Omni-iEEG: A Large-Scale, Comprehensive iEEG Dataset and Benchmark for Epilepsy Research
The 178-hour dataset with 36K expert annotations aims to standardize AI models for drug-resistant seizure localization.
A UCLA-led research team has released Omni-iEEG, a large-scale intracranial EEG (iEEG) dataset and benchmark for epilepsy research. It harmonizes data from 302 patients across 178 hours of recordings, includes over 36,000 expert-validated pathological event annotations, and defines standardized clinical evaluation tasks. This resource enables ML researchers to develop and benchmark models for localizing epileptogenic zones, directly addressing the challenge of drug-resistant epilepsy affecting one-third of patients.
Why It Matters
It provides the first standardized foundation for developing clinically translatable AI that could improve surgical outcomes for millions with drug-resistant epilepsy.