Processing of Compton events in the PETALO readout system

TL;DR

This paper explores the use of liquid xenon in PET imaging, employing neural networks to improve gamma ray event reconstruction affected by Compton scattering, aiming to enhance image quality.

Contribution

It introduces a neural network approach for tagging poorly reconstructed gamma events in liquid xenon PET, advancing event reconstruction techniques.

Findings

Neural network can identify poorly reconstructed gamma events.

Significant fraction of Compton-scattered events are not poorly reconstructed.

Further research needed for real-time implementation.

Abstract

PETALO (Positron Emission TOF Apparatus based on Liquid xenOn) exploits the unique characteristics of liquid xenon as a scintillator for use in a PET detector. Here initial simulation studies are detailed which highlight the potential of such a detector and outline the steps taken in the reconstruction of 511 keV gamma rays and in the full PET image reconstruction. In particular, a neural network-based approach is conceived in order to tag gamma rays that are poorly reconstructed due to Compton scattering. It is found that though a significant fraction of events undergo Compton scattering, not all of these events will necessarily be poorly reconstructed. Further study is necessary to determine whether or not an online implementation of the neural network-based event tagging will provide sufficient recovery of the image quality.