融合形状分类的血管造影图像多级分割

Multi-level Segmentation of Angiographic Images Based on Shape Classification

针对血管造影图像中血管分割难度较大的问题,提出一种融合血管形状分类的血管造影图像多级分割方法,旨在分割出带有血管类型信息的轮廓边界,辅助医生快速诊断。首先,根据血管的几何特征对血管进行分类,建立血管分类标准;其次,对血管进行粗分割,提取血管整体轮廓;然后,提出一种局部区域约束的水平集算法用于精准分割整体轮廓中不准确的血管段;最后,利用分类标准确定整体和局部血管分割结果中的血管段类型,自动生成包含血管分类信息的血管轮廓。实验结果表明,该方法能够快速提取出血管轮廓,识别其中包含的血管段类型,满足血管分割的实际需求。

This paper developes a method about multilevel segmentation of angiographic images with fused vessel shape classi⁃fication to segment the contour boundary with vessel type information and assist doctors in rapid diagnosis,which can solve the prob⁃lem that vessel segmentation is more difficult in angiographic images.Firstly,according to the geometric characteristics of blood ves⁃sels,blood vessels are classified,and the classification standard of blood vessels is established.Secondly,the blood vessel is rough⁃ly segmented to extract the overall contour of the blood vessel.Then,a local region constrained level set algorithm is proposed to seg⁃ment the inaccurate vascular segments in the overall contour.Finally,the classification criteria are used to determine the types of vessel segments in the global and local vessel segmentation results,and the vessel contour containing vessel classification informa⁃tion is automatically generated.Experimental results show that this method can quickly extract the vascular contour and identify the types of vascular segments,which can meet the actual needs of vascular segmentation.