Deep learning-based dose prediction in proton beam therapy for hepatocellular carcinoma: comparison of network architectures and loss functions

Clinical Snapshot

Key Findings

• P Value: Not explicitly reported

• Effect Size: Lowest prediction errors with U-Net + dose gradient-aware loss: MAE 35.4 cm³, RMSE 51.3 cm³

• Primary Outcome: Dose prediction accuracy using dose gradient-aware loss function

• Nnt Or Sensitivity: Dice similarity coefficient of 0.88 for 10% dose region prediction

• Confidence Interval: DSC: 0.88 ± 0.05 for dose gradient-aware models

Clinical Application

Requires significant computational infrastructure and specialized expertise in radiation physics Limited applicability given Australia has only one proton therapy facility (Adelaide); may support treatment planning optimization and patient selection for overseas referrals HCC patients suitable for proton beam therapy at specialized centers

Abstract

In proton beam therapy (PBT) for hepatocellular carcinoma (HCC), deep learning (DL)-based dose prediction offers clinical value by providing immediate reference dose distributions as a guideline tool for treatment planners and enabling virtual PBT dose assessment at institutions lacking PBT facilities to support clinical decision-making. This study proposes a dose gradient-aware DL training approach and a beam arrangement-free framework that predicts dose distributions from computed tomography (CT) images and target/organs-at-risk structures without requiring beam arrangement inputs. Using data from 172 HCC patients, we systematically compared 20 DL models combining five architectures with four loss functions, including a novel dose gradient-aware loss capturing PBT-specific dose distribution characteristics arising from the Bragg peak. Prediction accuracy was evaluated using dose metrics for clinical target volume, planning target volume, and normal liver, 10% dose region volume assessment, visual examination, Dice similarity coefficient (DSC) for 10% dose region, and dose-volume histogram (DVH) analysis. DL models employing dose gradient-aware loss demonstrated reduced prediction errors in target volume dose metrics, achieved the lowest errors in 10% dose region volume when combined with U-Net (MAE: 35.4 cm3, RMSE: 51.3 cm3), attained the highest DSC (0.88 ± 0.05), showed the closest DVH agreement with clinical plans for both target structures and normal liver. A comprehensive evaluation consistently confirmed the superiority of dose gradient-aware loss across all metrics, demonstrating the importance of physics-informed, dose gradient-aware learning for accurate PBT dose distribution prediction. Particularly, DL models combining simple architectures with dose gradient-aware loss predicted doses that closely approximated clinical doses in PBT.

References

1. 1.Ogawa, S., Kadoya, N., Kato, T., Kato, R., Tozuka, R., Narita, Y., Oyama, S., Murakami, M., & Jingu, K. (2026). Deep learning-based dose prediction in proton beam therapy for hepatocellular carcinoma: comparison of network architectures and loss functions. Journal of Radiation Research. https://doi.org/10.1093/jrr/rrag026