多阈值和标记分水岭相融合的肺部CT图像分割方法  

Segmentation method of CT lung image based on multi-threshold and marker watershed algorithm

作　　者：肖雪[1] 龚恒[2] 陈钟[2,3] 

机构地区：[1]重庆电子工程职业学院,重庆401331 [2]北京大学信息科学技术学院,北京海淀区100871 [3]北京大学网络与软件安全保障教育部重点实验室,北京海淀区100871

出　　处：《激光杂志》2014年第9期74-78,共5页Laser Journal

基　　金：国家自然科学基金(61170263)资助

摘　　要：针对肺部CT图像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,提出了一种多阈值和标记分水岭相融合的肺部分割方法。首先采用多阈值法对肺部CT图像进行粗分割,并去除图像中气管与主支气管;然后采用标记控制分水岭方法进行精分割,并利用形态学运算对肺实质边缘修补,最后采用临床肺CT图像在Matlab 2012平台上对算法性能进行仿真测试。结果表明,本文方法可以较好保留肺部CT图像的边界信息,提高了肺部CT图像分割精度,误分和错分概率大幅度下降,取得了十分理想的分割结果,为肺部疾病临床医学诊断提供了有价值的参考信息。Since segmentation is confined by low conditions such as intensity non-homogeneity within different orga-nizations, complex structures and other factors in lung CT image, it is difficult to segment lung boundary exactly. To improve the segmentation results, an integrated method of multi-threshold and marker controlled watershed segmen-tation is proposed. Firstly, multi-threshold method is used to solve image coarse segmentation and eliminate the resid-ual trachea and main bronchus of image, and then marker controlled watershed segmentation method is used to fine segment and morphological operations is used to repair lung edge, finally clinical CT cardiac images are used to test the performance on matlab 2012 platform. The experimental results show that, this proposed method can better pre-serve the image edge information, effectively improve the accuracy of lung CT image segmentation, has obtained the ideal segmentation effect, it can provides valuable reference information for the clinical diagnosis of pulmonary dis-eases.

关 键 词：肺部图像 多阈值 图像分割 记分水岭算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]