Comparison of Maximum-Likelihood List-Mode Reconstruction Algorithms in PET

In this work we compare images obtained from different list-mode reconstruction algorithms for PET data. Our analysis is based on list-mode data generated from a Geant4 simulation of the Jaszczak phantom. The first issue addressed is the influence of the raytracer used during reconstruction on the quality of the resulting images. We find that the Wu anti-aliased raytracer leads to better images than the Siddon algorithm at the expense of higher computational effort. The second question concerns a comparison of various modifications of the OSEM list-mode reconstruction algorithm. Here the convergent variant COSEM yields slightly better results at costs of much higher computational effort. However, the newer Enhanced COSEM does not need that effort to produce a similar image.