利用模糊识别算法分析冠状动脉造影图像中的血管结构

Analyzing the Structures of Coronary Artery Trees in Angiogram Images Based on Fuzzy Recognition Algorithm

定性与定量地描述冠状动脉血管,很大程度依赖于造影图像中的血管结构识别结果.对此,该文提出了一种多特征模糊识别算法判别血管结构.实现过程中,首先通过图像预处理获得血管初始特征,然后利用一圆周探测器沿血管路径扫描并获取多种局部测度;在定义各种局部测度的多特征模糊子集及其隶属度函数之后,通过构造模糊识别算子准确地判别血管的端、段、分支和交叉结构.该方法在仿真血管模型和多套实际冠状动脉造影图像上获得了较好的效果,对实际图像的结构识别平均正确率达到92.60%.

The qualitative and quantitative description of coronary artery depends largely on inferring the artery tree structure in the angiograms. In this paper, an algorithm of Multi-feature based fuzzy recognition is proposed to infer vessel structure in the angiograms. In the implementation, the initial vessel features are attained by preprocessing the original image, and then a circle detector is used to scan and calculate multi feature metrics along the vessel path. After defi- ning the membership degree of the multi-feature metrics, a fuzzy operator is constructed to infer the vessel structures, i. e. , the distal ends, segments, bifurcations and crossovers of the artery tree. The algorithms perform well in a simulated phantom, and the ratios of structure identification reach on average to 92.60% in the clinical angiograms.