Evidence Hub

Improving Compton imaging of 478 keV prompt gamma-ray in boron neutron capture therapy with neural network-based event filtering: a simulation study

Jiye Qiu, Makoto Sakai, Caixun Zhang, Nikolaos Voulgaris, Shingo Tamaki, Sachie Kusaka, Isao Murata

Physics in Medicine and Biology, 2026

Abstract

Abstract Objective. Boron neutron capture therapy (BNCT) is a promising radiotherapy modality that effectively targets cancer cells through boron neutron capture reactions. Real-time determination and localization of the in vivo boron dose during treatment play an important role in BNCT. This study aims to investigate the feasibility of neural network-based Compton event filtering to enhance 478 keV prompt gamma-ray imaging for BNCT under high-background radiation conditions with a Si/TlBr Compton camera. Approach. Monte Carlo simulations were performed using PHITS to generate signal and noise Compton events under clinically relevant neutron irradiation. Fully connected neural network models were trained to classify signal and noise events. The filtered events were then used for Compton image reconstruction, and quantitative imaging performance was evaluated using event retention rate with 10 B concentration and image quality metrics. Results. The neural network models achieved stable classification performance with area under the ROC curve values exceeding 0.8. The Compton event retention rate shows a strong linear relationship with tumor 10 B concentration, demonstrating the quantitative capability. Image reconstruction results indicate improved image quality in tumor regions, although additional background artifacts were observed. Significance. The proposed neural network-based Compton event filtering method provides an effective preprocessing strategy for data reduction…

Read the paper

Related studies

Go deeper in Evidence Hub

Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.

Open this study Sign up free