基于交互式图割算法的结肠组织提取  

Colon Tissue Extraction Based on Interactive Graph Cuts Algorithm

作　　者：苗语[1] 张丽媛[1] 杨华民[1] 闫飞[1] 赵建平[1] 师为礼[1] 蒋振刚[1] 

机构地区：[1]长春理工大学计算机科学与技术学院,长春130022

出　　处：《长春理工大学学报（自然科学版）》2014年第6期99-102,共4页Journal of Changchun University of Science and Technology(Natural Science Edition)

摘　　要：针对腹部CT医学图像中结肠组织自身局部特征的问题,研究了一种基于交互式Graph Cuts结肠组织的分割方法。首先人工标定一部分体素标记为"目标"和"背景"种子点。然后将图像映射成网络图,通过相邻像素间的灰度特征分配边的权重值,采用26邻域系统实现三维图像的分割。最后使用最大流/最小割方法最小化能量函数,得到结肠区域。实验结果表明,交互式的图割算法能够准确地从三维腹部CT医学图像中提取出结肠组织,体现了结肠数据的局部特征。算法能自动将所有腹部CT切片的结肠组织分割出来,实验获得的结果有利于结肠病变的发现和精确定位。In view of local features of colon tissue in the abdomen of CT medical image problem,this paper uses a segmentation method based on interactive Graph Cuts.First,we mark certain voxels as＂object＂and ＂background＂seeds.Then,the image is mapped to a network graph,the weight values of edges are distributed through between adjacent pixels gray features,using the 26 domain system to achieve the 3D image segmentation.Finally,using the max-flow/min-cut method to minimize the energy function,getting the colon region.The experimental results show that,the interactive graph cut algorithm can extract accurate colon tissue from 3D CT medical abdominal images and reflect the local features of colonic data.This algorithm can effectively separate the colon tissue of all abdominal CT section,The obtained experimental results benefit the identification of colon lesions and precise positioning.

关 键 词：三维医学图像分割 图割 结肠组织 灰度特征 最大流最小割 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]