List-mode reconstruction with system modeling derived from projections

Image reconstruction direct from list-mode data can preserve and utilize all the precision and accuracy of the substantial volume of data acquired by high resolution dynamic emission tomography systems. However, there remains a need for a full system modeling methodology to account for randoms, scatter, attenuation and normalization within list-mode reconstruction. Modeling and/or correcting for these effects has been substantially researched and developed for sinogram-based reconstructions, and so this work considers an expectation maximisation (EM) reconstruction algorithm which uses the list-mode data directly, but also accesses projection data to determine the randoms, scatter, attenuation and normalization components of the system model. Using the example of the high resolution research tomograph (HRRT, CTI PET Systems) scanner and the sinogram corrections available with this system, it is found that it is necessary to spatially align each list-mode event with the nearest sinogram bin in order to avoid a sampling mismatch when using the sinogram-data based corrections. Simulation results indicate that provided projections with high enough sampling are used, this compromise of the spatial precision of the list-mode data is not of significance for the HRRT, and moreover is a much less important effect compared to the inclusion or exclusion of an image space resolution model. The storage and temporal sampling advantages of the list-mode data are however retained.