Research involving 23 countries has demonstrated the safety and benefits of using artificial intelligence (AI) for contouring organs at risk, a critical and time-consuming step in cancer radiotherapy. The IAEA-coordinated ELAISA Study incorporated data from low- and middle-income countries (LMICs) to show how AI technology can improve radiotherapy access worldwide.
Contouring tumours and nearby healthy tissues is essential for the safe and effective use of radiotherapy. However, differences in how observers outline these areas—known as inter-observer variability—can affect the accuracy and consistency of treatment planning. Previous research has shown that instructor-led guidance workshops help reduce this variability.
Although nearly half of all cancer patients require radiotherapy at some point, global use of this treatment remains insufficient, partly due to a shortage of trained professionals. The IAEA-led Lancet Oncology Commission on Radiotherapy and Theranostics projects a need for over 84,000 radiation oncologists by 2050 to manage the estimated 35.2 million new cancer cases. May Abdel-Wahab, Director of the IAEA Division of Human Health and co-lead of the commission, noted that this figure represents a 60 percent increase over the 2022 workforce. She added that as cancer incidence and treatment complexity rise, radiation oncologists will need to allocate more time to contouring tasks.
To address these challenges, the IAEA studied how AI could assist with contouring head and neck cancers specifically in LMICs. While AI auto-segmentation algorithms have shown promise in retrospective studies, their clinical benefit in LMIC settings and impact on interobserver variability had been largely unexamined until now.
Abdel-Wahab emphasized that AI-assisted contouring can help improve the efficiency of radiation oncologists.
Nearly 100 radiation oncologists from 22 radiotherapy centers across Albania, Argentina, Azerbaijan, Bangladesh, Belarus, Costa Rica, Georgia, India, Indonesia, Jordan, Kazakhstan, Kenya, Kyrgyzstan, Malaysia, Moldova, Mongolia, Nepal, North Macedonia, Pakistan, Sudan, Tunisia, and Uganda participated in the study. Aarhus University Hospital in Denmark provided 16 head and neck cancer cases for analysis.
Participants were randomly divided into two groups: one used AI-assisted contouring and the other manual methods. After an online IAEA workshop on AI-assisted contouring, both groups continued contouring cases, first with their original method and then all using AI. A follow-up round using AI was conducted six months later.
The study found that AI assistance significantly improved contouring quality by reducing inter-observer variability and shortened contouring times, even without prior training. Instructional workshops further enhanced the time-saving benefits of AI-assisted contouring, although they only improved the quality for two specific organs at risk. These effects were sustained in short- and long-term follow-ups.
Jesper Grau Eriksen, clinical professor at Aarhus University and a lead investigator, stated that combining teaching with AI-assisted contouring was the most effective approach to reduce contouring time. He noted that appropriate implementation of AI tools can save resources and enable more radiation oncologists, especially in LMICs, to treat additional patients.
The study’s findings have been published in the Journal of Global Oncology and presented at the European Society for Radiotherapy and Oncology’s annual meetings.
Leave a Reply