A Monte Carlo model of noise components in 3D PET

Abstract

This work presents a new model, developed by Monte Carlo methods, to estimate noise components (scatter and random coincidences) in three-dimensional (3-D) positron emission tomography (PET). The model allows the amount, spatial, and temporal distribution of true, scattered, and random coincidences to be estimated independently for any radioactive source (both phantoms and real patients), taking proper, account of system dead time. The model was applied to a 3-D NaI(Tl) current-generation PET scanner for which there are no currently available methods for estimating scatter and random components in whole-body studies. The quantitative accuracy of the developed noise model was tested by comparing simulated and measured PET data in terms of physical parameters, count-rate curves, and spatial distribution profiles. Scatter and random components were assessed for phantoms representing brain, abdomen, and whole-body studies. Evidence was found of high scatter and random contribution in 3-D PET clinical studies. The clinical response of the PET system, in terms of signal-to-noise ratio, was assessed and optimized, confirming the suitability of the default energy window, although suggesting a possible improvement by setting a lower energy threshold higher than the current default: The proposed noise model applies to any current generation 3-D PET scanner and has been included in the Monte Carlo software package PET-EGS, devoted to 3-D PET and freely available from the authors.