Résumé
In list-mode PET reconstruction the accuracy of the projector model plays a key role in the reconstructed image quality. This projector must compute on the fly the contribution of each voxel from the field-of-view for a given line-of-response, including geometric and detector scattering effects. Recently proposed projectors consider the use of a Gaussian model to take into account detector geometric effects without explicitly modelling detector scatter. In this study we propose a new projector based on the intrinsic detector response function (IDRF), which represents the probability distribution function to record an interaction within each crystal detector. The advantage of this projector is to allow the modelling of all different detector characteristics (geometry, attenuation, scatter). An evaluation study was performed by comparing the projector results using a Monte Carlo Simulation (MCS) as ground truth. The results show that the proposed IDRF based projector performs best, leading to the closest agreement compared to the MCS results, in contrast to the Gaussian model based projector.