Quantitative detection of seizures with minimal-density EEG montage using phase synchrony and cross-channel coherence...

Area of Science:

• Biomedical Engineering
• Clinical Neurology
• Computational Neuroscience

Background:

• Sub-clinical seizures are common in pediatric critical care, complicating diagnosis and potentially worsening outcomes.
• Current gold-standard seizure detection via electroencephalograms (EEG) requires specialized personnel for electrode placement and interpretation.
• Delays in seizure detection and treatment negatively impact neurological outcomes in critically ill children.

Purpose of the Study:

• To develop and validate an automated seizure detection algorithm using limited EEG channels.
• To assess the feasibility of bedside seizure detection independent of expert input.
• To evaluate the accuracy of phase synchrony and cross-channel coherence for seizure identification.

Main Methods:

• Extracted phase synchrony and cross-channel coherence from 4 and 8 pre-selected scalp EEG signals.
• Employed binary classification to differentiate seizure from non-seizure signal segments.
• Compared algorithm predictions against gold-standard seizure onset markings from multi-channel EEGs.

Main Results:

• The algorithm achieved an average accuracy of 77.2% using 4 EEG channels.
• An average accuracy of 76.5% was obtained using 8 EEG channels.
• Demonstrated the feasibility of seizure detection with a simplified 4 or 8-channel EEG montage.

Conclusions:

• Automated seizure detection is feasible using a limited number of EEG channels (4 or 8).
• This approach allows for readily applicable bedside seizure detection, reducing reliance on expert interpretation.
• The findings support the potential for earlier and more accessible seizure detection in pediatric critical care settings.