Machine learning applications in radiation oncology: Current use and needs to support clinical implementation.

Area of Science:

• Radiation Oncology
• Medical Physics
• Artificial Intelligence
• Machine Learning

Background:

• Emerging use of AI/ML applications in radiation oncology.
• Absence of clear guidelines for commissioning ML-based applications.
• Need to understand current practices and requirements for clinical implementation.

Purpose of the Study:

• Investigate current AI/ML application usage in radiation oncology.
• Identify needs for supporting the implementation of ML applications in clinical practice.

Main Methods:

• Survey distributed to radiation oncology medical physicists.
• Survey covered clinical applications, model training, acceptance, commissioning, quality assurance (QA), and GDPR.
• Data collected from 213 medical physicists across 202 centers.

Main Results:

• 69% of centers use or prepare to use ML, primarily for contouring and treatment planning.
• Human oversight remains critical (86%) for ML output quality checks.
• Limited knowledge on ethics, legislation, and data sharing; need for guidelines, training, and multicenter data.

Conclusions:

• Survey highlights a significant need for education and guidelines.
• Implementation and QA protocols for ML applications are essential.
• Addressing these needs will facilitate beneficial clinical integration of ML in radiation oncology.