冠状动脉造影图像序列中估计二维运动及变形的弹性配准算法  

Elastic Registration Algorithm for Two-Dimension Motion and Deformation Estimation from Coronary Artery Angiographic Image Sequence

作　　者：孙正[1] 郁道银[1] 陈晓冬[1] 徐智[1] 黄家祥[1] 谢洪波[1] 

机构地区：[1]天津大学精密仪器与光电子工程学院,光电信息技术科学教育部重点实验室,天津300072

出　　处：《工程图学学报》2003年第4期90-96,共7页Journal of Engineering Graphics

摘　　要：提出一种由单面血管造影图像序列分析心脏冠状动脉运动和变形的方法,即利用弹性配准算法,将动脉运动的估计简化为对连续两帧图像的骨架像素的匹配。采用动态规划的方法进行配准,同时引入自回归建模,二者结合起来互相作用:动态规划为自回归模型参数的估计提供样本值;自回归模型为动态规划提供适当的成本函数。采用临床得到的单面冠状动脉造影图像序列对该算法进行了验证。实验结果表明应用该方法计算血管骨架图像中各点的光滑移动向量场,能够得到精确的结果。In this paper an approach for estimating the motion of coronary arteries in digital single-plane angiographic image sequences is proposed. Binary skeleton images are registered using an elastic registration algorithm in order to reduce the problem of estimating the motion of arteries to elastically matching the skeletons of two successive frames. The method is based on dynamic programming (DP) algorithm in conjunction with autoregressive (AR) modeling of the deformation. They are used jointly: DP provided several sequences of observations for AR modeling, and, at the same time, AR modeling furnished the convenient cost function that DP required. The performance of this motion estimation method is demonstrated on clinical single-plane angiographic image sequences of coronary artery. It is shown that motion estimation based on DP is successful especially with low contrast and noisy angiographic images.

关 键 词：冠状动脉造影图像序列 动态规划 弹性配准算法 成本函数 自回归模型 

分 类 号：R445[医药卫生—影像医学与核医学] O221.3[医药卫生—诊断学]