Impact of inter-crystal scattering on image quality of a dedicated brain PET scanner with DOI capability

<i>Objective</i>. Smaller scintillator crystals are commonly used in positron emission tomography (PET) to achieve higher resolution for preclinical research and clinical diagnosis, particularly in brain-dedicated PET scanners with depth-of-interaction (DOI) capability. Although DOI information provides parallax error correction, mispositioning of the line-of-response (LOR) induced by inter-crystal scattering (ICS) can lead to performance degradation, affecting both spatial resolution and image contrast. Previous studies primarily focused on ICS elimination or recovery on PET scanners that are equipped with one-to-one coupled scintillation detectors, with few studies proposing applicable methods for handling ICS events in brain-dedicated PET scanner equipped with light-sharing window detectors and DOI capability. This study specifically evaluates the impact of ICS on a system that have already undergone DOI correction.<i>Approach</i>. By positioning ICS events to their first interaction crystal in Monte Carlo simulations and removing ICS events through a contour-based mask in experimental studies, evaluations of the impact of ICS events on image quality in the context of DOI-corrected systems were conducted. Metrics including spatial resolution and peak-to-valley ratio (PVR) were used for quantitative analysis, while reconstructed images and line profiles provided qualitative comparisons.<i>Main results</i>. Three stages of images were compared in both simulation and experimental studies, revealing that although image quality could be improved by DOI correction, the presence of ICS still affects spatial resolution and image contrast. Applying a contour-based ICS removal method prior to image reconstruction showed a higher PVR, resulting in improved spatial resolution and better lesion detectability for the scanner, though at the cost of sensitivity loss.<i>Significance</i>. Through simulation and experimental evaluation, we have demonstrated that the presence of ICS can severely degrade the performance of high-resolution brain-dedicated PET scanners with DOI decoding capabilities, highlighting the necessity for accurate ICS event classification.