Artificial intelligence to improve radiotherapy for cancer patients

Around half of all cancer patients receive radiotherapy. Despite technological advances, many of the processes involved are still manual and time-consuming. This can result in inconsistent and, in some cases, suboptimal treatment pathways.
With funding from the Novo Nordisk Foundation’s Grand AI Challenge, a cross-disciplinary research team will, through the AIM@cancer project, develop artificial intelligence to support decision-making before, during and after radiotherapy.

Central role in the project

Christian Rønn Hansen, Associate Professor at the Department of Clinical Research at the University of Southern Denmark (SDU) and medical physicist at Odense University Hospital, heads the national database DcmCollab. This database collects large volumes of imaging data from Danish cancer patients and forms the basis for developing AI models that can later be tailored to specific tasks along the treatment pathway.

He is also one of four Danish principal investigators on the AIM@cancer project and is responsible for recruiting researchers to Odense University Hospital to develop targeted AI models for use in radiotherapy.

Large-scale models with broad applications

– The AI models we aim to develop in this project are so-called foundational models – large image recognition models trained on data from CT and MRI scans, among others. Whereas previous AI solutions have typically been designed for a single, narrow task, these new models can be applied across multiple clinical purposes – from identifying tumours and predicting treatment outcomes to supporting follow-up care, says Christian Rønn Hansen.

The models will be designed to handle different types of health data – not just imaging, but also, for instance, patient records and previous treatment histories – making them more widely applicable. Transparency is also a key priority, ensuring that clinicians can understand and assess how the AI generates its recommendations.

– In other words, we make the uncertainties in each prediction clear, he explains.

Accurate and equitable solutions

– We may not be able to collect the world’s largest datasets here in Denmark, but our health data is of uniquely high quality and reflects the entire patient population – not just those who can afford a particular treatment. This gives us a distinct advantage when it comes to developing AI models that benefit all cancer patients, says Christian Rønn Hansen.

The upcoming AI tools will, among other things, help delineate tumours prior to treatment, monitor patient response during therapy, and identify risks of late effects or relapse after treatment has ended.

– The aim is to develop AI models specifically for radiotherapy that can adjust treatment as it progresses and ensure that each individual patient receives the most appropriate care – before, during and after radiotherapy, he adds.

The AIM@cancer project will officially launch in January 2026, with recruitment of researchers expected to begin in late summer 2025. In the long term, the AI tools developed should also be applicable in health systems with fewer resources – for example, in low- and middle-income countries.