Study of ordered subsets expectation maximization (OS-EM) algorithm based on different subsets

Objective: Compare influence on positron emission tomography (PET) reconstruction image quality and convergence rate of different subsets. Methods: Study the reconstruction results of simulated phantom based on different selection of subset level and iteration times. Results: Approximately computation time is required with same product of level and iteration times, the quality of image with big level is better than small one under the same iteration time, but it does not increase monotonously, there is a tradeoff between subset- level and image quality. Conclusion: To improve computation speed while guaranteeing image quality, selecting proper level of subset is necessary in practice