Effect of noise in the probability matrix used for statistical reconstruction of PET data

Iterative reconstruction algorithms require previous computation of the system probability matrix P. This matrix is usually estimated through approximated calculations. The approach employed in this work is, however based on Monte Carlo simulations. This technique allows to describe P more accurately. Nevertheless. the amount of simulated events may limit the statistical quality of P. thus affecting the reconstructed image. The goal of this study was to quantify this effect for OSEM and PWLS applied to the small animal PET system MADPFT (O = 86 mm). The results showed that PWLS is more sensitive to the inaccurate description of P than OSEM. A mean relative error for P below ≈ 40% (obtained from simulations with more than ≈ 6,000 mean detected counts per detector pair) in combination with OSEM guaranteed a good image quality for 1 mm² pixels. For PWLS, however, a slight decrease of the mean relative error in the range 20%-30% might strongly affected the image properties. Simulations with more that ≈ 45,000 mean detected coincidences per detector pair slightly improved the accuracy of P.