Using AI for Diagnostic and Prognostic Prediction in Paediatric Brain Cancer

Childhood brain cancer is the second most common form of childhood cancer, though it holds
the highest mortality rate. The need for invasive surgeries, radiation therapy, and/or
chemotherapy is both traumatic for the child and can result in lifelong disabilities. These disabilities
can be intellectual, physical, and/or developmental. While radiologist and oncologists can
differentiate between broad tumour types from MRI imaging, precise diagnosis requires a
comprehensive understanding of a tumour’s genomics and epigenomics. Utilising the tumour’s
genomics and epigenomics, clinicians now diagnose brain cancer and risk stratify therapy
specific to a patient (precision medicine). A physical tumour sample is required to determine a
tumour’s genomics and epigenomics. In non-brain cancers, a biopsy can be taken without any
major risk to the patient. However, in brain cancers this is not possible given the requirement to
cut open the skull. Current best practice in the majority of cases is to remove as much of the
brain tumour as possible in an initial surgery in order to reduce the likelihood of further
surgeries. However, this surgery itself carries risks and may in some cases be futile.
Our research will investigate the possibility of using deep learning techniques to investigate
physical features of the tumour (impossible to discern with the human eye) derived from MRI
images, together with clinical, pathologic (including molecular data) and demographic data, to
build diagnostic AI models to identify brain tumour genomic and epigenomic information without
needing a physical tumour sample. This information can be further used in the development off
prognostic AI models of key outcomes among paediatric brain cancer patients, including
mortality and morbidity. A performant model would then allow a clinician to tailor treatment at an
earlier stage prior to surgery, aid in determining the surgical strategy, and provide more
information earlier to the child and their family to assist in informed decision making and surgical planning.