对角线剖面分析引导的全自动肺CT图像分割  

Diagonal profile analysis induced fully automatic lung CT image segmentation

作　　者：侯庆锋[1] 

机构地区：[1]泰山医学院放射学院,山东泰安271016

出　　处：《中国医学影像技术》2014年第11期1734-1738,共5页Chinese Journal of Medical Imaging Technology

基　　金：山东省医药卫生科技发展计划项目(2013WS0315);山东省自然科学基金项目(ZR2011HL027)

摘　　要：目的提出一种基于对角线剖面分析的全自动肺CT图像分割方法。方法首先构造待分割图像的对角线剖面图,然后自动分析该剖面,得到各组织结构的位置信息和像素值信息,引导区域增长算法分离患者身体和背景,再利用灰度阈值算法分离胸壁与肺区,再用数学形态学算法修正肺边缘,得到肺区掩模图像,最后利用肺区掩模图像与原图像运算提取肺区。结果利用从不同数据库选取的51幅CT图像检验提出的方法,所得结果与诊断医师手工分割结果进行比较,计算重叠率指标(OR),最低OR为95.86%,最高OR为99.25%,平均OR为97.85%。结论对角线剖面分析方法能高效地实现全自动肺CT图像分割。Objective To propose a fully automatic lung CT image segmentation method based on diagonal profile analysis.Methods Firstly,the diagonal profile was constructed,then an analysis program was implemented and the coordinates and pixel values of pixels corresponding to different anatomy structures were determined.Region growing algorithm was used to separate patient＇s body area and the background using the results of diagonal profile analysis.Then grey level threshold was used to separate thoracic wall and lung fields.A series of mathematic morphology operators were used to correct the lung contours and construct lung field mask.Finally,lung field was extracted from the lung field mask and raw image.Results Totally 51 CT images were used to test our proposed algorithm and the results were compared with those segmented by experienced radiologists manually.The overlap ratio（OR）was used to evaluate the performance.The minimum OR is 95.86%,the maximum OR is 99.25%and the average OR is 97.85%.Conclusion The results illustrated that our proposed diagonal profile analysis based algorithm can segment lung CT image effectively.

关 键 词：肺分割 计算机辅助检测 对角线剖面分析 

分 类 号：R814.42[医药卫生—影像医学与核医学]