机器人辅助术中的时间空间对准方法.docVIP

Spatio-temporal registration in Robot Assisted Surgery Jia Gu, Key Lab for Biomedical Informatics and Health Engineering, Chinese Academy of Sciences Abstract—This paper presents a method for spatio-temporal alignment of extracted vessels central axis in X-ray angiography sequences. This work is performed in a segmentation frame. Motion information is exploited to solve spatial ambiguities and improve the segmentation stage in presence of vessel superposition, crossing or discontinuities. The registration aims at establishing correspondences in time and in space between consecutive image frames relating the contrast product propagation inside the vessels. It allows finding a temporal mapping and a one-to-one correspondence between features points while rejecting non-homologies. It relies on an elastic thin-plate spline model to estimate the non-rigid motion induced by the heart and find the point correspondences between two images. An optimization process is led through a simulated annealing. Index Terms—Coronary angiography, spatio-temporal registration, feature-based method, thin-plate-spline, simulated annealing 机器人辅助手术中的时间空间对准方法 摘要：本文提出了一种基于X光的时间空间对准方法用于机器人辅助手术，并可以得到血管中心线的分割信息。在算法中运动信息被引入以消除空间中的模糊信息，并可以解决手术过程中的血管重叠、交叉或断裂情况。配准部分被引入可以求解帧间的对应关系。算法的中心思想是利用薄型样条模型求解组织的非刚体形变，并用模拟退火进行优化。 1. INTRODUCTION Conventional X-ray coronary angiography still remains the gold standard for the quantitative analysis of coronary diseases. It provides a sequence of two-dimensional projections of the 3-D vascular network, which contains vessel crossing, overlaps and loop. The visual interpretation by the physicians is subject to inter- and intra-individual variability. A segmentation stage is thus needed to extract the vessel contour and quantify the stenosis or perform a 3-D reconstruction from two views in order to improve the interpretation of the scene. This paper deals with spatio-temporal registration along the sequence in a segmentation frame. The objective is to take