Radiogenomic Analysis of Pediatric Brain Tumors

Brain and central nervous system (CNS) tumors are the most common solid tumors in children. While radiotherapy and chemotherapy have resulted in improving the survival outcomes, these aggressive therapies come at the cost of subsequent and significant long-term sequelae such as neurocognitive, psychosocial abnormalities, and hearing loss1. Most survivors, especially children under the age of 4, suffer from an impaired quality of life with growth deficits due to radiation over-exposure, and are prone to increased risk for secondary malignancies2. Currently, there has been limited work in the development of advanced machine learning and data integration tools to predict survival and predict which patients will respond to aggressive therapies and which ones need de-escalation in therapies. Hence, there is a pressing need to identify approaches using machine learning that can reliably integrate imaging, histopathological, and molecular signatures for improved subtype stratification and to identify alternative treatment options for pediatric tumors with poor prognosis.