Compton-ANN: Bringing Artificial Intelligence to Compton Image Reconstruction

The design of a reconstruction model for Compton events leveraging Artificial Intelligence is here presented. An Artificial Neural Network (ANN) approach by adapting a conventional Feed-Forward Network type with multi-layer perceptron is used. The ANN operates on a given ensemble of events described with a five parameter set, which preserves the maximum information incorporated via the Compton scattering. Training of the Network and several reconstruction results with GATE simulated data are shown. Achieved spatial resolution and possible model improvements are also discussed.