A theoretical model for EM-ML reconstruction algorithms applied to rotating PET scanners.

In this paper we show how to compute the normalizing and the system matrix terms of the EM-ML reconstruction algorithm for rotating planar detector PET scanners. The method introduced is valid for either pixelated or continuous scintillators. We base our computations in geometrical considerations, but other effects of the PET process can be easily included. In this regard, the intrinsic resolution of the detection system, the depth of interaction (DOI) of the incident gamma rays and the efficiency of the scintillators have been modeled in our development. The computation of the normalizing term and the system matrix is valid for any basis function used for the discrete approximation of the radionuclide concentration. We show that our computations are comparable to those of a Monte Carlo method at a small fraction of the computational cost.