Ordered spatial subsets for faster reconstruction in spect

Iterative reconstruction is now becoming increasingly used in clinical SPECT reconstruction. Ordered subsets (OS) algorithms have been particularly attractive for accelerating the convergence in iterative reconstruction. Traditionally, the ordered subsets in an OS algorithm are formed by grouping the sinogram data according to the angular positions of the projection. In this work, we propose an alternative approach to form the ordered subsets by sub-sampling the spatial locations in the detector array so as to exploit the depth-dependent blur in the detector response. We demonstrate the proposed approach in the context of EM reconstruction and the results in our experiment show that it can indeed accelerate the convergence of EM; more importantly, it can further improve the convergence rate of traditional OS reconstruction when used together with angular ordered subsets.