Artificial intelligence has the potential to significantly improve radiation protection and patient safety across medical imaging and cancer treatment applications, but its safe deployment will require strong human oversight, robust validation processes, and effective regulatory frameworks, according to experts at a recent meeting organised by the International Atomic Energy Agency (IAEA).

The meeting brought together 187 specialists from 65 countries, along with representatives from 18 international organisations, professional bodies, and safety initiatives, to discuss the growing role of AI in radiation protection across radiology, radiotherapy, and nuclear medicine.

Experts at the meeting emphasised the importance of what they described as “integrated intelligence”, a model that combines AI-driven technologies with human expertise and clinical oversight to ensure patient safety and responsible decision-making.

“The concept of ‘integrated intelligence’  combining artificial intelligence technology and tools with human judgement and oversight was highlighted as key to employing the many benefits that AI can bring to patient safety,” said Chadia Rizk, Radiation Protection Specialist at the IAEA.

According to the discussions, radiation protection is increasingly evolving toward a more adaptive, patient-specific, and data-driven approach. AI technologies are now becoming deeply integrated into clinical workflows, supporting healthcare professionals in tasks such as clinical decision-making, image interpretation, automated organ segmentation, treatment planning, and radiation dose monitoring.

Experts noted that AI systems could enable real-time radiation dose estimation and help identify potentially high exposure levels, thereby improving optimisation of protection and safety while reducing unnecessary repeat examinations and avoidable radiation exposure.

The technology also has the potential to enhance the justification and precision of medical exposures, supporting more efficient and personalised care pathways in diagnostic imaging and cancer treatment.

However, participants stressed that the successful adoption of AI in radiation protection depends heavily on the quality, reliability, and representativeness of the data used to train AI systems. They also highlighted the importance of continuous monitoring, clinical validation, and safe integration into healthcare environments.

“We need reliable, high-quality datasets that can be easily accessed, shared and applied across national, regional and global healthcare systems, to inform decision making using AI tools,” said Stine Sofia Korreman, Professor of Medical Physics at Aarhus University, Denmark, who chaired the meeting.

The discussions reflect growing international attention on the responsible deployment of AI in healthcare, particularly in high-risk clinical environments where patient safety, transparency, and regulatory compliance remain critical priorities.

As AI adoption accelerates across medical imaging and oncology, experts believe that balancing technological innovation with strong governance and human oversight will be essential to unlocking the full benefits of AI-enabled radiation protection systems.