Recent advances in applying machine learning to proton radiotherapy.

Area of Science:

• Radiation Oncology
• Medical Physics
• Artificial Intelligence in Healthcare

Background:

• Precision and timeliness are critical in radiation oncology patient care.
• Machine learning (ML) is established in photon radiotherapy but emerging in proton therapy.
• Proton therapy's clinical workflow applications of ML remain underexplored.

Purpose of the Study:

• To systematically review current and potential ML applications in the proton therapy clinical workflow.
• To identify and summarize key ML techniques and their impact on proton therapy processes.

Main Methods:

• Systematic literature search of PubMed and Embase (2019-2024).
• Search terms included 'proton therapy', 'machine learning', and 'deep learning'.
• Summarized and incorporated findings from 38 relevant studies.

Main Results:

• U-Net architectures are prevalent for patient pre-screening.
• Convolutional neural networks are key for dose and range prediction.
• ML models improve image quality, reduce radiation exposure, and enhance online dose/range monitoring.

Conclusions:

• ML models are increasingly vital for proton therapy treatment and discovery.
• ML significantly enhances patient screening, planning, image quality, and dose/range calculations.
• Machine learning is advancing the precision and personalization of proton therapy.