CT image,perform grayscale enhancement and filtering and denoising as preprocessing.Adjust the appropriate window width and window level to modify the pixel value of the pancreatic CT image.noise,while preserving edge features and making the image clearer.Semi-automatic segmentation algorithm based on level set.In order to solve the problem that the level set method is sensitive to initialization,the region growing is used as the pre-segmentation of the level set method,and the initial region is extracted and used as the initial level set function.equation,a new balloon force is proposed to control the evolution process and direction,slowing down the evolution process in regions with weak or no margins,thereby solving the problem of the evolution contour leaking in regions with weak or no margins and deviating from the border of the pancreas.Automatic segmentation algorithm based on UCTrans Net and DRLSE.Since the initialization process of the level set method is difficult to avoid manual interaction and has human subjectivity,in order to solve the problem of manual intervention,an automatic pre-segmentation algorithm based on UCTrans Net is proposed.More detailed feature information,so as to grasp more information about the boundary,better capture the geometric information of the pancreas,learn more valuable features of the pancreas,and construct the initial level set function as the initial pre-segmentation of the level set,and finally pass The improved level set DRLSE algorithm can achieve more accurate automatic segmentation of the pancreas.In this paper,the performance of the segmentation method is verified by the public NIH pancreas dataset.The experimental results show that the proposed algorithm can effectively segment pancreatic CT images.