Status: Funded - Open
Felix Richter, MD, PhD
BACKGROUND: Neurologic changes are assessed by physical exam, which is conducted at limited time points, can be delayed, is variable between examiners, and may not discern subacute changes. Neonatal seizures are especially subtle and can be challenging to differentiate from normal movements. GAP: Hypoxic ischemic encephalopathy (HIE) is a neonatal syndrome that occurs due to brain asphyxia, usually due to a sentinel peripartum or intrapartum event, and has high mortality and morbidity including quadriplegia, cerebral palsy, and epilepsy. Infants with HIE are monitored with video, EEG, telemetry, vitals, and serial labs, thus serving as an excellent patient population to study the potential of computer vision and multimodal neuromonitoring. HYPOTHESIS: We will apply computer vision and multimodal neuromonitoring to the largest neonatal video EEG database in the world, and we hypothesize that we will achieve significantly better neonatal seizure detection in HIE when compared to existing EEG-only algorithms. METHODS: We will train computer vision and integrative machine learning models on video EEG and other clinical data to predict seizures in neonates with HIE. RESULTS: Pending. IMPACT: The long-term potential of this work includes improved seizure monitoring in the NICU, quantified outcomes for research (e.g., phenotypes to study the genetic penetrance of epilepsy, clinical trial endpoints), and technology that could help the growth and development of all infants, even those outside the NICU.