EEG based system identifies brain regions linked to epilepsy activity
EEGBind: Detecting Source-Level Interictal Epileptiform Discharges via EEG-Centric Multimodal Binding
Machine LearningMultimedia
Summary
Epilepsy doctors need to know where in the brain abnormal signals come from to decide on surgery. The authors created EEGBind, a system that mainly uses brainwave data (EEG) and adds helpful video info to better find these signals and their brain locations. This method is careful to keep brainwave data clear and uses video only as support. Their system showed better accuracy than other leading methods on a public benchmark dataset.
What this means in practice
- •For clinical neurophysiology teams: Identify specific brain regions of epileptic activity using EEG supported by video to improve presurgical evaluation accuracy.
- •For neurology-focused medical device developers: Develop EEG analysis tools enhanced with video context to classify epilepsy-related brain activity sources with improved robustness.$Commercial implications: Enables selling diagnostic support devices combining EEG and video data to hospitals for epilepsy surgery planning.
Authors
Muchen Li, Anglin Liu, Xuetian Gao, Ruijian Xu, Jintai Chen
Abstract
Source-level analysis of interictal epileptiform discharges (IEDs) is relevant to presurgical evaluation and treatment planning because it helps characterize where epileptiform activity is likely to arise. Beyond detecting whether an IED is present, this setting requires assigning IED-positive activity to clinically meaningful brain-region categories. This setting is challenging because source-region evidence in short electroencephalography (EEG) windows can be subtle, partial, and affected by subject variability, class imbalance, and imperfect multimodal context. We present EEGBind, an EEG-centric multimodal binding framework for five-class source-level IED classification. EEGBind treats EEG as the primary modality and binds synchronized video-context features around an EEG-centric representation. Instead of relying on early or overly strong multimodal fusion, which may perturb the source-sensitive EEG representation, EEGBind uses video context as auxiliary evidence for robust classification. A view-consistent repair stage is further used to improve hidden-set robustness while preserving the learned source-class boundary. On the NeuroMM 2026 Grand Challenge Track 3 NMM-Source-IED benchmark, EEGBind achieves 0.8395 on weighted-F1 and outperforms strong competitors. These results support EEG-centric multimodal binding as a practical strategy for source-level IED classification. The open-source code is available at https://github.com/HKUSTGZ-ML4Health-Lab/NeuroMM2026_IED_Detection.