A simulation study of the convergence properties of the EM algorithm for image reconstruction in Positron Emission Tomography

We have studied certain characteristics of the Expectation — Maximization (EM) Algorithm for image reconstruction in Positron Emission Tomography (PET). The convergence and image quality have been studied using a series of simulation experiments. We. comment on the impact that different level of count rate has on the image quality, as it is expressed in terms of root mean square (RMS) error, and the region of feasible images. We observe the difference in the convergence rate of local regions in the field of view that have different activity levels in order to demonstrate the need of a new stopping criterion that will take into account the nonuniform recovery of the true activity resolution in the image plane.