CAJ | 학술논문

肺实质分割是基于CT图像的计算机辅助检测技术必不可少的步骤。针对现有活动轮廓模型对初始位置敏感、深度凹陷区域分割不准确和抗噪性差等缺点，提出了一种基于改进Live-Wire算法确定Snake模型初始轮廓的交互式分割方法。该方法结合并改进Live-wire算法和一般的阈值法对图像进行预分割，将得到的边缘作为Snake模型的初始轮廓，通过Snakes模型演化得到肺实质轮廓结果。实验结果表明该方法能快速地对肺部图像进行分割，与传统方法相比具有人工交互次数减少、抗噪音性好、更具鲁棒性和效率性的优点。
폐실질분할시기우CT도상적계산궤보조검측기술필불가소적보취。침대현유활동륜곽모형대초시위치민감、심도요함구역분할불준학화항조성차등결점，제출료일충기우개진Live-Wire산법학정Snake모형초시륜곽적교호식분할방법。해방법결합병개진Live-wire산법화일반적역치법대도상진행예분할，장득도적변연작위Snake모형적초시륜곽，통과Snakes모형연화득도폐실질륜곽결과。실험결과표명해방법능쾌속지대폐부도상진행분할，여전통방법상비구유인공교호차수감소、항조음성호、경구로봉성화효솔성적우점。
Lung segmentation is an indispensable step in pulmonary nodules computer-aided detection based CT images. Aimed at decreasing the Snake dependence on initial contour points, improving the accuracy of segmenting the depth sag area and resistance of noise, this paper proposes an interactive lung parenchyma segmentation algorithm, which uses improved Live-Wire algorithm to define Snake initial contour. It modifies the traditional Live-Wire algorithm by combining it with the threshold algorithm. After the pre-segmentation, the edge is obtained. Then taking the edge as the initial contour, Snake model can get lung parenchyma by evolution. Experimental results show that the method can extract the lung paren-chyma effectively and quickly. Compared with traditional method, this method can reduce the number of human-computer interactions and can well resist noise, which is more robust and efficient.