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 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 EEG recorded across three sessions (two on the same day and one one month later), while Dataset 2 involved 74 participants recorded one hour apart. Absolute, relative, and ratio-based spectral features were computed across five brain regions using fast Fourier transform. 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. 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 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. Test-Retest Reliability of Single Spectral Power and Spectral Power Ratios in Relative and Absolute Values Jinwon Changa a Williams College, 39 Chapin Hall Drive, Williamstown, Massachusetts 01267, United States Corresponding Author: Jinwon Chang Williams College 2567 Paresky, 39 Chapin Hall Drive, Williamstown, Massachusetts 01267 Phone: +1(857)268-3527 Fax: +820263390550 Email: [email protected]

The current study provides the first systematic evaluation of the test-retest reliability of 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 appli- cations. Two separate datasets of healthy adults were analyzed. Dataset 1 included 60 participants with 1 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. EEG recorded across three sessions (two on the same day and one one month later), while Dataset 2 in- volved 74 participants recorded one hour apart. Absolute, relative, and ratio-based spectral features were computed across five brain regions using fast Fourier transform. 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. 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 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.

Electroencephalography; Test-retest reliability; Spectral power; Resting state 1. Introduction Electroencephalography (EEG) is a widely utilized method for recording voltage fluctuations resulting from transmembrane ionic currents generated by synchronized neuronal populations. These fluctuations reflect diverse physiological processes including postsynaptic potentials, dendritic resonance, calcium spikes, and intrinsic membrane currents (Buzs´ aki et al., 2012). EEG’s high temporal resolution enables the detection of rapid neural dynamics, such as event-related potentials (ERPs), which can be time-locked to cognitive and behavioral events with millisecond precision. Consequently, EEG remains a powerful tool for study- ing brain–behavior relationships in both clinical and cognitive neuroscience. Despite these strengths, EEG remains underused in comparison to structural and functional neuroimaging techniques. One limitation is its traditional reliance on subjective visual interpretation, such as identifying epileptiform discharges, K- complexes, or sleep spindles. Quantitative EEG (qEEG) addresses this limitation by applying statistical and computational approaches to EEG signals, enhancing objectivity and reproducibility across studies. qEEG extends traditional EEG by enabling statistical comparisons across individuals and groups. By shifting focus from expert-dependent scoring to quantifiable neural markers, qEEG facilitates population-level analysis and supports its integration in large-scale cognitive and clinical investigations (Klimesch, 1999). Among its ana- lytic approaches, spectral power analysis and functional connectivity are two of the most widely employed. Functional connectivity captures temporal correlations between signals from spatially distinct electrode sites, often using coherence or covariance measures. However, these indices exhibit limited test-retest reliability and are highly sensitive to electrode configuration and frequency bands. In a study assessing test-retest reliability across ten repeated EEG recordings, spectral power metrics showed substantially greater reliability than co- herence, highlighting functional connectivity’s vulnerability to spatiotemporal variability (Gudmundsson et al., 2007). In contrast, spectral power analysis offers a more stable and interpretable index of brain activity. Using fast Fourier transform (FFT), EEG signals are decomposed into sinusoidal components, enabling quan- tification of power within canonical frequency bands such as delta, theta, alpha, beta, and gamma. These band-specific power values can then be statistically compared between conditions or groups, facilitating the identification of neural correlates of cognitive function or disease states. Nevertheless, spectral power analysis is not immune to limitations due to its susceptibility to signal artifacts. EEG data recorded at the scalp level are affected by numerous extraneous sources such as ocular, cardiac, muscle, line, and channel noise. Although various preprocessing techniques exist to mitigate these artifacts, their effectiveness remains incon- sistent (Delorme, 2023). As EEG signals are known to be non-stable and non-linear (Khanna et al., 2015), 2 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. identifying reliable spectral indices remains a central challenge. To improve the stability of spectral features, this study evaluates two strategies for expressing spectral power: (1) using relative power values, calculated by dividing the power of each frequency band by total power, and (2) using spectral power ratios, derived by dividing power in one frequency band by that of another (e.g., frontal theta/alpha ratio). The relative power approach minimizes individual differences in overall signal amplitude and mitigates the influence of broad- band noise. Spectral power ratios, in turn, capture inter-band dynamics that may better reflect underlying cortical interactions and are less sensitive to global power fluctuations. Numerous studies have employed spectral power ratios to quantify cognitive states and detect psychiatric abnormalities. For example, Chang and Choi (2023) demonstrated that the alpha/beta ratio across anterior frontal, frontal, central, parietal, occipital, and temporal regions significantly distinguished patients with depression from healthy controls, with area under the curve (AUC) values significantly higher than 0.5 in receiver operating characteristic (ROC) analyses. Similarly, decreases in the theta/beta ratio in central and parietal regions have also been reported in depressed individuals, reflecting impaired cortical regulation of attention and arousal systems. Importantly, the theta/beta ratio has been validated in attention-deficit hyperactivity disorder (ADHD) research and is recognized by the U.S. Food and Drug Administration (FDA) as a supportive diagnostic bio- marker when combined with clinical assessment (Stein et al., 2016). Beyond psychiatric disorders, spectral power ratios have also shown discriminative power in neurodegenerative diseases. The theta/alpha ratio, for instance, has consistently differentiated patients with Alzheimer’s disease or frontotemporal dementia from healthy older adults across multiple independent studies (Schmidt et al., 2013; ¨Ozbek et al., 2021; Chang & Chang, 2023), supporting its relevance as a non-invasive marker of cognitive decline. Despite these promising findings, no previous studies have systematically assessed the test-retest reliability of spectral power ratios. The current study fills this gap by evaluating the reliability of both absolute and relative spectral power and spectral power ratio measures across two independent datasets. We hypothesize that spectral power ratios will demonstrate superior test-retest reliability compared to single-band power measures, and that relative power values will outperform absolute values in terms of stability. 2. Methods 2.1. Dataset 1 2.1.1. Participants Dataset 1 was obtained from a publicly available EEG dataset on OpenNeu- ro originally designed to examine test-retest reliability across five mental states in healthy adults: eyes-closed rest, eyes-open rest, mental arithmetic (subtraction), autobiographical memory recall, and music imagery (Wang et al., 2022). Inclusion criteria were: (1) right-handedness, (2) body mass index (BMI) < 28, and (3) sleep onset prior to 00:30 AM on the day before recording. Exclusion criteria included: (1) current psych- iatric or neurological disorders, (2) use of psychoactive medication within the past three months, and (3) history of head trauma. Participants were instructed to refrain from alcohol and caffeine on the day of EEG recording. A total of 60 healthy adults (mean age = 20 years, SD = 2; 32 female) participated and were compensated (˜ $30) for their involvement. 2.1.2. Experimental procedure Each participant completed two laboratory visits separated by a one-month interval. On Visit 1, the first session began with administration of the Self-rating Anxiety Scale (SAS), Self-rating Depression Scale (SDS), and Epworth Sleepiness Sca- le (ESS) (Zung, 1971; Zung, 1965; Johns, 1991). Participants then completed 5-minute eyes-open resting EEG, followed by the Mini New York Cognition Questionnaire (mini NYC-Q; Gorgolewski et al., 2015). This sequence was repeated for 5-minute eyes-closed resting EEG. Participants subsequently completed the Amsterdam Resting-State Questionnaire (ARSQ 2.0), Karolinska Sleepiness Scale (KSS), and Positive and Negative Affect Schedule (PANAS) (Diaz et al., 2014; Akerstedt & Gillberg, 1990; Watson et al., 1988). Next, they underwent three EEG recordings during cognitive tasks (subtraction, memory recall, and music imagery), with the mini NYC-Q administered after each. After a ˜90-minute break, participants completed a second session on the same day following the same protocol, except without SAS, SDS, and ESS. The entire procedure was repeated in full during the second visit. 2.1.3. EEG acquisition and preprocessing Resting- state EEG recordings were used exclusively for the present analysis, focusing on the eyes-closed condition due to its widespread application and reduced susceptibility to ocular artifacts. EEG was recorded using 3 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. either a 63- or 64-channel Ag/AgCl electrode cap (Brain Products GmbH, Germany), based on the exten- ded 10–20 international system. Two channels recorded electrooculograms, and FCz served as the online reference. Data were sampled at 500 Hz and impedance was maintained below 5 k Ω. After resampling to 200 Hz, preprocessing was performed using EEGLAB (Delorme & Makeig, 2004), with the Artifact Sub- space Reconstruction (ASR) plugin. Burst artifacts were removed using a 0.5-second sliding window and a threshold of 20 standard deviations. Bad data segments were additionally excluded using RMS thresholds (-lnf7) with a 25% channel outlier limit. Following average re-referencing, independent component analysis (ICA) was applied, and artifact-related components (e.g., ocular, cardiac, muscle, line noise) were removed using BEM dipfit models with MNI-based electrode coordinates. Cleaned EEG data were re-referenced to the average reference. 2.1.4. Spectral analysis Spectral power was computed for canonical frequency bands: delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta ( >13–30 Hz), and gamma (30–70 Hz). The discrete fast Fourier transform (FFT) was applied in a 2.5-second FFT window length with a frequency resolution of 10 steps per Hz. Spectral power was then averaged within five anatomical regions: frontal (AF3, AF4, AF7, AF8, F1, F2, F3, F4, F5, F6, F7, F8, Fp1, Fp2, Fz), central (C1, C2, C3, C4, C5, C6, Cz, FC1, FC2, FC3, FC4, FC5, FC6), parietal (CP1, CP2, CP3, CP4, CP5, CP6, P1, P2, P3, P4, P5, P6, P7, P8, Pz), occipital (O1, O2, Oz, PO3, PO4, PO7, PO8, POz), and temporal (FT7, FT8, T7, T8, TP8, TP9, TP7, TP10). Spectral power ratios were calculated for each region: delta/theta, delta/alpha, delta/beta, delta/gamma, theta/alpha, theta/beta, theta/gamma, alpha/beta, alpha/gamma, and beta/gamma. For relative spectral power analysis, each ratio was normalized by dividing by the total spectral power across all frequency bands. 2.2. Dataset 2 2.2.1. Participants Dataset 2 was sourced from a separate OpenNeuro dataset (Cavanagh, 2021), originally designed to investigate EEG correlates of depression and anxiety (Cavanagh et al., 2019). Participants were university students aged 18–25 years, with no history of seizures or head trauma, and no current psychoactive medication use. From the 122 available participants, those with elevated Beck Depres- sion Inventory (BDI) scores (>9), inconsistent mood assessments, or missing recordings were excluded. The final sample included 74 healthy participants (mean age = 19 years, SD = 1; 39 female) with low, stable BDI scores (<7) and no reported anxiety disorders. 2.2.2. Experimental procedure Each participant underwent two EEG sessions, one prior to and one following a cognitive task (not analyzed here). In each session, participants completed the Beck Depression Inventory (BDI) and the Spielberger Trait Anxiety Inventory (STAI), followed by 6 minutes of EEG recording. The resting-state EEG consisted of 1-minute eyes-open and 1-minute eyes-closed segments. Only the eyes-closed EEG data were analyzed in this study. 2.2.3. EEG acquisition and preprocessing EEG was acquired using a 64-channel Synamps2 system (Compumedics Neuroscan) with a 10/10 electrode configuration. Data were sampled at 500 Hz with a bandpass filter of 0.5–100 Hz, and impedance was kept below 10 k Ω. The reference electrode was positioned between Cz and CPz. Data were downsampled to 256 Hz, and eyes-closed segments were extracted. Preprocessing followed the same pipeline as Dataset 1, except for the omission of ASR artifact rejection to preserve segment dura- tion. 2.2.4. Spectral analysis The same spectral analysis protocol was applied as the first dataset. Spectral power was averaged differently within five anatomical regions: frontal (AF3, AF4, F1, F2, F3, F4, F5, F6, F7, F8, Fp1, Fp2, Fpz, Fz), central (C1, C2, C3, C4, C5, C6, Cz, FC1, FC2, FC3, FC4, FC5, FC6, FCz), parietal (CP1, CP2, CP3, CP4, CP5, CP6, CPz, P1, P2, P3, P4, P5, P6, P7, P8, Pz), occipital (O1, O2, Oz, PO3, PO4, PO5, PO6, PO7, PO8, POz), and temporal (FT7, FT8, T7, T8, TP8, TP7). 2.3. Sta- tistical analysis All statistical analyses were conducted using MATLAB R2024b (MathWorks, Natick, MA) and MedCalc Statistical Software v20.218 (MedCalc Software Ltd., Ostend, Belgium). In Dataset 1, paired t-tests were used to assess differences in SAS, SDS, and ESS scores between sessions 1 and 3. Repeated measures ANOVA with Huynh-Feldt correction ( ε > 0.75) was used for KSS, PANAS, mini NYC-Q, and ARSQ scores across sessions 1, 2, and 3 (Girden, 1992; Huynh & Feldt, 1976). False discovery rate (FDR) correction using the Benjamini–Hochberg procedure was applied to control for multiple comparisons. Test- retest reliability of spectral features was assessed using intraclass correlation coefficients (ICC (2,k), two-way random effects, absolute agreement, average measures), following established guidelines (Koo & Li, 2016). ICCs were computed at two levels: (1) per cortical region and (2) per individual electrode. In Dataset 1, ICCs were calculated for three time spans: session 1 vs. 2 (short-term), session 1 vs. 3 (long-term), and across all three sessions (overall reliability). In Dataset 2, only short-term reliability (session 1 vs. 2) was 4 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. assessed. Reliability thresholds followed standard interpretations: ICC [?] 0.75 was considered ”good”, and ICC [?] 0.90 was ”excellent”. 3. Results 3.1. Stability of Behavioral Scores (Dataset 1) To evaluate the stability of cognitive and behavioral states across sessions, paired t-tests and repeated measures ANOVA were conducted on behavioral scores in Dataset 1. Paired t-tests showed no significant differences between session 1 and session 3 in SAS (p = 0.173), SDS (p = 0.124), and ESS (p = 0.272), indicating no major changes in anxiety, depression, or sleepiness levels over the 1-month interval (Table 1). Repeated measures ANOVA with Huynh-Feldt correction, followed by FDR-adjusted p-values, revealed no significant session effects across KSS, PANAS (positive/negative affect), ARSQ 2.0 items, or mini NYC-Q subscales (all adjusted p > 0.125; Table 2). These results suggest stable behavioral and cognitive states across all three sessions, supporting the interpretation that any EEG differences likely reflect measurement variability rather than state changes. Table 1. SAS, SDS, and ESS scores between session 1 and session 3 1st session 3rd session p value SAS 40.7 ±8.4 41.9 ±9.2 0.173 SDS 44.4 ±8.2 45.9 ±9.6 0.124 ESS 9.5 ±3.5 10.1 ±3.4 0.272 Table 2. KSS, PANAS, mini-NYC-Q, ARSQ 2.0, and cognitive task (subtraction, memory, and music) scores between session 1, session 2, and session 3 \received DD MMMM YYYY \acceptedDD MMMM YYYY session 1 session 2 session 3 p value adjusted p value KSS 5.4 ±0.2 5.3 ±0.2 5.4 ±0.2 0.705 0.757 PANAS P 30.5 ±0.8 29.2 ±1.0 29.1 ±0.8 0.215 0.443 PANAS N 20.4 ±0.9 18.2 ±0.9 19.4 ±0.9 0.036 0.261 Discontinuity of Mind 6.3 ±0.4 5.9 ±0.3 5.8 ±0.3 0.237 0.443 Theory of Mind 6.6 ±0.4 6.5 ±0.4 6.0 ±0.3 0.298 0.443 Self 8.8 ±0.2 8.4 ±0.3 8.2 ±0.3 0.061 0.352 Planning 9.3 ±0.3 8.4 ±0.4 8.0 ±0.3 0.011 0.125 Sleepiness 5.0 ±0.4 4.4 ±0.4 4.8 ±0.4 0.322 0.444 Comfort 7.2 ±0.3 6.5 ±0.4 7.5 ±0.3 0.009 0.125 Somatic Awareness 4.5 ±0.4 4.0 ±0.4 4.0 ±0.3 0.216 0.443 Health Concern 7.5 ±0.3 7.3 ±0.2 7.0 ±0.2 0.233 0.443 Visual Thought 8.8 ±0.4 9.1 ±0.3 8.6 ±0.3 0.486 0.563 Verbal Thought 5.4 ±0.4 4.5 ±0.4 5.0 ±0.3 0.075 0.352 Subtraction: Correct Not correct 0.4 ±0.1 0.5 ±0.1 0.5 ±0.1 0.193 0.443 Subtraction: final number 4547 ±35 4376 ±95 4465 ±39 0.145 0.443 Memory 1.5 ±0.1 1.5 ±0.1 1.7 ±0.2 0.204 0.443 Music 3.9 ±0.4 3.3 ±0.2 3.6 ±0.4 0.360 0.474 positive 6.1 ±0.4 5.8 ±0.4 5.8 ±0.4 0.841 0.871 negative 3.7 ±0.4 2.8 ±0.4 2.8 ±0.4 0.085 0.352 future 7.4 ±0.4 5.8 ±0.5 6.4 ±0.4 0.013 0.125 past 6.4 ±0.4 6.1 ±0.5 5.6 ±0.4 0.418 0.527 5 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. session 1 session 2 session 3 p value adjusted p value myself 7.7 ±0.4 7.9 ±0.4 7.2 ±0.4 0.255 0.443 people 7.6 ±0.4 7.2 ±0.4 7.0 ±0.4 0.281 0.443 surroundings 5.7 ±0.4 4.9 ±0.5 4.8 ±0.5 0.172 0.443 vigilance 6.4 ±0.4 6.9 ±0.4 6.3 ±0.4 0.302 0.443 images 7.4 ±0.4 7.1 ±0.4 6.9 ±0.4 0.552 0.615 words 3.7 ±0.4 3.5 ±0.4 3.7 ±0.4 0.907 0.907 specific-vague 6.7 ±0.4 6.9 ±0.3 6.3 ±0.4 0.306 0.443 intrusive 4.8 ±0.4 4.4 ±0.4 5.0 ±0.4 0.448 0.541 P values were adjusted by Benjamini-Hochberg FDR correction for multiple comparisons. Discontinuity of Mind, Theory of Mind, Self, Planning, Sleepiness, Comfort, Somatic Awareness, Health Concern, Visual Thought, Verbal Thought are items of ARSQ 2.0, while positive, negative, future, past, myself, people, surroundings, vigilance, images, words, specific-vague, and intrusive are items of mini NYC-Q. 3.2. Test-Retest Reliability of Spectral Power Features (Dataset 1) Intraclass correlation coefficients (ICCs) were computed for each spectral feature at the electrode level. Topographic maps (Figure 1) showed that reliability varied across both spatial regions and frequency bands. ICC values were also averaged by cortical region (frontal, central, parietal, occipital, temporal). ICCs were calculated in three comparisons: (Session 1 vs. 2 (short-term), Session 1 vs. 3 (long-term), and Sessions 1–3 (overall stability)) in dataset 1 (Table 3). F-alpha, C-theta, C-alpha, P-delta, P-theta, P-alpha, O-theta, C-alpha/beta, P-alpha/beta, O-alpha/beta, T-alpha/beta, R-C-beta, R-P-beta, R-O-beta, R-C-delta/theta, R-P-delta/theta, R-P-delta/beta, R-P-theta/beta, R-P-alpha/beta, R-O-delta/theta, R-O-delta/alpha, R- T-delta/theta, R-T-delta/alpha, and R-T-theta/alpha showed good ICCs ( > 0.75) for all three comparisons. Table 3. ICCs of each spectral power feature across each region in dataset 1 session 1, 2, 3 95% CI session 1, 2 95% CI session 1, 3 95% CI F Delta 0.7041 0.5467 - 0.8134 0.8487 0.7453 - 0.9099 0.2768 -0.2139 - 0.5687 F Theta 0.7547 0.6230 - 0.8456 0.9005 0.8334 - 0.9406 0.4617 0.09486 - 0.6793 F Alpha 0.9301 0.8893 - 0.9568 0.9692 0.9473 - 0.9818 0.8729 0.7822 - 0.9251 F Beta 0.552 0.3208 - 0.7153 0.4764 0.1392 - 0.6838 0.8332 0.7195 - 0.9006 F Gamma 0.3652 0.03117 - 0.5985 0.5697 0.2844 - 0.7420 -0.1997 -1.0171 - 0.2852 F All 0.6854 0.5185 - 0.8014 0.7443 0.5713 - 0.8474 0.5396 0.2351 - 0.7237 C Delta 0.8461 0.7642 - 0.9029 0.8432 0.7367 - 0.9065 0.6922 0.4834 - 0.8164 C Theta 0.8783 0.8134 - 0.9232 0.8691 0.7806 - 0.9219 0.8765 0.7931 - 0.9263 C Alpha 0.944 0.9106 - 0.9655 0.9672 0.9407 - 0.9813 0.9098 0.8475 - 0.9464 C Beta 0.548 0.3107 - 0.7140 0.4116 0.02534 - 0.6463 0.8791 0.7978 - 0.9278 C Gamma 0.01608 -0.5087 - 0.3797 0.03248 -0.6169 - 0.4215 0.2797 -0.2157 - 0.5718 C All 0.3728 0.04101 - 0.6038 0.2649 -0.2260 - 0.5599 0.9331 0.8875 - 0.9601 P Delta 0.9545 0.9302 - 0.9713 0.9580 0.9290 - 0.9750 0.9169 0.8612 - 0.9503 P Theta 0.922 0.8796 - 0.9510 0.9496 0.9121 - 0.9706 0.8701 0.7832 - 0.9223 P Alpha 0.9289 0.8890 - 0.9557 0.9731 0.9523 - 0.9845 0.8641 0.7723 - 0.9188 P Beta 0.8516 0.7621 - 0.9090 0.8368 0.7056 - 0.9067 0.8807 0.7996 - 0.9289 P Gamma 0.07667 -0.4014 - 0.4137 0.1009 -0.4836 - 0.4582 0.3876 -0.02699 - 0.6346 P All 0.6389 0.4497 - 0.7713 0.5495 0.2551 - 0.7288 0.8947 0.8221 - 0.9375 O Delta 0.8482 0.7673 - 0.9043 0.9469 0.9111 - 0.9683 0.7466 0.5755 - 0.8487 O Theta 0.9523 0.9256 - 0.9702 0.9421 0.8984 - 0.9664 0.9419 0.9029 - 0.9653 O Alpha 0.925 0.8818 - 0.9535 0.9762 0.9515 - 0.9872 0.8325 0.7202 - 0.8998 O Beta 0.894 0.8268 - 0.9358 0.883 0.7323 - 0.9409 0.856 0.7586 - 0.9141 6 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. session 1, 2, 3 95% CI session 1, 2 95% CI session 1, 3 95% CI O Gamma 0.6483 0.4635 - 0.7773 0.6271 0.3743 - 0.7775 0.7662 0.6080 - 0.8604 O All 0.888 0.8235 - 0.9306 0.9058 0.8179 - 0.9482 0.8476 0.7454 - 0.9089 T Delta 0.002526 -0.5324 - 0.3719 0.004488 -0.6677 - 0.4056 0.7627 0.6025 - 0.8583 T Theta -0.001221 -0.5371 - 0.3693 -0.0008405 -0.6754 - 0.4021 0.9157 0.8588 - 0.9497 T Alpha -0.006963 -0.5453 - 0.3655 -0.003873 -0.6800 - 0.4002 0.8923 0.8200 - 0.9356 T Beta 0.0004930 -0.5345 - 0.3704 -0.0001088 -0.6743 - 0.4026 0.7290 0.5481 - 0.8378 T Gamma 0.0003264 -0.5348 - 0.3703 0.0004481 -0.6734 - 0.4029 0.2792 -0.2065 - 0.5694 T All 0.0001063 -0.5351 - 0.3701 0.0003119 -0.6736 - 0.4029 0.8742 0.7900 - 0.9247 F delta theta 0.7615 0.6336 - 0.8498 0.8519 0.7157 - 0.9182 0.6077 0.3426 - 0.7658 F delta alpha 0.5666 0.3413 - 0.7248 0.8342 0.6931 - 0.9066 0.2939 -0.1542 - 0.5721 F delta beta 0.6458 0.4597 - 0.7759 0.817 0.6789 - 0.8937 0.2400 -0.2835 - 0.5484 F delta gamma 0.6584 0.4758 - 0.7847 0.7563 0.5872 - 0.8555 0.0242 -0.6518 - 0.4209 F theta alpha 0.6088 0.4044 - 0.7520 0.8927 0.8209 - 0.9358 0.4129 0.04122 - 0.6440 F theta beta 0.7537 0.6232 - 0.8444 0.8918 0.8027 - 0.9385 0.4974 0.1541 - 0.7007 F theta gamma 0.558 0.3262 - 0.7202 0.8322 0.6850 - 0.9061 0.2493 -0.2641 - 0.5532 F alpha beta 0.8932 0.8357 - 0.9328 0.8989 0.8220 - 0.9414 0.8531 0.7490 - 0.9133 F alpha gamma 0.6172 0.4141 - 0.7583 0.887 0.8011 - 0.9345 0.1215 -0.4641 - 0.4739 F beta gamma 0.5057 0.2438 - 0.6879 0.8995 0.8186 - 0.9426 -0.1251 -0.9080 - 0.3330 C delta theta 0.735 0.5908 - 0.8336 0.6544 0.3877 - 0.8005 0.5802 0.2944 - 0.7499 C delta alpha 0.6184 0.4160 - 0.7589 0.598 0.3339 - 0.7584 0.4663 0.1221 - 0.6778 C delta beta 0.7847 0.6634 - 0.8659 0.7745 0.5935 - 0.8710 0.5992 0.3287 - 0.7607 C delta gamma 0.6052 0.3920 - 0.7518 0.7771 0.6261 - 0.8670 0.2474 -0.2703 - 0.5527 C theta alpha 0.7044 0.5419 - 0.8147 0.9157 0.8593 - 0.9496 0.5742 0.2895 - 0.7451 C theta beta 0.9155 0.8704 - 0.9467 0.9681 0.9467 - 0.9809 0.8262 0.7100 - 0.8959 C theta gamma 0.7636 0.6374 - 0.8510 0.8516 0.7521 - 0.9112 0.5745 0.2847 - 0.7465 C alpha beta 0.9319 0.8940 - 0.9575 0.9388 0.8976 - 0.9635 0.8906 0.8073 - 0.9366 C alpha gamma 0.7958 0.6874 - 0.8711 0.8369 0.7278 - 0.9024 0.6694 0.4494 - 0.8019 C beta gamma 0.7662 0.6411 - 0.8527 0.8938 0.8227 - 0.9364 0.4878 0.1379 - 0.6950 P delta theta 0.8929 0.8261 - 0.9348 0.8921 0.6896 - 0.9509 0.8311 0.7170 - 0.8992 P delta alpha 0.742 0.5908 - 0.8407 0.9439 0.8839 - 0.9701 0.6423 0.3933 - 0.7880 P delta beta 0.8968 0.8129 - 0.9411 0.9101 0.6916 - 0.9619 0.8771 0.7943 - 0.9265 P delta gamma 0.5626 0.3365 - 0.7220 0.7594 0.4930 - 0.8739 0.3458 -0.1024 - 0.6108 P theta alpha 0.689 0.5200 - 0.8045 0.9441 0.9065 - 0.9666 0.5718 0.2863 - 0.7436 P theta beta 0.9238 0.8823 - 0.9522 0.9696 0.9454 - 0.9826 0.839 0.7313 - 0.9036 P theta gamma 0.6682 0.4933 - 0.7903 0.8593 0.7276 - 0.9226 0.3858 -0.03180 - 0.6339 P alpha beta 0.9289 0.8894 - 0.9556 0.9438 0.8969 - 0.9682 0.8728 0.7725 - 0.9269 P alpha gamma 0.7224 0.5755 - 0.8247 0.8389 0.7227 - 0.9053 0.5286 0.2200 - 0.7164 P beta gamma 0.6875 0.5226 - 0.8024 0.8443 0.7082 - 0.9128 0.4213 0.03378 - 0.6538 O delta theta 0.69 0.5266 - 0.8040 0.87 0.7047 - 0.9342 0.5842 0.3042 - 0.7516 O delta alpha 0.6326 0.4382 - 0.7677 0.9174 0.8484 - 0.9532 0.5401 0.2413 - 0.7228 O delta beta 0.7815 0.6552 - 0.8646 0.8411 0.6090 - 0.9227 0.7109 0.5178 - 0.8270 O delta gamma 0.5468 0.3106 - 0.7126 0.407 0.03080 - 0.6406 0.5242 0.1992 - 0.7167 O theta alpha 0.6545 0.4731 - 0.7813 0.9281 0.8797 - 0.9571 0.4933 0.1666 - 0.6940 O theta beta 0.8692 0.7994 - 0.9175 0.9365 0.8879 - 0.9632 0.7266 0.5417 - 0.8368 O theta gamma 0.5401 0.2930 - 0.7106 0.4861 0.1409 - 0.6928 0.4957 0.1527 - 0.6994 O alpha beta 0.9009 0.8476 - 0.9376 0.8692 0.7714 - 0.9238 0.8579 0.7615 - 0.9153 O alpha gamma 0.4695 0.1825 - 0.6667 0.2576 -0.2522 - 0.5585 0.7212 0.5351 - 0.8331 O beta gamma 0.5719 0.3423 - 0.7304 0.4917 0.1507 - 0.6960 0.5802 0.2984 - 0.7490 T delta theta 0.8172 0.6999 - 0.8895 0.7849 0.5409 - 0.8879 0.7595 0.5989 - 0.8560 T delta alpha 0.8059 0.6793 - 0.8831 0.8919 0.7970 - 0.9395 0.7178 0.5125 - 0.8345 7 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. session 1, 2, 3 95% CI session 1, 2 95% CI session 1, 3 95% CI T delta beta 0.8102 0.7081 - 0.8806 0.8763 0.7910 - 0.9265 0.6742 0.4575 - 0.8047 T delta gamma 0.6584 0.4758 - 0.7847 0.806 0.6757 - 0.8840 0.3238 -0.1273 - 0.5950 T theta alpha 0.8477 0.7625 - 0.9050 0.941 0.9011 - 0.9647 0.7586 0.5846 - 0.8581 T theta beta 0.9113 0.8639 - 0.9441 0.9587 0.9310 - 0.9753 0.8329 0.7211 - 0.9000 T theta gamma 0.7435 0.6074 - 0.8381 0.8564 0.7533 - 0.9156 0.5153 0.1862 - 0.7110 T alpha beta 0.9197 0.8769 - 0.9493 0.9297 0.8818 - 0.9581 0.8623 0.7692 - 0.9178 T alpha gamma 0.7407 0.6032 - 0.8363 0.8077 0.6670 - 0.8874 0.5152 0.1892 - 0.7102 T beta gamma 0.648 0.4607 - 0.7779 0.8556 0.7477 - 0.9159 0.2173 -0.3158 - 0.5336 R F Delta 0.7212 0.5648 - 0.8261 0.9174 0.7767 - 0.9613 0.4362 0.06795 - 0.6606 R F Theta 0.8007 0.6940 - 0.8745 0.9475 0.9121 - 0.9687 0.5718 0.2831 - 0.7443 R F Alpha 0.7406 0.6026 - 0.8364 0.939 0.8980 - 0.9636 0.5097 0.1819 - 0.7065 R F Beta 0.8422 0.7522 - 0.9020 0.9237 0.6592 - 0.9706 0.7089 0.5113 - 0.8264 R F Gamma 0.4915 0.2256 - 0.6779 0.9598 0.9049 - 0.9800 -0.4668 -1.4744 - 0.1278 R C Delta 0.6971 0.5183 - 0.8133 0.7473 0.5076 - 0.8617 0.5272 0.2163 - 0.7159 R C Theta 0.8672 0.7963 - 0.9163 0.9312 0.8847 - 0.9589 0.7731 0.6197 - 0.8646 R C Alpha 0.8799 0.8099 - 0.9258 0.9029 0.8378 - 0.9419 0.8233 0.6938 - 0.8966 R C Beta 0.9237 0.8831 - 0.9519 0.9479 0.9106 - 0.9693 0.8642 0.7727 - 0.9188 R C Gamma 0.5064 0.2439 - 0.6886 0.7058 0.5105 - 0.8236 0.1265 -0.4770 - 0.4813 R P Delta 0.8241 0.6762 - 0.9006 0.876 0.6144 - 0.9455 0.7734 0.6085 - 0.8672 R P Theta 0.8478 0.7636 - 0.9048 0.9226 0.8707 - 0.9537 0.751 0.5801 - 0.8519 R P Alpha 0.853 0.7736 - 0.9076 0.8838 0.8060 - 0.9305 0.7847 0.6261 - 0.8742 R P Beta 0.9445 0.9149 - 0.9650 0.9765 0.9535 - 0.9872 0.8987 0.8305 - 0.9394 R P Gamma 0.4886 0.2261 - 0.6745 0.6601 0.4185 - 0.7996 0.1946 -0.3581 - 0.5210 R O Delta 0.744 0.5944 - 0.8418 0.8112 0.5837 - 0.9032 0.7067 0.5120 - 0.8242 R O Theta 0.7877 0.6749 - 0.8660 0.9228 0.8697 - 0.9541 0.6034 0.3359 - 0.7631 R O Alpha 0.8794 0.8146 - 0.9241 0.9155 0.8580 - 0.9496 0.7953 0.6516 - 0.8789 R O Beta 0.9175 0.8735 - 0.9480 0.9264 0.8668 - 0.9580 0.8524 0.7537 - 0.9116 R O Gamma 0.7816 0.6645 - 0.8624 0.8286 0.6597 - 0.9070 0.6697 0.4485 - 0.8023 R T Delta 0.7467 0.6058 - 0.8417 0.8409 0.7098 - 0.9095 0.6078 0.3492 - 0.7645 R T Theta 0.9048 0.8540 - 0.9400 0.9462 0.9098 - 0.9679 0.8092 0.6809 - 0.8860 R T Alpha 0.881 0.8145 - 0.9257 0.915 0.8460 - 0.9515 0.8228 0.7044 - 0.8939 R T Beta 0.8656 0.7937 - 0.9154 0.9061 0.8428 - 0.9439 0.755 0.5891 - 0.8538 R T Gamma 0.7151 0.5622 - 0.8207 0.8747 0.7900 - 0.9252 0.4731 0.1151 - 0.6859 R F delta theta 0.8735 0.8029 - 0.9210 0.9005 0.8324 - 0.9408 0.8029 0.6702 - 0.8822 R F delta alpha 0.8895 0.8054 - 0.9360 0.9396 0.8974 - 0.9642 0.8456 0.6846 - 0.9175 R F delta beta 0.8323 0.7270 - 0.8981 0.8717 0.7481 - 0.9300 0.7909 0.6485 - 0.8754 R F delta gamma 0.5979 0.3852 - 0.7457 0.7304 0.4872 - 0.8505 0.5344 0.2270 - 0.7204 R F theta alpha 0.883 0.8163 - 0.9274 0.9367 0.8941 - 0.9622 0.8112 0.6737 - 0.8893 R F theta beta 0.8828 0.8190 - 0.9264 0.9051 0.8347 - 0.9446 0.7966 0.6601 - 0.8784 R F theta gamma 0.7346 0.5924 - 0.8327 0.837 0.6537 - 0.9147 0.5905 0.3121 - 0.7559 R F alpha beta 0.8186 0.7222 - 0.8856 0.8471 0.7329 - 0.9110 0.7317 0.5525 - 0.8393 R F alpha gamma 0.6371 0.4468 - 0.7702 0.7873 0.6062 - 0.8802 0.3793 -0.03586 - 0.6286 R F beta gamma 0.7445 0.6092 - 0.8386 0.8588 0.7534 - 0.9178 0.5390 0.2265 - 0.7250 R C delta theta 0.9386 0.9013 - 0.9624 0.9286 0.8486 - 0.9625 0.9125 0.8535 - 0.9477 R C delta alpha 0.8454 0.7431 - 0.9072 0.8929 0.8095 - 0.9383 0.7494 0.5689 - 0.8527 R C delta beta 0.8887 0.8167 - 0.9329 0.8762 0.7821 - 0.9282 0.8227 0.7002 - 0.8947 R C delta gamma 0.763 0.6371 - 0.8504 0.7866 0.6421 - 0.8727 0.5597 0.2684 - 0.7358 R C theta alpha 0.8619 0.7705 - 0.9171 0.9415 0.9023 - 0.9650 0.7589 0.5668 - 0.8618 R C theta beta 0.8974 0.8402 - 0.9359 0.9261 0.8764 - 0.9558 0.803 0.6657 - 0.8833 R C theta gamma 0.7917 0.6808 - 0.8686 0.8373 0.7286 - 0.9026 0.5889 0.3175 - 0.7533 8 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. session 1, 2, 3 95% CI session 1, 2 95% CI session 1, 3 95% CI R C alpha beta 0.8723 0.8042 - 0.9195 0.8391 0.7301 - 0.9040 0.8075 0.6789 - 0.8848 R C alpha gamma 0.7224 0.5746 - 0.8249 0.7474 0.5788 - 0.8488 0.5361 0.2206 - 0.7235 R C beta gamma 0.7948 0.6847 - 0.8708 0.9011 0.8349 - 0.9409 0.5456 0.2367 - 0.7291 R P delta theta 0.9419 0.9081 - 0.9641 0.9533 0.9188 - 0.9727 0.9205 0.8656 - 0.9528 R P delta alpha 0.8981 0.8269 - 0.9396 0.9731 0.9533 - 0.9842 0.8373 0.6957 - 0.9089 R P delta beta 0.9436 0.8901 - 0.9690 0.9534 0.9047 - 0.9751 0.9227 0.8460 - 0.9581 R P delta gamma 0.8288 0.7160 - 0.8972 0.7944 0.5900 - 0.8888 0.8114 0.6852 - 0.8872 R P theta alpha 0.8608 0.7747 - 0.9152 0.9562 0.9268 - 0.9738 0.7884 0.6172 - 0.8793 R P theta beta 0.926 0.8764 - 0.9557 0.9047 0.8409 - 0.9430 0.8939 0.7859 - 0.9429 R P theta gamma 0.7881 0.6680 - 0.8682 0.7756 0.5913 - 0.8724 0.7482 0.5802 - 0.8492 R P alpha beta 0.9201 0.8772 - 0.9497 0.8669 0.7753 - 0.9209 0.9425 0.9036 - 0.9656 R P alpha gamma 0.7175 0.5671 - 0.8217 0.5887 0.3082 - 0.7551 0.7337 0.5551 - 0.8408 R P beta gamma 0.8026 0.6954 - 0.8760 0.7502 0.5706 - 0.8531 0.7431 0.5691 - 0.8467 R O delta theta 0.9094 0.8611 - 0.9429 0.905 0.8384 - 0.9438 0.9005 0.8332 - 0.9406 R O delta alpha 0.9253 0.8832 - 0.9534 0.9575 0.9275 - 0.9749 0.8743 0.7888 - 0.9251 R O delta beta 0.892 0.8326 - 0.9324 0.8971 0.8098 - 0.9419 0.8125 0.6859 - 0.8881 R O delta gamma 0.7069 0.5513 - 0.8149 0.5688 0.2832 - 0.7414 0.6371 0.3899 - 0.7837 R O theta alpha 0.8876 0.8276 - 0.9291 0.9593 0.9320 - 0.9756 0.7897 0.6495 - 0.8741 R O theta beta 0.8607 0.7866 - 0.9121 0.8695 0.7821 - 0.9219 0.7315 0.5498 - 0.8397 R O theta gamma 0.6692 0.4947 - 0.7909 0.5273 0.2198 - 0.7152 0.6028 0.3321 - 0.7634 R O alpha beta 0.8746 0.8076 - 0.9210 0.7984 0.6581 - 0.8805 0.8823 0.8032 - 0.9297 R O alpha gamma 0.7219 0.5748 - 0.8244 0.3364 -0.09276 - 0.5996 0.8885 0.8133 - 0.9334 R O beta gamma 0.7442 0.6087 - 0.8384 0.6196 0.3692 - 0.7714 0.7506 0.5823 - 0.8511 R T delta theta 0.9282 0.8866 - 0.9555 0.9302 0.8707 - 0.9607 0.893 0.8214 - 0.9360 R T delta alpha 0.9378 0.8973 - 0.9626 0.9687 0.9462 - 0.9816 0.8946 0.8117 - 0.9394 R T delta beta 0.8705 0.8013 - 0.9183 0.8886 0.8140 - 0.9333 0.7916 0.6525 - 0.8752 R T delta gamma 0.7081 0.5515 - 0.8162 0.8046 0.6732 - 0.8832 0.4188 0.02617 - 0.6530 R T theta alpha 0.9325 0.8955 - 0.9577 0.9695 0.9489 - 0.9818 0.8761 0.7880 - 0.9270 R T theta beta 0.9008 0.8477 - 0.9375 0.9403 0.9003 - 0.9643 0.7993 0.6644 - 0.8800 R T theta gamma 0.7701 0.6480 - 0.8549 0.8729 0.7811 - 0.9253 0.5033 0.1642 - 0.7042 R T alpha beta 0.845 0.7625 - 0.9022 0.8302 0.7167 - 0.8984 0.7556 0.5900 - 0.8542 R T alpha gamma 0.7172 0.5674 - 0.8214 0.7642 0.6013 - 0.8600 0.4498 0.07468 - 0.6722 R T beta gamma 0.6717 0.4971 - 0.7929 0.852 0.7505 - 0.9120 0.2671 -0.2372 - 0.5644 R: relative; F: frontal; C: central; P: parietal; O: occipital; T: temporal; CI: confidence interval 3.3. Test-Retest Reliability in an Independent Dataset (Dataset 2) In Dataset 2, ICCs were computed between the two sessions (short-term reliability only). In the same way, ICCs across each spectral power feature were measured through individual electrode (Figure 2). While many absolute and relative power features showed near-zero or negative ICCs, several spectral power ratios demonstrated robust reliability. Also, ICCs of each spectral power feature across each region were mea- sured between session 1 and 2 in Dataset 2 (Table 4). F-theta-alpha, F-theta-beta, F-alpha-beta, F-alpha- gamma, C-theta-alpha, C-theta-gamma, C-alpha-beta, C-alpha-gamma, C-beta-gamma, P-theta-alpha, P- theta-beta, P-theta-gamma, P-alpha-beta, P-alpha-gamma, O-theta-alpha, O-theta-beta, O-theta-gamma, O-alpha-beta, T-theta-alpha, T-theta-gamma, T-alpha-beta, T-alpha-gamma, R-F-delta-alpha, R-P-theta- alpha, R-O-theta-alpha, and R-O-theta-beta showed good ICCs ( > 0.75) between session 1 and 2. Table 4. ICCs of each spectral power feature across each region in dataset 2 9 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. Session 1, 2 95% CI F Delta -3.2E-05 -0.5879 - 0.3702 F Theta -0.00002534 -0.5879 - 0.3702 F Alpha -0.00001182 -0.5879 - 0.3702 F Beta -0.00007978 -0.5880 - 0.3702 F Gamma -0.0001011 -0.5881 - 0.3702 F All -0.00004205 -0.5879 - 0.3702 C Delta -0.00003680 -0.5879 - 0.3702 C Theta -0.00003193 -0.5879 - 0.3702 C Alpha -0.00001718 -0.5879 - 0.3702 C Beta -0.0001012 -0.5880 - 0.3702 C Gamma -0.0001271 -0.5881 - 0.3702 C All -0.00005292 -0.5879 - 0.3702 P Delta -0.00003268 -0.5879 - 0.3702 P Theta -0.00003447 -0.5879 - 0.3702 P Alpha -0.00001617 -0.5879 - 0.3702 P Beta -0.0001164 -0.5881 - 0.3702 P Gamma -0.0001509 -0.5882 - 0.3701 P All -5.8E-05 -0.5880 - 0.3702 O Delta -0.00001141 -0.5879 - 0.3702 O Theta -0.000009448 -0.5879 - 0.3702 O Alpha -0.000005436 -0.5879 - 0.3702 O Beta -0.00002738 -0.5879 - 0.3702 O Gamma -0.00003365 -0.5879 - 0.3702 O All -0.00001493 -0.5879 - 0.3702 T Delta -0.00005020 -0.5880 - 0.3702 T Theta -0.00005160 -0.5879 - 0.3702 T Alpha -0.00002555 -0.5879 - 0.3702 T Beta -0.0001753 -0.5882 - 0.3701 T Gamma -0.0002265 -0.5883 - 0.3701 T All -0.00008736 -0.5880 - 0.3702 F delta theta 0.6114 0.3811 - 0.7557 F delta alpha 0.5625 0.3032 - 0.7250 F delta beta 0.4184 0.07341 - 0.6345 F delta gamma 0.4356 0.09990 - 0.6455 F theta alpha 0.8844 0.8169 - 0.9271 F theta beta 0.8509 0.7622 - 0.9063 F theta gamma 0.8312 0.7324 - 0.8936 F alpha beta 0.9008 0.8424 - 0.9376 F alpha gamma 0.9103 0.8574 - 0.9435 F beta gamma 0.7919 0.6691 - 0.8691 C delta theta 0.3831 0.01648 - 0.6125 C delta alpha 0.3374 -0.05620 - 0.5838 C delta beta 0.08487 -0.4559 - 0.4244 C delta gamma 0.2895 -0.1329 - 0.5538 C theta alpha 0.9005 0.8424 - 0.9373 C theta beta 0.8056 0.6922 - 0.8774 C theta gamma 0.8957 0.8348 - 0.9342 C alpha beta 0.8606 0.7790 - 0.9121 C alpha gamma 0.9083 0.8547 - 0.9422 C beta gamma 0.8564 0.7722 - 0.9095 10 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. Session 1, 2 95% CI P delta theta 0.202 -0.2749 - 0.4994 P delta alpha 0.2711 -0.1615 - 0.5420 P delta beta -0.03251 -0.6431 - 0.3506 P delta gamma 0.2629 -0.1746 - 0.5368 P theta alpha 0.8901 0.8259 - 0.9307 P theta beta 0.8619 0.7799 - 0.9132 P theta gamma 0.9028 0.8460 - 0.9387 P alpha beta 0.8661 0.7872 - 0.9157 P alpha gamma 0.9002 0.8406 - 0.9374 P beta gamma 0.8283 0.7271 - 0.8919 O delta theta 0.2066 -0.2679 - 0.5023 O delta alpha 0.3767 0.008584 - 0.6079 O delta beta 0.008339 -0.5797 - 0.3767 O delta gamma 0.3727 -0.001103 - 0.6062 O theta alpha 0.8449 0.7525 - 0.9026 O theta beta 0.912 0.8557 - 0.9457 O theta gamma 0.9034 0.8465 - 0.9391 O alpha beta 0.8616 0.7807 - 0.9127 O alpha gamma 0.8332 0.7172 - 0.8991 O beta gamma 0.7941 0.6425 - 0.8772 T delta theta 0.04457 -0.5267 - 0.4007 T delta alpha 0.07829 -0.4635 - 0.4195 T delta beta -0.05526 -0.6777 - 0.3359 T delta gamma 0.1033 -0.4283 - 0.4364 T theta alpha 0.9063 0.8492 - 0.9415 T theta beta 0.7818 0.6545 - 0.8623 T theta gamma 0.898 0.8384 - 0.9357 T alpha beta 0.8606 0.7785 - 0.9123 T alpha gamma 0.903 0.8460 - 0.9389 T beta gamma 0.7951 0.6745 - 0.8710 R F Delta 0.6714 0.4778 - 0.7931 R F Theta 0.7106 0.5422 - 0.8173 R F Alpha 0.8241 0.7203 - 0.8893 R F Beta 0.7114 0.5407 - 0.8185 R F Gamma 0.4039 0.04984 - 0.6255 R C Delta 0.5478 0.2812 - 0.7154 R C Theta 0.8038 0.6882 - 0.8765 R C Alpha 0.8014 0.6855 - 0.8747 R C Beta 0.748 0.6008 - 0.8411 R C Gamma 0.4756 0.1660 - 0.6701 R P Delta 0.4154 0.07179 - 0.6318 R P Theta 0.8266 0.7253 - 0.8906 R P Alpha 0.7244 0.5639 - 0.8261 R P Beta 0.7725 0.6398 - 0.8564 R P Gamma 0.413 0.06395 - 0.6313 R O Delta 0.4736 0.1647 - 0.6684 R O Theta 0.8414 0.7468 - 0.9004 R O Alpha 0.7535 0.6096 - 0.8445 R O Beta 0.7774 0.6470 - 0.8597 R O Gamma 0.5768 0.3293 - 0.7331 11 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. Session 1, 2 95% CI R T Delta 0.4448 0.1145 - 0.6513 R T Theta 0.7865 0.6610 - 0.8656 R T Alpha 0.7889 0.6650 - 0.8670 R T Beta 0.6515 0.4483 - 0.7801 R T Gamma 0.4903 0.1892 - 0.6794 R F delta theta 0.6486 0.4441 - 0.7781 R F delta alpha 0.8566 0.7724 - 0.9096 R F delta beta 0.6027 0.3681 - 0.7500 R F delta gamma 0.6573 0.4577 - 0.7837 R F theta alpha 0.7474 0.5992 - 0.8409 R F theta beta 0.3047 -0.1038 - 0.5621 R F theta gamma 0.5476 0.2823 - 0.7149 R F alpha beta 0.6362 0.4207 - 0.7713 R F alpha gamma 0.7763 0.6444 - 0.8592 R F beta gamma 0.4057 0.05218 - 0.6267 R C delta theta 0.7187 0.5525 - 0.8231 R C delta alpha 0.8048 0.6896 - 0.8771 R C delta beta 0.4607 0.1431 - 0.6605 R C delta gamma 0.7066 0.5335 - 0.8154 R C theta alpha 0.7551 0.6108 - 0.8458 R C theta beta 0.5812 0.3339 - 0.7366 R C theta gamma 0.4655 0.1494 - 0.6638 R C alpha beta 0.6731 0.4830 - 0.7936 R C alpha gamma 0.6671 0.4737 - 0.7898 R C beta gamma 0.5176 0.2328 - 0.6965 R P delta theta 0.6841 0.4975 - 0.8013 R P delta alpha 0.8141 0.7043 - 0.8830 R P delta beta 0.2911 -0.1256 - 0.5535 R P delta gamma 0.6252 0.4051 - 0.7639 R P theta alpha 0.8512 0.7637 - 0.9063 R P theta beta 0.7136 0.5454 - 0.8196 R P theta gamma 0.5209 0.2373 - 0.6988 R P alpha beta 0.7523 0.6066 - 0.8441 R P alpha gamma 0.6477 0.4434 - 0.7774 R P beta gamma 0.5528 0.2901 - 0.7184 R O delta theta 0.5828 0.3391 - 0.7368 R O delta alpha 0.8305 0.7307 - 0.8933 R O delta beta 0.553 0.2893 - 0.7187 R O delta gamma 0.7568 0.6134 - 0.8469 R O theta alpha 0.8662 0.7874 - 0.9157 R O theta beta 0.8615 0.7801 - 0.9128 R O theta gamma 0.7165 0.5489 - 0.8217 R O alpha beta 0.7623 0.6234 - 0.8501 R O alpha gamma 0.6303 0.4128 - 0.7672 R O beta gamma 0.6114 0.3864 - 0.7544 R T delta theta 0.7102 0.5397 - 0.8175 R T delta alpha 0.8424 0.7497 - 0.9008 R T delta beta 0.5833 0.3382 - 0.7376 R T delta gamma 0.6947 0.5149 - 0.8079 R T theta alpha 0.8196 0.7133 - 0.8864 12 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. Session 1, 2 95% CI R T theta beta 0.5281 0.2490 - 0.7033 R T theta gamma 0.7655 0.6277 - 0.8523 R T alpha beta 0.63 0.4117 - 0.7672 R T alpha gamma 0.7521 0.6066 - 0.8438 R T beta gamma 0.5367 0.2641 - 0.7083 R: relative; F: frontal; C: central; P: parietal; O: occipital; T: temporal; CI: confidence interval By comparing results across both datasets, C-alpha/beta, P-alpha/beta, O-alpha/beta, and T-alpha/beta consistently showed good ICCs ( >0.75) for all measurements. 4. Discussion 4.1. Test-retest reliability of spectral power features The present study assessed the test-retest reliability of various spectral EEG features, including absolute and relative single-band power and spectral power ratios, across two independent datasets. Dataset 1 incorporated both short-term (within-day) and long-term (1- month interval) comparisons, while Dataset 2 provided an independent replication with a shorter interval. Across all analyses, spectral power ratios demonstrated higher reliability than single-band spectral power, consistent with our primary hypothesis. In Dataset 1, behavioral scores remained statistically stable across sessions, confirming that EEG differences were unlikely to result from cognitive or emotional fluctuations. ICC analyses showed that several ratio features, especially alpha/beta ratios in central (C), parietal (P), oc- cipital (O), and temporal (T) regions, consistently yielded good reliability (ICC > 0.75) across all timepoints. This pattern was replicated in Dataset 2, despite differences in acquisition systems, preprocessing, and par- ticipant population. These findings support the growing consensus that spectral power ratios, particularly the alpha/beta ratio, are more stable over time than traditional single-band metrics. The higher reliability of ratios likely stems from their ability to normalize inter-individual variability and suppress non-neural artifacts. As both numerator and denominator are drawn from the same EEG signal, shared sources of noise (e.g., movement, impedance variation, general arousal) may be reduced through cancellation, resulting in a more stable metric across time. Furthermore, the robustness across spatial regions suggests that alpha/beta ratio is not restricted to a single functional network, but instead reflects a distributed neural property, po- tentially related to cortical excitability and cognitive control. 4.2. Clinical and Cognitive Relevance of the Alpha/Beta Ratio Although this is the first study to directly evaluate the test-retest reliability of spectral power ratios, the alpha/beta ratio has already been widely used in both cognitive and clinical EEG research. For instance, Chang and Choi (2023) demonstrated that reduced alpha/beta ratios across frontal, central, and parietal regions were significantly associated with depression when compared to healthy controls, with area under the curve (AUC) values exceeding 0.7 in ROC curves. Frontal and parietal alpha/beta showed significant negative correlation with pain scores of patients with lumbar disk herniation (Li et al., 2025). Also, frontal alpha/beta showed significant differences (p value < 0.05) in between patients with frontotem- poral dementia and healthy controls (Chang & Chang, 2023). In healthy individuals, alpha/beta ratio was significantly related to stress (Yi & Mohd, 2020). Collectively, these studies highlight the potential of al- pha/beta ratio not only as a state-sensitive index of cortical functioning, but also as a reliable biomarker for longitudinal or diagnostic applications. The present study adds an essential dimension to this body of evidence by establishing that alpha/beta ratios can be measured reproducibly over both short- and long- term intervals, supporting their use in clinical monitoring and treatment evaluation. 4.3. Limitations One of the strengths of this study is the use of two independent EEG datasets with different experimental proto- cols, recording hardware, and preprocessing pipelines. The replication of key findings across these datasets increases confidence in the generalizability and robustness of spectral ratio reliability. The consistency of alpha/beta ratios in particular suggests that this feature is less sensitive to methodological variation than 13 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. many traditional EEG metrics. However, several limitations should be acknowledged. First, while Dataset 1 included extensive behavioral assessments confirming stable cognitive states across sessions, Dataset 2 lacked such measures. As a result, we cannot fully exclude the possibility of unmeasured cognitive variation contributing to spectral fluctuations in Dataset 2. Nonetheless, the short time interval between sessions (approximately 1 hour) reduces the likelihood of significant state changes. Second, although preprocessing was kept consistent across datasets where possible, the choice of preprocessing pipeline can strongly affect spectral power values (Delorme, 2023). Future studies should systematically assess how different artifact rejection techniques (e.g., ICA variants, ASR thresholds), referencing schemes, and frequency decomposition

affect the reliability of spectral features. The high reliability of spectral ratios observed here may not generalize to all analytic pipelines. Finally, while ICC provides a robust measure of test-retest reliability, it does not capture sensitivity to cognitive or clinical change. Thus, future work should evaluate both relia- bility and discriminability—the ability of features to distinguish meaningful cognitive or clinical states—in parallel. 5. Conclusions This study is the first to systematically evaluate the test-retest reliability of EEG spectral power ratios alongside absolute and relative single-band power across both short-term and long-term intervals. Using two independent datasets of healthy individuals, we demonstrate that the alpha/beta ratio, particularly in central, parietal, occipital, and temporal regions, exhibits consistently high reliability (ICC> 0.75) regardless of time interval or dataset origin. Compared to traditional spectral power metrics, spectral power ratios provided a more robust and stable measure across sessions. This suggests that ratio-based indices may help mitigate noise and inter-individual variability, making them especially well-suited for clinical and cognitive applications where reproducibility is critical. The findings support the use of spectral power ratios, especially the alpha/beta ratio, as promising biomarkers for longitudinal EEG studies and as candidate features in the development of diagnostic tools for neurological and psychiatric conditions. Future work should expand on this foundation by exploring how preprocessing choices affect reliability and by assessing these features in clinical populations where tracking changes over time is essential. Funding Source This research received no external funding. Declaration of Competing Interest The author declares no conflict of interest.

The author declares no acknowledgments. Author Contributions The corresponding author contributes to the study. 14 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. Data Availability Statement The original dataset for first dataset could be found in OpenNeuro: doi:10.18112/openneuro.ds004148.v1.0.1 The original dataset for second dataset could be found in OpenNeuro: doi:10.18112/openneuro.ds003478.v1.1.0

Akerstedt T, Gillberg M. Subjective and objective sleepiness in the active individual. Int J Neurosci. 1990;52(1-2):29-37. doi:10.3109/00207459008994241 Buzs´ aki G, Anastassiou CA, Koch C. The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nat Rev Neurosci. 2012;13(6):407- 420. doi:10.1038/nrn3241 Cavanagh JF. EEG: Depression rest [data set]. OpenNeuro. 2021. Available from: https://doi.org/10.18112/openneuro.ds003478.v1.1.0 Cavanagh JF, Bismark AW, Frank MJ, Allen JJB. Multiple dissociations between comorbid depression and anxiety on reward and punishment proces- sing: evidence from computationally informed EEG. Comput Psychiatr. 2019;3:1-17. doi:10.1162/cpsy - a 00024 Chang J, Chang C. Quantitative electroencephalography markers for an accurate diagnosis of frontotemporal dementia: a spectral power ratio approach. Medicina (Kaunas). 2023;59(12):2155. doi:10.3390/medicina59122155 Chang J, Choi Y. Depression diagnosis based on electroencephalography power ratios. Brain Behav. 2023;13(8):e3173. doi:10.1002/brb3.3173 Delorme A. EEG is better left alone. Sci Rep. 2023;13(1):2372. doi:10.1038/s41598-023-27528-0 Delorme A, Makeig S. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J Neurosci Methods. 2004;134(1):9-21. doi:10.1016/j.jneumeth.2003.10.009 Diaz BA, Van Der Sluis S, Benjamins JS, et al. The ARSQ 2.0 reveals age and personality effects on mind-wandering experiences. Front Psychol. 2014;5:271. doi:10.3389/fpsyg.2014.00271 Girden ER. ANOVA: Repeated Measures. Sage University Papers Series on Quantitative Applications in the Social Sciences, No. 84. Thousand Oaks, CA: Sage; 1992. Gor- golewski KJ, Mendes N, Wilfling D, et al. A high resolution 7-Tesla resting-state fMRI test-retest dataset with cognitive and physiological measures. Sci Data. 2015;2:140054. doi:10.1038/sdata.2014.54 Gudmunds- son S, Runarsson TP, Sigurdsson S, Eiriksdottir G, Johnsen K. Reliability of quantitative EEG features. Clin Neurophysiol. 2007;118(10):2162-2171. doi:10.1016/j.clinph.2007.06.018 Huynh H, Feldt LS. Estima- tion of the Box correction for degrees of freedom from sample data in randomized block and split-plot designs. J Educ Stat. 1976;1:69-82. doi:10.3102/10769986001001069 Johns MW. A new method for measu- ring daytime sleepiness: the Epworth sleepiness scale. Sleep. 1991;14(6):540-545. doi:10.1093/sleep/14.6.540 Khanna A, Pascual-Leone A, Michel CM, Farzan F. Microstates in resting-state EEG: current status and future directions. Neurosci Biobehav Rev. 2015;49:105-113. doi:10.1016/j.neubiorev.2014.12.010 Kli- mesch W. EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain Res Brain Res Rev. 1999;29(2-3):169-195. doi:10.1016/S0165-0173(98)00056-3 Koo TK, Li MY. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J Chiropr Med. 2016;15(2):155-163. doi:10.1016/j.jcm.2016.02.012 Li R, Shao W, Zhao S, et al. EEG al- pha/beta features as a biomarker for quantifying pain in patients with lumbar disk herniation. Front Neu- rosci. 2025;19:1507245. doi:10.3389/fnins.2025.1507245 ¨Ozbek Y, Fide E, Yener GG. Resting-state EEG alpha/theta power ratio discriminates early-onset Alzheimer’s disease from healthy controls. Clin Neu- rophysiol. 2021;132(9):2019-2031. doi:10.1016/j.clinph.2021.05.012 Schmidt MT, Kanda PA, Basile LF, et al. Index of alpha/theta ratio of the electroencephalogram: a new marker for Alzheimer’s disease. Front Aging Neurosci. 2013;5:60. doi:10.3389/fnagi.2013.00060 Stein MA, Snyder SM, Rugino TA, Hornig M. Commentary: Objective aids for the assessment of ADHD—a response to Arns et al. (2016). J Child Psychol Psychiatry. 2016;57(6):770-771. doi:10.1111/jcpp.12534 Wang Y, Duan W, Dong D, Ding L, Lei X. A test-retest resting, and cognitive state EEG dataset during multiple subject-driven states. Sci Data. 2022;9(1):566. doi:10.1038/s41597-022-01607-9 Watson D, Clark LA, Tellegen A. Development and validation of brief measures of positive and negative affect: the PANAS scales. J Pers Soc Psychol. 1988;54(6):1063- 1070. doi:10.1037//0022-3514.54.6.1063 Yi Wen T, Mohd Aris SA. Electroencephalogram (EEG) stress 15 Posted on 7 Aug 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.175458291.12752718/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. analysis on alpha/beta ratio and theta/beta ratio. Indones J Electr Eng Comput Sci. 2020;17(1):175-182. doi:10.11591/ijeecs.v17.i1.pp175-182 Zung WW. A rating instrument for anxiety disorders. Psychosomatics. 1971;12(6):371-379. doi:10.1016/S0033-3182(71)71479-0 Zung WW. A self-rating depression scale. Arch Gen Psychiatry. 1965;12:63-70. doi:10.1001/archpsyc.1965.01720310065008 Figure 1. Topographic maps of reliability across spectral power features in Dataset 1. Blue represents low ICCs and red represents high ICCs. ICCs of relative and absolute single spectral power and relative and absolute spectral power ratios were analyzed. Figure 2. Topographic maps of reliability across spectral power features in Dataset 2. Blue represents low ICCs and red represents high ICCs. ICCs of relative and absolute single spectral power and relative and absolute spectral power ratios were analyzed. 16