Cortical activation, gait performance, and muscle activation during dual-task walking in older adults with mild cognitive impairment

Abstract

Background

Dual-task (DT) walking poses great challenges to gait control in older adults with mild cognitive impairment (MCI). However, the integrated adaptive characteristics of cortical, muscular, and gait responses under DT interference remain insufficiently elucidated.

Objective

This study aimed to compare gait performance, cortical activation, and muscle activation characteristics between older adults with MCI and cognitively intact controls during DT walking.

Methods

A total of 25 MCI older adults and 25 cognitively intact controls were recruited and completed single-task (ST) and DT walking tests. The DT condition involved performing a continuous subtraction task (subtracting 3 from a random number between 400 and 500) while walking. Data were collected using functional near-infrared spectroscopy, a 3D motion capture system, and surface electromyography to quantify cortical activation, gait parameters, and lower-limb muscle activation.

Results

Under the DT walking condition, the MCI group showed significant gait impairments, including decreased gait speed, increased step width, and altered gait variability. Among the examined bilateral cortical regions, significant Group × Task interaction effects in cortical activation were observed in the left prefrontal cortex, left primary motor cortex, and left supplementary motor area, indicating altered task-related cortical activation patterns in older adults with MCI during DT walking. In addition, rectus femoris activation was significantly higher in the MCI group than in controls during DT walking.

Conclusion

DT walking induced noticeable gait impairments in older adults with MCI, which were accompanied by adjustments in cortical control patterns and increased muscular activation. However, these potential compensatory responses, characterized by increased muscle activation and altered cortical regulation, appeared insufficient to fully mitigate the negative effects of cognitive load on gait performance, suggesting reduced gait automaticity in older adults with MCI.

Trial registration

ChiCTR2500105909, 7/14/2025.

Introduction

Mild cognitive impairment (MCI) is a clinical syndrome defined by objective cognitive deficits that represent a transitional stage between normal aging and dementia [1, 2]. It affects approximately 6.7% to 25.2% of older adults [3]. MCI is primarily characterized by impairments in learning and memory, language, visuospatial ability, complex attention, executive functioning, or social functioning [4].

Maintaining mobility through regular physical activity, such as safe walking, is essential for healthy aging, functional independence, and quality of life in older adults [5]. Declining cognitive function may impair gait performance during walking in older adults with MCI. In daily life, walking is frequently performed under dual-task (DT) conditions, such as walking while speaking [6], performing mental calculations, or avoiding obstacles [7]. With age-related declines in brain function and motor control capacity, older adults may require greater cognitive effort to maintain walking performance under DT conditions [8]. Compared with single-task (ST) walking, DT walking imposes greater cognitive demands and is associated with poorer gait performance, and a higher risk of falls in older adults [9, 10]. Previous studies have shown that DT walking can lead to reductions in gait speed and stride length, particularly in individuals with cognitive decline [11, 12]. Similar findings have also been reported in stroke, where proprioceptive training combined with DT exercises improved gait speed in individuals with chronic stroke [13]. Therefore, we used a DT paradigm to investigate the effect of cognitive impairment on gait performance in older adults.

DT walking requires a high level of cognitive function. This involves regulating gait control and allocating attention efficiently [14, 15]. This process depends on effective muscular function, sensory feedback, and the integration of central nervous system resources. These factors coordinate motor commands and regulate multiple muscle groups during walking under increased task demands [16, 17]. Given that cognitive decline is often accompanied by degeneration of neural and muscular systems, maintaining stable and efficient lower-limb muscle activation during DT walking becomes particularly challenging [18]. Previous studies have reported DT walking with smartphone-related cognitive or motor tasks reduced gait speed and stride length, while increasing muscle activation in healthy young adults [19]. However, the characteristics of lower-limb muscle activation during challenging walking tasks in older adults with MCI have not been fully investigated.

Since gait in daily life often occurs under DT conditions, understanding how cognitive impairment affects walking requires examining both cortical and muscular control. According to the capacity-sharing theory, cognitive resources are limited and must be allocated between concurrent tasks, which may lead to performance decline under DT conditions [20]. Functional near-infrared spectroscopy (fNIRS) is a non-invasive neuroimaging technique that enables repeated monitoring of cortical hemodynamic responses during naturalistic motor tasks, with relatively high tolerance to movement-related artifacts [21, 22]. Based on neurovascular coupling, fNIRS measures task-related changes in oxyhemoglobin and deoxyhemoglobin concentrations and has been widely used to investigate cortical activation during locomotion [23–26]. The prefrontal, motor, and sensory cortices jointly contribute to the cognitive control, motor planning, and coordination required for walking, particularly in complex environments [10, 25, 27]. Evidence suggests that DT walking leads to task-related modulation in cortical activation, especially within prefrontal and motor networks [7]. However, findings remain inconsistent, with both increased and decreased cortical activation reported in older adults with cognitive impairment, which may reflect differences in task demands or neural efficiency [28–30]. This study combines fNIRS with surface electromyography (sEMG) to simultaneously examine cortical hemodynamic responses, lower-limb muscle activation, and gait performance in older adults with MCI.

In summary, this study aims to investigate differences in cortical and muscle activation characteristics as well as gait performance between older adults with MCI and cognitively intact controls during DT walking. It also aims to explore cortical-muscle coordination and control patterns under the DT conditions. The findings may provide neurophysiological and biomechanical evidence for fall risk assessment and exercise program development for older adults with MCI. We hypothesized that: (1) compared with cognitively intact older adults, DT walking would lead to greater impairments in gait performance, increased gait variability, and elevated muscle activation in older adults with MCI, whereas group differences would be minimal under ST walking; (2) older adults with MCI would exhibit reduced cortical activation compared with cognitively intact controls during DT walking.

Methods

Participants

Participants were recruited from the community. Recruitment was conducted through advertisements on WeChat, flyers, and posters. A priori power analysis was conducted using G*Power 3.1.9.7 software to estimate the minimum required sample size. Based on an effect size (partial eta squared, η²ₚ = 0.55 [31]), an alpha level of 0.05, and a statistical power of 0.80, the minimum total sample size was established as 20. Considering potential attrition, the final cohort comprised 25 older adults with MCI and 25 age-matched cognitively intact controls.

Prior to enrolment, written informed consent was obtained from all participants. The study protocol was approved by the Ethics Committee of Sports Science, Shandong Sport University (approval No. 2025045) and was conducted in accordance with the Declaration of Helsinki. The study was also registered with the Chinese Clinical Trial Registry (registration No. ChiCTR2500105909).

Cognitive status was screened using the Montreal Cognitive Assessment (MoCA). A cutoff score of < 26 was used to identify participants with MCI risk in community-based older adults [32]. Inclusion criteria [33]: MCI group: age ≥ 65 years; MoCA score < 26 [32]; independent ambulation without assistive devices; right-hand dominance. Control group: age ≥ 65 years; MoCA score ≥ 26; independent ambulation without assistive devices; right-hand dominance. Exclusion criteria: musculoskeletal diseases or diseases affecting walking [7]; severe cardiovascular, respiratory, or neurological diseases [27]; severe visual or hearing impairment.

The MCI and control groups demonstrated comparable baseline characteristics, with no significant between-group differences in age, sex, height, body weight, leg length, and MoCA score (Table 1).

Table 1.

Participants’ basic information (mean ± SD)

	Control group (n = 25)	MCI group (n = 25)	p value
Age(year)	68.84 ± 3.90	70.16 ± 2.64	0.168
Sex (male\female)	9\16	9\16	1.000
Height (cm)	161.30 ± 7.57	158.07 ± 6.98	0.123
Weight (kg)	66.66 ± 11.99	63.07 ± 12.27	0.300
Leg length (cm)	81.80 ± 3.41	82.14 ± 4.64	0.781
MoCA score	27.26 ± 1.10	22.20 ± 1.98	< 0.001

Experimental protocol

Study design

A cross-sectional, repeated-measures design was employed. All participants conducted walking tests under ST and DT conditions in a randomized order using a computer-generated randomization protocol to minimize potential order effects. Participants then changed into standardized attire (consisting of comfortable sportswear and athletic shoes), wore the required wearable devices, and were familiarized with the experimental procedures. Demographic information was collected prior to testing. Each task condition was performed for three trials, which resulted in a total of six trials. Each trial lasted 60 s and was followed by a rest period of 30s.

Task conditions

ST walking: Participants walked along a 10 m walkway at a self-selected speed for 60 s without any additional task. DT walking: Participants walked along walkway while simultaneously completing a cognitive task. The cognitive task required continuous serial subtraction of 3 from a randomly selected number between 400 and 500 [34]. A new starting number (400–500) was verbally provided immediately before each trial, and all verbal responses were recorded to calculate task accuracy (Fig. 1).

Fig. 1.

Schematic diagram of experimental procedure and environment

Data collection

Cortical activation

fNIRS data were collected using Nirsmart software (Huichuang, Zhenjiang, China) at a sampling frequency of 11 Hz. The optical system comprised 24 emitters and 16 receivers, with a photodetector-emitter distance of 3 cm, which resulted in a total of 48 measurement channels. fNIRS channels were mapped to the Montreal Neurological Institute standard space according to the international EEG 10–20 system. Channels were subsequently classified into specific Brodmann areas based on spatial overlap, with the highest proportion of channels representing functional regions [35]. Participants wore a full-head fiber-reinforced test cap during testing. Staff then parted hair beneath the emitters to expose the scalp for checking and adjusting signal intensity. The head cap was covered with a black cloth cover to effectively avoid the interference of ambient light and other potential light sources. Staff must be proficient in operating fNIRS equipment, which minimizes preparation time to avoid skin pressure from photoelectrodes. Throughout task execution, the start of each test—whether walking or resting—is synchronously calibrated within the fNIRS time-domain signal, which enables simultaneous acquisition of cerebral blood flow alongside kinematic and kinetic data.

Channel numbers (e.g., CH1) correspond to specific optode positions and cortical regions. Channels were grouped into anatomically defined regions of interest (ROIs) associated with cognitive-sensorimotor control during walking, including the prefrontal cortex (PFC), primary motor cortex (M1), primary somatosensory cortex (S1), and supplementary motor area (SMA). The final ROI-based analysis included 37 channels covering the PFC, M1, S1, and SMA. Each ROI was further divided into left and right hemispheres. Channels located outside the predefined ROIs, including CH26 and CH34, were excluded from the ROI-based analyses. The ROI-based analysis and the use of a single hemodynamic indicator reduced the number of statistical comparisons and helped control the risk of Type I error. (Fig. 2; Table 2)

Fig. 2.

Schematic diagram of data collection and processing. Notes. BF, long head of the biceps femoris; RF, rectus femoris; MG, medial gastrocnemius; TA, tibialis anterior. MVIC, maximum voluntary isometric contraction; PFC, prefrontal cortex; M1, primary motor cortex; S1, primary somatosensory cortex; SMA, supplementary motor area; OC, occipital cortex; ST, single-task; DT, dual-task. The OC channels and channels 26 and 34 were excluded from data analysis

Table 2.

fNIRS channel assignments for each region of interest (ROI)

ROIs	Channel numbers (CH)
lPFC	8, 9, 10, 11, 12, 20, 22, 23, 24, and 25
rPFC	3, 4, 5, 6, 7, 17, 18, 19, and 21
lM1	31 and 44
rM1	15, 16, and 42
lS1	30, 32, and 33
rS1	1 and 35
lSMA	13, 14, 27, and 43
rSMA	2, 28, 29, and 45

PFC Prefrontal cortex, M1 Primary motor cortex, S1 Primary somatosensory cortex, SMA Supplementary motor area

Gait parameters and muscle activation

Gait data were acquired using a 3D motion capture (Vicon, Oxford Metrics Ltd., Oxford, UK) with 12 cameras at 100 Hz. The sampling frequency of the AMTI force measuring platform was 1,000 Hz. Muscle activation was recorded at 2,000 Hz using a wireless EMG telemetry system (Noraxon Inc. Scottsdale, USA) from the following muscles on the dominant limb: the tibialis anterior (TA), medial gastrocnemius (MG), rectus femoris (RF), and the long head of the biceps femoris (BF). The electrodes were placed in the acquisition position of the target muscle, and the acquisition signal was synchronized with the fNIRS system.

Data processing

fNIRS preprocessing

fNIRS analysis was performed using MATLAB with the NIRS_KIT toolbox [36]. Raw optical recordings exported from Nirsarmt acquisition software in nirs format were pre-processed (filtering, artefact rejection) and analyzed within NIRS_KIT. Optical density signals were first converted to oxygenated hemoglobin (HbO) and deoxyhemoglobin (HbR) concentrations using the modified Beer–Lambert law. Task-related cortical activation is generally reflected by an increase in HbO accompanied by a decrease in HbR due to neurovascular coupling [37]. Because changes in oxygenated hemoglobin (ΔHbO₂) generally exhibit greater sensitivity to task-related cortical activation and higher signal-to-noise ratios than HbR, ΔHbO₂ was used as the primary indicator for subsequent analyses [38]. An infinite impulse response band-pass filter (0.01–0.1 Hz) removed physiological noise associated with respiration (0.4 Hz), heartbeat (1–2 Hz), and blood pressure (0.1 Hz), and motion artifacts were corrected using the temporal derivative distribution repair method [39]. The preprocessed HbO signals were segmented based on the task markers, and data from the beginning (0 s) to the end (60 s) of each walking trial were extracted to obtain task-state data across 6 trials.

Gait processing

After performing operations such as naming, complementing, modeling, segmenting, and deleting in the Vicon Nexus system, the data were imported into Visual 3D (USA, Visual Version 6.0) for analysis. Kinematic data were filtered using a Butterworth low-pass filter with a cutoff frequency of 10 Hz [40]. Moreover, the start and end points of gait cycles were defined using labels, with a complete gait cycle being from the time the heel of one foot strikes the ground until that heel strikes the ground again. Foot strikes were defined as the moment when the heel contact force table was greater than 20 N [40]. The mean and variability of gait cycle, stride length, stride width, stride frequency, and stride speed were calculated. The coefficient of variation (CV) [CV= (standard deviation / mean) × 100%] was used to calculate variability [7].

DT cost of gait performance

The DT cost of gait speed was calculated as (ST-DT) / ST × 100% to quantify the degree of DT interference [41].

Muscle activation processing

sEMG signals were recorded at 2,000 Hz with the Noraxon surface EMG system and exported as c3d files. All processing was completed in Visual 3D: band-pass filtering with a 4th-order zero-lag Butterworth filter (10–400 Hz); full-wave rectified, and smoothed with a 100-ms root-mean-square (RMS) sliding window to generate linear envelopes. For normalization, the 1–4 s epoch exhibiting the highest EMG amplitude was extracted from each of three maximal voluntary isometric contractions (MVICs) and averaged to obtain the reference average surface EMG. The RMS envelope recorded during walking was then normalized to the MVIC value and expressed as %MVIC. The integral electromyogram was calculated using the formula.

Cognitive task data processing

The outcome measure of cognitive task performance was calculated as (the number of correct calculations / the total number of calculations) × 100%. The results revealed no significant differences in accuracy across groups and task conditions.

Statistical analysis

Statistical analysis was performed using SPSS 26.0 software. The mean values of three trials for each condition were used for the data analysis. All variables were presented as mean ± SD. The Shapiro–Wilk test was conducted to verify the normality of the data. If a variable significantly deviated from a normal distribution (p < 0.05), the corresponding comparisons were performed using nonparametric tests. Categorical data (gender) were analyzed using chi-square tests. T-tests were conducted on continuous variables (height, age, weight, education duration, MoCA score). The two-way mixed-design analysis of variance (ANOVA) was conducted for each outcome measure, with groups (MCI group and control group) and task conditions (ST and DT) as factors. The aim was to examine the main effects of group, task, and their interaction. If significant interaction effects were present, Bonferroni-corrected post-hoc analyses were performed. Channels were grouped into anatomically defined ROIs. The HbO signals within each ROI were averaged, and statistical analyses were conducted at the ROI level to reduce the number of comparisons and improve interpretability. The significance level was set at 0.05.

Partial eta squared (η²_p) was used to represent the effect size of main effects and interactions in ANOVA. The thresholds for η²_p [42] were as follows: <0.06 (small), 0.06–0.14 (medium), > 0.14 (large), which were in accordance with behavioral science standards. Cohen’s d (d) [43] was used to represent the effect size of post hoc analysis, with the following thresholds: <0.20 (trivial); 0.21–0.50 (small); 0.51–0.80 (medium); >0.81 (large).

Results

Cortical activation

Two-way mixed-design ANOVA revealed significant Group × Task interaction effects for the ΔHbO₂ in the lPFC (p = 0.043, η²_p = 0.083), lM1 (p = 0.007, η²_p = 0.143), lSMA (p = 0.027, η²_p = 0.098). Post-hoc analyses showed that, in the control group, the ΔHbO₂ in the lPFC (p = 0.006, d = 0.617), lM1 (p = 0.001, d = 0.809), and lSMA (p = 0.009, d = 0.639) were significantly lower under the DT condition than under the ST condition. Furthermore, under the ST condition, the ΔHbO₂ in the MCI group was significantly lower than that in the control group in the lPFC (p = 0.005, d = 0.825), lM1 (p = 0.006, d = 0.818), and lSMA (p = 0.028, d = 0.640) (Figs. 3 and 4; Table 3).

Fig. 3.

3D plots of cortical activation levels between MCI and control groups under ST and DT conditions. Notes. A Con-ST, cognitively intact older adults during ST walking; B Con-DT, cognitively intact older adults during DT walking; C MCI-ST, older adults with MCI during ST walking; D MCI-DT, older adults with MCI during DT walking

Fig. 4.

Cortical activation levels in channels showing significant Group × Task interaction effects between the MCI and control groups under ST and DT conditions. Notes. A–H correspond to ΔHbO₂ in the left PFC, right PFC, left M1, right M1, left S1, right S1, left SMA, and right SMA, respectively. PFC, prefrontal cortex; M1, primary motor cortex; S1, primary somatosensory cortex; SMA, supplementary motor area. * p < 0.05

Table 3.

Comparison of cortical activation between MCI and control groups under single-task and dual-task conditions (ΔHbO₂, ×10⁻⁷ mmol/L; mean ± SD)

Cortical activation	Control group		MCI group		group		task		group × task
	Con-ST	Con-DT	MCI-ST	MCI-DT	p	η2p	p	η2p	p	η2p
lPFC	0.732±0.80	0.276±0.67	0.102±0.73	0.113±0.56	0.018*	0.111	0.054	0.075	0.043*ab	0.083
rPFC	0.668±0.93	0.341±0.58	0.141±0.52	-0.053±0.63	0.003*	0.174	0.047*	0.080	0.607	0.006
lM1	0.504±0.62	0.033±0.55	0.028±0.55	0.103±0.33	0.075	0.065	0.045*	0.081	0.007*ab	0.143
rM1	0.351±0.53	0.155±0.52	0.116±0.49	0.079±0.40	0.170	0.039	0.165	0.040	0.342	0.019
lS1	0.480±0.63	0.270±0.57	-0.016±0.51	-0.003±0.51	0.007*	0.143	0.213	0.032	0.160	0.041
rS1	0.416±0.58	0.213±0.63	0.015±0.55	-0.035±0.44	0.007*	0.144	0.247	0.028	0.479	0.010
lSMA	0.361±0.55	0.030±0.48	0.037±0.46	0.100±0.45	0.017*	0.113	0.129	0.047	0.027*ab	0.098
rSMA	0.207±0.60	0.036±0.54	0.058±0.46	0.027±0.39	0.471	0.011	0.274	0.025	0.449	0.012

Con-ST Cognitively intact older adults during single-task walking, Con-DT Cognitively intact older adults during dual-task walking, MCI-ST Older adults with mild cognitive impairment during single-task walking, MCI-DT Older adults with mild cognitive impairment during dual-task walking, PFC Prefrontal cortex, M1 Primary motor cortex, S1 Primary somatosensory cortex, SMA Supplementary motor area. ^* p < 0.05 for group × task interaction. ^a significant difference between groups. ^b significant difference between tasks

The ΔHbO₂ in the rPFC, lS1 and rS1 showed significant main effects of group, with significantly lower ΔHbO₂ in the MCI group compared with the control group. The ΔHbO₂ in the rPFC showed significant main effects of task, with significantly lower ΔHbO₂ in the DT condition compared with the ST condition (Figs. 3 and 4; Table 3).

Gait performance

Significant Group × Task interaction effects were found for stride width (p = 0.032, η²_p = 0.100) and gait speed (p = 0.007, η²_p = 0.152). Post-hoc analysis revealed that, under the DT condition, older adults with MCI exhibited significantly longer stride width (p = 0.037, d = 0.397) and slower gait speed (p < 0.001, d = 4.52) than the control group. Within the control group, stride width was significantly shorter during the DT condition than during the ST condition (p = 0.041, d = 0.104); within the MCI group, gait speed was significantly slower during the DT condition than during the ST condition (p < 0.001, d = 4.50). In addition, stride frequency, gait speed, and gait cycle showed a significant main effect of task: participants exhibited lower stride frequency, lower gait speed, and longer gait cycle under the DT condition than under the ST condition (all p < 0.001, η²_p = 0.544, 0.287, and 0.509, respectively) (Fig. 5; Table 4).

Fig. 5.

Spatiotemporal gait parameters between MCI and control groups under ST and DT conditions. Note. A–E correspond to gait cycle, stride frequency, stride length, stride width, and gait speed, respectively. * p < 0.05

Table 4.

Gait performance between MCI and control groups under ST and DT conditions (mean ± SD)

Gait performance measures	Control group		MCI group		group		task		group × task
	Con-ST	Con-DT	MCI-ST	MCI-DT	p	η2p	p	η2p	p	η2p
Gait cycle (m)	1.134±0.11	1.208±0.10	1.149±0.10	1.239±0.114	0.462	0.012	<0.001*	0.509	0.508	0.010
Stride frequency (steps/min)	106.3±9.61	100.2±10.7	105.4±8.59	97.7±10.2	0.530	0.009	<0.001*	0.544	0.400	0.016
Stride length (m)	0.623±0.04	0.603±0.06	0.614±0.04	0.594±0.06	0.502	0.010	0.010	1.390	0.990	<0.001
Stride width (m)	0.119±0.03	0.103±0.22	0.114±0.27	0.122±0.38	0.295	0.025	0.451	0.013	0.032*ab	0.100
Gait speed (m/s)	0.963±0.32	0.929±0.34	0.935±0.33	0.763±0.35	0.022*	0.112	<0.001*	0.287	0.007*ab	0.152
Gait cycle variability (%)	0.021±0.10	0.034±0.03	0.031±0.02	0.037±0.016	0.058	0.083	0.024*	0.116	0.349	0.021
Stride frequency variability (%)	0.212±0.01	0.035±0.02	0.025±0.02	0.043±0.02	<0.001*	0.367	0.156	0.052	0.500	0.012
Stride length variability (%)	0.038±0.03	0.021±0.01	0.037±0.03	0.041±0.02	0.078	0.072	0.223	0.035	0.043*ab	0.094
Stride width variability (%)	0.041±0.02	0.039±0.02	0.041±0.02	0.052±0.04	0.279	0.028	0.391	0.018	0.240	0.033
Gait speed variability (%)	0.124±0.12	0.099±0.10	0.116±0.10	0.098±0.10	0.849	0.001	0.281	0.026	0.849	0.001

^*p < 0.05 for group × task interaction. ^a significant difference between groups. ^b significant difference between tasks

A significant Group × Task interaction effect was found for stride length variability (p = 0.043, η²_p = 0.094). Post-hoc analysis revealed that, under the DT condition, older adults with MCI exhibited significantly higher stride length variability (p = 0.001, d = 0.158) than the control group. Within the control group, stride length variability was significantly lower during the DT condition than during the ST condition (p = 0.026, d = 3.79). Stride frequency variability and gait cycle variability showed a significant main effect of task: participants exhibited lower stride frequency variability and greater gait cycle variability under the DT condition than under the ST condition (p < 0.001, η²_p = 0.367; p = 0.024, η²_p = 0.116). No significant main effect of group was found (Fig. 6; Table 4).

Fig. 6.

Gait variability between MCI and control groups under ST and DT conditions. Note. A–E correspond to gait cycle variability, stride frequency variability, stride length variability, stride time variability, and gait speed variability, respectively. * p < 0.05

A significantly higher DT cost was found in the MCI group than in the control group (control group: 1.55 ± 19.39%; MCI group: 18.37 ± 16.06%; t (45) = − 3.23, p = 0.002).

Muscle activation

A significant Group × Task interaction effect was found for RF (p = 0.036, η²_p = 0.094). Post-hoc analysis revealed that, under the DT condition, older adults with MCI exhibited significantly greater muscle activation (p = 0.020, d = 0.715) than the control group. Within the control group, muscle activation was significantly greater during the DT condition than during the ST condition (p = 0.006, d = 0.199). No significant main effects of group and task were found in the other muscle (Fig. 7; Table 5).

Fig. 7.

Muscle activation between MCI and control groups under ST and DT conditions. Notes. TA, tibialis anterior; MG, medial gastrocnemius; RF, rectus femoris; BF, biceps femoris long head. * p < 0.05

Table 5.

Muscle activation between MCI and control groups under ST and DT conditions (mean ± SD)

Muscle activation measures	Control group		MCI group		group		task		group × task
	Con-ST	Con-DT	MCI-ST	MCI-DT	p	η2p	p	η2p	p	η2p
Tibialis anterior (mV·s)	0.249±0.11	0.253±0.12	0.290±0.15	0.312±0.18	0.223	0.034	0.064	0.076	0.168	0.043
Medial gastrocnemius (mV·s)	0.476±0.24	0.504±0.27	0.708±0.50	0.678±0.45	0.975	0.000	0.073	0.073	0.307	0.024
Rectus femoris (mV·s)	0.232±0.16	0.228±0.16	0.328±0.20	0.371±0.23	0.035*	0.095	0.077	0.068	0.036 *ab	0.094
Long head of the biceps femoris (mV·s)	0.225±0.16	0.219±0.15	0.267±0.20	0.246±0.17	0.490	0.011	0.421	0.015	0.663	0.005

^* p < 0.05 for group × task interaction. ^a significant difference between groups. ^b significant difference between tasks

Cognitive task accuracy

No significant between-group difference was observed in cognitive task accuracy during DT walking (control group: 91.50 ± 8.87%; MCI group: 86.91 ± 12.34%; t (43) = 1.45, p = 0.155).

Discussion

This study showed differences in gait performance, gait variability, cortical activation, and muscle activation between older adults with MCI and cognitively intact controls during ST and DT walking. DT walking significantly affected gait performance and variability, which was accompanied by distinct cortical and muscle activation patterns, with more pronounced effects in the MCI group. During ST walking, older adults with MCI exhibited overall lower cortical activation than controls, whereas gait differences were relatively limited. During DT walking, the MCI group showed greater gait impairments, including reduced gait speed, increased stride width, and altered gait variability. Cortical activation remained generally lower in the MCI group than in controls, consistent with the significant main effects of group. These findings suggest that DT walking imposes greater demands on gait control and neural regulation in older adults with MCI. This deduction partially supports Hypotheses 1 and 2.

Effects of DT walking on gait performance and cortical activation in older adults with MCI

The first and second hypotheses were partially supported by the results of this study. This study showed Group × Task interactions in cortical activation and gait performance under ST and DT walking conditions. In the control group, reduced activation in the prefrontal and sensorimotor-related cortical regions was observed during DT walking, while gait performance remained stable. This suggests that cognitively intact older adults were able to maintain relatively stable gait control despite increased cognitive demands. In contrast, older adults with MCI showed overall lower levels of cortical activation, and under DT condition, showed poorer gait performance, including reduced gait speed, increased stride width, and altered gait variability.

These findings suggest that older adults with MCI may have a reduced capacity to flexibly mobilize cognitive–motor cortical resources in response to increased task demands during DT walking. Although no significant differences in cortical activation were observed within the MCI group across task conditions, the additional cognitive demands introduced by DT walking may have further exacerbated the decline in gait control, given the generally lower cortical activation. This finding is consistent with previous studies indicating that when cognitive load increases or task difficulty rises, older adults with MCI exhibit lower activation in the prefrontal and sensorimotor cortices than cognitively intact older adults, accompanied by declines in motor performance [44, 45]. In this study, this interpretation is supported by pronounced gait impairments in the MCI group, despite preserved cognitive task accuracy and no corresponding increase in cortical activation. This phenomenon aligns with the capacity-sharing theory [46, 47], which suggests that limited cognitive resources under DT conditions may result in compromised motor performance. Previous studies have highlighted that increased competition between cognitive and motor resources under DT conditions can lead to altered cortical activation patterns and declines in gait performance [8].

Interestingly, we observed that older adults with MCI showed generally lower cortical activation across several regions and did not exhibit increased activation under the DT condition. This finding differs from a previous study reporting increased prefrontal activation during DT conditions in individuals with MCI, which has been interpreted as compensatory recruitment [46]. However, increased activation does not necessarily indicate more efficient neural processing, and greater cortical recruitment may instead reflect neural inefficiency [48]. This interpretation is consistent with the Compensation-Related Utilization of Neural Circuits Hypothesis (CRUNCH) [49, 50], which proposes that neural activation may plateau or decline once task demands exceed compensatory capacity. The reduced cortical activation observed during DT walking, together with the significant gait impairments in the MCI group, may therefore suggest that the cognitive-motor demands of the DT condition exceeded the available neural resources for gait control [51]. In addition, reduced activation was primarily observed in left-lateralized regions. Although aging has been associated with reduced hemispheric asymmetry, as described by the Hemispheric Asymmetry Reduction in Older Adults (HAROLD) model [52], the predominantly left-lateralized pattern observed here may reflect task-specific cortical involvement during DT walking rather than generalized bilateral compensation [53].

Effects of DT walking on prefrontal-motor cortical activity in older adults with MCI

The second hypothesis was partially supported by the results of this study. This study found significant Group × Task interaction effects in the lPFC, lM1, and lSMA, suggesting lower cortical activation and task-related modulation in the MCI group. This result suggests that coordinated engagement of multiple cortical networks plays an important role in gait control during DT walking. DT walking relies not only on motor execution but also on coordinated engagement of the prefrontal and sensorimotor cortical networks to integrate executive control, attentional allocation, and motor processing [7, 54]. It involves task-related reorganization across multiple regions, including motor and parietal cortices, which is closely associated with the maintenance of gait performance under DT condition [55]. Although some studies have primarily focused on PFC activation, accumulating evidence indicates that the premotor cortex and SMA also play important roles in regulating gait speed and postural control [55]. The relatively lower activation across multiple cortical regions observed in older adults with MCI may reflect altered coordination between prefrontal and sensorimotor cortical processes, which may limit effective gait regulation under complex task conditions.

The reduced activation across multiple cortical regions observed during DT walking may be influenced by structural and functional degeneration associated with MCI. Previous studies have reported that reductions in gray and white matter density are significantly associated with declines in gait performance, particularly slower gait speed [29, 56]. Even in cognitively intact middle-aged and older adults, DT-related gait deterioration is related to widespread white matter degradation [57]. These structural changes may reduce the efficiency of information transfer between the PFC and sensorimotor regions, which affects the integration of cognitive and motor resources during DT walking.

The significantly higher DT cost of gait speed in the MCI group indicates that gait performance was more susceptible to interference. Altered cortical activity in older adults with MCI may reflect differences in task prioritization during DT walking [58]. Previous studies have shown that, when cognitive and motor tasks are performed simultaneously and compete for limited resources, older adults and individuals with cognitive impairment may adjust their task priorities; often, they tend to prioritize the cognitive task under increased cognitive load at the expense of motor performance [59, 60]. When cognitive task performance remains relatively preserved under DT condition, individuals may allocate greater attentional resources to the cognitive task, while motor performance declines, as reflected by deteriorations in temporal gait parameters and dynamic balance [10, 61]. In this study, cognitive task accuracy did not differ significantly between groups during DT walking, despite pronounced gait impairments in the MCI group. The higher DT cost observed in MCI participants, along with the lack of a corresponding increase in cortical activation, may suggest a tendency toward prioritizing the cognitive task at the expense of gait performance. However, because cognitive performance during ST condition was not assessed, task prioritization strategies cannot be definitively determined.

Effects of DT walking on gait control and increased muscle activation in older adults with MCI

The first hypothesis was supported by the results of this study. Under the DT condition, older adults with MCI exhibited more pronounced impairments in gait regulation than cognitively intact older adults, characterized by reduced gait speed, increased stride width, and concurrent alterations in gait variability. According to established models of gait control, when the automatic execution of gait is compromised, such as due to normal aging, central nervous system pathology, or increased contextual cognitive demands, individuals may increasingly rely on cognitive control processes to regulate gait for maintaining functional performance [62, 63]. In this study, older adults with MCI showed a more pronounced reduction in gait speed during DT walking, which suggests greater cognitive-motor interference as task demands increase. When gait-related neural networks are tasked with supporting executive functions concurrently, the heightened cognitive load may disrupt gait regulation, which contributes to poorer performance [61, 64, 65]. These results suggest that MCI older adults are more vulnerable to cognitive load, impairing their gait regulation. An increase in stride width under the DT condition was observed in the MCI group. An increased stride width is commonly interpreted as a stability-oriented strategy, which reflects an expansion of the base of support in response to compromised gait control [66]. The increased stride length variability observed in older adults with MCI during DT walking suggests reduced consistency in gait regulation, which likely reflects impaired gait control adjustments under increased cognitive demands.

During DT walking, older adults with MCI exhibited a decline in gait automaticity accompanied by an increased reliance on cognitive resources. In this study, DT walking in older adults with MCI was characterized by slower gait speed, wider stride width, and direction-specific changes in gait variability, including reduced stride-to-stride variability and increased gait cycle variability. Previous studies have reported that the DT interference on gait and cognitive performance is attenuated during treadmill walking [41, 63]. This finding has been attributed to the rhythmic characteristics of the treadmill environment; in this environment, externally imposed, constant belt speed constrains spontaneous adjustments of gait parameters and enhances gait automaticity through rhythmic movement; these conditions reduce the attentional resources required for walking [41, 63]. In contrast, the overground walking paradigm adopted in this study lacks external rhythmic cues and therefore relies more heavily on endogenous neuromuscular control. As a result, this condition may be more sensitive in revealing the regulatory challenges associated with reduced gait automaticity under DT condition.

Declines in gait automaticity are typically accompanied by increased demands on muscular regulation [63]. The MCI group exhibited a significantly higher DT cost of gait speed, which indicates greater susceptibility of gait performance to cognitive interference under DT condition. In older adults with cognitive impairment, gait control under DT condition may rely more heavily on the modulation of cortical and muscular activation, which reflects a further decline in gait automaticity [41, 67]. Previous studies have shown that, even in cognitively intact older adults, aging markedly alters the neural mechanisms underlying gait control; and although secondary motor areas and the PFC are increasingly recruited to support gait, the efficiency of muscle activation may already be compromised [67, 68]. Under DT condition, this altered neural control may further influence neuromuscular regulation in older adults with MCI. In this study, RF activation was significantly higher in the MCI group during DT walking. As a biarticular knee extensor, the RF plays a critical role in facilitating knee joint control and limb propulsion during the swing phase [69]; thus, its increased activation may indicate greater demands on neuromuscular regulation under elevated task demands. However, given that both cortical activation and gait performance were reduced in the MCI group, the elevated RF activation more likely reflects reduced neuromuscular efficiency rather than an effective compensatory mechanism. These findings suggest that impaired cortical control during DT walking may contribute to altered neuromuscular regulation and reduced gait automaticity in older adults with MCI.

Clinical implications

These findings highlight the importance of maintaining physical activity and safe mobility, such as walking, to support functional independence and healthy aging in older adults with MCI. Older adults with MCI are more susceptible to gait regulation impairments during DT walking in this study, which suggests an elevated risk of falls when cognitive demands are imposed in daily life. Incorporating DT walking assessments into routine clinical functional evaluations is important. DT gait training may be considered part of intervention programs for older adults with MCI to enhance cortical activation and support gait stability. Task prioritization strategies, such as encouraging individuals to prioritize gait stability over concurrent cognitive tasks during walking, may further contribute to reducing fall risk in everyday activities. Clinically relevant cortical–muscular evidence to inform fall prevention and individualized rehabilitation planning is also needed in older adults with MCI.

Limitation

First, our fNIRS measurements were confined to cortical hemodynamic responses. Although short-separation channels can help further reduce superficial physiological influences, the fNIRS setup focused on captured task-related cortical activation. Future studies may benefit from applying multimodal neuroimaging approaches to provide a more comprehensive characterization of the neural mechanisms underlying DT walking. Second, the DT paradigm in this study was restricted to a single cognitive task (serial subtraction by 3); a broader range of DT conditions can be incorporated to better reflect the cognitive demands encountered during daily walking. Third, gait performance and muscle activation were assessed only in the dominant lower limb, and bilateral measurements should be considered to explore potential asymmetries and their relationship with hemispheric cortical activation. Finally, the cross-sectional nature of this study limits interpretation of neural and neuromuscular responses during DT walking. Future longitudinal studies may help clarify adaptive changes in neural and neuromuscular control strategies under increasing cognitive demands.

Conclusion

This study found that, under the DT condition, older adults with MCI showed reduced activation in multiple brain regions, poorer gait performance, and increased RF activation. These findings suggest that increased muscle activation despite reduced cortical activation may be insufficient to compensate for impaired gait performance under DT load. This pattern may indicate reduced gait automaticity. These results highlight the importance of incorporating DT gait performance into the clinical evaluation and functional assessment of older adults with cognitive impairment.

Abbreviations

DT

Dual-task

MCI

Mild cognitive impairment

ST

Single-task

fNIRS

Functional near-infrared spectroscopy

sEMG

Surface electromyography

MoCA

Montreal Cognitive Assessment

ROIs

Regions of interest

PFC

Prefrontal cortex

M1

Primary motor cortex

S1

Primary somatosensory cortex

lSMA

Supplementary motor area

TA

Tibialis anterior

MG

Medial gastrocnemius

RF

Rectus femoris

BF

Long head of the biceps femoris

HbO

Oxyhemoglobin concentration

HbR

Deoxyhemoglobin

ΔHbO₂

Changes in oxygenated hemoglobin

RMS

Root-mean-square

MVICs

Maximal voluntary isometric contractions

Con-ST

Cognitively intact older adults during single-task walking

Con-DT

Cognitively intact older adults during dual-task walking

MCI-ST

Older adults with mild cognitive impairment during single-task walking

MCI-DT

Older adults with mild cognitive impairment during dual-task walking

Authors’ contributions

XL, YZ and SX are responsible for participating in the conception, data collection, writing and submission of the manuscript. ZZ and DM are responsible for the data management and monitoring of this study. WS, and JW are responsible for the study design and revision of the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the Science and Technology Innovation Project of the General Administration of Sport of China (No. 24KJCX065; 25KJCX047), Shandong Provincial Sports and Fitness Technology Innovation Center (Open Solicitation Project under Jinan’s “20 Measures for New Higher Education Institutions” (No.SJSCXZX2025005) and the Graduate Student Scientific Research Innovation Program of Shandong Sport University (No. 250101).

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to privacy concerns related to human participants but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study protocol was approved by the Ethics Committee of Shandong Sport University (approval number: 2025045), and all participants signed the informed consent form before the experiment.

Consent for publication

Written informed consent for publication was obtained from the participant.

Competing interests

The authors declare no competing interests.