Validation of an Automatic Arousal Detection Algorithm for Whole-Night Sleep EEG Recordings

Area of Science:

• Neuroscience
• Sleep Medicine
• Biomedical Engineering

Background:

• Sleep arousals, transient EEG accelerations, indicate sleep perturbations and reduced sleep quality.
• Manual detection of arousals is time-consuming, subjective, and lacks inter-rater reliability.
• Objective and automated methods are needed for accurate arousal detection.

Purpose of the Study:

• To develop and validate a fully automatic algorithm for detecting artefact and arousal events in whole-night EEG recordings.
• To compare the performance of the automated algorithm against human visual detection.
• To provide a reliable, time-efficient alternative to manual arousal scoring.

Main Methods:

• Developed an automatic algorithm using time-frequency analysis with individual-specific thresholds.
• Applied the algorithm to 35 whole-night sleep EEG recordings from healthy young and older adults.
• Compared automated detection results with visual scoring from two independent research centers.

Main Results:

• Automated detection identified more events than human scorers, who showed high variability.
• The algorithm demonstrated high agreement with human raters, evidenced by strong correlations and excellent Cohen's kappa values.
• Detection performance was not significantly influenced by participant sex or sleep stage, but age showed potential impact depending on the reference rater.

Conclusions:

• The developed automatic algorithm provides a reliable and time-sparing method for arousal detection in sleep EEG.
• This tool can enhance comparability across scorers, studies, and research centers.
• The freely available algorithm supports objective sleep quality assessment.