Longitudinal test-retest reliability of quantitative EEG in healthy individuals using an automated preprocessing approach

Electroencephalography (EEG) is a widely accessible and cost-effective technique, making it suitable for large-scale and longitudinal monitoring of brain activity. The evaluation of the reliability of EEG measurements is a necessary step for their use in diagnosis and monitoring. Reliability can be affected by manual cleaning and analysis of the signal; therefore, automated or semi-automated preprocessing approaches increase reproducibility and allow larger datasets to be analyzed. This study evaluated the reliability of relative EEG band powers using a longitudinal resting-state dataset. EEG signals were processed with an automated pipeline that included independent component analysis, wavelet-enhanced component correction, and normalization. Power estimation was performed using a multitaper spectral approach to obtain robust power spectral density estimates. Data were collected from 43 healthy participants aged 21–64 across four sessions spanning two years using 64 scalp electrodes. Reliability was quantified using the intraclass correlation coefficient (ICC), and associations between age and relative band power were evaluated to assess neurophysiological relevance. Relative power was examined across eight frequency bands and summarized for eight neural independent components (ICs) and four scalp regions of interest (ROIs). The average ICC across all bands was 0 . 91 ± 0 . 04 for ICs and 0 . 92 ± 0 . 03 for ROIs ( p < 0 . 05 ), indicating strong and statistically significant reliability. These findings demonstrate that relative EEG band power derived from automated preprocessing is highly replicable, supporting its potential use in the longitudinal assessment of brain-signal alterations in neurological and psychiatric conditions.