An ECG-based algorithm for the automatic identification of autonomic activations associated with cortical arousal
Mathias Basner1, Barbara Griefahn, Uwe Müller
1German Aerospace Center, Institute of Aerospace Medicine, Köln, Germany. mathias.basner@dlr.de
Sleep
|November 1, 2007
Summary
Heart rate changes can predict cortical arousals, offering a potential automated method to supplement traditional EEG scoring. This ECG-based approach is objective, reproducible, and time-saving for nonclinical populations.
Area of Science:
- Neuroscience
- Cardiology
- Sleep Medicine
Background:
- Electroencephalography (EEG) arousals are linked to autonomic nervous system activity.
- Visual EEG arousal scoring is laborious and has low interobserver agreement.
- Cortical arousal prediction using heart rate changes is hypothesized.
Purpose of the Study:
- To determine if heart rate variability alone can predict cortical arousals.
- To develop and validate an electrocardiogram (ECG)-based algorithm for arousal detection.
- To assess the feasibility of supplementing visual EEG scoring with an automated method.
Main Methods:
- Analysis of 56 healthy subject nights with dual visual AASM EEG arousal scoring.
- Calculation of likelihood ratios (LRs) for heartbeat differences following EEG arousals.
- Validation of an ECG algorithm against visual EEG scoring using ROC analysis.
Main Results:
- The ECG algorithm achieved an Area Under the Curve (AUC) of 0.91 in ROC analysis.
- Sensitivity was 68.1% and specificity was 95.2% at the chosen threshold.
- ECG and EEG arousal indexes showed moderate agreement, with the algorithm performing well in nonclinical populations.
Conclusions:
- The current ECG algorithm cannot replace visual EEG arousal scoring but can supplement it.
- Improved sensitivity is needed for detecting shorter EEG arousals (<10 seconds).
- The automated ECG method offers an objective, reproducible, cost-effective, and time-saving alternative for specific applications.
