一种基于对称结构优化的OSEM快速重建算法

基金项目:

北京市重点建设学科——应用数学(71D0911003)

国家自然科学基金(60972115)

详细信息

An Symmetric Ordered Subset Expectation Maximization Accelerated Algorithm

1. Laboratory of Applied Mathematics, Beijing Information Science & Technology University, Beijing 100101, China;
2. School of Science, Beijing Jiaotong University, Beijing 100044, China

摘要: 由投影重建图像的EM重建算法中,有序子集最大期望值法能够对经典EM算法加速,提高收敛速度,受到广泛地关注。本文提出了一种基于对称性结构的OSEM快速重建算法,在更好地满足子集平衡的条件和优化有序子集的迭代顺序的同时,压缩了尺度窗口的计算。数据实验结果表明:当合适选取子集的数量时,S-OSEM算法有效地提高了重建速度和成像精度。
- 对称结构 /
- 最大期望值法 /
- 代数迭代 /
- 最大似然估计
Abstract: The ordered subsets EM algorithm has excited considerable interest for image reconstruction from projection due to its acceleration of the standard EM algorithm.In this paper,we introduce an update algorithm called Symmetric Ordered Subset Expectation Maximization Algorithm(S-OSEM)which accelerates the OSEM as well as satisfies subset balance and orders the subset effectively by the Symmetric structure of projection lines. Furthermore,standard EM also can be accelerated by the Symmetric structure of projection lines.The results indicate that with the appropriate number of subsets,the S-OSEM algorithm can improve the reconstruction speed and imaging precision effectively.
- symmetric structure /
- expectation maximization /
- algebraic iterative algorithm /
- maximum likelihood estimation