Unilateral and bilateral digital mirror visual feedback with different movement types modulate mu oscillations in individuals with stroke

Abstract

Objective

Mirror visual feedback (MVF) simulates motor execution of the affected hand and therefore potentially recruits motor areas in the affected hemisphere for individuals with stroke. However, whether and how digital form of MVF modulates oscillatory activities in individuals with stroke remain unclear. We investigated the neural mechanisms underlying digital MVF combined with different movement types in healthy controls and individuals with stroke using electroencephalography (EEG).

Methods

The participants performed a motor task with different combinations of MVF forms and movement types, including: (1) the bimanual training (BT) condition, (2) the unilateral MVF with unimanual training (UMUT) condition, (3) the unilateral MVF with bimanual training (UMBT) condition, and (4) the bilateral MVF with bimanual training (BMBT) condition. We investigated cortical excitability in motor regions by examining mu (9–14 Hz) power attenuation.

Results

Both the healthy and stroke groups showed pronounced mu power suppression in the central regions contralateral to unilateral MVF in the UMUT condition, indicating enhanced cortical excitability induced by digital MVF. Additionally, we found larger mu power attenuation in the affected hemisphere in both the UMUT and BMBT conditions than in the UMBT condition in the stroke group. Moreover, the individuals with stroke with better motor ability of the affected hand showed larger mu power attenuation in the affected hemisphere, and a higher degree of mu-power lateralization shifted toward the affected hemisphere in the BMBT condition.

Conclusions

Our electrophysiological findings provide neural evidence that digital MVF can facilitate cortical recruitment in the affected motor regions in individuals with stroke.

Keywords

stroke mirror visual feedback digital mirror therapy mu oscillations electroencephalography

Introduction

Impaired motor function in the hemiparetic upper extremity is a leading cause of long-term physical disability in individuals with stroke.¹ Various intervention strategies have been developed to improve motor function in individuals with stroke. Some treatments involve actual movement training to facilitate patients’ motor performance through overt execution (i.e., external production of motor output), such as bimanual training.^2,3 Other approaches aim to engage action representations to improve cortical reorganization in the damaged brain regions of patients through covert action simulation (i.e., internal rehearsal or mental manipulation of movements).^4,5 These include action observation,⁶ motor imagery,⁷ and mirror visual feedback (MVF) techniques.^8–10

MVF utilizes the reflected image of one hand’s movement to produce a visual illusion of movement of the other hand.^10,11 During traditional mirror therapy (MT), individuals with stroke are usually instructed to perform unimanual movements with the unaffected hand while observing the reflected image (i.e., MVF) in a mirror positioned along the individual’s midsagittal plane. MVF generated by the movement of the unaffected hand is perceived as the movement of the affected hand. This MVF therefore simulates motor execution of the affected hand and potentially recruits motor areas within the affected hemisphere in individuals with stroke.^9,12,13

Ample electrophysiological studies have investigated the neural mechanisms underlying motor control in healthy individuals using electroencephalography (EEG).^14,15 Previous research has demonstrated mu (9–14 Hz) power attenuation in the central electrodes during both motor execution^14,15 and observation of others’ actions,^16–20 indicating enhanced cortical excitability in motor regions¹⁵ during overt and covert movements, respectively. Recent research has also reported mu power decreases in the hemisphere ipsilateral to the moving hand when healthy participants observed the reflected hand in a mirror during unimanual movement.^21–23 These findings suggest that MVF activates motor regions contralateral to MVF, even without the presence of actual motor output from the hand simulated by MVF. In individuals with stroke, neurophysiological studies have found cortical engagement of motor regions in the affected hemisphere during covert motor simulation, as manifested by oscillatory activities during action observation²⁴ and during observation of MVF that simulated movements of the affected hand in traditional MT.²⁵ Taken together, accumulating evidence has revealed that MVF facilitates cortical recruitment in motor regions contralateral to MVF in both healthy individuals and individuals with stroke.

Recently, digital-based MT uses a camera to capture the movements of the unaffected hand for individuals with stroke and transforms it in real time into a mirrored image that is perceived as the affected hand to generate MVF on a screen.^8,26 The screen is placed along the coronal plane in front of the patients, offering several benefits over traditional MT. These advantages include enhanced immersive MVF and better perception of limb ownership during rehabilitation.²⁷ Notably, during traditional MT, individuals with stroke are required to observe MVF by turning their heads or trunks toward the mirror placed along the midsagittal plane (i.e., toward the affected side), making it challenging to concurrently observe the movement of the unaffected hand. In contrast, digital-based MT allows patients to observe both hands simultaneously and allocate more balanced visual attention across both hands through the monitor. Therefore, digital-based MT is able to provide two forms of MVF: unilateral and bilateral MVF. Unilateral MVF simulates the movement of the affected hand only. Bilateral MVF presents the movements of both hands: MVF on the affected side simulates the movement of the affected hand, symmetrical and synchronous to the unaffected hand, while MVF on the unaffected side displays movements identical to the unaffected hand captured by the camera.

Considering the potential benefits of digital-based MT, it may serve as a viable alternative to traditional MT. However, whether and how digital MVF modulates neural activities in individuals with stroke remain unclear. In this study, we investigated the neural mechanisms underlying digital MVF combined with different movement types for healthy controls and individuals with stroke using EEG. The participants performed a motor task with different combinations of MVF forms and movement types, including: (1) the bimanual training (BT) condition, (2) the unilateral MVF with unimanual training (UMUT) condition, (3) the unilateral MVF with bimanual training (UMBT) condition, and (4) the bilateral MVF with bimanual training (BMBT) condition. This experimental design provided a novel approach to investigate the underlying neural mechanisms of overt motor execution and covert motor simulation through manipulating movement types and MVF forms, respectively.

Materials and methods

Participants

The demographic data of the participants are listed in Table 1. In the stroke group, 14 participants were recruited from 2 physical medicine and rehabilitation departments in Taiwan and completed the EEG experiment. Three participants were excluded due to excessive noise in the EEG signals, and the EEG data of the remaining 11 participants were included in the final analysis. The inclusion criteria for the stroke group were as follows: (1) diagnosis of unilateral stroke; (2) stroke onset occurring more than 6 months before participation; (3) age between 20 and 80 years; (4) a baseline Fugl-Meyer Assessment of the Upper Extremity (FMA-UE) score greater than 40; and (5) the ability to follow the study instructions. The exclusion criteria were as follows: (1) diagnosis of global or receptive aphasia; (2) severe neglect, defined as a mean deviation of ≥ 1.5 inches on the 8-inch horizontal line in the line bisection subtest of the Behavioral Inattention Test; or (3) major medical problems or comorbidities that could interfere with upper limb use or cause severe pain. The screening and the assessments were completed in the hospitals.

Table 1.

Demographic data of participants in the healthy and the stroke groups.

	Healthy group (n = 16)	Stroke group (n = 11)	p-value
Age (years old)	47.50 ± 6.18	54.43 ± 11.76	.10
Gender (Male/Female)	10/6	10/1	.10
Duration from stroke onset (Months)	--	18.78 ± 16.33
MMSE-2: SV	--	27.36 ± 1.91
Hemiplegic side (right/left)	--	6/5
Brunnstrom stage: UE proximal	--	4.73 ± 0.52
Brunnstrom stage: UE distal	--	4.68 ± 0.51
FMA-UE	--	58.82 ± 4.51
MIQ-RS	90.13 ± 7.40	85.55 ± 13.24	.32

In the healthy group, 20 right-handed subjects completed the EEG experiment. Four participants were excluded due to excessive noise in the EEG signals, and EEG data of 16 participants were included in the final analysis. The inclusion criteria of healthy controls were as follows: (1) age over 18 years; (2) no history of neurologic or psychiatric diagnosis; and (3) no upper-limb orthopedic diagnosis or conditions within the past 6 months. Eligible healthy adults were excluded if they had major medical problems or comorbidities that could cause severe upper-limb pain or interfere with brain neural activity.

Procedures

To unify the terminology, the hand captured by the camera to create the mirrored image is defined as the leading hand, corresponding to the unaffected hand in individuals with stroke. The other hand is defined as the following hand (i.e., the hand that followed the movement of the reflected hand image on the monitor), corresponding to the affected hand in individuals with stroke. Healthy controls were instructed to use their dominant hand (i.e., right hand) as the leading hand. The electrode sites contralateral to the leading hand and those contralateral to the following hand corresponded to the unaffected and the affected hemispheres in the stroke group, respectively.

The EEG experiment was conducted at Chang Gung University in Taiwan, from April 2022 to March 2024. Before the EEG experiment, participants completed the assessment to evaluate their motor-related function in the following upper extremity (i.e., the left extremity in the healthy group and the affected extremity in the stroke group). Motor imagery ability was evaluated with the Motor Imagery Questionnaire-Revised, Second Edition (MIQ-RS)^28,29 in both the control and stroke groups, and upper extremity motor function was evaluated using the Fugl-Meyer Assessment of the Upper Extremity (FMA-UE)³⁰ in the stroke group.

The digital system consisted of a laptop computer, a 24-inch LCD monitor, an adjustable monitor arm, and a USB webcam (Logitech C270, Logitech). The 24-inch monitor was mounted on an adjustable arm and positioned in front of the participant. The viewing distance between the participants and the monitor was about 40 to 50 cm. The webcam was used to capture upper-limb movements of the participants, and it supports a maximum resolution of 1280 × 720 pixels and a frame rate of 30 frames per second. The webcam has a fixed-focus lens with a 55-degree field of view and automated light correction. Each research session took place in a laboratory with consistent ambient lighting. The system software was self-developed by using HTML5, CSS, and JavaScript to allow flexible deployment in clinical settings. Real-time webcam streaming and recording were implemented via the RecordRTC library (version 5.5.3). The real-time image of the leading hand was horizontally flipped using a CSS transform and displayed on the screen to create a mirrored image that simulates the following hand. No additional image filtering or enhancement was applied. At the start of every session, a quick calibration process was carried out to match the participant’s limb position with the visual display. With browser-based processing, the system was operated in real time without intentional delays. User interaction data were stored locally in the browser’s Local Storage and could be exported in CSV or text format for research documentation.

During the EEG experiment, participants were comfortably seated in front of the digital rehabilitation system and performed an upper extremity motor task (Figure 1(a)). The monitor was placed in front of the participants and the camera was positioned above the leading hand. The following hand of the participants was positioned behind the monitor. The camera captured the movement of the leading hand and transformed it into a mirrored image. This reflected image, displayed on the screen, was intended to be perceived as the movement of the following hand.

Figure 1.

Schematic illustration of task design. (a) Upper: In the conditions with unilateral MVF, a camera above the leading hand (i.e., the unaffected hand for individuals with stroke) captured its movement and displayed a mirrored image on the monitor, which was perceived as the movement of the following hand (i.e., the affected hand for individuals with stroke). Lower: The consistent topographic representation across all the participants: the right side on the topography corresponds to the contralateral hemisphere (CH) of the following hand (i.e., the affected hand in the stroke group), while the left side corresponds to the CH of the leading hand (i.e., the unaffected hand in the stroke group). (b) Participants performed a motor task with four experimental conditions: UMUT, UMBT, BMBT and BT. White, grey, and orange hands represent active movement hand, resting hand, and MVF display, respectively.

The experimental conditions are illustrated in Figure 1(b). Participants performed a motor task with varying types of MVF (unilateral on the following side vs. bilateral) and manual movement (unimanual on the leading side vs. bimanual). The experimental design followed a 2 (group: healthy controls, individuals with stroke) × 4 (motor pattern: UMUT, UMBT, BMBT and BT) mixed-design, with group as the between-subject factor and motor pattern as the within-subject factor.

Each trial began with an auditory cue. Participants were instructed to perform wrist flexion with a fisted hand and return to a neutral wrist position within 2 seconds after hearing the cue. The interval between auditory cues was 6 seconds with 1 second jitters. Prior to the formal experiment, participants were trained with a metronome to ensure the consistency of performing each movement within the 2-second timeframe. Short breaks of 1 to 2 minutes were provided every 10 trials. There were 50 trials in each experimental condition.

The four conditions with different motor patterns were counterbalanced across the participants in each group. In the BT condition, the participants were instructed to perform wrist flexion with both hands, with no MVF on the monitor screen. In the UMUT condition, the participants performed wrist flexion with the leading hand and concurrently observed the unilateral MVF simulating the movement of the following hand. In the UMBT condition, the participants performed wrist flexion with both hands; meanwhile, unilateral MVF on the following side served to guide motor execution of the following hand. Lastly, in the BMBT condition, the participants performed wrist flexion with both hands and simultaneously observed bilateral MVF on both the leading and following sides.

EEG acquisition and recording parameters

The EEG data were recorded continuously using a Brain Products amplifier with a 32electrode elastic cap (Ag/AgCl) arranged according to the 10–20 international system. The montage included six midline sites (Fz, FCz, Cz, CPz, Pz, and Oz) and twelve sites over each hemisphere (FP1/2, F3/4, F7/8, FC3/4, FT7/8, C3/4, T7/8, CP3/4, TP7/8, P3/4, P7/8, and O1/2). Electrooculography (EOG) was used to record vertical and horizontal eye movements. Ground and reference electrodes were positioned at FPz and the mastoid sites (A1 and A2), respectively. Surface electromyogram (EMG) electrodes were attached to the flexor carpi radialis muscles of the left and right forearms for each participant to detect the movement onsets. Electrode impedances were kept below 5 KΩ throughout the recording. Ongoing EEG signals at each electrode site were sampled at a rate of 1000 Hz. The activity was filtered with a low-pass filter of 100 Hz, and no high-pass filter was employed.

EEG preprocessing

Data processing and analyses were performed using Brain Vision Analyzer and the Fieldtrip toolbox³¹ in MATLAB (MathWorks), supplemented by in-house MATLAB scripts. Offline, the continuous EEG signals were high-pass filtered at 0.1 Hz and re-referenced to the algebraic average of the left and right mastoids (A1 and A2). The EMG signals were high-pass filtered at 10 Hz. A trigger was sent during the recording when the participant flexed their wrist to touch a button with their hand. EMG signals from 2500 to 500 ms before this trigger in each trial served as the baseline period and were used to calculate the mean and the standard deviation (SD) of baseline activities. EMG data from 500 to 0 ms before the trigger were used to detect the EMG onset. We determined the EMG movement onset of the leading hand when the signals exceeded the threshold (mean ± 4 SD).

We epoched the EEG signals based on the onsets of movement of the leading hand according to EMG signals. The continuous data were segmented into epochs starting 2000 ms before and ending 3000 ms after the EMG onset. Artefact rejection was then performed on the epoched data. In the first step, epochs with sudden bursts or drastic drifts were examined by visual inspection and excluded. In the second step, independent component analysis (ICA) was conducted to extract and remove artifacts including eye blinks, eye movements and muscle activity. The data were then baseline corrected using the 100 ms period before the EMG onset. The number of components removed during ICA was 3.56 ± 1.17 and 3.91 ± 1.38 for the healthy and stroke groups, respectively. The average number of retained clean trials (out of 50 experimental trials) after artifact rejection for the UMUT, UMBT, BMBT, and BT conditions was 40.30 ± 8.30, 40.30 ± 9.92, 41.42 ± 10.03, and 42.04 ± 11.06 trials for the healthy group, and 41.00 ± 7.72, 41.27 ± 9.70, 41.55 ± 8.90, and 42.36 ± 7.05 trials for the stroke group.

EEG time–frequency analysis

The epochs were then entered into a time–frequency decomposition using Morlet wavelet transformation with a length of seven cycles. Time–frequency representations of power were estimated for each trial, electrode, and participant across frequencies from 4 to 20 Hz in 1-Hz steps. For each participant, the resulting time–frequency power was averaged across trials according to each condition. The power estimates were rescaled to decibels (dBs), with a baseline from -800 ms to -500 ms relative to the EMG onset. For the stroke group, we flipped the channels along the midline for the participants with the affected hemisphere on the left side (i.e., individuals with right hemiplegia). This procedure ensured consistent topography across participants, with the right and the left sides of the EEG topography corresponding to the contralateral hemisphere (CH) of the following hand and that of the leading hand, respectively. Namely, the right and left sides of the EEG topography corresponded to the affected and unaffected hemispheres of individuals with stroke, respectively (Figure 1(a)).

First, statistical analyses were performed on power estimates using analysis of variance (ANOVA). Mu power was averaged across the interval from 0 to 1100 ms relative to the movement onset. To investigate the oscillatory activities in motor regions, we examined mu power (9–14 Hz) at the central electrodes (C3 and C4) contralateral to the leading hand (i.e., the unaffected hemisphere in individuals with stroke) and contralateral to the following hand (i.e., the affected hemisphere in individuals with stroke). The following methods were used for the hypothesis testing. (1) A 2 (group: healthy, stroke) × 2 (hemisphere: CH of leading hand, CH of the following hand) × 4 (motor patterns: UMUT, UMBT, BMBT and BT) mixed ANOVA was conducted for the omnibus testing. (2) To investigate the effects specific to MVF and manual movement, planned comparisons by dependent t-test were performed to compare two specific conditions within each group. Specially, we compared the UMBT versus BT conditions to test the effect of adding unilateral MVF to bimanual training, and the BMBT versus BT conditions to test the effect of adding bilateral MVF to bimanual training. We also compared the UMBT versus UMUT conditions to test the effect of bimanual training compared to unimanual training when unilateral MVF was provided. Lastly, we compared the BMBT versus UMBT conditions to test the effect of bilateral MVF compared to unilateral MVF during bimanual movement.

Considering the temporal dynamics of oscillatory activities, we also conducted cluster-based nonparametric permutation tests.³² This analysis controls for type I errors without making prior assumptions by identifying clusters showing significant differences at consecutive time points based on t-tests at each time-point sample. We conducted two analyses. (1) We first used one-sample t-tests to investigate whether mu power significantly smaller than zero (p < .05, one-tailed) for each hemisphere in each condition for each group, using the averaged mu power (9–14 Hz) in the CH of the leading hand and the CH of the following hand separately. (2) We used dependent samples t-tests (p < .05, two-tailed) to investigate time-resolved difference in mu power between various motor patterns, using the averaged mu power (9–14 Hz) in motor regions contralateral to the following hand (i.e., the affected hemisphere in individuals with stroke). Specifically, we compared the UMBT versus BT conditions, BMBT versus BT conditions, UMBT versus UMUT conditions, and BMBT versus UMBT conditions for each group separately. Samples with p-values below .05 were selected and grouped into clusters based on temporal adjacency. For each cluster, a cluster-level t-statistic was computed by summing the t-values within the cluster. Monte Carlo p-values were estimated using 10,000 permutations, in which condition labels were randomly shuffled to generate a null distribution of cluster-level statistics under the null hypothesis. The observed cluster-level t-statistic was compared with the null distribution to obtain a corrected p-value. A cluster was considered significant if the corrected p-value was smaller than .05, indicating that the observed effect was unlikely to occur by chance. Corrected p-values were reported, and effect sizes were calculated using Cohen’s d.³³

Lateralization index of mu power

To investigate lateralization in motor areas (C3/4 electrodes) across the two hemispheres, we calculated the lateralization index by subtracting mu power in the central electrode in the CH of the leading hand from that of the CH of the following hand (i.e., subtracting mu power of the unaffected hemisphere from that of the affected hemisphere in individuals with stroke). A negative value of the lateralization index indicated larger mu power attenuation for the CH of the following hand compared to the CH of the leading hand, thus reflecting dominance of the CH of the following hand over the CH of the leading hand (i.e., dominance of the affected over the unaffected hemisphere in individuals with stroke).

We first conducted a cluster-based permutation test (p < .05, two-tailed) to examine whether there was significant lateralization in each motor pattern condition. To this end, we compared mu power (9–14 Hz) between the CH of the following hand and that of the leading hand during the interval between 0 and 1100 ms after the movement onset. Additionally, we investigated whether lateralization differed across the entire 1100-ms interval among the various motor patterns within each group. We averaged lateralization index across 0–1100 ms and conducted dependent t-tests comparing the UMBT versus BT conditions, the BMBT versus BT conditions, the UMBT versus UMUT conditions, and the BMBT versus UMBT conditions within each group.

Correlation analyses

We investigated the relationship between covert simulation and overt movement. First, we examined whether motor imagery ability (i.e., MIQ-RS scores) for the affected hand in individuals with stroke was positively associated with their corresponding motor capability (i.e., FMA-UE scores) by computing correlation coefficient between the two scores (p < .05, onetailed). We also examined the correlations in neural activities between covert simulation and overt motor execution. To this end, we computed correlation coefficient (p < .05, one-tailed) in mu power averaged across 0–1100 ms after the movement onset in the CH of the following hand between the UMUT and BT conditions for both groups.

Moreover, we investigated the relationship between the neural indices and motor capability of upper extremity (MIQ-RS for both groups, and FMA-UE for the stroke group). We examined the correlation between mu power at the central electrode (C3/4) in the CH of the following hand for the 1100-ms interval after movement onset in each motor pattern and the assessment score by cluster-based permutation tests. We also examined the correlations between the lateralization index in each motor pattern and each assessment score. Given that each correlation analysis between an assessment score and a neural index was conducted for four motor patterns (i.e., UMUT, UMBT, BMBT, and BT), the family-wise error rate was controlled by dividing .05 by 4, yielding a significance threshold of .0125 at each sample point (i.e., p < .0125, one-tailed). Clusters were considered significant if the corrected p-value was smaller than .0125 (one-tailed), and the corrected p-values were reported.

Pearson’s correlation coefficient (r) was computed for the healthy group. Given the relatively small sample size in the stroke group, correlations were assessed using Spearman’s rho (ρ), which is more robust to non-normality and outliers and does not assume a specific data distribution. One-tailed tests were applied to assess statistical significance, because we had a priori hypotheses regarding the expected direction of the effects (i.e., the individuals with larger mu power suppression and more negative lateralization index would demonstrate better motor ability).

Statistical analysis

The details of the statistical tests used in each analysis are described in the corresponding sections above. In summary, for the mu power averaged across 0–1100 ms after EMG onset, we used a 2 (group) × 2 (hemisphere) × 4 (motor pattern) mixed-design ANOVA for the omnibus analysis, followed by paired-sample t-tests to compare mu power between specific conditions within each group. Considering the temporal dynamics of oscillatory activity, we also conducted cluster-based nonparametric permutation tests. These included one-sample t-tests to determine whether mu power significantly differed from zero, and paired-sample t-tests to examine time-resolved differences in mu power between different motor patterns in the CH of the following hand (i.e., the affected hemisphere in individuals with stroke). For the lateralization index, we conducted cluster-based permutation tests to examine whether significant lateralization was present in each motor pattern condition, and paired-sample t-tests to compare the lateralization index averaged across 0–1100 ms between conditions. For the correlation analysis, we computed Pearson’s correlation coefficients for the healthy group and Spearman’s rho for the stroke group.

Results

Time–frequency results

The time–frequency results for the CH of the leading hand and the CH of the following hand for both groups are illustrated in Figure 2(a) and (b). The three-way repeated-measures ANOVA for mu power across the entire 0–1100 ms interval showed a significant main effect of condition [F (2.08, 75) = 6.72, p = .002, η_p² = .21]. No other significant effects were found.

Figure 2.

Time frequency results of mu power in the two groups. (a) Time–frequency results of mu power in the healthy (left) and the stroke (right) groups for four motor patterns. 0 ms represents the EMG onset of the leading hand. The topography represents mu power (averaged across 9–14 Hz from 0 to 1100 ms). The temporal window for the significant effect is indicated by the grey squares. Bottom: Time course of mu power trajectory (averaged across 9–14 Hz) for each motor pattern. Shaded areas represent the standard error of the mean (SEM). (b) Mu power (averaged across 9–14 Hz from 0 to 1100 ms) for the healthy (left) and the stroke (right) groups. Error bars represent SEM.

Considering that no effect of group or hemisphere was found, we conducted planned comparisons to test the differences between two conditions of interest in the CH of the following hand (i.e., the affected hemisphere in the stroke group) for each group. For the healthy controls, the paired t-test showed a significant difference between the UMBT and UMUT conditions [t(15) = 2.43, p = .028, d = 0.61], indicating a smaller decrease in mu power in the UMBT (-1.24 ± 1.10 μV) than in the UMUT (-1.46 ± 1.60 μV) condition (Figure 2(b)). This result indicated reduced cortical engagement of motor regions contralateral to the following hand for bimanual compared to unimanual movement when unilateral MVF was provided. No other effects were found.

The cluster-based permutation test examining the dependent t-test showed mu power significantly smaller than 0 for both groups (Figure 2(a)). For the healthy controls, significant mu power attenuation was found in the UMUT condition (the CH of the leading hand: from 215 to 1100 ms relative to EMG onset (corrected p < .001, d = 0.97), the CH of the following hand: from 128 to 1100 ms (corrected p < .001, d = 1.14)), UMBT condition (the CH of the leading hand: from 248 to 1100 ms (corrected p < .001, d = 1.28); the CH of the following hand: from 272 to1100 ms (corrected p < .001, d = 1.13)), BMBT condition (the CH of the leading hand: from 406 to 1100 ms (corrected p = .002, d = 0.72); the CH of the following hand: from 349 to 1100 ms (corrected p = .003, d = 0.77)) and BT condition (the CH of the leading hand: from 921 to1086 ms (corrected p = .008, d = 0.51); the CH of the following hand: from 604 to 1100 ms (corrected p = .004, d = 0.66)). For the individuals with stroke, significant mu power attenuation was found in the UMUT condition (the CH of the leading hand: from 482 to 1100 ms (corrected p = .005, d = 0.91); the CH of the following hand: from 443 to 1100 ms (corrected p = .003, d = 0.87)), UMBT condition (the CH of the leading hand: from 3 to159 ms (corrected p = .005, d = 0.94), from 642 to 719 ms (corrected p = .037, d = 0.59) and from 1007 to 1049 ms (corrected p = .049, d = 0.55)), and BMBT condition (the CH of the leading hand: from 110 to 1100 ms (corrected p < .001, d = 1.02); the CH of the following hand: from 197 to 1100 ms (corrected p = .003, d = 1.11)). No other significant effects were found.

The results of the cluster-based permutation between motor patterns in the central electrodes (C3/4) in the CH of the following hand are illustrated in Figure 3. For the healthy controls (Figure 3(a)), we found larger decreases in mu power in the UMBT than in the BT condition (from 273 to 372 ms (corrected p = .019, d = 0.57) and from 564 to 729 ms (corrected p = .011, d = 0.63) relative to EMG onset), indicating enhanced cortical engagement of motor regions when unilateral MVF was provided during bimanual movement. We also found a larger decrease in mu power in the BMBT condition than in the BT condition (from 381 to 396 ms (corrected p = .021, d = 0.53)), indicating enhanced cortical engagement of motor regions when bilateral MVF was provided during bimanual movement. Finally, we found smaller decreases in mu power in the UMBT condition than in the UMUT condition (from 348 to 465 ms (corrected p = .009, d = 0.62) and from and 753 to 1016 ms (corrected p = .002, d = 0.84)), indicating reduced cortical engagement of motor regions during bimanual compared to unimanual movement when unilateral MVF was provided.

Figure 3.

The differences in mu power in the CH of the following hand between conditions. (a) For the healthy controls, we found larger mu power attenuation after movement onset in the UMBT versus BT and the BMBT versus BT conditions, while smaller mu power attenuation was found in the UMBT than in the UMUT condition. (b) For the individuals with stroke, we found larger mu power attenuation for the BMBT versus BT and the BMBT versus UMBT conditions, while smaller mu power attenuation was found in the UMBT than in the UMUT condition. The topographic map shows the cluster t-statistics in the significant time window. The temporal window for the significant effect is indicated by the grey squares.

For the individuals with stroke (Figure 3(b)), we found larger decreases in mu power in the BMBT than in the BT condition (from 243 to 369 ms (corrected p = .021, d = 0.73) and from 982 to 1100 ms (corrected p = .004, d = 0.95)), indicating enhanced cortical engagement when bilateral MVF was provided during bimanual movement. We found a smaller decrease in mu power in the UMBT than in the UMUT condition (from 850 to 951 ms (corrected p = .019, d = 0.76)), reflecting reduced cortical engagement during bimanual compared to unimanual movement on the unaffected side when unilateral MVF was provided to the affected side. Finally, we found larger decreases in mu power in the BMBT than in the UMBT condition (from 267 to 389 ms (corrected p = .011, d = 0.84) and from 846 to 991 ms (corrected p = .004, d = 0.86)), indicating enhanced cortical engagement during the bilateral compared to the unilateral MVF on the affected side during bimanual movement.

Lateralization index

The results of permutation tests are illustrated in Figure 4(a) and (b). For the control group, the permutation test demonstrated a larger decrease in mu power in the CH of the following hand than in the CH of the leading hand in the UMUT condition (from 921 to 1100 ms (corrected p = .001, d = 0.81)), as reflected by negative lateralization indices during this interval. In contrast, we found smaller decreases in mu power in the CH of the following hand than in the CH of the leading hand in the UMBT condition (from 383 to 553 ms (corrected p = .013, d = 0.60) and from 673 to 749 ms (corrected p = .020, d = 0.57)), resulting in positive lateralization indices during the intervals. No other significant effects were found.

Figure 4.

The results of the lateralization index of mu power. Lateralization index was calculated as mu power in the CH of the following hand minus mu power in the CH of the leading hand (i.e., contralateral (co) versus ipsilateral (ip) hemisphere to the following hand). (a) We found negative lateralization indices in the UMUT condition for the healthy controls. Conversely, we found positive lateralization indices in the UMBT condition for both groups. The temporal window for the significant effect is indicated by the grey squares. The topographic map shows the cluster t-statistics in the significant time window. (b) Time course of trajectory for lateralization index (averaged across 9–14 Hz) in each condition. Shaded areas represent the standard error of the mean (SEM). Colored horizontal lines indicate the time window where mu power significantly differed between the two hemispheres. (c) Lateralization index (averaged 9–14 Hz from 0 to 1100 ms) in each condition for the healthy and the stroke groups. We found a more positive lateralization index for the UMBT than the UMUT and BMBT conditions for the healthy controls. Error bars represent SEM. Asterisks indicate statistical significance (*p < .05).

For the stroke group, the permutation test demonstrated smaller decreases in mu power in the affected hemisphere compared to the unaffected hemisphere in the UMBT condition (from 8 to 78 ms (corrected p = .018, d = 0.76) and from 201 to 317 ms (corrected p = .012, d = 0.84)), resulting in positive lateralization indices during these intervals. No other significant effect was found.

Finally, we investigated whether the lateralization in motor regions during the entire 1100-ms interval differed between various motor patterns for each group (Figure 4(c)). The dependent sample t-tests showed a more positive lateralization index for the UMBT condition than for the UMUT [t(15) = 2.52, p = .024, d = 0.63] and BMBT [t(15) = 2.57, p = .021, d = 0.64] conditions in the healthy controls.

Correlation analyses

We found a significant positive correlation (ρ = 0.90, p < .001) between MIQ-RS and FMAUE scores (Figure 5(a)) in the individuals with stroke. We also found a positive correlation in mu power in the CH of the following hand between the UMUT and BT conditions (r(14) = 0.65, p = .003) in the healthy controls (Figure 5(b)).

Figure 5.

The results of correlation analyses. (a) We found a positive correlation between MIQ-RS and FMA-UE scores for the affected hand in the stroke group. (b) We found a positive correlation in mu power in the CH of the following hand between the UMUT and BT conditions in the control group. (c) The individuals with stroke demonstrated negative correlations between mu power (averaged 9–14 Hz from 0 to 340 ms) in the affected hemisphere and MIQ-RS scores (top), as well as negative correlations between the lateralization index of mu power (i.e., contralateral (co) versus ipsilateral (ip) hemisphere of the affected hand) (averaged 9–14 Hz from 20 to 300 ms) and FMA-UE scores (bottom), in the BMBT condition. The temporal window showing the significant effect is indicated by the grey squares. The topographic map shows the cluster t-statistics for correlation during the significant interval. Asterisks indicate statistical significance (*p < .05, ** p < .01, *** p < .001).

For the correlation between the assessment scores (MIQ-RS, FMA-UE) and the neural indices (mu power in the CH of the following hand, the lateralization index), significant negative correlations between mu power in the affected hemisphere and MIQ-RS scores (from 0 to 340 ms, corrected p = .008) were found in the BMBT condition for the stroke group (Figure 5(c), top). Additionally, negative correlations between the lateralization index and FMA-UE scores (from 20 to 300 ms, corrected p = .012) were found in the BMBT condition for the stroke group (Figure 5(c), bottom). No other significant correlations were found.

Discussion

The current study aimed to investigate neural mechanisms underlying digital MVF combined with movements in heathy controls and individuals with stroke. Digital MVF provides bilateral MVF that cannot be achieved via traditional MT, thereby providing a novel approach for the manipulation of MVF forms (i.e., unilateral vs. bilateral). We recorded EEG while participants performed a motor task with different motor patterns.

Both the healthy and the stroke groups showed pronounced mu power suppression contralateral to the unilateral MVF when the hand simulated by MVF (i.e., the following hand) was at rest (i.e., the UMUT condition). The novel finding was that individuals with stroke benefited from bilateral (i.e., the BMBT condition) compared to unilateral MVF (i.e., the UMBT condition) during bimanual movement, whereas healthy controls did not. This was manifested by a larger decrease in mu power in the CH of the following hand in the BMBT condition than in the UMBT condition only in the stroke group. These findings suggest that individuals with stroke may require visual feedback of the leading hand (i.e., the unaffected hand) during bimanual movement. Moreover, the individuals with stroke with better motor function of the affected hand showed larger mu power attenuation in the affected hemisphere, and a higher degree of mu-power lateralization shifted toward the affected hemisphere in the BMBT condition. The correlations between behavioral assessments and neural indices demonstrated that mu power suppression served as a neural signature of motor capability. Overall, our findings showed how digital MVF effectively induced cortical engagement in the affected motor regions in individuals with stroke.

Covert motor simulation induces mu power attenuation

Previous research found mu power suppression in the motor regions contralateral to unilateral MVF displayed in a mirror when the hand simulated by MVF was at rest, accompanied by mu power suppression in the CH of the moving hand, in healthy participants²¹. These findings suggested that both observing MVF and performing motor execution induced cortical engagement, as reflected by mu power suppression.^21,22 Consistent with the prior findings, our results of the UMUT condition demonstrated mu power attenuation in bilateral central regions in both the control and stroke groups. These results indicated that covert motor simulation evoked by unilateral MVF and overt motor execution facilitated cortical excitability in motor areas in the CH of the following hand and in the CH of the leading hand, respectively. The results also demonstrated that unilateral MVF provided by digital-based device activated action representations in motor regions similar to traditional MT,^25,34,35 and that the neural recruitment through covert simulation was effective in both healthy controls and individuals with stroke.

Bilateral MVF during bimanual movement facilitates the engagement of the affected motor areas in individuals with stroke

When investigating the modulation of oscillatory activities by different types of MVF and manual movement, we focused on the motor regions in the CH of the following hand (i.e., the affected hemisphere in the stroke group). We found a larger decrease in mu power when unilateral MVF was introduced during bimanual movement (i.e., the UMBT condition) compared to when unilateral MVF was absent (i.e., the BT condition) in the healthy controls, but not in the individuals with stroke. The findings demonstrated that unilateral MVF facilitated the recruitment of motor regions during bimanual movement only in the control group. Additionally, although both groups showed larger mu power suppression when bilateral MVF was provided during bimanual movement (i.e., the BMBT condition) compared to when MVF was absent (i.e., the BT condition), only the stroke group demonstrated larger mu power suppression during bimanual movement when bilateral MVF was provided (i.e., the BMBT condition) compared to when unilateral MVF was provided (i.e., the UMBT condition). Taken together, these results suggested that the individuals with stroke benefited from visual feedback of the unaffected hand during bimanual movement, while the healthy controls demonstrated similar degrees of cortical excitability in the CH of the following hand irrespective of the MVF form (i.e., UMBT or BMBT) during bimanual movement.

Previous studies have demonstrated the benefits of bimanual movements,^36,37 especially symmetrical and synchronous movements of the upper extremities, compared to unimanual movements in motor rehabilitation owing to the following reasons. First, mirrored and symmetrical movements represent a typical coordination mode within the repertoire of action representations.^36,38 Bimanual synchronous movements activate homologous muscle groups on both sides of the body in a coordinated manner by concurrently recruiting similar cortical networks in both hemispheres.³⁹ Second, in individuals with stroke, the phenomenon of ‘learned nonuse’ of the affected hand may lead to hyperexcitability of the unaffected hemisphere, thereby enhancing interhemispheric inhibition from the unaffected hemisphere to the affected hemisphere.^36,40 The symmetrical bimanual task may help normalize the excessive interhemispheric inhibition and facilitate appropriate motor output from the affected hemisphere in individuals with stroke.

Previous findings have shown that MVF may facilitate motor recovery in individuals with stroke, since covert action simulation may recruit the shared neural substrates and action representations with overt execution of the corresponding movements.^4,41–43 These neural substrates include the mirror neuron system (MNS), primarily encompassing regions in the inferior frontal gyrus and the inferior parietal lobule.^41,42,44 Through providing action observation with correct motor patterns of the affected hand, MVF potentially activates the MNS in the affected hemisphere of individuals with stroke^12,25,35,45

Taken together, considering the beneficial effects of bimanual movements and MVF, the BMBT condition in our experiment may facilitate the cortical engagement of the MNS in the affected hemisphere by combining overt execution of symmetrical bimanual movements and covert observation of ‘corrected’ visual feedback of the affected hand.

Bimanual movement reduces the cortical engagement of motor regions when only unilateral MVF was provided

In both the control and stroke groups, we found smaller mu power attenuation when unilateral MVF was provided during bimanual movement (i.e., the UMBT condition) than when it was provided during unimanual movement (i.e., the UMUT condition). The results may stem from a conflict between perceptual (i.e., visual feedback from unilateral MVF) and kinesthetic feedback (i.e., sensorimotor and proprioceptive feedback of bimanual motor execution) in the UMBT condition. In the UMUT condition, participants covertly simulated the movement of the following hand through action observation. In contrast, in the UMBT condition, participants imitated the movement simulated by MVF with actual physical output of the following hand. The mismatch between visual feedback and actual movement may therefore reduce cortical excitability in motor regions^9,46,47 in the UMBT condition.

Lateralization index reflects dominance of hemisphere

The healthy controls demonstrated negative lateralization indices of mu power in the UMUT condition. This finding suggested that the cortical engagement of motor regions in the CH of the following hand induced by covert motor simulation was stronger than that in the CH of the leading hand activated by overt motor execution in the control group.

Both groups demonstrated positive lateralization indices in the UMBT condition, suggesting that cortical engagement was dominated by the CH of the leading hand. These findings further supported that unilateral MVF during bimanual movement (i.e., the UMBT condition) reduced the cortical engagement in the CH of the following hand in both the control and stroke groups. Previous neuroimaging research has shown that mismatch between visual input and motor output affects event-related fields⁴⁶ and neural activity⁴⁷ in healthy subjects. Our results complement the previous findings by showing that inconsistency between visual feedback and actual movements modulates oscillatory activities.

The relationship between covert motor simulation and overt motor execution

In the stroke group, we found that their motor imagery ability, as assessed by the MIQ-RS, was positively correlated with their motor ability of the upper extremity evaluated by the FMA-UE. Previous studies have demonstrated a correlation between the capability of covert simulation and overt motor function in individuals with stroke.^48,49 In accordance with the prior findings, our results revealed that individuals with stroke with better motor ability showed better motor imagery performance.

In the healthy controls, we found a positive correlation of mu power suppression in the CH of the following hand between the UMUT and BT conditions. In the UMUT condition, the healthy participants observed the action simulated by MVF without producing physical movement with the following hand; while in the BT condition, they performed overt motor execution of both hands. The positive correlation therefore suggested that the healthy participants with larger cortical excitability in motor regions during overt movement (i.e., BT condition) showed a higher degree of cortical engagement during covert action simulation (i.e., UMUT condition). Previous research has demonstrated a correlation in mu power attenuation between covert action simulation (i.e., motor imagery) and motor execution.⁵⁰ Our findings replicated and extended the previous results by showing a similar correlation in mu power suppression between overt motor execution and covert simulation induced by MVF.

Mu suppression correlates with motor ability for individuals with stroke

In the stroke group, we found that the participants with better abilities of motor imagery in the affected upper extremity demonstrated a higher degree of cortical engagement of their affected motor regions in the BMBT condition, as reflected in the larger attenuation of mu power. Additionally, their motor control ability predicted the lateralization of mu power attenuation between the two hemispheres, suggesting that the participants with better motor function of the affected upper extremity showed stronger dominance of the affected hemisphere over the unaffected one. Previous research has shown that assessment scores of hand dexterity and motor imagery are correlated with mu power suppression induced by motor imagery⁵¹ in healthy individuals. Our electrophysiological data extended these findings by showing similar relationships between motor function and mu power suppression induced by the combination of MVF and motor output in the individuals with stroke.

The underlying mechanism of the MNS

Converging findings have revealed pronounced attenuation in mu power in healthy individuals during both overt motor execution¹⁵ and covert motor simulation, including action observation^17,20 and motor imagery.^50,52 Neuroimaging studies have suggested that such mu power suppression in motor regions reflects the activation of the MNS.^18,53 The MNS enables individuals to simulate observed actions through automatic and implicit matching between visual perception and internal action representations.⁵⁴ Concerning the dual characteristics of the MNS (i.e., activation during both overt and covert actions), accumulating evidence has supported the utilization of covert motor simulation as a potential rehabilitation approach to facilitate functional restoration of the affected hemisphere in individuals with stroke.^5,13

Previous neurophysiological research has shown that MVF in traditional MT enhances cortical excitability in the MNS, as evidenced by oscillatory activities observed in EEG^34,35 and magnetoencephalography (MEG) recordings.²⁵ According to these studies, MVF facilitates cortical engagement of the affected hemisphere in individuals with stroke mainly by diminishing the asymmetry of neural activation across the two hemispheres (i.e., reducing the dominance of the unaffected hemisphere).¹³ Our results extend these previous findings by showing that digital MVF facilitates cortical engagement of the affected hemisphere in individuals with stroke.

Limitations and future direction

A limitation of the current study is the small sample size of the stroke group, which can be attributed to the following reasons. First, it was challenging to recruit individuals with sufficient motor ability to perform the manual task and to meet the inclusion criteria. Second, imprecise motor control or compensatory movements in individuals with stroke often produced large artifacts during the motor task, resulting in more exclusion of EEG data from the analysis. We addressed this issue by examining effect size. Most of our electrophysiological results showed medium and large effect sizes, supporting the strength of our findings. Another limitation is that the individuals with stroke were relatively young and were in the chronic stage, which may limit the generalizability of the findings to older individuals or those in the acute or subacute stages.

Some future directions could be extended based on the current findings. First, we aligned the EEG epochs according to the EMG onset of the leading hand (i.e., the unaffected hand in the individuals with stroke). We chose the EMG onset of the leading hand, rather than that of the following hand, for the following reasons: (1) MVF is triggered by the leading hand, inducing covert motor simulation in the CH of the following hand; aligning EEG epochs to the following hand’s EMG onset would consider only the overt motor execution and overlook the cortical engagement induced by covert simulation; (2) a consistent reference time point across conditions is necessary for valid comparisons of mu power attenuation, particularly in conditions where the following hand produces no overt movement (i.e., the UMUT condition); (3) aligning both hemispheres to the same reference time allows calculation of the lateralization index and preserves the temporal dynamics of MVF-induced cortical activity across two hemispheres; (4) EMG signals from the affected hand in the stroke group were relatively noisy and intermittent due to the involuntary compensatory movements, making precise determination of EMG onset less feasible in some participants. Future studies may consider aligning EEG epochs to the EMG onset of the following hand if feasible. Second, motor imagery ability was assessed using a self-report assessment (i.e., subjective report in the MIQ-RS). Future studies could consider incorporating more objective measures, such as the hand rotation task,^55,56 to evaluate motor imagery ability and examine its correlation with neural indices.

Conclusions

In conclusion, our electrophysiological data showed larger mu power attenuation in the affected hemisphere in both the UMUT and BMBT conditions than in the UMBT condition in the individuals with stroke. From a clinical perspective, the UMUT pattern may benefit individuals with stroke who are unable to produce motor output by providing motor intervention through covert simulation, while the BMBT pattern allows individuals with stroke with residual motor ability to actively engage in motor training through overt execution. Notably, digital-based MT is more feasible for providing the BMBT form compared to traditional MT, highlighting the potential and advantages of digital MVF. Our electrophysiological findings provide neural evidence that digital MVF can facilitate cortical activation in the affected motor regions of individuals with stroke, and suggest that mu power attenuation may serve as a neural signature for evaluating the intervention effects.

Footnotes

ORCID iDs

Ya-Ting Chen

Chih-Chi Chen

Yu-Wei Hsieh

Ethical considerations

This study was approved by the Institutional Review Board of the Chang Gung Memorial Hospital (202002234A3) and the Taipei Hospital, Ministry of Health and Welfare (TH-IRB-00210030).

Consent to participate

The participants provided informed written consent prior to the experiment and received financial compensation for their time.

Consent for publication

All authors have agreed to publish this article. All figures presented in this study were created by the authors and do not contain any materials reproduced from other sources.

Author contributions

YTC: Conceptualization, Data curation, Formal analysis, Methodology, Software, Visualization. Writing – original draft, and Writing – review & editing. CHC: Formal analysis, Investigation, Methodology, Software, and Writing – review & editing. LLC: Investigation, and Writing – review & editing. CCC: Investigation, Resources, and Writing – review & editing. YWH: Conceptualization, Methodology, Resources, Supervision, Project administration, Funding acquisition, Writing – original draft, and Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by grants from the National Science and Technology Council (NSTC 112-2628-B-182-006 and NSTC 112-2326-B-182-004-MY3) and Chang Gung Memorial Hospital (BMRPD25) in Taiwan.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

All data produced in this study are available upon reasonable request to the corresponding author.*

Appendix

References

Stig Jørgense

Nakayama

Otto Raaschou

, et al. Functional and neurological outcome of stroke and the relation to stroke severity and type, stroke unit treatment, body temperature, age, and other risk factors: the Copenhagen stroke study. Topics in stroke rehabilitation 2000; 6(4): 1–19. https://doi.org/10.1310/bt7j-2n6u-vd53-e1qu

Cauraugh

Kang

. Bimanual movements and chronic stroke rehabilitation: looking back and looking forward. Appl Sci 2021; 11(22): 10858. https://doi.org/10.3390/app112210858

Blanco-Mora

Almeida

Vieira

, et al. (eds). A study on EEG power and connectivity in a virtual reality bimanual rehabilitation training system. In: 2019 IEEE International Conference on Systems, Man and Cybernetics (SMC), 2019. IEEE. https://doi.org/10.1109/SMC.2019.8914190

Jeannerod

. Neural simulation of action: a unifying mechanism for motor cognition. NeuroImage 2001; 14(1): S103–S109. https://doi.org/10.1006/nimg.2001.0832

Mulder

. Motor imagery and action observation: cognitive tools for rehabilitation. J Neural Transm 2007; 114: 1265–1278. https://doi.org/10.1007/s00702-007-0763-z

Boni

Galluccio

Baroni

, et al. Action observation therapy for arm recovery after stroke: a preliminary investigation on a novel protocol with EEG monitoring. J Clin Med 2023; 12(4): 1327. https://doi.org/10.3390/jcm12041327

Guo

Wang

, et al. Variation in brain connectivity during motor imagery and motor execution in stroke patients based on electroencephalography. Front Neurosci 2024; 18: 1330280. https://doi.org/10.3389/fnins.2024.1330280

Chang

C-S

Y-Y

Chen

C-L

, et al. Alternative motor task-based pattern training with a digital mirror therapy system enhances sensorimotor signal rhythms post-stroke. Front Neurol 2019; 10: 1227. https://doi.org/10.3389/fneur.2019.01227

Deconinck

Smorenburg

Benham

, et al. Reflections on mirror therapy: a systematic review of the effect of mirror visual feedback on the brain. Neurorehab Neural Repair 2015; 29(4): 349–361. https://doi.org/10.1177/1545968314546134

10.

Altschuler

Wisdom

Stone

, et al. Rehabilitation of hemiparesis after stroke with a mirror. The Lancet 1999; 353(9169): 2035–2036. https://doi.org/10.1016/s0140-6736(99)00920-4

11.

Ramachandran

Altschuler

. The use of visual feedback, in particular mirror visual feedback, in restoring brain function. Brain 2009; 132(7): 1693–1710. https://doi.org/10.1093/brain/awp135

12.

Michielsen

Selles

Van Der Geest

, et al. Motor recovery and cortical reorganization after mirror therapy in chronic stroke patients: a phase II randomized controlled trial. Neurorehab Neural Repair 2011; 25(3): 223–233. https://doi.org/10.1177/1545968310385127

13.

Zhang

Fong

Welage

, et al. The activation of the mirror neuron system during action observation and action execution with mirror visual feedback in stroke: a systematic review. Neural plasticity 2018; 2018(1): 2321045. https://doi.org/10.1155/2018/2321045

14.

Alegre

Labarga

Gurtubay

, et al. Movement-related changes in cortical oscillatory activity in ballistic, sustained and negative movements. Exp Brain Res 2003; 148: 17–25. https://doi.org/10.1007/s00221-002-1255-x

15.

Pfurtscheller

Da Silva

. Event-related EEG/MEG synchronization and desynchronization: basic principles. Clin Neurophysiol 1999; 110(11): 1842–1857. https://doi.org/10.1016/s1388-2457(99)00141-8

16.

Wang

Liu

, et al. Neural activity and decoding of action observation using combined EEG and fNIRS measurement. Front Hum Neurosci 2019; 13: 357. https://doi.org/10.3389/fnhum.2019.00357

17.

Muthukumaraswamy

Johnson

McNair

. Mu rhythm modulation during observation of an object-directed grasp. Cognit Brain Res 2004; 19(2): 195–201. https://doi.org/10.1016/j.cogbrainres.2003.12.001

18.

Pineda

. The functional significance of mu rhythms: translating “seeing” and “hearing” into “doing”. Brain Res Rev 2005; 50(1): 57–68. https://doi.org/10.1016/j.brainresrev.2005.04.005

19.

Perry

Bentin

. Mirror activity in the human brain while observing hand movements: A comparison between EEG desynchronization in the μ-range and previous fMRI results. Brain Res 2009; 1282: 126–132. https://doi.org/10.1016/j.brainres.2009.05.059

20.

Frenkel-Toledo

Bentin

Perry

, et al. Dynamics of the EEG power in the frequency and spatial domains during observation and execution of manual movements. Brain Res 2013; 1509: 43–57. https://doi.org/10.1016/j.brainres.2013.03.004

21.

Rizzo

Petrini

Del Percio

, et al. Mirror visual feedback during unilateral finger movements is related to the desynchronization of cortical electroencephalographic somatomotor alpha rhythms. Psychophysiology 2022; 59(12): e14116. https://doi.org/10.1111/psyp.14116

22.

Rizzo

Del Percio

Petrini

, et al. Cortical sources of electroencephalographic alpha rhythms related to the anticipation and experience of mirror visual feedback-induced illusion of finger movements. Psychophysiology 2023; 60(6): e14281. https://doi.org/10.1111/psyp.14281

23.

Bartur

Pratt

Dickstein

, et al. Electrophysiological manifestations of mirror visual feedback during manual movement. Brain Res 2015; 1606: 113–124. https://doi.org/10.1016/j.brainres.2015.02.029

24.

Tani

Ono

Matsubara

, et al. Action observation facilitates motor cortical activity in patients with stroke and hemiplegia. Neurosci Res 2018; 133: 7–14. https://doi.org/10.1016/j.neures.2017.10.002

25.

Rossiter

Borrelli

Borchert

, et al. Cortical mechanisms of mirror therapy after stroke. Neurorehab Neural Repair 2015; 29(5): 444–452. https://doi.org/10.1177/1545968314554622

26.

Lee

M-T

Chen

C-C

H-L

, et al. Comparisons of three different modes of digital mirror therapy for post-stroke rehabilitation: Preliminary results of randomized controlled trial. Digit Health 2024; 10: 20552076241260536. https://doi.org/10.1177/20552076241260536

27.

Ding

Wang

Guo

, et al. Camera-based mirror visual feedback: potential to improve motor preparation in stroke patients. IEEE Transactions on Neural Systems and Rehabilitation Engineering 2018; 26(9): 1897–1905. https://doi.org/10.1109/TNSRE.2018.2864990

28.

Butler

Cazeaux

Fidler

, et al. The movement imagery questionnaire-revised, (MIQ-RS) is a reliable and valid tool for evaluating motor imagery in stroke populations. Evidence-Based Complementary and Alternative Medicine 2012; 2012(1): 497289. https://doi.org/10.1155/2012/497289

29.

Gregg

Hall

Butler

. The MIQ-RS: a suitable option for examining movement imagery ability. Evidence-Based Complementary and Alternative Medicine 2010; 7(2): 249–257. https://doi.org/10.1093/ecam/nem170

30.

Platz

Pinkowski

van Wijck

, et al. Reliability and validity of arm function assessment with standardized guidelines for the Fugl-Meyer test, action research arm test and box and block test: a multicentre study. Clinical rehabilitation 2005; 19(4): 404–411. https://doi.org/10.1191/0269215505cr832oa

31.

Oostenveld

Fries

Maris

, et al. FieldTrip: Open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Comput Intell Neurosci 2011; 2011: 9. https://doi.org/10.1155/2011/156869

32.

Maris

Oostenveld

. Nonparametric statistical testing of EEG- and MEG-data. J Neurosci Methods 2007; 164(1): 177–190. https://doi.org/10.1016/j.jneumeth.2007.03.024

33.

Meyer

Lamers

Kayhan

, et al. Enhancing reproducibility in developmental EEG research: BIDS, cluster-based permutation tests, and effect sizes. Developmental Cognitive Neuroscience 2021; 52: 101036. https://doi.org/10.1016/j.dcn.2021.101036

34.

Fong

Ting

Zhang

, et al. Event-related desynchronization during mirror visual feedback: a comparison of older adults and people after stroke. Front Hum Neurosci 2021; 15: 629592. https://doi.org/10.3389/fnhum.2021.629592

35.

Bartur

Pratt

Frenkel-Toledo

, et al. Neurophysiological effects of mirror visual feedback in stroke patients with unilateral hemispheric damage. Brain Res 2018; 1700: 170–180. https://doi.org/10.1016/j.brainres.2018.09.003

36.

Cauraugh

Summers

. Neural plasticity and bilateral movements: a rehabilitation approach for chronic stroke. Prog Neurobiol 2005; 75(5): 309–320. https://doi.org/10.1016/j.pneurobio.2005.04.001

37.

Stewart

Cauraugh

Summers

. Bilateral movement training and stroke rehabilitation: a systematic review and meta-analysis. J Neurol Sci 2006; 244(1-2): 89–95. https://doi.org/10.1016/j.jns.2006.01.005

38.

Kelso

. Dynamic patterns: The self-organization of brain and behavior. MIT press, 1995.

39.

Bernstein

. The Co-ordination and Regulation of Movements. Press, 1967.

40.

Liepert

Bauder

Miltner

, et al. Treatment-induced cortical reorganization after stroke in humans. Stroke 2000; 31(6): 1210–1216. https://doi.org/10.1161/01.str.31.6.1210

41.

Buccino

Solodkin

Small

. Functions of the mirror neuron system: implications for neurorehabilitation. Cogn Behav Neurol 2006; 19(1): 55–63. https://doi.org/10.1097/00146965-200603000-00007

42.

Iacoboni

Dapretto

. The mirror neuron system and the consequences of its dysfunction. Nat Rev Neurosci 2006; 7(12): 942–951. https://doi.org/10.1038/nrn2024

43.

Jeannerod

. Motor cognition: What actions tell the self. OuP Oxford, 2006.

44.

Rizzolatti

Craighero

. The mirror-neuron system. Annu Rev Neurosci 2004; 27(1): 169–192. https://doi.org/10.1146/annurev.neuro.27.070203.144230

45.

Michielsen

Smits

Ribbers

, et al. The neuronal correlates of mirror therapy: an fMRI study on mirror induced visual illusions in patients with stroke. J Neurol Neurosurg Psychiatry 2011; 82(4): 393–398. https://doi.org/10.1136/jnnp.2009.194134

46.

Wasaka

Kakigi

. Conflict caused by visual feedback modulates activation in somatosensory areas during movement execution. NeuroImage 2012; 59(2): 1501–1507. https://doi.org/10.1016/j.neuroimage.2011.08.024

47.

Fink

Marshall

Halligan

, et al. The neural consequences of conflict between intention and the senses. Brain 1999; 122(3): 497–512. https://doi.org/10.1093/brain/122.3.497

48.

Morioka

Osumi

Nishi

, et al. Motor-imagery ability and function of hemiplegic upper limb in stroke patients. Ann Clin Transl Neurol 2019; 6(3): 596–604. https://doi.org/10.1002/acn3.739

49.

Martínez-Rolando

Martín-Casas

Pérez-Pérez

. Assessment of Motor Imagery and Its Relationship to Functional Impairment in Post-Stroke Patients: A Descriptive CrossSectional Study. J Clin Med 2024; 13(19): 5929. https://doi.org/10.3390/jcm13195929

50.

Zich

Debener

De Vos

, et al. Lateralization patterns of covert but not overt movements change with age: An EEG neurofeedback study. NeuroImage 2015; 116: 80–91. https://doi.org/10.1016/j.neuroimage.2015.05.009

51.

Wang

Chen

, et al. Study of the correlation between the motor ability of the individual upper limbs and motor imagery induced neural activities. Neuroscience 2023; 530: 56–65. https://doi.org/10.1016/j.neuroscience.2023.08.032

52.

Chen

Bin

Daly

, et al. Event-related desynchronization (ERD) in the alpha band during a hand mental rotation task. Neurosci Lett 2013; 541: 238–242. https://doi.org/10.1016/j.neulet.2013.02.036

53.

Arnstein

Cui

Keysers

, et al. μ-suppression during action observation and execution correlates with BOLD in dorsal premotor, inferior parietal, and SI cortices. J Neurosci 2011; 31(40): 14243–14249. https://doi.org/10.1523/JNEUROSCI.0963-11.2011

54.

Rizzolatti

Fogassi

Gallese

. Neurophysiological mechanisms underlying the understanding and imitation of action. Nat Rev Neurosci 2001; 2(9): 661–670. https://doi.org/10.1038/35090060

55.

Chen

Y-T

Tsou

K-S

Chen

H-L

, et al. Functional but inefficient kinesthetic motor imagery in adolescents with autism spectrum disorder. Journal of autism and developmental disorders 2018; 48(3): 784–795. https://doi.org/10.1007/s10803-017-3367-y

56.

ter Horst

Van Lier

Steenbergen

. Mental rotation strategies reflected in eventrelated (de) synchronization of alpha and mu power. Psychophysiology 2013; 50(9): 858–863. https://doi.org/10.1111/psyp.12076