三维网格引导的癌变病人CT图像的腹腔分割  

Mesh-Based Segmentation of Abdominal Cavity for Patient with Cancer in CT Images

作　　者：卢涤非[1,2] 叶修梓[3] 吴吟 蔡文立 

机构地区：[1]浙江警察学院计算机与信息技术系,杭州310053 [2]Department of Radiology,Massachusetts General Hospital and Harvard Medical School [3]温州大学计算机学院,温州325035

出　　处：《计算机辅助设计与图形学学报》2015年第6期1017-1023,共7页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金面上项目(61272304);浙江省中青年学科带头人学术攀登项目(pd2013435)

摘　　要：针对癌变CT图像中各器官变形严重,传统分割算法无法有效、完整地分割整个腹腔的问题,提出基于三维网格的分割算法.首先借助肋骨和脊椎等骨架获得初始腹腔骨架;然后构造一个球形初始三角网格,并建立与腹腔骨架间的关联;再通过3个目标函数变形网格;最后在网格附近的边界点优化网格,获得腹腔分割结果.实验结果表明,该算法能够有效地对病变严重的CT图像进行腹腔分割,对噪声和器官变形有良好的鲁棒性.The organs with cancers in CT image deform seriously and the traditional segmentation methods cannot segment the entire chest correctly. This paper presents a 3D mesh based segmentation method. Firstly, an initial chest skeleton is found using ribs and back bones; secondly a sphere triangle mesh is constructed and the correspondence between the mesh and the chest skeleton is created; thirdly the mesh is deformed according to three objective functions; finally the mesh is refined by edge voxels near the mesh and the segmentation result is refined. Experimental results show that the proposed method can efficiently segment the seriously-deformed chest in CT image, which is robust against the noise and deformation of organs.

关 键 词：胸部CT图像 三维图像分割 网格变形 仿射变换 图像边界 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]