Artificial Intelligence in Particle Therapy Treatment Planning: Current Advances and Future Directions
James C.L. Chow and Huan Giap
Particle therapy, including proton and heavy-ion therapy, offers highly conformal dose distributions through the Bragg peak and, particularly for heavier ions, enhanced relative biological effectiveness. However, treatment planning is technically complex because dose deposition is sensitive to anatomical variations, range uncertainties, and biological factors such as linear energy transfer (LET) and relative biological effectiveness (RBE). The reviewed article describes how artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), is increasingly being incorporated to address these challenges.
AI applications include automated target and organ-at-risk segmentation, automated beam arrangement and plan generation, rapid three-dimensional dose prediction, and optimization. Reinforcement learning can iteratively optimize beam parameters, while DL models can predict dose distributions and potentially LET/RBE maps. AI also facilitates adaptive treatment planning by predicting anatomical changes such as tumor shrinkage, organ motion, and weight loss, enabling more rapid replanning. Hybrid AI–physics approaches, combining data-driven models with Monte Carlo or analytical dose calculations, are particularly promising because they aim to combine computational efficiency with physical accuracy.
In conclusions, although AI-based approaches can reduce planning time and improve consistency and adaptability, clinical implementation remains limited by insufficient and heterogeneous datasets, limited model generalizability, black-box behavior, regulatory requirements, and the need for rigorous quality assurance. Multi-institutional validation, standardized datasets, explainable AI, and human-in-the-loop workflows are therefore essential.
Future development should focus on hybrid AI–physics models, reinforcement learning, federated learning, and potentially quantum machine learning. The authors conclude that AI could become an integral component of precision particle therapy, but successful translation requires robust clinical validation, transparency, multidisciplinary oversight, and appropriate regulatory frameworks.
Published by International Journal of Particle Therapy.
