基于欧氏距离变换的肺部粘连肿瘤分割算法  

Segmentation of attached tumor in lung based on Euclidean distance transform

作　　者：彭江[1] 张建州[1] 喻忠全[1] 

机构地区：[1]四川大学计算机学院,四川成都610065

出　　处：《计算机工程与设计》2008年第21期5516-5518,共3页Computer Engineering and Design

基　　金：国家自然科学基金项目(60371024)

摘　　要：以肺部CT图像为研究对象,针对肺部粘连肿瘤图片本身的特点,提出了一种基于边缘跟踪的二维欧氏距离变换算法。从目标区域的最外层边界开始,自外向内对目标区域进行边缘跟踪、腐蚀,直到肿瘤区粘连部分与肺部边界分离。算法能够计算精确的欧氏距离。通过实验分割出的肿瘤和放射科医生手工勾画的肿瘤轮廓对比,5幅病例图像重叠率达到了75%左右,实验结果表明该方法对于中晚期肺部粘连肿瘤的分割有一定的效果。Targeting at lung CT image, aiming at the basis characteristics of attached tumor, a 2D Euclidean distance transform algorithm based on contour tracking is presented. Beginning with the outest layer, the algorithm tracks and erodes the contour of the object from outer to inner, until to the division between the zone ofattached tumor and the contour oflung. This algorithm produces perfect Euclidean distance. By comparing the segmentation tumor by experiments and the tumor outlined by radiation doctors, the overlapping rate of five cases images can be around 75%. The experimental result proves the certain effect of applying this method to the segmentation of the attacked tumors at the medium or advanced stages.

关 键 词：肺 粘连肿瘤 欧氏距离变换 分割 腐蚀 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]