An advanced analytic method incorporating the geometrical properties of scanner and radiation emissions into the system model for the true component of 3D PET data

An approach to an analytic system model for the true component of 3D PET data offering the accuracy of a Monte Carlo model run for an infinite time is presented. The system model, which is used in the expectation maximisation (EM) algorithm, accounts for the 3D nature of the emission process in which the geometric sensitivity to a point source varies along and across the tube of each line of response (LOR). Therefore the system model cannot be based just on line or tube integrals as it is for the model of transmission scanning. By accounting for the varying sensitivity within each LOR in the reconstruction process unbiased and quantitatively exact images can be achieved without using calibration or scaling factors after the reconstruction process. Two approaches to the model are investigated, i.e., LOR- and voxel-driven which are validated using high statistics Monte Carlo simulation (SimSET).