论文部分内容阅读
提出了由多层螺旋计算机层析(MSCT)数据估计冠脉三维运动的算法。首先对多期相数据进行基于Hessian矩阵的局部血管增强,随后采用自适应阈值区域生长方法分割出血管并进行细化,得到不同时刻的冠脉骨架。血管进行分段后,利用连贯点漂移(CPD)点配准算法对不同期相的各段血管配准,计算点对之间的对应关系矩阵及空间变换,从而估计三维运动场。采用运动场已知的模拟数据评估算法精度,结果表明,对超过50mm的大幅度运动,配准误差小于1voxel,运动场估计误差小于1%。对实际的全期相数据,估计左右冠脉的运动场,也获得了较为均匀和平滑的结果。
An algorithm for estimating the three-dimensional coronary motion from multislice spiral computed tomography (MSCT) data was proposed. Firstly, local blood vessel enhancement based on Hessian matrix was carried out on multiphase data, then the blood vessels were segmented by adaptive threshold area growth method and the coronary arteries were obtained at different time. After the blood vessels are segmented, the CPD point registration algorithm is used to register the blood vessels in different phases, and the corresponding relationship matrix and spatial transformation between point pairs are calculated to estimate the 3D motion field. The accuracy of the algorithm was evaluated using the known simulation data of the stadium. The results showed that the registration error was less than 1 voxel for the large-scale motion over 50 mm and the error of the motion field estimation was less than 1%. For the actual phase data of all phases, it is also estimated that the exercise fields of the left and right coronary arteries have a relatively uniform and smooth result.