Validation of an automated seizure detection algorithm for term neonates

Area of Science:

• Neonatal neurology
• Clinical neurophysiology
• Biomedical engineering

Background:

• Epileptic seizures in infants are challenging to detect.
• Accurate seizure detection is crucial for timely intervention and treatment monitoring.
• Existing detection methods may have limitations in prolonged, unedited EEG recordings.

Purpose of the Study:

• To validate the performance of a novel seizure detection algorithm (SDA).
• To assess the SDA's accuracy on unedited, prolonged infant EEG data.
• To evaluate the SDA's effectiveness across different sensitivity settings and seizure durations.

Main Methods:

• Expert annotation of 70 infant EEGs (35 seizure, 35 non-seizure) as the gold standard.
• Testing the SDA across a range of sensitivity settings.
• Comparing SDA annotations with expert annotations using event-based and epoch-based metrics (Cohen's Kappa Index).

Main Results:

• The SDA achieved detection rates of 52.6-75.0% with false detection rates of 0.04-0.36 FD/h.
• A sensitivity setting of 0.4 yielded the best performance with a Kappa Index of 0.630.
• Algorithm performance improved with longer seizure durations.

Conclusions:

• The developed seizure detection algorithm demonstrates promising performance in validating infant EEG recordings.
• The SDA warrants further investigation in live clinical settings.
• The algorithm has the potential to enhance seizure detection and aid in comparing treatment efficacy.