ESTRO 2025 - Abstract Book

S2183

Interdisciplinary – Education in radiation oncology

ESTRO 2025

2524

Digital Poster IPEM AI Survey 2024: A focus on radiotherapy responses Paul J Doolan 1 , Sofia Michopoulou 2 , Virginia Marin Anaya 3 , Alison Starke 4 , Richard Meades 5

1 Medical Physics, German Oncology Center, Limassol, Cyprus. 2 Medical Physics, University Hospital Southampton NHS Foundation Trust, Southampton, United Kingdom. 3 Radiation Protection, University College London Hospitals NHS Foundation Trust, London, United Kingdom. 4 Radiotherapy, NHS Greater Glasgow and Clyde, Glasgow, United Kingdom. 5 Nuclear Medicine, Royal Free London NHS Foundation Trust, London, United Kingdom Purpose/Objective: For eight weeks, between May and July 2024, the IPEM AI Group launched a survey to gauge the attitudes and opinions of UK medical physics and clinical engineering (MPCE) staff towards AI. The survey was open to all MPCE professionals working in the clinic, academia or in industry. This work extracts the radiotherapy responses and compares them to other disciplines and the wider cohort. Material/Methods: Questions were divided into eight sections: background demographics; usage; change of role; job fears; training; preparedness; concerns; and open-ended. Questions were multiple choice and free text. Results: Of the 410 responses, 152 (22.6%) were from radiotherapy. Radiotherapy was among the highest users of AI at work (75.7%), which is a similar figure to MRI (72.2%) and much higher than Nuclear Medicine (49.4%) or Diagnostic Radiology (42.9%). Despite this high usage, the number of radiotherapy respondents that stated they had any prior AI knowledge (9%) was not higher than the overall cohort (10%). In radiotherapy, the most common usage is for segmentation (69.7%), with MVision (n=44), Limbus (27) and RayStation (19) cited as the most common applications used (Figure 1). Radiotherapy respondents showed no more concern about their change of role or job loss than any other discipline. Around half (n=85) feel at least partly under pressure to learn about AI tools, nearly three-quarters (n=112) show some concern about keeping pace with AI advancements, while more than 80% (n=122) are investing in their own AI knowledge and skills. Despite this investment, only 22% (n=33) of radiotherapy respondents positively stated they knew where to look for resources to develop their AI knowledge and skills. Alongside Clinical Scientific Computing (88%), Radiotherapy was among the disciplines that felt most prepared for AI integration (79%) – the value for the overall cohort (64%) or some individual disciplines, such as ultrasound (54%), was much lower. Radiotherapy generally felt their employer was following good practice guidelines (74%), which was significantly higher than the overall cohort (53%).

The proportion of radiotherapy respondents having, at least some, concerns about quality (87%), consistency (70%) and bias (93%) of AI results were of a similar magnitude to the whole cohort.

Conclusion: This survey shows that while radiotherapy is among the highest users of AI, its members lack formal training or know where to find appropriate resources. There are also concerns about the results produced by AI. These results can inform national or local AI policies.

Keywords: AI, survey, physics