Machine detection of spike-wave activity in the EEG and its accuracy compared with visual interpretation

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Machine detection of epileptiform activity in EEG aids seizure monitoring through consistency and data reduction.
• Existing devices face challenges with reliability, speed, or portability due to limited or complex algorithms.
• A need exists for a reliable, portable, and efficient machine for detecting epileptiform EEG activity.

Purpose of the Study:

• To describe a novel, largely analog device for detecting epileptiform activity in EEG.
• To evaluate the device's performance in identifying spike and wave activity compared to human experts.
• To demonstrate a cost-effective and reliable solution for automated EEG analysis.

Main Methods:

• Development of a portable, largely analog device utilizing multi-criteria detection for EEG.
• The device recognizes spikes by shape and waves by frequency, with inter-channel comparisons.
• Testing against 3 certified electroencephalographers using 18 hours of EEG data with 769 spike-wave events.

Main Results:

• The device detected 96.5% of consensus spike and wave activity.
• A low false positive rate of 0.56% was observed, though higher with chewing artifact.
• Machine detection consistency over 6 months was significantly higher than human reader variability.

Conclusions:

• A multi-criteria detection algorithm implemented in special-purpose circuitry offers a cost-effective solution for reliable machine detection of spike and wave activity.
• The developed analog device provides a portable and efficient method for EEG analysis in seizure monitoring.
• The device demonstrates high accuracy and consistency, outperforming human reader variability in long-term analysis.