Automated single channel seizure detection in the neonate

Area of Science:

• Neonatal neurology
• Medical instrumentation
• Signal processing

Background:

• Neonatal seizures are a critical neurological issue with significant long-term consequences.
• Electroencephalography (EEG) is the standard for seizure identification but requires extensive electrode placement.
• Reducing electrode count can improve patient handling and data acquisition speed.

Purpose of the Study:

• To develop and validate an automated neonatal seizure detection algorithm using a minimal number of EEG electrodes.
• To evaluate the efficacy of different bipolar channel derivations for seizure detection.

Main Methods:

• A novel automated seizure detection algorithm was developed, trained on multi-channel EEG data.
• The algorithm was tested using a dataset of 411 seizures from 251.9 hours of EEG recordings in 17 neonates.
• Performance was assessed using various bipolar channel derivations, focusing on the C3-C4 channel.

Main Results:

• The C3-C4 channel configuration achieved 90.77% correct seizure detection with a 9.43% false detection rate.
• This performance favorably compares to a multi-channel detection method (81.03% detection, 3.82% false detection).

Conclusions:

• Automated neonatal seizure detection is feasible using a simplified two-electrode (C3-C4) approach.
• This method offers a promising alternative for efficient and effective neonatal seizure monitoring.
• Further research may optimize this reduced-electrode system for clinical application.