Radiotherapy Professionals Embrace AI as a Clinical Support Tool

Radiotherapy planning requires precise identification of organs at risk near a tumor, but manual contouring is repetitive and time-consuming, often delaying plan completion and straining specialist capacity. Some clinicians also worry that artificial intelligence could erode skills or displace roles. A new study shows that AI contouring tools can streamline early planning steps while preserving expert oversight.

Researchers at King’s College London report that radiotherapy professionals across five regional cancer treatment centers in England were overwhelmingly positive about AI tools that generate first-draft contours of organs at risk (OAR) during treatment planning. The study was published in Sociology of Health & Illness. Clinicians used the software to propose organ boundaries and then reviewed and refined those outputs before finalizing plans.

image Credit: iStock

The research assessed the lived experiences of 32 health care professionals, including clinical scientists, radiographers, dosimetrists, and oncologists. Participants described the technology as removing the most routine stage of delineation while keeping expert control of quality, safety, and accountability. Professionals emphasized that they retained responsibility for corrections and for the final treatment decision.

Interviewees reported that manual OAR drawing typically took between half an hour and two hours per patient, whereas AI produced a first draft in five to 10 minutes. They also noted that the systems could make basic mistakes, especially when anatomy or local practice diverged from training expectations, underscoring the continued need for human judgment. This “partial discard” of work helped explain why the tools were accepted so positively.

Clinicians said time saved by AI-supported contouring was redirected toward complex planning, service improvement, research, and more personalized patient care. The authors concluded that acceptance of AI in oncology depends on how well it fits specific tasks and workflows and whether it supports, rather than competes with, professional expertise.

“People worry about it replacing clinicians. I don't think our expertise is drawing around lumps and bumps,” said one oncologist.

“The research is timely as employers across health care and other professional sectors face growing pressure to introduce AI into complex, skilled work. Its central lesson is clear: AI may be more readily accepted when it is designed and implemented as a tool for experts, rather than an expert system that competes with them,” said Dr. Juan Baeza, reader in health policy at King’s Business School.

Related Links
King’s College London