肺实质CT图像分割方法  

Segmentation method for CT image of lung parenchyma

作　　者：崔颖[1] 王小鹏[1] 于挥 陈璐[1] 

机构地区：[1]兰州交通大学电子与信息工程学院,甘肃兰州730070

出　　处：《计算机工程与设计》2015年第5期1274-1278,1310,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61261029);金川公司预研基金项目(JCYY201309)

摘　　要：针对肺部CT图像中因各组织灰度不均匀、结构复杂等因素造成双肺边界难以准确分割的问题,提出一种多阈值和基于投影的标记控制分水岭分割方法。运用多阈值法对图像进行粗分割,结合形态学开运算去除图像中残余的气管与主支气管;对于左右肺区未完全分离的情况,采用基于投影的标记控制分水岭分割方法进行分离;利用形态学开闭运算对粗分割结果进行细化。实验结果表明,该方法能够对肺实质进行较准确的分割。Aiming at the problems that segmentations are confined by the intensity in homogeneity within different organizations, complex structures and other factors in lung CT image,and it is difficult to segment lung boundary exactly.To improve the seg-mentation results,an integrated method of multi-threshold and marker controlled watershed segmentation based on projection was proposed.Firstly,multi-threshold method was used to solve image coarse segmentation,and morphological opening opera-tion was combined to eliminate the residual trachea and main bronchus of image.Then the method of marker controlled water-shed segmentation based on proj ection was used to solve uncompleted separation of left and right pulmonary area.Finally,mor-phological opening and closing operators were utilized to refine the coarse segmentation.Experimental results show that using this method can segment pulmonary parenchyma accurately.

关 键 词：肺实质 计算机辅助诊断 CT图像 多阈值 图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]