摘要
高精度的肺组织分割是肺功能定量分析的前提和基础,有利于慢性阻塞性肺病的辅助诊断.传统的肺组织分割方法没有去除肺轮廓内部的细支气管,且无法处理严重粘连的肺气肿病例.本文提出一种全自动的三维肺组织分割方法:首先采用带有错误检测机制的二维图像阈值选取和三维区域增长进行粗分割,得到肺气道和肺实质组成的肺部充气区域;然后结合阈值和气管形态分析分割肺气道树,在防止泄露的同时提取更多肺内细支气管;最后通过扫描线粘连定位和动态规划,实现前、后联合粘连的定位及左右肺分离.实验中分别采用20组EXACT09数据和20组VESSEL12数据对所提出的方法进行评价,结果显示提取的气道树平均分支数为131个,与医生标记的金标准比较分支检出率为54.09%;肺实质分割结果与数据集提供肺轮廓M ark比较,Jaccard系数为95.35%,平均绝对边界距离为0.89mm.结果表明,本文方法能够有效的提取肺组织,运行时间与传统方法相比在临床实际应用中具有一定优势.
High precision segmentation of pulmonary tissue is the premise and foundation for quantitative analysis of pulmonary function,which is useful to aided diagnosis of chronic obstructive pulmonary disease( COPD). Traditional methods do not remove the bronchioles inside the lung contours,and fails in cases with severe adhesion of emphysema. A fully automatic 3D pulmonary tissue segmentation method is presented in this paper. First,a rough segmentation is performed using a threshold selection with error detection mechanism in 2D image and 3D region growing and pulmonary airspace is gained which consist of pulmonary airways and pulmonary parenchyma. Then airways tree is segmented based on thresholds and morphological analysis,with more bronchioles extracted within the lung and leakage prevented. Finally,through adhesion positioning based on scanning line and dynamic programming,the anterior and posterior adhesions are located,and thus the left and right lungs are separated. 20 sets of EXACT09 data and 20 sets of VESSEL12 data were utilized to evaluate the proposed method. The average branch count of airway tree was 131,compared with the gold standard marked by doctors,the branch detection rate was 54. 09%; pulmonary parenchyma segmentation results were compared with lung contour mask that dataset provides,the average Jaccard coefficient was 95. 35%,and the average absolute boundary distance was 0. 89 mm. Results showthat the proposed method is able to extract pulmonary tissue effectively,and running time has some advantages in clinical practice compared with traditional methods.
出处
《小型微型计算机系统》
CSCD
北大核心
2016年第3期581-587,共7页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61172002
61302012)资助
国家科技支撑计划项目(2014BAI17B01)资助
国家"八六三"高技术研究发展计划项目(2012AA02A607)资助
中央高校基本科研业务费(N130418002
N140402003
N140407001)资助
关键词
肺组织分割
肺实质分割
肺气道分割
肺功能定量分析
pulmonary tissue segmentation
pulmonary parenchyma segmentation
pulmonary airways segmentation
quantitative analysis of the pulmonary function