Direct Annihilation Position Classification Based on Deep Learning Using Paired Cherenkov Detectors: A Monte Carlo Study

Abstract

To apply deep learning to estimate the three-dimensional interaction position of a Cherenkov detector, an experimental measurement of the true depth of interaction is needed. This requires significant time and effort. Therefore, in this study, we propose a direct annihilation position classification method based on deep learning using paired Cherenkov detectors. The proposed method does not explicitly estimate the interaction position or time-of-flight information and instead directly estimates the annihilation position from the raw data of photon information measured by paired Cherenkov detectors. We validated the feasibility of the proposed method using Monte Carlo simulation data of point sources. A total of 125 point sources were arranged three-dimensionally with 5 mm intervals, and two Cherenkov detectors were placed face-to-face, 50 mm apart. The Cherenkov detector consisted of a monolithic PbF2 crystal with a size of 40 × 40 × 10 mm3 and a photodetector with a single photon time resolution (SPTR) of 0 to 100 picosecond (ps) and readout pitch of 0 to 10 mm. The proposed method obtained a classification accuracy of 80% and spatial resolution with a root mean square error of less than 1.5 mm when the SPTR was 10 ps and the readout pitch was 3 mm.