Project 465489

Pediatric low-grade gliomas (pLGG) are the most common brain tumours in children. With surgical excision as the main treatment, and chemotherapy and radiation as additional treatment options, death from these tumours is relatively rare, but the 10-year progression-free survival is less than 50%. Affected children commonly suffer multiple recurrences requiring multi-modal therapy leading to considerable morbidity. A robust imaging-based pLGG detection and prediction of treatment response is crucial for guiding the type, timing, and intensity of therapy to achieve an optimal outcome for patients. Recently, it has been shown that the genetic markers of pLGG can significantly impact patient outcome with improved prognosis and individualized treatment strategies. pLGG genetic markers are assessed by the analysis of the tumour tissue, which is obtained through surgery or biopsy. This invasive method only provides limited information on the tumour at a given point in time. Noninvasive magnetic resonance imaging (MRI) can provide longitudinal information on the tumour in its entirety. In this research, we will develop MRI-based Artificial Intelligence (AI) algorithms to predict pLGG genetic markers as early as possible. Currently, the response to therapy for pLGG patients is assessed by neuroradiologists measuring the change in tumour size between frequent follow-up MRI scans. This is prone to inconsistent and potentially inaccurate measurements and also puts a strain on the healthcare system. We will develop MRI-based AI algorithms that go beyond existing size-based approach, to predict treatment response as early as possible with the fewest post-treatment MRI examinations for individual patients. With optimal usage of healthcare resources, MRI-based AI tools could assist in early detection and clinical decision making for individualized treatment strategies for patients with pLGG.