从CT图像中自动分割出肺部区域的算法研究  

Research on Algorithm for Automated Lung Segmentation in CT Images

作　　者：吴龙海[1] 周荷琴[1] 张鹿[1] 李传富[1] 

机构地区：[1]中国科学技术大学自动化系,安徽合肥230027

出　　处：《航天医学与医学工程》2008年第5期425-429,共5页Space Medicine & Medical Engineering

基　　金：安徽省教委重点课题基金资助(2006KJ097A)

摘　　要：目的为用于肺部疾病的计算机辅助诊断,研究设计从CT图像中提取肺部区域的自动分割算法。方法在最优阈值分割的基础上,用自动区域生长去除气管/支气管区域,对边界跟踪法进行改进以快速去除背景干扰和获得肺部边界,最后进行肺部边界修补得到完整的肺部图像。算法采用迭代法寻找最优阈值解决了阈值选取的敏感性问题,提出了基于前层图像中气管/支气管位置的气管/支气管提取方法,避免了种子点的人工选取,基于前次搜索方向改进了八邻域搜索方法来提高边界跟踪的速度。结果用该算法对不同病人的4组胸部CT序列进行处理,能自动、快速地分割出肺部区域且精度较高。结论提出的算法能有效地从CT图像中自动提取肺部区域。Objective To design an automatic segmentation algorithm for lung region abstraction from CT images in computer-aided diagnosis（CAD） of lung diseases. Methods Based on the optimal threshold segmentation, an automatic region-growing method was adopted to eliminate the trachea and bronchi, the boundary tracking algorithm was modified for background elimination and lung boundary abstraction. Then, lung boundary repair was performed to obtain a fine boundary. To reduce the sensitivity of threshold selection, an iterative process was employed to find the optimal threshold. A trachea/bronchus extraction method based on position of trachea/bronchus in previous slice was introduced, which avoided selecting seed-point by handle in region-growing. Based on previous searching direction, 8-neighborhood searching method was improved to increase its efficiency. Results Experiments with four chest CT data sets showed that this algorithm was able to abstract the lung region automatically, quickly and with better precision. Conclusion The proposed algorithm is quite efficient for automated lung segmentation in the computed-aided diagnosis of lung diseases.

关 键 词：图像分割 边界跟踪 肺部区域 CT成像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] R445.3[自动化与计算机技术—计算机科学与技术]