基于Snake模型的病变CT肺部图像分割

Pathological Lung Segmentation for CT Images Based on Snake Model

传统的肺实质分割算法大多基于阈值法、区域增长、边缘提取以及形态学等方法,这对健康肺部CT图像可以得到比较理想的分割结果,但临床上所面临的CT图像大多具有各种病变,严重时会导致肺实质大面积缺失或呈区域性离散块状,这使得传统的肺实质分割算法效果不理想。为此,提出一种以生理解剖学知识为基础并基于Snake模型的肺实质分割算法。通过多组CT图像实验表明,无论肺内有无病变,该算法对CT肺部图像分割效果理想,而且速度快、完全自动。

Traditional lung segmentation algorithms are mainly based on threshold, region growing, edge extraction, morphology and so on, all these segmentation algorithms are suitable to healthy lung CT images. But there are many pathological changes in clinical CT images, which cause vast lose of lung parenchyma or regional discrete blocks. Segmentation results are unsatisfactory when applying traditional segmentation algorithms on pathological lung CT images. Therefore, a new lung segmentation algorithm was put forward based on anatomical knowledge and Snake model. Experiments show that no matter whether the CT images are pathological or not, this segmentation algorithm has good results, high speed, and total automation.