Robust and low complexity algorithms for seizure detection.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Clinical Neurology

Background:

• Automated seizure detection is crucial for epilepsy management.
• Existing methods often generate excessive false alarms, impacting clinical utility.
• Subject-specific seizure detection remains a challenge.

Purpose of the Study:

• To develop and evaluate two novel, low-complexity, and robust automated methods for seizure detection.
• To minimize false alarm rates in epilepsy monitoring.
• To assess the performance of these methods using intracranial EEG (iEEG) recordings.

Main Methods:

• Utilized power ratios between frequency bands from iEEG recordings as key features.
• Calculated features from both monopolar and bipolar iEEG derivations for comparison.
• Implemented individually optimized thresholds for alarm generation.
• Tested the detector on long-term continuous iEEG data from 5 epilepsy patients.

Main Results:

• Achieved an average sensitivity of 88.9% for seizure detection.
• Demonstrated a very low false detection rate of 0.041 per hour.
• Reported a mean detection latency of 9.4 seconds.
• Showcased the effectiveness of power band ratios for seizure identification.

Conclusions:

• The proposed methods offer a robust and efficient approach to automated seizure detection.
• The low false alarm rate enhances the reliability of iEEG-based epilepsy monitoring.
• These findings have potential implications for improving seizure management and patient outcomes.