For patients with head and neck cancer, radiotherapy can be life-saving. However, it can also cause complications that may affect patients long after treatment has ended. ‘Side effects such as swallowing difficulties or a dry mouth can have a huge impact on the quality of life,’ says Lisanne van Dijk, Associate Professor Radiotherapy.
Accurately predicting who will develop these complications remains challenging. As van Dijk explains: ‘Two patients may receive exactly the same radiation treatment and yet respond completely differently.’ Consequently, it is currently difficult to tailor treatment in a way that minimises harm while maintaining or maximizing treatment effectiveness.
‘Conventional toxicity models typically rely on a single average radiation dose per organ. But in reality, radiation damage is a three-dimensional process and is not distributed evenly throughout the body,’ Van Dijk explains. As a result, different parts of an organ receive different amounts of radiation. ‘Swallowing, for example, involves over 50 muscles that work together in symphony. That kind of complexity does not fit into a simple model,’ Van Dijk adds.
Moreover, patient anatomy and physiology differs. This means that even when patients receive comparable radiation treatment, how radiation is distributed and how the body responds may vary. As a result, conventional models cannot accurately predict the risk of side effects for individual patients.
To address this gap, van Dijk and her colleagues developed an AI model capable of capturing the complex factors that influence individual toxicity risk before treatment. With support from the Impact Accelerator Grant, the research team was able to translate these models into a practical clinical decision-support tool, designed to select, optimise and personalise treatment planning in clinical practice.
The tool enables clinicians to generate individual risk profiles and evaluate alternative radiotherapy strategies. This helps to identify the most suitable treatment plan for each patient. In addition, it includes an ‘inform patient’ module. This feature automatically generates a patient-friendly report outlining the predicted risks of specific side effects, and where possible, options to prevent or manage them. This facilitates more informed conversations between doctor and patient and helps patients to prepare emotionally for life after treatment.
The tool is currently being finalised for clinical implementation at UMCG. In the future, it may also be adapted for other tumour types, expanding its potential beyond head and neck cancer. These developments show how AI research can lead to tangible improvements in patient care. As van Dijk concludes: ‘It is not just about survival, but about how someone lives after treatment. By better predicting who is at risk of serious side effects, we can prevent a great deal of suffering, improve quality life for our patients and consequently reduce health care burden and cost.’
The UMCG offers the Impact Accelerator Grant to help researchers translate their findings into real-world impact. The grant is intended for ongoing or recently completed projects without funding for impact-oriented activities outside of academia. It provides funding for activities such as focus groups, public engagement, developing business models or other activities that help bridge the gap from research findings to impact. For more information about the Grant and the conditions, contact the Impact Team.
