Artificial Intelligence in Proton Therapy: What Works, What Does Not, and What Is Next

Area of Science:

• Medical Physics
• Oncology
• Computer Science

Background:

• Proton therapy faces challenges due to range uncertainty, anatomical variations, and biological heterogeneity.
• Traditional algorithms struggle with complex data integration and relationships across the clinical workflow.

Purpose of the Study:

• To review current and emerging artificial intelligence (AI) applications in proton therapy.
• To discuss AI's role in addressing proton therapy's unique challenges and improving workflows.

Main Methods:

• Review of current and emerging AI applications in proton therapy.
• Exploration of AI's impact on imaging, treatment planning, quality assurance, and adaptive strategies.
• Discussion of AI's role in robustness, uncertainty management, and biological optimization.

Main Results:

• AI methods, including machine learning and deep learning, integrate heterogeneous data for complex relationship capture.
• AI applications span image reconstruction, synthesis, segmentation, dose prediction, and adaptive treatment strategies.
• AI enhances quality assurance and workflow coordination, moving towards workflow-level intelligence.

Conclusions:

• AI is pivotal in advancing proton therapy, offering solutions for range uncertainty and biological variability.
• Clinical validation, safety, interpretability, and accessibility are crucial for AI adoption in proton therapy.
• AI integration promises to enhance efficiency and effectiveness in proton therapy delivery and patient care.