Simulating whole-body PET scanning with rapid analytical methods

摘要

We present a 3D whole-body PET simulator that generates multiple, yet statistically accurate, realizations of projection data within a computation time that is short enough to enable the measurement of the task performance of image reconstruction algorithms. The whole-body simulator takes into account the following effects: the seperate amounts of statistical noise contributed by true, scattered, and random coincidences, activity outside the field of view, activity decay between bed positions, detector efficiencies and resolution, and noise arising from the transmission scan. The principle of the simulation is to analytically calculate projections based on the geometrical specification of the emission and attenuation objects that comprise the MCAT phantom. A pre-determined level of statistical noise is then added to the projection data, which is then inputted to the same data correction and image reconstruction procedures used in practice. We compare the results of multiple realizations of simulated and measured phantom studies at statistical noise levels similar to those encountered in 3D whole-body PET scanning. We obtain comparable statistical noise properties for the fully corrected emission sinograms, and also show that a simple model accurately predicts the statistical noise added by random and scattered coincidences generated by activity outside the field of view.