Seizure onset detection based on one sEMG channel.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Epilepsy diagnosis and monitoring often rely on complex EEG systems.
• Surface electromyography (sEMG) offers a potential alternative for seizure detection due to its sensitivity to muscle activity during seizures.
• Developing accessible, low-resource seizure detection methods is crucial for widespread adoption.

Purpose of the Study:

• To develop and validate a novel, single-channel sEMG-based method for detecting the onset of tonic-clonic epileptic seizures.
• To create a generic detection algorithm suitable for small, wearable monitoring devices.

Main Methods:

• Utilized high-pass filtered (150 Hz Butterworth) single-channel sEMG data.
• Extracted a single feature: the number of zero-crossings within a hysteresis band (± 50 μV).
• Employed a leave-one-out cross-validation method on 6 epilepsy patients, analyzing counts within 1-second windows.

Main Results:

• Achieved 100% sensitivity in detecting tonic-clonic seizure onsets across all patients.
• Demonstrated a median detection latency of 7.6 seconds.
• Reported a low median false detection rate of 0.04 per hour.

Conclusions:

• The proposed sEMG-based method is highly sensitive and accurate for detecting tonic-clonic epileptic seizures.
• The method's simplicity and reliance on a single channel make it suitable for developing compact, low-cost seizure detection devices.
• This approach holds promise for improving remote epilepsy monitoring and patient care.