Advancing radiomic tools for brain tumours: Enriching data sets and developing novel AI approaches

Brain tumours are the most common cause of cancer death in children. Despite impressive advances in our understanding of these tumours, it can often take several weeks to reach a diagnosis, and the ability to predict those who will survive is still limited. Biological studies have increasingly defined a larger number of tumour types and subtypes, so that all too often doctors are left saying to children and their families that the tumour is rare.
Scans, particularly Magnetic Resonance Imaging (MRI), are used to diagnose tumours. MRI scans contain thousands of pieces of data but are often summarised into a few simple pieces of information (E.g. size of tumour). Modern types of scans can also tell us about a tumour’s biology. Through advanced computing (radiomics), it is possible to extract much more information from MRI images than may be visible to the eye, and this can be combined with artificial intelligence (AI) approaches to help diagnose tumours and predict how they may behave. This works well for the more common tumour types, but more data is needed to improve performance and applicability and for rare tumour types, which overall make up about a quarter of all children’s brain tumours.
In addition, there are mathematical challenges for AI in addressing rare tumours. In this project, a researcher will help bring together MRI data from the CBTN with an existing large study in the UK (Imaging of Tumours Study) which contains and continues to collect imaging data of brain tumours across the . This resource will be used to improve how current AI tools can diagnose tumours. New mathematical/AI approaches will also be developed to improve the AI tools. The data generated will also support other and future projects to identify imaging features that can help predict the behaviour of a tumour.