基于光学相干断层成像含支架的冠脉血管三维重建

摘要

Optical coherence tomography (OCT) has become a new method for detecting coronary artery disease,which is widely used for its high resolution.In this article,we study algorisms for 3D images reconstructing from 2D cross-sectional coronary images.The images are firstly denoised by self-adapted median filter algorithm.According to the characteristics of OCT images,an improved Otsu algorithm is used to segment the coronary vessels.A new full automatic algorithm is then applied for stent detection.All 2D OCT images are converted into polar coordination.Based on the visible stent features shown in the OCT images,with global intensity histogram and intensity contours,we can identify candidate pixels and remove the artifacts to locate the stent on the lower boundary.By merging adjacent candidate pixels to a point on the stent,we can then accurately delineate the stent position.This process does not require a priori knowledge of stent status and the contours of blood vessels.It is immune from interferences induced by artifacts and avoids the accuracy problem of clustering algorithm.The final results show that the proposed method is able to conduct 3D reconstruction and fusion effectively.%光学相干断层成像(OCT)以其具有高分辨率,实时成像等特征,已经成为冠状动脉疾病检测的一个新方法.对含支架的OCT冠脉血管图像进行三维重建研究,以期直观地显示三维空间结构图.通过比较选择合适的降噪算法对图像预处理,根据OCT图像的特性,利用改进的最大类间方差算法(Otsu)对冠脉血管进行分割,并利用一个新的自动算法对冠脉支架进行检测.由于图像存在大量可视支架,在对极坐标图像序列预处理之后,用全局的强度轮廓检测候选的像素,去除伪影,确定支架上下边界和位置,并合并相邻的候选像素为一个支架点,成功标记出检测到的支架点,可有效地排除伪影造成的干扰,提升聚类算法的精确度.在此基础上,利用Amira软件实现了对冠脉血管和支架进行三维重建和融合.