AI for Radiotherapy Auto-Contouring: Current Use, Perceptions of, and Barriers to Implementation
Sumeet Hindocha,
United Kingdom
PO-1582
Abstract
AI for Radiotherapy Auto-Contouring: Current Use, Perceptions of, and Barriers to Implementation
Authors: Sumeet Hindocha1, Kieran Zucker2, Raj Jena2, Kathryn Banfill2, Katherine Mackay2, Gareth Price3, Delia Pudney2, James Wang2, Alexandra Taylor2
1The Royal College of Radiologists, AI in Clinical Oncology (AICO), London, United Kingdom; 2Royal College of Radiologists, AI in Clinical Oncology (AICO), London, United Kingdom; 3Royal College of Radiologists , AI in Clinical Oncology (AICO), London, United Kingdom
Show Affiliations
Hide Affiliations
Purpose or Objective
Manual contouring of tumour and OARs is laborious but crucial a component of radiotherapy (RT). Errors here can lead to underdosing of the tumour or increased toxicity impacting on survival and quality of life. Commercial AI auto-contouring tools are becoming increasingly available. Their clinical use raises important considerations including quality assurance, validation, education, training and job planning. Despite this, there is little in the literature capturing the views of Clinical Oncologists regarding these factors. The Royal College of Radiologists (RCR) surveyed UK Clinical Oncologists on their perceptions, current use of and barriers to using AI-based auto contouring for RT.
Material and Methods
Survey questions consisted of close-ended questions and open-ended statements with options to select. Free-text comments were invited where necessary. The survey was conducted through the RCR’s Insights Panel and was sent to trainees, SAS doctors and consultants in the Faculty of Clinical Oncology.
Results
The survey was sent to 236 subscribers of the Insights Panel. 51 responded (response rate = 22%, considered acceptable by the Insights Panel). 40 respondents (78%) were consultants, 7 (14%) were trainees and 3 (6%) were of another grade.
40 (78%) felt the impact of AI on RT would be positive. None felt that AI would replace their role. 25 (49%) felt that use of AI in RT would decrease risk for patients and 11 (22%) felt there would be no change in risk to patients. 2 (4%) felt AI would increase the risk to patients.
We asked about the percentage of OAR contouring that was performed by AI or various staff (Fig 1). 19 (37%) reported that consultants undertook ≥60% of all OAR contouring. 2 reported that AI auto-contouring was used for the majority of OAR contouring.
23 respondents (45%) reported that AI auto-contouring is being used clinically in their departments. This is predominantly for OAR contouring for head and neck, brain, thorax and prostate radiotherapy (Table 1).
25 respondents replied to the question on how much time AI auto-contouring saved in a typical week. 15 (60%) reported a time saving, with 9 (36%) reporting this to be >1 hour per week.
Respondents were also asked what the key priorities regarding AI should be for the RCR and to share their views or experience with AI for auto-contouring. 5 categories emerged:
1) Validation and Quality Assurance
2) Education, Training and Guidance
3) Understanding the Impact on Clinical Oncologists
4) Patient Engagement
5) Rolling Out Clinical AI tools for RT
Conclusion
AI auto-contouring tools are in current clinical use and their use expected to grow. There is consensus on validation and quality assurance and there is a marked appetite for urgent guidance, education and training. Careful coordination is required to ensure that all RT departments and the patients they serve, may enjoy the benefits of AI in RT. Professional organisations such as the RCR have a key role to play in delivering this.