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

Diffuse midline glioma (DMG), a primary tumour within the most sensitive part of the brain, is a fatal disease primarily affecting children between 4-7 years of age. The overarching aim of this project is to build a treatment decision support platform facilitating personalized radiotherapy (RT) optimization based on MRI for afflicted pediatric patients. The researchers will develop a novel analytical pipeline bridging mechanistic modelling and data-driven machine learning. The platform will ultimately help selecting treatment options, and monitor response to treatment using non-invasive image-based biomarkers to treatment. It will also inform RT scheduling and dosing with an individualised radiosensitivity score. The ultimate goal will be to develop a digital health tool that is readily translatable to clinics worldwide to guide doctors in designing optimal treatment strategies for affected children and their families.

Pediatric diffuse midline glioma (DMG) is a rare fatal disease with currently no curative treatment. Owing to the delicate location of these tumours in the brain, treatment options and surgical interventions are greatly limited. Radiotherapy (RT) is one of the few life-prolonging treatments, but its therapeutic efficacy varies between individuals. Also, the current one-size-fits-all therapy is mostly based on clinical experience in adults. Hence, the efficacy of RT may be improved by tailoring it to the clinical needs of individual paediatric patients.

Since tumour anatomical location in the brain hinders an invasive biopsy, the researchers will investigate whether non-invasive magnetic resonance imaging (MRI) could be used to infer the tumour's radiosensitivity or response to RT. They will use a combinatorial approach by employing two powerful tools originating from different research fields: machine learning and modelling of tumour growth with differential equations. This combination of models facilitates not only to identify which child will benefit from conventional RT but also if and how the impact of RT can be maximised by changing its scheduling and dosing. By evaluating the criteria on which the algorithm made the decision for effective or non-effective RT, the researchers can infer information on the underlying tumour characteristics and how these are represented in the images. This will improve our biological understanding of DMG and could provide indications of which other therapies to combine with RT in order to improve the tumour's radiosensitivity.

As such, the researchers aim to create a digital health tool to guide doctors in their choice of the optimal treatment strategy to improve the quality of life for children and their families faced with the devastating diagnosis of DMG. Since the approach investigates cost-effective and broadly available RT and does not require biopsy information, it is readily translatable to clinics worldwide without the need for a specialised treatment centre.

Banner image above: The combination of machine learning and the modelling of tumour growth with differential equations will enable the creation of a digital health tool that will improve the quality of life of children diagnosed with diffuse midline glioma.

Additional Information

  • The research is part of the Postdoctoral Excellence Programme.
  • PEP fellow Dr Sarah Brüningk is hosted by BRCCH-funded Principal Investigator Prof Catherine Jutzeler (ETH Zurich), in collaboration with Prof Karsten Borgwardt (ETH Zurich) and Prof Javad Nazarian (University Children's Hospital Zurich).

Consortium Members
Related News