Test-Retest Reliability of Single Spectral Power and Spectral Power Ratios in Relative and Absolute Values

The current study provides the first systematic evaluation of the test-retest reliability of electroencephalography (EEG) spectral power ratios alongside absolute and relative spectral power using two independent datasets. This approach aims to identify reproducible and clinically meaningful EEG features for longitudinal and diagnostic applications.

Two separate datasets of healthy adults were analyzed. Dataset 1 included 60 participants with EEGs recorded across three sessions (two on the same day and one month later), while Dataset 2 involved 74 participants recorded 1 h apart. Absolute, relative, and ratio-based spectral features were computed across five brain regions using fast Fourier transform (FTP). Intraclass correlation coefficients (ICCs) were calculated to assess short- and long-term test-retest reliability.

Spectral power ratios, particularly the alpha/beta ratio, consistently outperformed single-band absolute and relative power features in reliability. Absolute alpha/beta ratios in the central, parietal, occipital, and temporal regions showed ICC values exceeding 0.75 across both datasets. These results were stable across different intervals and independent recording systems, demonstrating strong generalizability.

This study demonstrates that spectral power ratios, especially the absolute alpha/beta ratio, are highly reliable EEG features. By leveraging two independent datasets, we establish their reproducibility across different populations and conditions. Our findings position spectral power ratios as robust, reproducible biomarkers for clinical EEG applications. This is the first study to rigorously validate their reliability across independent datasets, laying a foundation for their use in diagnostic and longitudinal monitoring tools.