SPECT simulation using a space variant PSF model

A fast single photon emission computed tomography (SPECT) simulator using an analytical object shape dependent point spread function (PSF) model has been developed as an alternative to Monte Carlo simulation (MCS). The parameters of the PSF model are estimated using line source measurements in a rectangular reservoir. The PSF is able to predict the effect of object curvature on the scatter function. The accuracy of scatter fraction calculation of a point source in the center of a Jaszczak phantom appears to agree better with measurements than MCS. The method is especially suited for simulation of patient studies in cases where the emitting object may be considered homogeneous to a good approximation. Simulation of SPECT projections from segmented magnetic-resonance-imaging (MRI) brain data only takes a few hours on an HP700 architecture, which is about a thousand times faster than MCS. The simulator uses a few megabytes memory for its PSF table, compared to tens of gigabytes when an MCS generated transition matrix is used. Therefore, the simulator is efficient in performing the forward step in object shape dependent true 3-D SPECT reconstruction