Singles and coincidence processing for a digital PET/MR system using SiPM detectors

This paper presents a software-based singles and coincidence processing for a digital PET/MR system using SiPM detectors coupled to preclinical crystal arrays. Compared to traditional PET detectors our system outputs detector raw data of the individual detector elements instead of singles or coincidences. On the one hand this requires a higher bandwidth on the other hand the availability of raw detector data offers unique opportunities to employ more accurate and computationally complex algorithms to enable highly accurate PET images. Several processing algorithms are evaluated in terms of image quality and data throughput. We show that using a ML-based processing algorithm a spatial resolution of 1.6mm or better can be achieved. For our system a Center-of-Gravity(CoG)-based processing sequence the overall data throughput becomes sufficient enough for (soft) real-time processing.