In recent years, optical coherence tomography (OCT) has developed into a popular coronary imaging technology at home and abroad. The segmentation of plaque regions in coronary OCT images has great significance for vulnerable plaque recognition and research. In this paper, a new algorithm based on K-means clustering and improved random walk is proposed and Semi-automated segmentation of calcified plaque, fibrotic plaque and lipid pool was achieved. And the weight function of random walk is improved. The distance between the edges of pixels in the image and the seed points is added to the definition of the weight function. It increases the weak edge weights and prevent over-segmentation. Based on the above methods, the OCT images of 9 coronary atherosclerotic patients were selected for plaque segmentation. By contrasting the doctor's manual segmentation results with this method, it was proved that this method had good robustness and accuracy. It is hoped that this method can be helpful for the clinical diagnosis of coronary heart disease.
光学相干断层成像技术（OCT）现已发展成为国内外较热门的冠状动脉内影像技术，其中冠脉 OCT 图像的斑块区域分割对易损斑块的识别和研究有着重大意义。本文提出了一种基于 K-means 聚类与改进随机游走的新算法，实现了对冠脉钙化、纤维化斑块和脂质池的半自动化分割。本文主要创新点为改进了随机游走算法的权函数，将图像中像素间的边与种子点之间的距离加入到了权函数定义中，增加了弱边界的权值，防止了过分割现象的发生。本文基于以上方法对 9 名冠状动脉粥样硬化患者的 OCT 图像进行了斑块区域分割。通过对比医生手动分割结果，证明了本文方法具有良好的精度和鲁棒性，以期本文方法可对冠心病的临床诊断起到一定的辅助作用。.
Keywords: K-means clustering; optical coherence tomography; plaque region segmentation; random walk.