上下文窄带约束下腹主动脉瘤外轮廓分割

Automatic abdominal aortic aneurysm outer contour segmentation based on context information and narrow band constraint

针对腹主动脉瘤(Abdominal aortic aneurysm,AAA)外轮廓在医学图像中显著性较低,现有算法难以实现精确自动分割的问题,提出一种新型的基于上下文窄带约束的计算机断层扫描血管造影(Computed tomography angiography,CTA)图像序列连续分割算法.首先采用改进的局部二元拟合(Local binary fitting,LBF)水平集方法获取初步腹主动脉瘤外轮廓,然后利用上下文窄带约束实现外轮廓高精度分割.单层的分割结果作为相邻层分割先验实现迭代分割.相关实验表明:提出的算法性能优于传统方法,可以实现图像序列中腹主动脉瘤外轮廓的高精度自动分割.

The outer contour of abdominal aortic aneurysm （AAA） is fuzzy, which make it difficult for recent method getting its accurate segmentation. A novel context information and narrow band constraint based segmentation method for computed tomography angiography （CTA） sequence is provided. Firstly, improved LBF based level set method is used to get the preliminary outer contour of AAA. Then the context information and the narrow band constraint are used to get more accurate outer contour in current slice which would be used as prior of adjacent slice in iterative segmentation. The experimental result on the dataset shows the method proposed in this paper can get better segmentation result than existing method. So this method can get the outer contour of AAA with high accuracy.