AI designs radiotherapy 'gold-standard' plans
- On May 18, UCL and LSHTM said an international ARCHERY trial found AI-generated radiotherapy plans matched human gold-standard planning for cervical and prostate cancers. - The key result was 95%-plus high-standard planning in cervical cancer and 85% in prostate cancer, with findings presented at ESTRO 2026. - Head-and-neck cancer results are due later in 2026 from ARCHERY, which recruited 990 patients across six hospitals.
University College London and the London School of Hygiene & Tropical Medicine said on May 18 that an international clinical trial found AI software could generate radiotherapy treatment plans judged equivalent to gold-standard human plans for cervical and prostate cancer. The results were presented at the European Society for Radiotherapy and Oncology congress in Stockholm, according to UCL. The study, called ARCHERY, was run across hospitals in India, South Africa, Jordan and Malaysia and was designed to test whether automated planning could meet international best-practice standards in routine care. The work focused on a narrow but labor-intensive part of cancer treatment: planning where radiation should go, how strong it should be, and how to avoid healthy tissue. ### What exactly did the AI do in the radiotherapy workflow? Radiotherapy planning requires clinicians to outline tumors and nearby organs on CT scans, then configure radiation beams to hit the cancer while limiting damage to surrounding tissue, UCL said. Those steps are usually handled by oncologists and medical physicists and can take many hours, or in some settings days to weeks. In ARCHERY, the software automated two tasks: identifying target structures and organs at risk on CT images, and determining beam position, size and shape for treatment planning. (ucl.ac.uk) The Radiation Planning Assistant, the AI software used in the study, was built to produce plans suitable for cervical, prostate and head-and-neck cancers, according to the ARCHERY study site. Participants still received standard clinical care, and the study was observational rather than a trial of AI-directed treatment decisions at the bedside. ### How strong were the results for cervical and prostate cancer? (ucl.ac.uk) UCL said the AI planned radiotherapy to a high standard in more than 95% of cervical cancer cases. For prostate cancer, the figure was 85%, which UCL said would still be considered suitable for routine clinical use. Head-and-neck cancer results were not included in the May 18 announcement and are expected later this year. (archery.mrcctu.ucl.ac.uk) Professor Ajay Aggarwal, the study’s chief investigator at LSHTM and Guy’s and St Thomas’ NHS Trust, said the findings showed the technology reached “a very high standard” for cervical cancer and could support routine use in hospitals globally. He also said it could support prostate cancer treatment delivery “in any country setting.” (ucl.ac.uk) ### Why was this tested across lower-resource settings? ARCHERY was set up around a specific bottleneck: access to radiotherapy planning expertise in low- and middle-income countries. UCL said 94% of cervical-cancer deaths occur in low- and middle-income countries, and only 10% of people who need radiotherapy in low-income countries receive it; the figure is 40% in middle-income countries. The trial materials say shortages of equipment and specialized staff are major barriers to expanding access. (ucl.ac.uk) The study planned to recruit 990 patients across six public-sector hospitals in India, Jordan, Malaysia and South Africa, according to UCL’s trial unit. UCL said the broader trial involved more than 1,000 patients across cervix, prostate and head-and-neck cancers, reflecting the study’s international, multi-center design. (ucl.ac.uk) ### Does this mean AI is replacing radiation oncologists? The trial materials describe the software as automating bounded planning tasks rather than replacing the full radiotherapy pathway. UCL said the aim was to achieve international best-practice planning standards and make expert capacity go further, while the ARCHERY site says patients continued on the standard treatment pathway without extra imaging or additional visits. (ucl.ac.uk) The practical implication in the study documents is speed and consistency. ARCHERY says automation could reduce planning time from hours to minutes and reduce variation between clinicians, while UCL said the software can complete tasks that traditionally take days or weeks in some settings. ### What happens next? UCL said the next reported ARCHERY readout will cover head-and-neck cancer later in 2026. (ucl.ac.uk) The trial website says the study is also evaluating whether AI-based planning is faster and more cost-efficient than traditional approaches, with funding from the Rising Tide Foundation and the U.S. National Cancer Institute. (innovative-ctu.ucl.ac.uk)
