Identification of Post-Ictal Generalised EEG Suppression with Two-Channel EEG

Area of Science:

• Epilepsy research
• Neuroscience
• Medical device technology

Background:

• Scalp electroencephalography (EEG) is standard for epilepsy monitoring but can be invasive and limited for long-term use.
• Postictal generalized electrographic seizures (PGES) are critical indicators for Sudden Unexpected Death in Epilepsy (SUDEP) risk.
• Minimally invasive, ultra-long-term monitoring solutions are needed for comprehensive epilepsy management.

Purpose of the Study:

• To evaluate the feasibility of a two-channel subcutaneous EEG (SubQ) device for detecting and monitoring PGES.
• To assess the performance of a Variational Auto-Encoder (VAE) model for PGES detection and duration prediction using SubQ data.
• To compare SubQ EEG performance against traditional scalp EEG for PGES monitoring.

Main Methods:

• Utilized annotated scalp EEG and SubQ EEG data from UNEEG Medical A/S.
• Implemented a pre-processing pipeline including channel reduction, resampling, filtering, and feature extraction.
• Employed a VAE for anomaly detection to identify PGES, followed by post-processing for duration prediction.

Main Results:

• The VAE achieved 100% PGES detection rates for both scalp and SubQ datasets.
• Predicted PGES durations showed an average offset of 35.67s (scalp EEG) and 26.42s (SubQ).
• False positive rates were 59% (scalp EEG) and 56% (SubQ), indicating a need for model refinement.

Conclusions:

• Subcutaneous EEG is a potentially valuable tool for epilepsy studies and PGES monitoring.
• The SubQ device offers a minimally invasive approach for ultra-long-term epilepsy monitoring.
• Further research is required to optimize algorithms for reduced false positives and improved duration prediction in SubQ EEG for SUDEP risk assessment.