Foundation Models for Epileptogenic Zone Identification in Drug-Resistant Epilepsy

The article presents EpiiSLM, a dual foundation model system designed for identifying the epileptogenic zone (EZ) in drug-resistant epilepsy using stereo-electroencephalography (sEEG). EpiiSLM is trained on 104,990 minutes of sEEG data, achieving a contact-level positive predictive value (PPV) of 0.978 and 100% region-level accuracy, significantly surpassing the seizure onset zone baseline. This model's capability to operate with just one night of interictal sleep data could reduce the need for extensive sEEG monitoring, thereby enhancing surgical outcomes for patients.