A computational study of the focus-of-attention EM-ML algorithm for PET reconstruction

Abstract

The expectation-maximization maximum-likelihood (EM-ML) algorithm for image reconstruction in positron emission tomography (PET) essentially solves a large linear system of equations. In this paper, we study computational aspects of a recently developed preprocessing scheme for focusing the attention, and thus the computational resources, on a subset of the equations and unknowns in order to reduce the storage, computation, and communication requirements of the EM-ML algorithm. The approach is completely data-driven and uses no prior anatomic knowledge. The experimental results are obtained from runs on a small network of workstations using simulated phantom data as well as data obtained from a clinical ECAT 921 PET scanner.