NeuroMM 2026: Interictal Epileptiform Discharge Detection and Localization in Multimodal Neuro-Signals
Abstract
Interictal epileptiform discharges (IEDs) support epilepsy diagnosis, but automated analysis is challenged by sparse, heterogeneous events that resemble normal variants and physiological or acquisition artifacts. NeuroMM 2026 advances multimodal neurophysiological intelligence by combining EEG-centered physiology with synchronized multimedia evidence of behavior and recording context. Its public release contains 25,426 four-second epochs from 84 subjects, while held-out evaluation spans physiological IED detection, context-enhanced detection, and five-class spatial-distribution classification. Controlled baselines show that visual context can substantially improve compatible models but is not consistently beneficial, and validation-to-hidden rank shifts expose generalization limits. The benchmark supports reproducible IED research and motivates reliability-aware fusion, artifact robustness, and deployment-efficient evaluation.