Answers are grounded in this paper and cite their sources.
Prompt gamma rays emitted by cells during proton beam radiotherapy can be detected by Compton cameras and used to reconstruct images to verify accurate dose delivery and Bragg peak location. However, image reconstruction is complicated by misclassified scattering events because of the limited temporal resolution of the cameras. To address this, machine learning models were trained on simulated datasets generated using Geant4 and MCDE and implemented through the Big Data REU Integrated Development and Experimentation (BRIDE) platform to classify and filter scatter events for more accurate image reconstruction. After identifying the bestperforming models, predictions using a deep Fully Connected Network (FCN) were created for 24 cases of experimental phantom triples data. Images were reconstructed on CORE using Kernel Expectation-Maximization (KEM) and Simple Back Projection (SBP) algorithms on both raw and repaired data, then compared to assess the accuracy of the Bragg peak estimation and noise reduction. SBP reconstructions had more complete beam reconstructions and Bragg peak profiles than KEM reconstructions; however, they still included noise in 2-D beam profiles. KEM 1-D profile reconstructions improved after a reorientation script was implemented to transform the scatter data before classification, which reoriented the data such that it was consistent with training, and reversed before reconstruction. Index Terms—Proton beam therapy, Compton cameras, machine learning, image reconstruction, deep learning models
No takes yet. Share an insight, caveat, or question.
Lopez et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: