Morphology-based automatic seizure detector for intracerebral EEG recordings

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Epilepsy diagnosis relies heavily on interpreting electroencephalogram (EEG) recordings.
• Reviewing prolonged EEG data is time-consuming and requires expert analysis.
• Objective quantification of EEG features can aid in efficient seizure detection.

Purpose of the Study:

• To develop and evaluate a new automated system for detecting seizures in intracerebral EEG recordings.
• To quantify waveform sharpness as a key feature for seizure identification.
• To compare the system's performance against expert review and a commercial system.

Main Methods:

• The system quantifies seizure events by measuring the sharpness of EEG waveforms, specifically the slope of half-waves.
• The algorithm was optimized using 145 hours of single-channel EEG data and tested on 158 hours from seven patients each.
• Multichannel performance was assessed using 725 hours of EEG data from 21 patients.

Main Results:

• Single-channel testing achieved 87% sensitivity and 71% specificity.
• Multichannel testing yielded 81% sensitivity and 58.9% specificity.
• The system successfully detected diverse seizure patterns, including those overlooked by experts.

Conclusions:

• The developed seizure detection system effectively utilizes EEG waveform sharpness for automated seizure identification.
• The system demonstrates capability in detecting a broad spectrum of seizure types, complementing expert analysis.
• This approach offers a valuable tool for rapid review of prolonged EEG recordings, potentially improving diagnostic efficiency.