Research And Realization Of Liver Segmentation Algorithm For CT Images

Liver segmentation is the most important part of computer-aided diagnosis and planning of liver transplantation and partial hepatectomy.Precise extraction of liver contours and liver segmentation can help doctors judge the health of the liver,and perform surgical planning before surgery,to have a more clear understanding of the tumor to ensure that both can cut off the lesion and try to keep the healthy part of the liver.A segmentation algorithm based on the fast global minimization active contour model is proposed to segment the liver,combining the edge information and intensity information of the image.The energy function is constructed,decreasing from the initial contour curve to the position that the accurate liver contours is found.And then use the improved region growing method to extract the complete liver parenchyma.According to Couinaud segmentation method,an interactive method for liver segmentation is proposed.The liver is divided into seven segments ignoring the caudate lobe firstly.Reasonable landmarks are manually selected to form planes to segment the liver from the extracted hepatic vein tree and portal vein tree.Then the watershed method based on the distance map is firstly used to automatically segment the hepatic caudal lobe,and manual method is secondly used to make up the accurate results of the hepatic caudal lobe.And then the results of the hepatic caudal lobe assigned to the results of the seven segments of the liver,resulting in liver eight segments.The method is applied to test cases of abdominal CT datas,the test results show that basically meet the Couinaud segmentaion requirements,also adapted to the mutation cases.So in theory,the method has a strong feasibility.And the method is real-time,can be provided to the doctors to use.