Optimizing Proton Range Verification on Compton Camera Experimental Data with Neural Network Algorithms

This study presents a novel approach for signal event selection in Compton camera imaging, focusing on proton range verification. Experimental data were collected from a clinical cyclotron facility by employing the MACACO III Compton camera prototype. An event selection method based on neural networks was applied and compared with a traditional energy-cut method. Results demonstrate that the NN-based approach significantly enhances signal percentage and improves the precision in detecting millimetric shifts in depth-dose distribution. Ongoing work involves implementing the NN approach on experiments performed with an improved Compton camera prototype for further refinement and application in hadron therapy monitoring.