Validates AI radiotherapy treatment plans

University College London and the London School of Hygiene & Tropical Medicine said on May 18 that an international trial found artificial intelligence could generate radiotherapy treatment plans for cervical and prostate cancers that met predefined quality standards. The findings were presented at the European Society for Radiotherapy and Oncology congress in Stockholm, according to a UCL release. The study, called ARCHERY, tested whether software could carry out parts of treatment planning usually done by oncologists and medical physicists. Researchers said the work was designed to assess whether the system could be used across both high-income and lower-resource settings. ### Which parts of radiotherapy planning did the software take over? Radiotherapy planning involves outlining tumors and nearby organs on CT scans and then deciding the position, shape and size of radiation beams, UCL said. The ARCHERY trial tested AI software that automatically identifies target structures and determines beam configurations, tasks the researchers said can otherwise take days or weeks. ClinicalTrials.gov describes the same two steps as the core functions under evaluation in the study record for NCT05653063. (news-medical.net) The Radiation Planning Assistant, or RPA, is the software named on the ARCHERY study website. The site says the trial’s primary aim is to determine whether the tool can produce suitable plans for cervical, head and neck, and prostate cancers, and whether it is faster and more cost-efficient than traditional approaches. ### What did the trial report for cervical and prostate cancer? The new results showed the AI produced radiotherapy plans to a high standard in more than 95% of cervical cancer cases, according to UCL’s release. (news-medical.net) For prostate cancer, the figure was 85%, which the release said would still be considered suitable for routine clinical use. The release did not provide a full methods paper or breakdown of the benchmark in the text available online, but it said the trial assessed whether the technology could achieve international best-practice planning. (archery.mrcctu.ucl.ac.uk) More than 1,000 patients were included across three disease groups — cervix, prostate and head and neck — at hospitals in India, South Africa, Jordan and Malaysia, the researchers said. ClinicalTrials.gov lists an estimated enrollment of 990 and names recruiting sites including Tata Medical Centre in Kolkata, Tata Memorial Hospital in Mumbai, King Hussein Cancer Center in Amman and a site in Kuala Lumpur. (news-medical.net) ### Why were lower-resource settings central to this study? UCL said 94% of cervical cancer deaths occur in low- and middle-income countries and cited 350,000 deaths in 2022. The same release said only 10% of people who need radiotherapy in low-income countries receive it, compared with 40% in middle-income countries. Researchers identified shortages of skilled professionals as one of the main barriers. (e3.eurekalert.org) Ajay Aggarwal, the trial’s chief investigator and a professor at LSHTM and Guy’s & St Thomas’ NHS Trust, said the results support routine use for cervical cancer in hospitals globally and could support prostate cancer treatment “in any country setting,” according to the release. Aggarwal also said standard planning can take many hours over several weeks because it depends on the availability of specialized staff. (news-medical.net) ### Where were these results presented, and what is still pending? ESTRO’s 2026 program lists “The ARCHERY trial: Driving global AI implementation through evaluation” in a Stockholm session on May 16, with Jeannette Parkes named as speaker. UCL said the cervical and prostate findings were presented at the congress and that head-and-neck cancer results will be available later this year. (e3.eurekalert.org) ClinicalTrials.gov lists the study’s estimated primary completion date as December 1, 2025 and estimated study completion as September 1, 2026. The trial record remains posted as recruiting, and the identifier is NCT05653063. (clinicaltrials.gov) (estro.org)