Jeremi Chabros, MA, MPH, MD

Brain stimulation for medication-resistant epilepsy yields dramatically varied outcomes, with up to 50% of patients failing to respond. Current practice lacks tools to predict success, revealing a critical gap in understanding stimulation mechanisms. We will reverse engineer successful therapy by analyzing the largest multi-institutional dataset of chronic brain recordings from responsive neurostimulation (RNS) devices (>200 patients). First, using explainable AI, we will integrate multimodal biomarkers to predict who will respond and when. Crucially, we then employ causal inference to move beyond correlation and discover why therapy works, identifying stimulation patterns that drive lasting seizure reduction. This project will deliver a scientifically robust, clinically-meaningful tool for patient selection and temporal phenotyping, and an evidence-based protocol to optimize treatment, reducing non-responder rates and establishing a blueprint for precision neuromodulation.