Advancing radiotherapy with deep learning: A review of dose prediction models

Clinical Snapshot

Key Findings

• P Value: Not reported at the review level

• Effect Size: Not quantified — no pooled effect size reported; individual study metrics (e.g., DVH parameter agreement, dose difference indices) are described qualitatively as improved but not synthesised numerically

• Primary Outcome: Deep learning models demonstrated improvements in radiotherapy dose prediction accuracy and treatment planning efficiency compared to conventional approaches across multiple cancer sites and architectural paradigms

• Nnt Or Sensitivity: Not applicable — no NNT, sensitivity/specificity, or hazard ratio data synthesised; technical performance metrics such as mean absolute dose error and DVH fidelity are referenced descriptively but not pooled

• Confidence Interval: Not reported — no meta-analytic confidence intervals provided

Clinical Application

Clinical implementation remains contingent on institutional infrastructure, availability of large curated training datasets, regulatory approval of specific software tools, and integration with existing treatment planning systems (e.g., Eclipse, RayStation). Real-time adaptation and uncertainty quantification — identified as future priorities — are not yet clinically mature. Medical physicist oversight remains essential. In Australia, radiation oncology is delivered across public and private sectors under Medicare Benefits Schedule item numbers for treatment planning. The Therapeutic Goods Administration (TGA) regulates AI-based medical software as Software as a Medical Device (SaMD) under the ARTG framework — any deep learning dose prediction tool would require TGA conformity assessment before clinical deployment. The Australian Clinical Dosimetry Service (ACDS) and RANZCR provide quality frameworks relevant to adopting novel planning technologies. The RACGP has no direct guideline on radiotherapy planning AI, but referral pathways to radiation oncology remain unchanged. PBS funding is not directly implicated. Australian centres participating in TROG (Trans-Tasman Radiation Oncology Group) trials may be early adopters of validated deep learning planning tools, but multi-institutional Australian validation data are absent from this review. Adult patients with solid tumours undergoing external beam radiotherapy or brachytherapy where computerised treatment planning is employed; most directly applicable to high-volume sites including head and neck, prostate, lung, and breast cancers where deep learning dose prediction tools are most developed

Abstract

BACKGROUND: Artificial intelligence (AI), especially deep learning algorithms, holds significant promise in transforming radiation therapy by enabling more precise, personalized, and efficient treatment planning. In particular, deep learning models have demonstrated their utility in critical tasks such as dose prediction, helping to streamline complex processes and improve patient outcomes. METHODS: This paper provides a comprehensive review of recent research on deep learning applications for radiotherapy dose prediction, highlighting the architectures, datasets, and methodologies used. RESULTS: The review highlights the advancements in accuracy, efficiency, and robustness achieved by deep learning-based dose prediction models. These approaches help reduce treatment planning time and minimize human error, thereby improving clinical workflows. CONCLUSION: Deep learning offers promising solutions to enhance radiotherapy dose delivery. Future research should focus on developing the generalized dose prediction models, multi-institutional validation frameworks, explainable AI approaches for clinical trust, real-time adaptation capabilities, uncertainty quantification methods, and strategies for handling rare anatomical variations.

References

1. 1.Darkhor, P., Hashemi, B., Farajollahi, A., Jodeiri, A., & Ghasemi, A. (2026). Advancing radiotherapy with deep learning: A review of dose prediction models. Physica Medica: PM: An International Journal Devoted to the Applications of Physics to Medicine and Biology, 105827. https://doi.org/10.1016/j.ejmp.2026.105827