基于圆形约束C-V水平集的肺部CT图像病灶分割

Pulmonary Nodule Candidates Segmentation Based on Circle Dependent C-V Level Set for CT Images

针对肺部CT图像中圆形病灶区域的分割问题,对Chan-Vese水平集图像分割方法进行了分析和改进,提出了基于圆形约束的C-V水平集模型,进而提出了基于圆形约束水平集的肺部图像病灶分割算法,解决了图像中大小不同的多圆检测问题。对合成图像和实际临床肺部CT图像进行了分割实验,结果表明,该方法可以较好地分割出图像中的多个圆形区域,算法具有较好的抗噪性,实现速度较快,有利于实现肺部CT图像肺结节自动检测。

Problems of segmentation of pulmonary nodule candidates from CT images were studied in paper, the image segmentation method proposed by Chan-Vese was analysed and improved. A new level set method based on circle dependent had been deduced, and then an algorithm for segmentation of pulmonary nodule candidates in CT images had been presented, it solved the problem of detecting multi-circle areas with different sizes in an image. Experiments had been done for segmentation of synthetic and clinical pulmonary images by the proposed algorithm, the experiment results show that it can detect multi-circles areas correctly and efficiently, and it is robust to resist noise disturbance. This algorithm has advantages for the realization of automatic detection for lung nodules in pulmonary CT images.