A deep neural network for positioning and inter-crystal scatter identification in multiplexed PET detectors

Abstract

Objective: Conventional event positioning algorithms in light-sharing PET detectors are often limited by edge effects and the impact of inter-crystal scattering (ICS). This study explores the feasibility of deep neural network (DNN) techniques for more precise event positioning in finely segmented and highly multiplexed PET detectors with light-sharing. Approach: A DNN was designed for crystal localisation, and trained/tested with light distributions of photoelectric (P) and Compton/photoelectric (CP) events simulated using optical GATE and an efficient analytical method. Using the statistical properties of ICS events from simulation, an energy-guided positioning algorithm was built into the DNN, enabling selection of the unique or first crystal of interaction in P and CP events, respectively. Performance of the DNN was compared with Anger logic using light distributions from simulated 511-keV point sources near the PET detector. Results: Despite coarse photodetector data due to signal multiplexing, the DNN demonstrated a crystal classification accuracy of 90% for P events and 82% for CP events. For crystal positioning, the DNN outperformed Anger logic by at least 34% and 14% for P and CP events, respectively. Further improvement is somewhat constrained by the physics, specifically, the ratio of backward to forward scattering of gamma rays within the crystal array being close to 1. This prevents selecting the first crystal of interaction in CP events with a high degree of certainty. Significance: Light-sharing and multiplexed PET detectors are common in high-resolution PET, yet event positioning can be poor due to edge effects and ICS events. Our study shows that DNN-based event positioning can enhance 2D coincidence event positioning accuracy by nearly a factor of 2 compared to Anger logic. However, further improvements are difficult to foresee without timing information.