Harnessing Machine Learning and Mechanistic Modelling for Personalized Radiotherapy of Paediatric Diffuse Midline Glioma

Diffuse midline glioma (DMG) is a tumour that localizes in delicate and inaccessible regions of the brain, primarily affecting children aged four to seven and accounting for the highest mortality rate among paediatric brain tumours. Due to the precarious location of these tumours, treatment options are restricted to radiotherapy, which aims to alleviate the patient’s symptoms and prolong their life. However, the current one-size-fits-all therapy has been calibrated based on clinical experience in adults, and its efficacy when treating paediatric tumours varies among patients.

The goal of this research consortium was to establish a digital prognostic model to support clinicians in developing an effective treatment plan against DMG in paediatric patients. By combining mathematical modelling and machine learning, more than 1,000 magnetic resonance imaging scans were used to predict the volume growth and spread of DMGs. Additionally, the researchers sought to identify digital markers to predict tumour response to drugs based solely on imaging-derived information, allowing for personalized treatments for children with DMG.