三维肺部图像中疑似肺结节结构的自动提取

Auto Extraction of the Structures Similar to Pulmonary Nodules in Three-Dimensional Chest Images

本文提出了一种肺部CT图像三维数据中自动提取疑似结节区域的方法。首先结合闽值分割、种子填充等方法,在三维体数据上分割出肺实质。进而利用改进的模糊C均值聚类,提取出结节及具有结节特征的血管、支气管等感兴趣区域。该工作对感兴趣区域的特征提取有重要意义,是早期肺癌计算机辅助诊断重要的一步。

In this paper, we present a method for the automatic extraction of the pulmonary nodules or the structures like pulmonary nodules in three-dimensional （3-D） chest thin-sectlon images of computed tomography （CT）. Firstly, by combining global-threshold segmentation, mathematical morphology and seed fill algorithm together, the pulmonary parenchyma is segmented from the images. Then we improve the fuzzy c--means （FCM） algorithm. And with the algorithm we extract the nodules, vessels and bronchial tree which are the region of interest （ROI）. The work is significant to get the characteristic of ROI. So it is an important stage of subsequent Computer-Aided Diagnosis （CAD） for early detection of pulmonary cancer.