Related Experiment Videos
Computer rejection of EEG artifact. II. Contamination by drowsiness
Electroencephalography and Clinical Neurophysiology
|July 1, 1977
Summary
New algorithms automatically detect drowsiness from electroencephalogram (EEG) signals. These algorithms show high accuracy in identifying drowsy events, supporting integration into real-time EEG analysis systems.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Objective measurement of electroencephalogram (EEG) baselines is crucial for evaluating therapeutic interventions.
- Automated detection of drowsiness is needed for real-time EEG analysis.
Purpose of the Study:
- To develop and objectively evaluate algorithms for automatically detecting EEG patterns indicative of drowsiness.
- To integrate drowsiness detection into the ADI-EEG system for comprehensive real-time EEG analysis.
Main Methods:
- Algorithms analyze EEG signal features, including delta/alpha and theta/alpha band spectral intensity ratios.
- Drowsiness detection uses thresholds determined from a waking calibration period and heuristic criteria.
- Performance was evaluated on 20 clinical EEG recordings using consensus scoring from 5 expert scorers.
Main Results:
- The system identified 84% of drowsy episodes agreed upon by at least 3 out of 5 experts.
- It detected 89% of episodes identified by all 5 experts.
- Minimal false positives were observed, with only one event detected by the system not recognized by any scorer.
Conclusions:
- The developed algorithms demonstrate adequate performance for detecting drowsiness in EEG signals.
- These algorithms can be integrated into real-time EEG analysis systems like ADI-EEG.
- The integration allows for a more comprehensive analysis by combining artifact, sharp transient, and drowsiness detection.