胸部CT图像肺区域边界凹陷自动修补

摘要

After the lung regions are segmented in chest CT images, there exist some concavities or gaps on lung borders. The scheme of connection between boundary curve local minimum points is proposed to repair the lung boundary for gaining the entire lung parenchyma segmentation. Different matching templates are selected to detect the local minimum points in the border curve under different coordinates in order to determine locations that need to repair. Two contiguous local minimum points are connected across the concavity to repair the segmented lung border flaw. The experimentation shows that the scheme in this paper reduces the fault segmentation and enhances the efficiency of lung parenchyma segmentation efficiently.%提出用边界曲线局部极小值点连线法修补阈值分割后的CT横断面图像肺区域边界处血管和胸膜结节型凹陷。设置不同的匹配模板，实现在不同坐标系下寻找边界曲线的局部极小值点，从而确定需要修补的位置。将肺区域边界线上的点分为局部极小值点和非局部极小值点两类，连接凹陷缺口处两边的两个邻近局部极小值点，修补肺边界凹陷部分。实验表明，与通过计算边界点的曲率找边界凸点方法相比，该方法不仅有效地降低了肺实质分割的计算量，而且减少了由于过度分割造成的分割错误。