基于节点拓扑一致性的2D/3D冠脉血管非刚性配准  

Non-rigid registration based on node topological consistency for 2D/3D coronary artery images

作　　者：李致勋[1] 张英涛[1] 唐降龙[1] 

机构地区：[1]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001

出　　处：《智能计算机与应用》2016年第6期20-24,共5页Intelligent Computer and Applications

基　　金：国家自然基金(61100097)

摘　　要：目前,冠状动脉介入手术主要采用传统冠脉成像CCA加冠脉CT成像CCTA辅助的方式。但是,CCTA下的3D图像获取的是某一时刻的静态图像,在心跳和呼吸作用下,CCA下动态2D图像和CCTA下静态3D图像的血管可能并不处于同一位置,彼此之间会具有很大的形态及位置差异。本文在2D和3D中心线基础上,利用血管曲线特征对2种维度的血管进行粗糙和精细两级配准操作。在粗糙配准中,通过考虑节点间的拓扑关系一致性取消以往算法的双向一对一的对齐约束,对血管间的分支点进行多对多对齐配准。在精细配准中,基于节点对齐的配对血管段间采用伸缩特性的DTW技术进行像素间配准。实验结果表明研究提出的方法具有良好的效果。Currently,coronary artery interventional therapy mainly employs Conventional Coronary Angiography（ CCA） and Coronary Computed Tomography Angiography（ CCTA）. However,CCTA＇s image is static image in the moment. Moreover,in the role of heart beating and breathing,blood vessels under the 2D and 3D may not be with the same shape,and their forms and positions may thus be with great difference. Based on the extraction of the center line of the 2D and 3D,the proposed method made rough and fine two-stage registration through vascular curve characteristics of two dimensions. In rough registration,by keeping the consistency of the topological structure of the tree node,the node＇s many-to-many alignment could be made instead of the bijection one-to-one alignment constraint in existing algorithms. In fine registration, vessel segments are aligned by the DTW technology. Finally, the experimental results demonstrate the accuracy and efficiency of the proposed method.

关 键 词：树节点拓扑一致性 非刚性配准 2D/3D冠脉成像 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]