用于计算机辅助诊断的肺实质自动分割方法

Automated lung segmentation for chest CT images used for computer aided diagnostics

在肺结节的计算机辅助诊断过程中,肺实质的分割是关键的步骤。传统的肺部分割方法都是基于肺实质与周围组织对X线的衰减强度不同而进行的。这些方法对正常的肺部能得到理想的结果,但当肺部存在肺结节等异常征象时,这些方法会出现错误。该文提出了一套完整肺实质分割流程,首先用阈值法和边界跟踪算法得到初始的肺部轮廓,然后提出一种新的基于计算局部二维凸包的方法对原始的肺部轮廓进行修正。该算法能将与肺部周围组织相连的肺结节包括在肺实质中,从而在肺部存在与胸膜相连的结节的情况下也能得到满意的分割结果。采用该算法对6个病人的约400张肺部CT图像进行了肺实质提取,实验结果显示:该算法对正常和异常征象的肺部CT图像进行分割的正确率均能达到83%以上,为肺结节的计算机诊断提供了良好的条件。

Segmentation of lungs in chest computed tomography（CT） is often performed as a preprocessing step for computer aided diagnosis（CAD） for lung nodules.Traditional methods are based on threshold strategies which often incorrectly exclude some very important regions such as juxtapleural nodules and blood vessels when some abnormalities exist in the lung.This paper presents a method for lung parenchyma extraction including the initial border extraction,the modification of the initial border based on a 2-D convex hull and the final parenchyma extraction.Experiments on over 400 clinical chest CT images of 6 patients show that the segmentation accuracies are all above 83% in both normal and lung cancer cases.This segmentation method effectively includes all juxtapleural nodules and blood vessels in chest CT images for detection of lung nodules in CAD systems.