Resting-state EEG dual biomarker for motor–cognitive function in elderly individuals Authors: Yerim Huh, Jimin Jung, Woohee Han, Hyunjin Kim, Roneel V Sharan, Jungsoo Lee, Minji Lee Published: 2025 ABSTRACT AbstractAging is often accompanied by declines in both motor and cognitive functions, yet neurophysiological biomarkers that jointly reflect these changes remain poorly understood. Previous electroencephalography (EEG) studies have largely focused on either motor or cognitive impairment separately, limiting insights into their combined neural underpinnings. To address this gap, we examined spectral and spatial EEG features associated with both motor and cognitive function in elderly individuals during eyes-open and eyes-closed resting-states. Using power spectral density metrics across canonical frequency bands and rank-based analyses controlling for age and sex, we identified significant motor cognition interactions, predominantly in the beta band. Post-hoc and partial correlation analyses revealed that beta activity was positively associated with motor performance but negatively with cognitive performance after covariate adjustment. These findings suggest that beta oscillations capture overlapping motor–cognitive alterations in aging and may serve as potential electrophysiological biomarkers for early detection and intervention.Keywords: Electroencephalography (EEG), Motor–cognitive function, Resting-State EEG, Aging, Dual biomarkerSubject terms: Biomarkers, Neuroscience Introduction Increasing life expectancy has shifted attention toward preserving both cognitive and motor functions, which are essential for maintaining independence in elderly individuals. These two domains often decline together in aging populations; motor impairments frequently co-occur with deficits in attention, executive function, and memory1. This co-occurrence suggests shared neurophysiological mechanisms underlying both cognitive and motor decline2. Neuroimaging findings support this hypothesis, showing that structural and functional brain alterations in aging can affect both motor and cognition systems simultaneously3. In addition, clinically, incorporating motor indicators such as gait speed into cognitive assessments has been shown to enhance early detection of dementia4. Together, these findings underscore the need for integrative assessment strategies that jointly evaluate motor and cognitive function in elderly individuals. Electroencephalography (EEG) is a widely used, noninvasive technique for assessing age-related changes in brain function5. Through the decomposition of signals into canonical frequency bands, EEG enables band-specific analysis of neural oscillations that are associated with functional decline in the elderly6. Alterations in resting-state EEG—such as the slowing of peak alpha frequency and increased power in low-frequency bands (delta and theta)—have been linked to cognitive impairment and neurodegenerative processes7. Likewise, elevated theta and delta activity has been associated with motor deficits, including gait variability and postural instability8. Additionally, recent advances in electrophysiological research have highlighted the importance of brain-wide network analyses in understanding neural aging and disease. Studies using EEG and magnetoencephalography have increasingly moved beyond regional analyses to adopt comprehensive connectivity frameworks, enabling a more precise delineation of age-related alterations across distributed neural systems9–11. These findings suggest that EEG may serve as a sensitive biomarker for both motor and cognitive decline, offering potential utility for early detection and targeted intervention12. However, most prior studies have examined motor and cognitive outcomes in isolation, leaving it unclear whether a shared set of EEG features can simultaneously capture both domains of age-related decline. Addressing this gap may yield mechanistic insight into how aging alters the shared neural substrates of motor and cognitive function. From a neurophysiological perspective, aging-related declines in motor and cognitive performance are thought to arise from disrupted coordination across distributed cortical networks rather than isolated regional dysfunction13. This view aligns with evidence that spectral changes14 and spatial or network-level interactions15 jointly underpin efficient motor and cognitive processing. Therefore, a combined spectral-spatial analytical framework provides a biologically grounded means to characterize how aging alters both local neural excitability and inter-regional communication. Resting-state EEG exhibits distinct characteristics under eyes-open (EO) and eyes-closed (EC) conditions, even in healthy elderly individuals16. However, few studies have directly examined how EO and EC EEG relate to combined motor–cognitive function in aging. Given that EO and EC represent distinct neurophysiological states, analyzing both conditions may yield complementary insights into the neural mechanisms underlying functional decline. A direct comparison of EO and EC EEG thus offers a novel framework for identifying spatially and spectrally specific biomarkers of motor–cognitive aging. In this study, we propose a novel EEG-based biomarker approach to characterize combined motor and cognitive function in elderly individuals under two distinct resting conditions: EO and EC. Cognitive ability was assessed using the Korean Mini-Mental State Examination-2 (K-MMSE2)17,18, and motor function was evaluated using the Timed Up and Go (TUG) test19,20. Based on established clinical thresholds, participants were categorized into four motor–cognitive groups: GMGC (good motor and good cognition), BMGC (bad motor and good cognition), GMBC (good motor and bad cognition), and BMBC (bad motor and bad cognition). EEG power spectral density (PSD) features—including peak and median power across five canonical frequency bands—were extracted separately for EO and EC conditions. In addition, we divided the entire brain into five regions and conducted additional correlation analysis. Furthermore, to examine group-level differences, a topographical analysis was performed, and global network analysis were conducted to investigate motor and cognition related alterations in functional brain connectivity. This approach enables the identification of candidate dual biomarkers sensitive to concurrent functional decline. Our findings may serve as a robust biomarker for neurodegenerative disorders such as Alzheimer’s disease and idiopathic normal pressure hydrocephalus, which are characterized by concomitant declines in motor and cognitive function. The proposed framework is designed to sensitively and simultaneously capture both aspects. This application is particularly promising when implemented through wearable EEG devices, enabling unobtrusive at-home monitoring prior to clinical diagnosis. Such early, home-based monitoring could substantially reduce both personal and societal healthcare costs. Furthermore, the biomarker could be used as a single input for artificial intelligence algorithms to predict motor–cognitive decline, thereby improving computational efficiency and supporting scalable screening systems. Results Group-level differences in EEG spectral power To account for potential confounding effects of age and sex, we applied a nonparametric rank analysis of covariance (rank ANCOVA) with motor (good/bad) and cognition (good/bad) levels as between-group factors. The dependent variables were spectral peak and median power across canonical frequency bands. After controlling for covariates, significant main effects and interactions were observed in specific frequency bands. Eyes-open condition As shown in Table 1, no significant main effects were observed in the delta, theta, or alpha bands. In the delta band, however, sex showed significant effects in both spectral features (, for peak; , p < 0.001 for median). In the beta band, a significant interaction between motor and cognitive levels was found in the peak power (), indicating that beta oscillations are jointly influenced by motor and cognitive performance. In addition, age exerted a significant effect on both peak and median beta power (, respectively). In the gamma band, no significant effects of motor or cognition levels were found, whereas sex showed a significant effect on peak power (). Table 1Results of rank ANCOVA for EEG spectral power in the eyes-open condition.PeakMedianFrequencyFactorF statisticsp-valueFrequencyFactorF statisticsp-valueDeltaMotor1.180.280DeltaMotor0.260.612Cognition0.160.692Cognition0.300.587Motor Cognition3.010.086Motor Cognition<0.010.981Age0.660.420Age<0.010.993Sex6.450.018Sex13.80<0.001ThetaMotor0.270.604ThetaMotor0.110.741Cognition3.930.050Cognition3.790.055Motor Cognition0.550.462Motor Cognition0.800.372Age0.310.580Age0.670.416Sex0.030.870Sex<0.010.967AlphaMotor0.760.387AlphaMotor0.640.427Cognition1.070.305Cognition1.200.276Motor Cognition0.230.633Motor Cognition0.330.570Age0.310.577Age0.270.606Sex2.770.099Sex2.600.101BetaMotor<0.010.977BetaMotor0.040.847Cognition3.250.074Cognition3.280.073Motor Cognition4.000.048Motor Cognition3.280.073Age4.580.035Age4.500.036Sex0.650.421Sex0.050.830GammaMotor0.050.825GammaMotor0.080.776Cognition0.380.540Cognition0.110.737Motor Cognition0.660.419Motor Cognition0.760.385Age0.010.913Age0.150.697Sex4.450.038Sex3.690.058Note. Bold values indicate statistical significance at . Results of rank ANCOVA for EEG spectral power in the eyes-open condition. Note. Bold values indicate statistical significance at . Eyes-closed condition As shown in Table 2, no significant main effects were observed in the delta band for either peak or median power, whereas sex showed a significant effect for median power (). In the theta band, cognition level showed significant effects in both peak and median power ( for peak; for median), suggesting that theta activity primarily reflects cognitive performance. In the alpha band, cognition level again exhibited a significant main effect in peak power (), while no other main or interaction effects were found. Table 2Results of rank ANCOVA for EEG spectral power in the eyes-closed condition.PeakMedianFrequencyFactorF statisticsp-valueFrequencyFactorF statisticsp-valueDeltaMotor0.490.485DeltaMotor<0.010.949Cognition0.160.692Cognition0.540.466Motor Cognition0.290.592Motor Cognition0.640.424Age0.410.523Age0.400.530Sex2.870.094Sex5.620.020ThetaMotor0.490.487ThetaMotor0.700.404Cognition6.300.014Cognition6.060.016Motor Cognition0.740.393Motor Cognition0.530.468Age0.330.567Age0.200.653Sex3.900.051Sex3.880.052AlphaMotor0.630.431AlphaMotor0.700.405Cognition4.040.047Cognition2.220.139Motor Cognition0.720.397Motor Cognition0.420.520Age0.360.548Age0.150.699Sex0.020.900Sex0.750.388BetaMotor2.000.160BetaMotor0.730.396Cognition15.78<0.001Cognition12.280.001Motor Cognition8.590.004Motor Cognition5.260.024Age0.020.899Age0.030.862Sex0.270.604Sex0.030.872GammaMotor0.120.727GammaMotor<0.010.995Cognition1.920.169Cognition0.990.322Motor Cognition2.930.090Motor Cognition0.810.371Age0.430.512Age0.350.558Sex4.160.044Sex5.570.020Note. Bold values indicate statistical significance at . Results of rank ANCOVA for EEG spectral power in the eyes-closed condition. Note. Bold values indicate statistical significance at . Most notably, in the beta band, a significant interaction between motor and cognition levels was observed in both peak and median power ( for peak; for median). This result demonstrates that beta oscillatory activity is not solely driven by either motor or cognitive factors independently, but rather by their joint modulation, reflecting the integrative motor–cognitive interaction in the resting brain. In addition, cognition level showed a significant main effect in both features ( for peak; for median). In the gamma band, no significant main effects were found, although sex showed significant effects for both features ( for peak; for median). Even after controlling for age and sex as covariates, a significant interaction between motor and cognitive levels was consistently observed in the beta band under both eyes-open and eyes-closed conditions, suggesting that beta oscillations robustly reflect the integrated modulation of motor and cognitive functions. Post-hoc comparisons of motor and cognitive effects To further interpret the significant motor–cognition interaction observed in the beta band, a post-hoc simple-effect analysis was performed using rank ANCOVA while controlling for age and sex. As illustrated in Fig. 1, this analysis examined the simple effects of each factor at fixed levels of the other, thereby clarifying how motor and cognitive factors jointly modulate beta-band power. Specifically, four comparisons were conducted: (i) cognition effects under good and bad motor conditions, and (ii) motor effects under good and bad cognitive conditions. The detailed statistical results are summarized in Table S1. Violin plots of post-hoc simple-effect analyses for each beta-band feature. Each subplot depicts the interaction between motor and cognitive performance levels, showing the distribution of beta frequencies across groups (Good vs. Bad). Significant simple effects indicate that participants with better motor or cognitive performance exhibited higher beta median and peak frequencies in the EO or EC conditions. Asterisks (*) denote statistically significant results at . Abbreviations: EO, Eyes Opened; EC, Eyes Closed. Eyes-open condition As shown in Figure 1, under the EO condition, a significant effect was found in one of the four detailed comparisons. When cognitive function was intact, beta peak power significantly differed according to motor level (), indicating that individuals with better motor performance exhibited stronger beta activity. An additional age-related influence was also observed (), suggesting partial modulation of beta oscillations by aging. Furthermore, when motor function was fixed at a good level, cognitive level showed a significant age effect (), implying that aging may interact with cognitive status in shaping beta-band dynamics. Eyes-closed condition Under the EC condition, both peak and median beta features exhibited consistent interaction patterns. When cognitive function was preserved, significant differences in beta power were found between motor levels ( for peak; for median), highlighting the sensitivity of beta oscillations to motor performance. Conversely, among participants with impaired motor function, beta power significantly differed according to cognitive level ( for peak; for median), suggesting that beta activity also reflects cognitive decline under reduced motor capacity. Taken together, these findings demonstrate that beta oscillations reliably differentiate motor and cognitive function levels across both EO and EC conditions, with particularly strong modulation by cognitive ability in the resting condition. No other frequency bands showed consistent motor–cognitive interaction after covariate adjustment. Topographical patterns of group differences To further explore the spatial distribution of group effects, channel-wise rank ANCOVA was performed for all frequency bands, followed by pairwise group comparisons among the four groups. Distinct topographical patterns of F-statistics emerged across frequency bands (Figures S1–S4), with significant effects primarily in the theta, alpha, and beta ranges and most prominently over central, parietal and occipital electrodes. In particular, the BMGC vs. GMBC and BMGC vs. BMBC contrasts showed clusters of elevated F values in central–parietal regions, suggesting that both motor and cognitive performance levels contribute to region-specific modulation of resting-state oscillatory power. Delta and gamma bands exhibited comparatively weaker spatial effects. All statistical maps were corrected for multiple comparisons using the false discovery rate (FDR), and electrodes surviving significance after correction (FDR-corrected p <0.05) are denoted by white markers in the corresponding figures. Functional connectivity patterns across motor–cognitive function Rank ANCOVA was conducted on the weighted Phase Lag Index (wPLI) across frequency bands to control for age- and sex-related confounds. Because motor and cognitive functions rely on the coordinated interactions among multiple brain regions, network-based analyses were conducted to capture the functional connectivity underlying these processes. The analysis compared mean connectivity differences between motor and cognitive levels and their interaction. In the EO condition, the alpha band showed a significant main effect of motor level (F = 13.74, p < 0.001), as shown in Table S2. No other frequency bands exhibited significant differences for either cognition or the motor × cognition interaction under this condition. Similarly, in the EC condition, a significant effect of motor level was again observed in the alpha band (F = 10.44, p = 0.002); Table S2 summarizes the alpha-band connectivity results across conditions. However, no frequency band demonstrated a significant motor × cognition interaction, indicating that the combined influence of motor and cognitive status on connectivity was not statistically evident. These results indicate that alpha-band functional connectivity consistently differentiates motor performance across both EO and EC conditions, independent of age and sex. Associations between EEG features and behavioral performance Partial correlation analyses were performed while controlling for age, sex, and the alternate functional domain (motor or cognitive score) to examine whether EEG spectral power was independently associated with motor and cognitive performance. The analyses included five spatially distinct brain regions (frontal, central, right temporal, left temporal, and parieto-occipital) across all frequency bands. In the EO condition, no significant correlations were observed in any frequency band or region after covariate control, suggesting that EEG power did not simultaneously reflect both motor and cognitive performance. As shown in Table S3, all correlation coefficients failed to reach statistical significance across regions and frequency bands, indicating the absence of shared EEG markers between motor and cognitive domains. In contrast, under the EC condition, significant partial correlations were consistently observed in the theta and beta bands. All reported p-values were corrected for multiple comparisons using the FDR method. As presented in Table S4, theta power exhibited positive correlations with both TUG and MMSE across all regions (rho = 0.22–0.30, ). Specifically, strong associations were observed in the frontal (rho = 0.24, p = 0.017 in motor; rho = 0.23, p = 0.019 in cognition) and central (rho = 0.24, p = 0.016 in motor; rho = 0.24, p = 0.016 in cognition) regions, while the right temporal (rho = 0.29, p = 0.003 in motor; rho = 0.27, p = 0.006 in cognition), left temporal (rho = 0.22, p = 0.024 in motor; rho = 0.26, p = 0.010 in cognition), and parieto-occipital (rho = 0.25, p = 0.011 in motor; rho = 0.21, p = 0.034 in cognition) regions showed similar trends. In the beta band, EEG power exhibited negative correlations with both behavioral measures across all regions (–0.24to–0.32). The beta peak power (Fig. 2) showed significant negative correlations with both motor and cognitive performance in the frontal (–0.21 in motor; rho = –0.27, in cognition) and central (rho = –0.27, p = 0.012 in motor; rho = –0.24, p = 0.026 in cognition) regions. In addition, the beta median power (Fig. 3) showed significant negative correlations across all regions, including the frontal (rho = –0.25, p = 0.019 in motor; rho = –0.32, p = 0.008 in cognition), central (rho = –0.32, p = 0.046 in motor; rho = –0.28, p = 0.011 in cognition), right temporal (rho = –0.29, p = 0.010 in motor; rho = –0.26, p = 0.023 in cognition), left temporal (rho = –0.29, p = 0.010 in motor; rho= –0.26, p = 0.023 in cognition), and parieto-occipital (rho = –0.32, p = 0.008 in motor; rho = –0.28, p = 0.011 in cognition) regions. Eyes-closed Beta Peak Partial Correlation. The left panels show relationships between beta peak power and TUG scores, while the right panels depict correlations with MMSE scores across two cortical regions (frontal, central). Each dot represents one participant; the solid gray line indicates the fitted trend after controlling for age, sex, and the alternate functional domain, and the shaded area denotes the 95% confidence interval. Significance was determined at p < 0.05. Eyes-closed Beta Median Partial Correlation. (a) The left panels show relationships between beta median power and TUG scores, while the right panels depict correlations with MMSE scores across five cortical regions. Each dot represents one participant; the solid gray line indicates the fitted trend after controlling for age, sex, and the alternate functional domain, and the shaded area denotes the 95% confidence interval. (b) Summary of Spearman’s correlation coefficients for unadjusted (gray) and partial (red) analyses across cortical regions. In each region, two bar pairs represent associations with the Timed Up and Go (TUG; left) and Mini-Mental State Examination (MMSE; right) scores, respectively. Within each pair, gray bars indicate unadjusted correlations, whereas red bars indicate partial correlations after controlling for age, sex, and the alternate functional domain. The height and direction of each bar represent the magnitude and sign of the correlation coefficient (). Numerical values for unadjusted correlations are provided in Tables S5 and S6. Statistical significance was determined at . Overall, these results demonstrate that under the EC condition, theta power was positively associated, while beta power was negatively associated with both motor and cognitive performance, highlighting a clear oscillatory pattern that was not observed in the EO condition. Discussion Maintaining both motor and cognitive functions is essential for healthy aging, as their decline severely impacts independence and quality of life. This study aimed to identify EEG biomarkers that concurrently reflect motor and cognitive function during resting states with EO and EC. Through combined spectral, spatial, and correlation analyses controlling confounding factors, we uncovered distinct oscillatory patterns that differentiate groups based on their motor–cognitive function. When it comes to the spectral analysis, consistent and significant interaction between motor and cognition levels was observed in the beta band under both resting states. In fact, the post hoc results to identify whether there were significant changes in beta features according to differences in motor function in the group with good cognitive status showed that there were significant changes in the features of EO peak, EC peak, and EC median. The consistent differences in beta power across motor levels may reflect age-related variability in motor cortical inhibition. This interpretation is consistent with prior evidence suggesting that beta oscillations index inhibitory control in the aging motor system21. Altered beta dynamics may reflect compensatory or maladaptive adjustments in cortical excitability among elderly individuals. Interestingly, significant cognitive effects were observed only in the EC condition, specifically in beta peak and median power, when motor function was low. This finding suggests that the influence of cognitive performance on cortical oscillations becomes more pronounced in conditions where external sensory input is minimized and intrinsic neural dynamics dominate. During the EC condition, beta activity is closely related to internally directed cognitive processing and thalamocortical synchrony22,23, which reflect the brain’s capacity for maintaining neural coherence and cognitive engagement in the absence of external stimuli. Hence, individuals with better cognitive function may preserve stronger beta synchronization even when motor performance is reduced, indicating functional resilience or compensatory recruitment within cortical networks24,25. In contrast, the lack of cognitive effects in the EO condition implies that visual and sensory engagement may mask subtle cognition-related oscillatory differences, as all participants maintain a baseline level of alertness and external attention. These spatially localized differences, derived from the F-statistic–based topographical analysis, were predominantly observed in the central, parietal and occipital regions. They suggest that motor- and cognitive-related neural processes jointly modulate oscillatory activity in region-specific ways during rest. The convergent patterns across theta, alpha, and beta bands further support the notion that the motor–cognitive coupling underlying resting-state rhythms reflects integrated cortical network mechanisms rather than isolated regional effects. Consistent motor-related effects in the alpha-band functional connectivity were observed across both EO and EC conditions, even after controlling for age and sex. This suggests that alpha oscillatory networks play a stable and reproducible role in supporting motor control under varying sensory conditions26. The absence of cognition or interaction effects further indicates that motor-related connectivity changes occurred independently of cognitive performance rather than through a synergistic interaction, reflecting an independent motor modulation of large-scale network organization27. Partial correlation analyses were conducted to identify frequency bands that independently associate with motor and cognitive performance, controlling for shared variance between domains. These analyses revealed opposite patterns of association across frequency bands: positive correlations in the theta band and negative correlations in the beta band, both independently linked to motor and cognitive abilities. This complementary pattern suggests that distinct oscillatory mechanisms support motor–cognitive integration in aging, with theta activity reflecting compensatory control processes and beta activity indexing sensorimotor inhibition and efficiency. In the theta band, higher power was associated with better motor and cognitive performance across all regions, including frontal, central, temporal, and parieto-occipital areas. This finding supports prior evidence that enhanced theta synchronization reflects attentional control, working memory engagement, and executive integration in aging populations28. Such increases in theta activity, particularly in the frontal–central regions, may represent a compensatory mechanism supporting network-level coordination between motor and cognitive domains29. In contrast, significant beta band correlations emerged exclusively EC condition, suggesting that beta oscillations are particularly sensitive to intrinsic, internally driven neural dynamics. Across all regions, beta power showed significant negative correlations with both TUG and MMSE scores. Because lower TUG values indicate better motor performance, this pattern implies that higher beta power was associated with faster and more efficient motor execution30. Meanwhile, reduced beta power was related to lower MMSE scores, suggesting that decreased cortical beta activity accompanies cognitive decline. These findings are consistent with previous evidence that beta oscillations reflect sensorimotor integration and inhibitory control in the resting motor network21. Thus, greater beta synchronization under the EC condition may facilitate motor control efficiency, whereas diminished beta activity in elderly individuals reflects both weakened cortical inhibition and cognitive deterioration21,31. Taken together, these findings highlight the pivotal role of beta-band oscillations as dual biomarkers of motor–cognitive function in aging. They capture both cortical inhibitory control and cognitive resilience mechanisms, providing a neurophysiological basis for early detection and longitudinal monitoring of motor–cognitive decline using EEG-based markers. Furthermore, by enabling objective EEG-based quantification of motor–cognitive status, this dual-biomarker framework holds translational potential for clinical screening, precision rehabilitation planning, and early therapeutic intervention in aging populations. Several limitations of this study should be acknowledged. First, the cross-sectional design limits causal interpretations of the observed associations between EEG features and motor–cognitive function. Longitudinal studies are warranted to investigate whether the identified biomarkers can predict the trajectory of functional decline or the effects of interventions over time. Second, although our sample size was relatively large and well-characterized, further validation in more diverse populations, including clinical cohorts with specific neurodegenerative diseases, is necessary to establish broader applicability. Future research incorporating multimodal neuroimaging and longitudinal designs will be critical to deepen understanding and enhance the clinical utility of EEG biomarkers in aging. Conclusion This study identifies distinct EEG biomarkers that reflect combined motor and cognitive function in elderly individuals during resting-state conditions with EO and EC. Our findings highlight the value of integrating spectral and spatial analyses to uncover oscillatory patterns associated with motor-cognition functional decline. Using a non-parametric rank ANCOVA, we further confirmed that both the peak and median power of the beta band showed significant group differences across motor and cognitive levels, underscoring their potential as robust dual biomarkers of functional aging. Partial correlation analyses further revealed opposite association patterns, with theta power showing positive and beta power showing negative correlations with both motor and cognitive scores, reinforcing their complementary roles in motor–cognitive aging. This bidirectional association pattern between beta and theta bands has not been previously demonstrated in elderly resting-state EEG. These results advance our understanding of the neurophysiological basis of motor–cognitive interactions in aging and offer potential targets for early detection and personalized intervention strategies aimed at preserving independence and quality of life in the elderly population. Methods Participants A total of 120 elderly individuals were initially enrolled in the study. Of these, 16 individuals were excluded due to the presence of neurological or psychiatric disorders, unstable physiological measurements, or missing EEG data, resulting in a final sample of 104 participants (mean age = 72.56 ± 5.72 years, range 57–84; female = 64). The inclusion criteria for this study were as follows: (i) an age range of 55 to 84 years and (ii) individuals demonstrating either cognitive decline, defined as a score of 6 or higher on the Korean Dementia Screening Questionnaire-Cognition (KDSQ-C)32, or physical impairment, defined as a score of 9 or lower on the Short Physical Performance Battery (SPPB)33. The exclusion criteria were as follows: (i) severe neurological or musculoskeletal conditions affecting independent ambulation and the inability to provide informed consent, and (ii) those with a history of musculoskeletal disorders or major psychiatric disorders. All procedures were approved by the Institutional Review Board of Kumoh National Institute of Technology (approval number: 202306-HR-004) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants. To examine EEG differences by motor–cognitive function, participants were classified into four groups based on cognitive and motor performance. Cognitive function was assessed using the Korean version of the K-MMSE2 (mean = 26.27 ± 2.93, range: 17–30), with a cut-off of 24 indicating impaired cognition17. Motor function was evaluated using the TUG test (mean = 7.86 ± 2.00 s, range: 5.08–13.90), with sex-specific thresholds (9.21 s for males; 9.33 s for females20) used to define reduced motor performance. Accordingly, participants were divided into four groups: cognitively and physically normal (GMGC), cognitively impaired but physically normal (BMGC), cognitively normal but physically impaired (GMBC), and both cognitively and physically impaired (BMBC). Detailed group characteristics are presented in Table 3. Table 3Demographic factors by motor–cognition group.TotalGood motor & good cognitionGood motor & bad cognitionBad motor & good cognitionBad motor & bad cognitionSex (Male/Female)40/6437/331/122/80/11Age72.56 ± 5.72 [57, 84]70.51 ± 4.92 [57, 82]74.92 ± 6.08 [64, 83]77.90 ± 4.95 [70, 84]77.91 ± 2.63 [74, 82]K-MMSE226.27 ± 2.93 [17, 30]27.76 ± 1.47 [25, 30]22.38 ± 1.66 [18, 24]26.30 ± 1.64 [25, 30]21.36 ± 2.46 [17, 24]TUG maximum7.86 ± 2.00 [5.08, 13.90]6.92 ± 1.03 [5.11, 9.26]7.53 ± 1.02 [5.08, 8.71]11.31 ± 1.04 [9.58, 13.31]11.12 ± 1.52 [9.34, 13.90]Note. Variables are summarized as mean ± SD [min, max]. Demographic factors by motor–cognition group. Note. Variables are summarized as mean ± SD [min, max]. EEG recording and preprocessing Resting-state EEG data were recorded in two 5-minute sessions (EO and EC). Participants were seated comfortably in a quiet room with minimal external distractions and were instructed to maintain a relaxed posture while minimizing movement. During the EO condition, they fixated on a central cross at eye level, remained relaxed but alert, and avoided excessive blinking or eye movements. During the EC condition, they closed their eyes, relaxed, and stayed awake without engaging in deliberate mental activity. EEG signals were acquired using a 19-channel dry-electrode system (CGX Quick-20r, Cognionics Inc., San Diego, CA, USA) following the international 10–20 system. According to the CGX Quick system’s standard configuration, both the reference and ground electrodes were attached to the left earlobe using an ear clip with two separate snap connectors. This configuration provides stable contact and minimizes motion artifacts while maintaining electrical independence between reference and ground, as stated by the manufacturer. Signals were digitized at 500 Hz. Preprocessing was conducted using the EEGLAB toolbox34 in MATLAB. Data were segmented into 5-second non-overlapping epochs35, filtered (0.5–50 Hz) using a Hamming-windowed sinc FIR filter, and re-referenced to the average signal36. Independent component analysis with ICLabel37 was used to remove non-neural components, including muscle, eye, and line-noise artifacts. Bad channels were identified and interpolated using the MNE-Python toolbox38, and trials with severe artifacts were further excluded based on visual inspection of the EEG data to ensure data quality and reproducibility. EEG spectral analysis PSD was estimated using Welch’s method across five canonical EEG frequency bands: delta (1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), and gamma (30–50 Hz). Welch’s method is widely adopted in EEG research due to its ability to reduce variance in spectral estimates through segment averaging, making it suitable for non-stationary biological signals. PSD was calculated from each non-overlapping 5 s epoch (500 Hz; 2,500 samples) using Welch’s method with a 2,500-point window corresponding to the full epoch length and no within epoch overlap39. For each 5-second epoch and EEG channel, two spectral features were extracted within each frequency band40. These two complementary measures characterize both the dominant and the central tendencies of neural oscillatory activity, which have been shown to be sensitive to age, cognitive state, and neurological conditions22. Feature extraction was implemented in Python using the MNE-Python library. EEG connectivity analysis Functional connectivity was quantified using the wPLI, which measures phase synchronization between EEG channels while minimizing the influence of volume conduction and common sources41. Connectivity analysis was conducted separately for five canonical frequency bands. For each subject, preprocessed epoch data were analyzed using a frequency-domain connectivity estimation method implemented in MNE-Python (v1.10.1) with the multitaper approach and 3 Hz bandwidth smoothing. The wPLI was computed as below: where denotes the complex cross-spectral density between signals x and y at frequency f, extracts its imaginary part representing non-zero phase-lag interactions, and indicates the expectation operator used to normalize the measure between 0 and 141. The resulting complex cross-spectral density estimates were converted into full 19 × 19 connectivity matrices, from which the mean wPLI value was computed after excluding diagonal elements. Subsequently, rank ANCOVA was performed to evaluate differences in wPLI across motor (good/bad) and cognitive (good/bad) levels while controlling for age and sex. For each frequency band, F-statistics and corresponding p-values were extracted to determine the effects of motor and cognitive performance on large-scale functional connectivity. Statistical analysis To examine the effects of motor and cognitive functions on EEG features while controlling for potential confounders, we employed a non-parametric statistical framework. The Shapiro–Wilk test indicated that most EEG feature distributions significantly deviated from normality. Furthermore, because age and sex differed across the four groups, a rank ANCOVA42 was adopted to statistically adjust for these covariates. Analyses were conducted separately for each frequency band and for the two spectral features. In the rank ANCOVA, motor level (good/bad) and cognitive level (good/bad) were treated as between-group factors, while age and sex were entered as covariates. Rank-transformed EEG power values were analyzed to ensure robustness against non-normality and outliers, and all tests were computed using Type III sums of squares to accommodate unbalanced group sizes and interaction effects43. This procedure enabled evaluation of the main and interaction effects (motor cognition) on EEG spectral features after controlling for covariates. When a significant interaction between motor and cognitive factors was detected, post-hoc simple-effect analyses were conducted within the same rank ANCOVA framework. Four comparisons were performed: (i) the effect of cognition under the good motor condition, (ii) the effect of cognition under the bad motor condition, (iii) the effect of motor under the good cognitive condition, and (iv) the effect of motor under the bad cognitive condition. These tests clarified how one factor influenced EEG power at different levels of the other while maintaining the same covariate structure. Because each simple-effect test was conducted within an independent subgroup rather than across multiple comparisons, additional FDR correction was not applied. To further examine the spatial distribution of the observed group effects across scalp channels, rank ANCOVA was applied at each electrode while controlling for age and sex. Pairwise comparisons among the four groups were then conducted, and multiple-comparison correction was applied using the FDR method to control Type I errors without being overly conservative. For the statistical evaluation of functional connectivity, mean wPLI values for each subject and frequency band were analyzed using the same rank ANCOVA framework. Motor level (good/bad) and cognitive level (good/bad) served as between-group factors, with age and sex included as covariates. Interaction effects between motor and cognitive levels (motor cognition) were also examined to determine whether large-scale connectivity varied as a function of motor or cognitive performance after accounting for demographic covariates. In all analyses, statistical significance was set at . Partial correlation analysis To assess the independent relationships between EEG spectral features and motor–cognitive performance, partial correlation analyses were performed while controlling for age, sex, and the alternate functional domain (motor or cognition). All reported p-values in the partial correlation analyses were corrected for multiple comparisons using the FDR method to ensure statistical rigor. Regional grouping was applied to reduce spatial variability across subjects and to enable robust statistical comparison across functionally coherent cortical areas, and electrodes were grouped into five anatomically and functionally meaningful regions: frontal (Fp1, Fp2, F3, F4), parieto-occipital (P3, P4, O1, O2), central (Fz, C3, Cz, C4, Pz), right temporal (F8, T8, P8), and left temporal (F7, T7, P7). The five-region configuration reflects large-scale cortical networks associated with motor execution (central and frontal regions), visual–attentional processing (parieto-occipital regions), and language- and association-related integration (temporal regions), thereby enabling a focused assessment of spatial patterns linked to motor and cognitive functions44. Partial correlation allows estimation of the unique association between two variables after removing shared variance explained by covariates. We used a non-parametric (Spearman) partial correlation as follows. Let and denote the average-rank transforms of variables X and Y (ties assigned average ranks), and let be the design matrix of average-ranked covariates (age, sex, and the alternate functional domain). Residuals were obtained via , i.e., and . The partial Spearman correlation was computed as the Pearson correlation between residuals: where and denote the residualized average ranks for subject i, and N is the number of subjects. Because the closed-form formula below assumes no tied ranks, we computed Spearman’s as the Pearson correlation of average ranks: where represents the Spearman correlation coefficient between ranked variables X and Y. Positive values of indicate a monotonic increase, whereas negative values indicate a monotonic decrease. Supplementary Information Below is the link to the electronic supplementary material. Supplementary material 1 Author contributions Y.H., J.J. - Conceptualization, Data Curation, Methodology, Formal Analysis, Software, Writing - Original Draft, Visualization; W.H., H.K. - Investigation, Data Curation, Methodology; R.V.S. - Validation, Writing - Review & Editing; J.L. - Data Curation, Resources, Conceptualization, Methodology, Writing - Review & Editing, Supervision, Project administration; M.L. - Conceptualization, Methodology, Writing - Review & Editing, Supervision, Project administration. Funding This work was partly supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. RS-2023-00208884; RS-2024-00336880), the IITP(Institute of Information & Communications Technology Planning & Evaluation)-ICAN(ICT Challenge and Advanced Network of HRD) grant funded by the Korea government(Ministry of Science and ICT) (IITP-2025-RS-2024-00438207, 25%). Data availability The datasets generated or analyzed during the current study are not publicly available due to institutional and ethical restrictions, but are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally to this work: Yerim Huh and Jimin Jung. Contributor Information Jungsoo Lee, Email: jungsoo@kumoh.ac.kr. Minji Lee, Email: minjilee@catholic.ac.kr. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-27883-0. References Associated Data Supplementary Materials Supplementary material 1 Data Availability Statement The datasets generated or analyzed during the current study are not publicly available due to institutional and ethical restrictions, but are available from the corresponding author on reasonable request.