Machine learning approaches for improving EEG data utility in SUDEP research

点击翻译按钮获取中文摘要

Project Summary The parent R01 project will test the hypothesis that Sudden Unexpected Death in Epilepsy (SUDEP) cases exhibit different clinical, electroclinical and imaging features that can be identified and validated (Aim 1) and then incorporated into an individualized Bayesian risk prediction model (Aim 2). The study will compare SUDEP cases with age/sex-matched living epilepsy patients to identify clinical features and biomarkers, focusing on electroencephalography (EEG), electrocardiogram (ECG), and magnetic resonance imaging (MRI) data that are easily obtained during routine clinical visits. Potential biomarkers include postictal generalized EEG suppression, interictal ECG heart-rate variability, and decreased volume in limbic and brainstem regions on structural MRI scans. To leverage state-of-the-art computational tools for biomarker discovery, the parent R01’s Aim 3 employs artificial intelligence (AI) and machine learning (ML) techniques to uncover novel biomarkers from interictal EEG data. The proposed supplemental project is closely aligned with the parent R01’s Aim 3 and builds on the base of augmented datasets and new AI/ML techniques. Our research team consists of SUDEP and AI/ML experts with complementary expertise who are uniquely qualified to develop innovative analytic tools for EEG data AI/ML- readiness. In Aim 1, we will develop ML models to enhance data interpretation. In Aim 2, we will employ data augmentation techniques to improve the consistency of labeled EEG data from both SUDEP cases and living epilepsy patient controls. Overall, this administrative supplemental proposal will further enrich the research aims in our parent grant, and promote research rigor, transparency and reproducibility. Accomplishing these aims will maximize the data utility and improve AI/ML-readiness in epilepsy research.