A total of 102 healthy adults were recruited from Beijing Bo’ai Hospital, the China Rehabilitation Research Center (CRRC), Capital Medical University, and local communities from September 2022 to April 2024. A screening evaluation was conducted for each participant at the CRRC. After the screening phase, 81 participants meeting the eligibility criteria were randomly assigned to the control group or experiment group using a block randomization protocol. The random allocation sequence was generated by the R package blockrand v1.5 (https://cran.r-project.org/web/packages/blockrand/index.html); the block sizes were 6, 8 and randomly ordered; the allocation ratio was 1:1. We printed the randomization cards created by the blockrand package and enclosed them in sequentially numbered, opaque, sealed, and stapled envelopes. The allocation sequence was concealed from the study coordinator responsible for enrolling and evaluating participants. The corresponding envelope was opened only after all baseline evaluations had finished.

The inclusion criteria for this study were as follows: healthy adults (18–60 years old); right-handed, as assessed by a seven-item Short Form of Edinburgh Handedness Inventory (EHI) [46]; Mandarin language; absence of visual impairment; education level above junior high school; absence of drug and alcohol abuse history; absence of neurological or mental illness history. We excluded participants with skin injuries or skin diseases in the right arm; acute infections; metal implants; poor fNIRS signal quality caused by thick and strong hair; not willing to undergo testing because the fNIRS cap is too tight.

This study was approved by the Medical Ethics Committee of CRRC (Approval number: 2021-053-1). This study was registered in the China Clinical Trial Registration Center (Registration number: ChiCTR2200064082). All participants signed the appropriate informed consent form according to the Helsinki declaration.

The flow diagram of participant enrollment, randomized allocation, and analysis is shown in Fig. 1. The data of 72 participants was included in the final report, the demographic and head size information are shown in Table 1. Except for the length from nasion (Nz) to inion (Iz) (Nz-Cz-Iz), all the baseline characteristics between the two groups had no significant difference.

Fig. 1.

Flowchart of participant enrolment, randomized allocation and analysis. Abbreviation: fNIRS, functional near-infrared spectroscopy; ES, electrical stimulation.

A block design experiment paradigm was utilized in our study. The task duration was 15 s and the inter-task rest duration was 20 s. During the ME condition, the participants performed right hand and wrist extension actions while watching scenery video clips. During the MI condition, the participants performed the same hand action while watching related first-person perspective action videos. The conditions and the video clips in each block were presented randomly. In one session, each task was repeated eight times, each block included 5 trials, each trial lasted 3 s, and the total duration of the assessment was approximately 10 min (Fig. 2).

Fig. 2.

Schematic illustration of the experimental paradigm. Abbreviation: ME, motor execution; MI, motor imitation; ES, electrical stimulation.

Two computer programs were used and synchronized via Lab Streaming Layer (LSL) to conduct the experiment: (1) a Python program based on the PsychoPy 3.8 package (Open Science Tools Ltd., Nottingham, Nottinghamshire, UK) presenting the visual stimulus and sending makers to record tasks onsets; and (2) the Aurora recording software v2021.9.0.6 (NIRx Medical Technologies, Minneapolis, MN, USA) to record the fNIRS raw data with the markers from the PsychoPy 3.8 program [47]. A fixation cross was presented during the rest phase; 5 scenery videos randomly selected from 10 clips were displayed during ME phases; 5 hand action videos randomly selected from 10 clips were played during MI phase.

The experiment was conducted in a quiet and dark evaluation room. Participants were instructed to sit in a chair and relax in a comfortable position while wearing the fNIRS cap. Two 50 $\times$ 50 mm ES electrodes (Axelgaard Mfg. Co., Ltd., Fallbrook, CA, USA) were attached to the skin over the right extensor digitorum communis (EDC). Before the test, the experimenter introduced the test process to the participants, helping them to practice the tasks until they fully understand the experiment. The experiment group performed ME and MI with ES, while the control group performed ME and MI with sham ES. To maintain consistency between the two groups, except in the use of ES, the same experimental program was used, with the only difference being the intensity of the electrical current. In the experiment group, the current was set for each individual to induce finger extension action; whereas, in the control group, the intensity was set at 0.5 mA for all participants, as sham stimulation. In this study, the mean value of ES intensity in the experiment group was 10.70 $\pm$ 2.86 mA. All the assessments were performed by YC.

The fNIRS data were recorded using a portable continuous wave Nirsport2 (NIRx, Minneapolis, MN, USA) device with eight dual-wavelength light-emitting diodes (LED) centered at 769 and 850 nm sources and eight avalanche photodiode detectors. The data were collected with a sample rate of 10 Hz. The custom montage was designed using the fOLD toolbox (https://github.com/nirx/fOLD-public) and used in the fNIRS cap configuration, which included eight sources and eight detectors, forming a total of 20 source-detector pairs (channels) (3 cm) (Fig. 3) [48]. Based on the international electroencephalogram (EEG) 10–20 system, sources were placed at the positions F3, FC5, FC1, C3, CP5, CP1, P3, and Pz; detectors were placed at the positions AF3, F5, FC3, FCz, C1, CP3, CPz, and P1. We positioned the optodes in the fNIRS cap holders before it was worn by the participant, according to Aurora recording software’s (v2021.9.0.6, NIRx Medical Technologies, Minneapolis, MN, USA) signal quality evaluation results, and manually adjusted the sources and detectors to achieve an efficient optode-scalp coupling [8].

Fig. 3.

The custom fNIRS montage is shown in 2D (A) and 3D format (B). (A) Location of sources and detectors over 10–20 map. Sources are shown as red circles, detectors are shown as blue circles, and channels are shown as black lines. (B) Location of sources-detector pairs and channels over 3D brain surface and marked brain areas. Sources are shown as red dots, detectors are shown as black dots, and channels are shown as white lines with an orange dot (the midpoint of source-detector pairs). The brain areas covered by the montage are shown in different colors: Inferior Frontal Cortex, red; Premotor Cortex, green; Dorso Lateral Prefrontal Cortex, yellow; Somatosensory and Motor Cortex, blue; Superior Parietal Cortex, orange; Inferior Parietal Cortex, purple.

The EN-STIM 4 (ENRAF-NONIUS, Rotterdam, South Holland, Netherlands) apparatus was utilized to generate electrical current output during the task periods and drop to 0 mA during the rest periods (Fig. 2). A custom ES protocol was designed: rectangular waveform, 200 µs phase duration, 50 Hz pulse frequency; 1 Hz burst frequency, 1 s ramp up time, 15 s hold time, 1 s ramp down time, and 18 s interval time. The intensity of the electrical current for each participant in the experiment group was individualized to minimize evoked muscle contractions and avoid discomfort.

Block averaged HbO and HbR time series data of all the channels for each group and condition is shown in Fig. 5. The HbO concentration increases and HbR concentration decreases, which represents the cortical activation of the whole MNN. Averaged HbO and HbR time series data for each ROI and both conditions and groups are shown in Fig. 6.

Fig. 5.

Morphology of fNIRS responses of the whole MNN using the block averaging approach for different conditions and groups. Each column represents a different group. Each row represents a different condition. Red represents HbO, and blue represents HbR. Shaded lines indicate SE. Abbreviation: HbO, oxy-hemoglobin; HbR, deoxy-hemoglobin; SE, standard error; MNN, Mirror neuron network.

Fig. 6.

Morphology of fNIRS responses using the block averaging approach for all ROIs, conditions and groups. Each column represents a different condition. Each row represents a different ROI. Red represents HbO, and blue represents HbR. Solid line represents experiment group, and dotted line represents experiment group. Abbreviation: HbO, oxy-hemoglobin; HbR, deoxy-hemoglobin.

According to the observation of time to peak, we extracted Hb data from 10 s to 20 s after the onset of the tasks and computed the mean value thereof to reflect the cortical activation level. The distribution of Hb concentration mean value from 10 s to 20 s after the onset for each group and condition is shown in Fig. 7. The data for each ROI is shown in Fig. 8.

Fig. 7.

Descriptive statistics of Hb concentration mean value from 10 s to 20 s after the onsets of the whole MNN for different conditions and groups. HbO and HbR data are shown in red and blue, respectively. Each column represents a different group. Each row represents a different condition. Abbreviation: HbO, oxy-hemoglobin; HbR, deoxy-hemoglobin.

Fig. 8.

Descriptive statistics of Hb concentration mean value from 10 s to 20 s from onsets per group, condition and ROI. HbO and HbR data are shown in red and blue, respectively. Each column represents a different condition. Each row represents a different ROI. Abbreviation: HbO, oxy-hemoglobin; HbR, deoxy-hemoglobin.

Regarding cortical activation during ME condition, the estimated hemodynamic activation in different ROIs of MNN evoked by ME with and without ES are shown in Table 2 and Fig. 9. Regarding cortical activation during MI condition, the estimated hemodynamic activation in different ROIs of MNN evoked by MI with and without ES are shown in Table 3 and Fig. 9.

Fig. 9.

Estimates of fNIRS response per group, condition and ROI calculated by the LMM model using HbO concentration mean value from 10 s to 20 s after onsets. Control and experiment group responses are shown in green and orange, respectively. The presence of a response (statistical difference to 0, q 0.05) is indicated by a triangle. Error bars represent the 95% CI of the mean. Significant level: ns, not significant; * 0.05. LMM, linear mixed-effects models.

To test the difference between two groups, the mean HbO change data of the experiment group minus the control group during the MI and ME condition were used as indexes. The results are shown in Table 4 and Fig. 9. During MI condition, ES can significantly improve the activation of the following ROIs: Frontal_Inf_Tri_L (t = 2.022, p = 0.044), Frontal_Mid_2_L (t = 2.314, p = 0.022), and Precentral_L (t = 2.314, p = 0.022). There was no significant difference in all the ROIs between the control and experiment groups during ME condition.

For core MNN brain regions, our results revealed that, compared with the control group. except for IFG where activation increased significantly during MI but showed a tendency of decrease during ME, the other two core ROIs of MNN, PMC and IPL, showed increased activation during both ME and MI (Figs. 6,9). The MNN plays an important modulation role in the neural control of ME and MI, the results of this study suggest that it is possible to enhance this effect by ES [2]. In addition, ES may be helpful to improve motor initiation, because it can make muscles activate faster to start the desired action, and PMC is functionally related to motor initiation, the induced changes of brain and muscle both facilicate the initation of movement [58]. In the MI condition, the overall (Figs. 5,7) and most ROIs (Figs. 6,8) of MNN showed increased activation trends with ES, and there were more significantly activated ROIs (five ROIs activated with ES, two ROIs activated without ES, respectively) (Fig. 9, Table 4). There was no statistically significant difference between the two groups in the ME condition, nonetheless, we still observed a trend towards an increase in the experiment group in the overall (Figs. 5,7) and most ROIs (Figs. 6,8) of MNN.

For other brain regions, the MFG showed increased activation during both conditions in both groups (Figs. 6,8,9). The MFG is related to working memory, attention, and planning, it receives and sends widespread connections as part of network in higher cognitive functions [59]. These increases may be attributed to the MNN-based rehabilitation strategies including more cognitive process and ES enhances the sensory inputs to improve the connections of different brain areas [60]. The postcentral gyrus oversees somatotopic organization and sensory motor integration, for this area, an increased trend was observed during ME, whereas a decreased trend was shown during MI (Figs. 6,8,9). This may be attributed to the absence of motion visual inputs during ME, which may reduce the feelings of ES and decrease the attentional level [61]. The activation of the precentral gyrus increased during MI but decreased during ME. One of the possible reasons is that comparing with action videos with visual information of biological actions, the visual information of scenery videos can’t present action information, which may reduce the attention level, resulting in a weaker motor performance or a smaller range of motion (ROM) than in MI [62]. No motor performance data such as speed and ROM were recorded in this experiment, to further study the relationships of motor performance and motor visual inputs, video analysis or data glove with sensors are needed to compare the ME quality and amplitude of the produced motion in future studies [63]. Besides, the decrease of SPG during our experiments may be attributed to less visuo-motor transformation and spatial processing workload; the PFC may also take part in this process to handle visuo-spatial working memory [47].

The reason for no significant difference being found between the experiment and control groups in some ROIs may be attributed to our choice of participant. For healthy individuals with no motor dysfunctions, it is easy to finish all the ME and MI tasks. In some ROIs, the activation level was lower with ES than without ES, a possible reason is that ES moving the hands passively, thus affecting the active participation and attention level of the participants. However, for patients with neurological conditions, it is hard to finish the full range hand movement without the help of ES. Therefore, activation of the MNN may be obviously different with or without ES in patients with CNS injuries. Another possible reason may be the relationship of visual and sensory information, for MI, ES is a promoting factor to provide consistent visual and sensory information; whereas, for ME, ES is sometimes a interference factor, because of the inconsistency between visual and sensory information. In addition, the technical characteristics of fNIRS may be a factor contributing to the lack of statistically significant results, because of its relatively poor spatial resolution and relatively shallow penetration depths [64]. Further investigations using different technologies and more clinical trials need to be conducted to achieve a deeper understanding of these factors.

Without ES, most ROIs’s activation patterns are similar between ME and MI conditions [65]. Our previous study has shown that the activation of MNN was obviously higher during NMES + MI than during NMES + ME, which revealed that NMES will enhance the brain activation when there is the executed action’s visual information [1]. The visual system plays a part in restoring motor cognitive functions through neuroplasticity in stroke rehabilitation [66]. Motor function is related to visual and spatial processing, which is important to rehabilitation strategies aiming to enhance upper extremity motor function after stroke [67]. Compared with ME, MI is a more complex task, including more cognitive components, which means a high energy and resource load [68]. In our study, some participants reported that it was easy to relax when looking at the scenery videos, even feeling fatigue and boredom in the latter half of the experiment, thereby reducing the range of movements. This may have resulted in decreased attentional level and motivation.

It is necessary to try different kinds of visual information presentation forms to enrich the MNN based rehabilitation options. Some studies have focused on this. For example, mirror therapy (MT) is an important rehabilitation therapy based on MNN theory [5, 69]. The mirror visual feedback to the CNS can sustain promising therapeutic potential as part of the treatment for pain reduction and hand function [69]. In recent years, researchers keep searching for new kinds of MT visual information presentation forms. Compared with conventional MT, the parieto-frontal network showed greater activation in video based MT, which shows that greater focus on visual feedback could promote neuroplasticity [5]. Immersive virtual reality based MT can also improve cortical activation and enhance rehabilitation outcomes [70]. In addition, integrated application of MT and robotic training can enhance the functional gains of activities of daily living (ADLs) [71]. Taken together, these studies suggest that technologies used to present visual information may play a crucial role in MNN-based rehabilitation treatments.

The above findings suggest that ES may help to enhance the activation of MNN, it is conceivable that the effects of central-peripheral synchronous stimulation on cortical activation may be contributed to this phenomenon. In addition, multi-sensory integration may also play a key part in this process. The consistency of different sensory inputs, including visual, tactile, ES evoked motor proprioception, and motor perception reinforce each other, thus improving brain excitability (Fig. 10).

Fig. 10.

Schematic diagram of the possible neural mechanism using central and peripheral stimulus synchronously to improve neural plasticity, particularly the combination of MNN-based rehabilitation methods and ES.

By sensory-motor coupling and integration, the central and peripheral synchronous stimulus influence each other, attaining synergistic effects. Some studies have confirmed this assumption, especially in BCI studies. For example, PES combined with action observation helps to improve BCI signal quality and feature classifier accuracy [47]. In addition, BCI action observation feedback combined with PES is capable of facilitating sensorimotor cortical activation in the affected hemispheres [72], thus enhancing task-driven corticospinal plasticity and treatment induced neural plasticity [73]. Furthermore, adding other treatment methods to MNN-based methods may be helpful to perform motor tasks, for example, the combination of NIBS and MNN-based rehabilitation strategies are important study directions in neurorehabilitation [74]. These strategies are helpful to modulate neural oscillations and enhance human motor and cognitive function [75]. In another study, MNN strategies paired with conventional rehabilitation resulted in significant improvements in wrist and hand impairment [76]. This has been expanded to include more forms of central and peripheral stimulus synchronous application, including combination of different central stimuli, such as TMS, tDCS, TUS, tBPM combined with MNN-based rehabilitation strategies.

In addition to stimulating muscles and peripheral nerves, ES could also activate the spinal cord and brain. On the one hand, the visual stimulation and biological kinematics information provided by action videos, as well as the cognitive activities of observation, imagery, and understanding of the actions, help to enhance brain excitability [77]. On the other hand, the somatic and tactile sensation of electrodes attached to the skin, the contraction caused by ES acting over muscles and nerves, the stimulation of joint kinesiology and positional sensation driven by muscle contraction, and the increase of motor amplitudes assisted by ES may form feedback loops in a comprehensive way [78]. When the spinal cord obtains ES-evoked sensory responses, both intraspinal networks and ascending pathways can be influenced, thus enhancing the excitability of the peripheral nerves, spinal cord, and brain [39]. Furthermore, the combination of the two forms of comprehensive training improves the interest of treatment. The participant’s attention is further concentrated, and the motivation is enhanced, especially when the action scenes appears, and the activation of brain regions may be further enhanced. In summary, central-peripheral synchronous stimulation may enhance the activation level of brain regions with MNN as the core, improve the degree of participation in rehabilitation training, thus playing a role in improving the rehabilitation efficacy.

The findings of our study have prosperous potential applications in neurorehabilitation treatments, they have unique values in design and manufacture of the related rehabilitation devices. The combination of action observation, imagination, comprehension, imitation based on MNN and conventional ES in daily physical therapy can help enhance the brain activation of the existing treatment techniques, thus enhancing clinical effects and efficacy. The theory of integrating stimulus from both central and peripheral directions to promote each other may play an important role in the neurology and neurorehabilitation fields [79]. By combining active movement and assist movement organically, ME + ES and MI + ES will hopefully be new rehabilitation interventions, especially for patients with manual muscle test (MMT) grade 2 to 4 and who need active-assist training. In addition, by modifying the ES parameters and visual inputs (for example, relaxing sceneries with light music), the program we used may be beneficial to reduce pain, thus playing a part in pain management [80]. Moreover, this kind of treatment is easy to utilize in group training, therefore, it is possible to enhance brain connectivity among different individuals and bring some positive effects to the physiotherapy sessions, thus providing positive impacts on mood and motivation [81].

In the rehabilitation engineering field, this study verifies the feasibility of enhancing central stimulation by superimposing peripheral stimulation (ES, peripheral magnetic stimulation, and mechanical stimulation) to improve the signal intensity and SNR of brain activity signals. This has important reference value for the development of instruments and equipments, such as fNIRS signal based neurobiofeedback and BCI devices [44]. ES can be utilized to augment the activation signal generated by CNS and the sensory feedback during an action, and improve classification accuracy of the BCI signal decoding results [39]. To further improve the accuracy of the classification algorithm, large samples of data need to be collected and analyzed in a uniform way. In addition, the cortical activation patterns can be used as trigger signals to control ES devices to conduct BCI rehabilitation trainings [82]. The implementation of the related therapies can be integrated into a all-in-one device, thus helping new rehabilitation device development, and can be used in rehabilitation institute, hospital, community, and home rehabilitation programs, especially in tele-rehabilitation settings, which has good socioeconomic value.

Additionally, accurate evaluation of brain activation can be a neuroimaging biomarker to estimate the recovery potential of patients with CNS conditions, thus producing individualized rehabilitation plans. Activation of MNN may be a potential neuroimaging biomarker for motor function recovery [83]. For example, an electroencephalogram (EEG) study has revealed that event-related desynchronization during action observation is an early predictor of recovery in sub-cortical stroke [84]. Another fNIRS study showed that the correlation between FC of MNN and autistic traits can be used in clinical outcome evaluations [85]. In the future, the correlation between the activation level of the MNN and the results of the patient functional ability scales can be combined to build predictive models.

To sum up, synchronization of MNN based rehabilitation strategies (central intervention) and ES (peripheral intervention) may produce synergistic effects in ehancing brain activation. The fNIRS signals detected during these conditions could be used as potential neural imaging biomarkers to predict rehabilitation outcomes.

There are some limitations of our study. First, as the Nz-Cz-Iz distance has a statistically significant difference between the two groups, a 3D digitizer should be used to measure the individualized channel location data for more accurate brain modeling. Second, we did not assess the performance of the motor trials, therefore, it was impossible to use the quality of motor outcome as a variable to build the models. In future studies, motion tracking systems, data gloves or video analysis software can be used to record the motor quality, such as range of motion, mean speed, mean distance, normal path length, spectral arc length, number of peaks, and task time metrics [86]. Third, other brain regions also play important roles during ME and MI, and both hemispheres are engaged in the process of related information. Therefore, it is reasonable to detect whole brain information using an fNIRS system with more channels in the future. Another possible solution is to use different brain imaging metrics to further confirm the experiment results, for example, using EEG + fNIRS or functional magnetic resonance imaging (fMRI) + fNIRS to take full advantage of the temporal and spatial resolution of different technologies. Fourth, patients with different disease should be investigated to verify these findings, especially in patients with motor and cognitive impairments. Fifth, there are many indexes in fNIRS data analysis, including peak, time to peak, area under the curve, mean, slope from stimulus onset to peak, minimum, peak-to-peak and various frequency domain indicators [47]. Further analyses should be performed in future studies, especially in the BCI signal patterns extraction and neurobiofeedback signal augments studies [62]. Resting state inter- and intra-hemispheric FC, brain signal complexity, and frequency domine indexes could be utilised to explore the brain change evoked by MNN-based approaches and it’s clinical relevance [87]. Last, with the development of artificial intelligence, related fNIRS preprocessing and processing approaches may play more important roles, which include machine learning, deep learning, and hybrid models [88].