Evaluation of different random estimation methods for the MADPET-II small animal PET scanner using GATE

The validity of random estimation techniques has been studied for various low energy thresholds (LETs) for singles list-mode simulated data sets of the MADPET-II small animal PET scanner. MADPET-II is a dual layer PET scanner prototype consisting of LSO crystals read individually from avalanche photo-diodes (APDs). The simulations have been performed using GATE simulation toolkit. Several sources have been simulated, such as point, planar, cylindrical and Derenzo. We have evaluated the randoms using three standard methods, the delayed window (DW), singles rate (SR) and time histogram fitting (THF). In addition, we know from simulations how to discriminate between random and true coincidences. Therefore, we compare the randoms estimated with the above methods with the number of randoms obtained using GATE information. An overestimation in the number of randoms has been observed for DW and SR methods. We pointed out that this mismatch is due to the inter-crystal scatter (events that take place between the crystals) since is reduced for LETs higher than 255 keV. Nevertheless, this discrepancy does not have a significant effect to the quality of the reconstructed images.