Semi-monolithic meta-scintillator simulation proofof-concept, combining accurate DOI and TOF

Abstract

Metascintllators arrangements have shown to achieve an equivalent CTR of 200 ps for BGO-plastic and 140 ps for LYSO-plastic sets. In this paper, we examine a novel architecture: The idea is to slice a slow scintillator (BGO or LYSO) into thin slabs read out by an array of SiPM in the semimonolithic manner providing depth-of-interaction (DOI) information; and interleave them with thin segmented fast scintillators (plastic EJ232 or EJ232Q) read out by single SiPMs as in pixelated designs, providing pixelated level coincidence time resolution (CTR), in what we call a semi-monolithic meta-scintillator (SMMS). We thus combine layers of slow scintillator of dimension 0.3x25.5x(15 or 24) mm3 and layers of fast scintillator of dimensions 0.1x3.1x(15 or 24) mm3 in a Monte Carlo Gate platform to investigate the performance of this new type of semimonolithic detector. It is shown that the time resolution of SMMS is equivalent to that of single metapixels of the same configuration. In particular, 15 mm deep LYSO based SMMS led to CTR 121 ps, before implementation of timewalk correction (107 ps CTR). Same dimensions for BGO based SMMS led to CTR of 241 ps, a 15% deviation from metapixel experimental results. Further to this timing study, we expand the study to application of neural networks on the photon distributions and timestamps recorded at the SiPM array. This leads to determination of the DOI with < 3mm precision and 0.85 confidence level in the best scenario and more than 2 standard deviations precision in reconstructing energy sharing and energy of interactions. Overall, neural network prediction capabilities, taking advantage of enhanced understanding of the photon distribution, exceed those of the standard energy calculation through addition of numbers of detected photons or any analytic approach on energy sharing.