Towards Long-term Non-invasive Monitoring for Epilepsy via Wearable EEG Devices

TL;DR

This paper develops and optimizes seizure detection algorithms for wearable EEG devices, achieving high sensitivity and zero false positives, enabling long-term, low-power epilepsy monitoring in a wearable form.

Contribution

It introduces a parallel ultra-low-power implementation of seizure detection algorithms using minimal EEG channels for long-term wearable epilepsy monitoring.

Findings

Zero false positives with 100% sensitivity on subject-specific data.

Achieves 300 hours of continuous monitoring on a small battery.

Optimized algorithms for ultra-low-power embedded platforms.

Abstract

We present the implementation of seizure detection algorithms based on a minimal number of EEG channels on a parallel ultra-low-power embedded platform. The analyses are based on the CHB-MIT dataset, and include explorations of different classification approaches (Support Vector Machines, Random Forest, Extra Trees, AdaBoost) and different pre/post-processing techniques to maximize sensitivity while guaranteeing no false alarms. We analyze global and subject-specific approaches, considering all 23-electrodes or only 4 temporal channels. For 8s window size and subject-specific approach, we report zero false positives and 100% sensitivity. These algorithms are parallelized and optimized for a parallel ultra-low power (PULP) platform, enabling 300h of continuous monitoring on a 300 mAh battery, in a wearable form factor and power budget. These results pave the way for the implementation of affordable, wearable, long-term epilepsy monitoring solutions with low false-positive rates and high sensitivity, meeting both patient and caregiver requirements.