Abstract
157
Objectives We assess the impact of coincidence resolving time (CRT) on convergence and noise performance in time-of-flight (TOF) PET aiming at optimized reconstruction parameters.
Methods Speed of maximum likelihood expectation maximization (MLEM) convergence, contrast, and signal and noise power are assessed on NEMA PET phantoms simulated by GATE, a Monte Carlo simulation tool developed at CERN, and evaluated using a research time-of-flight reconstruction code including resolution recovery. Results are explained by an analytical model previously developed for SPECT reconstruction.
Results A short coincidence resolving time in TOF-PET induces fast convergence, allowing the number of iterations to be reduced. For 500 ps CRT this results in up to 4 times lower noise power for an object of 30 cm diameter compared to non-TOF PET, at comparable contrast and signal power. A further decrease in CRT will reduce the noise power proportionally. The effect is the same as for SPECT imaging where we find a linear dependence of the noise power on object size. Further noise reduction by factors in the range of 2-5 is achieved by resolution recovery when applied in forward and back-projection.
Conclusions We have found an explanation for the excellent noise properties of TOF-PET with MLEM. Reconstruction parameters have to be adapted relative to standard PET. The low noise level is in line with an analytical model suggesting a linear dependence of noise power and object size