Research On Liver Segmentation Algorithm Based On Multiphase CT Images And Construction Of Visualization Platform

Liver cancer is one of the most common malignant tumors.It has the characteristics of high morbidity and high mortality,and has become a major killer of human health and life.The current treatment methods for liver cancer are mainly interventional resection,but inaccurate surgical resection will increase the risk of recurrence of the disease and bring greater pain and burden to patients.With the concept of precision medicine and the advancement of medical information technology,3D visualization technology is gradually applied to medical image analysis.By performing three-dimensional reconstruction of the two-dimensional liver CT sequence,the physician can accurately and quantitatively analyze the condition of the lesion,and can design and optimize the surgical scheme on the virtual liver model,ultimately achieving the goal of improving the accuracy and reliability of clinical surgical treatment.In this paper,the algorithm of image segmentation and the construction of software platform in the process of 3D visualization of liver are studied:The algorithm of liver segmentation in plain scan image is studied.By using support vector machine to sift out the invalid slices in the sequence,the fuzzy c-means cluster algorithm with spatial constraints and morphological reconstruction method are used to realize the automatic segmentation of liver contour.A post-processing method based on convex hull detection and nearest neighbor connection is proposed.The results show that the proposed method can segment liver CT image sequence without human interaction,with an average segmentation accuracy of 92.8%and an average segmentation time of 1.7s.The segmentation algorithm of intrahepatic vessels in venous phase images is studied.A three-dimensional region growing method with moving seed and dynamic threshold is proposed.In view of the difficulty of seed point picking and the difficulty of fine structure resolution,Hessian matrix is used to filter and enhance the vascular structure.Using this method to segment blood vessels can achieve 93.2%segmentation accuracy,and the average segmentation time of a single image is 0.53s.The results of 3D reconstruction show that this method can effectively enhance the spatial continuity of the segmented blood vessels.The segmentation algorithm of intrahepatic lesions in arterial phase images is studied.This paper analyzes the advantages and disadvantages of the two active contour models based on boundary and region,and proposes a hybrid active contour model based on local region fitting and gradient information.The experimental results show that the active contour model can effectively segment the liver cancer and intrahepatic nodular hyperplasia lesions with the features of uneven gray level and fuzzy edge contour.Three dimensional visualization platform of liver.C++is used to program and implement the algorithm studied,and the three-dimensional visualization platform is built by integrating the above results.The functions of image sequence reading,browsing,liver parenchyma segmentation,intrahepatic vascular segmentation and three-dimensional reconstruction are preliminarily realized.