The illusion caused by mirror visual feedback, or mirror illusion (MI), is a neurophysiological phenomenon that translates observed mirror-reflected movements of a limb into a feeling of corresponding movements in the contralateral limb that is behind the mirror or absent [1, 2].

It has been shown that mirror visual feedback activates the system of mirror neurons and motor structures of the brain in the hemisphere ipsilateral to the active arm, stimulating neuroplasticity [2, 3] and facilitating restoration of interhemispheric functional balance [4, 5]. These physiological mechanisms underpin mirror therapy, which employs MI to treat a number of chronic neurological pathologies. The clinical efficacy of mirror therapy has been established for postamputation phantom limb pain [1, 6, 7] and stroke hemiparesis [8–10]. There is evidence for the effectiveness of the method in the treatment of a complex regional pain syndrome [11–14] and functional motor disorders [15]. Researchers and specialists are currently searching for new mirror therapy protocols to enhance its efficacy; one of the approaches under consideration is Graded Motor Imagery [16].

Electrophysiological correlates of the brain activity accompanying MI were studied in healthy individuals and patients using electroencephalography (EEG) [5, 17, 18], transcranial magnetic stimulation [19, 20] and magnetic encephalography [21, 22], and hemodynamic activity parameters have been assessed using functional magnetic resonance imaging [4, 23] and near-infrared spectroscopy (NIRS) [24, 25]. However, there are no reported studies that would have investigated electrophysiological and hemodynamic parameters simultaneously. We expect that simultaneous recording of EEG and NIRS will yield a better understanding of the physiology of MI and allow a comparison of the informative value of the two methods in identification of the MI's correlates. Moreover, the neurophysiology of MI during sensory stimulation remains largely unexplored.

This study aimed to determine the neurophysiological correlates of MI in healthy individuals by simultaneously registering brain activity using EEG and NIRS during a mirror procedure involving movement and tactile stimulation.

METHODS

Participants of the study

The study was conducted at N.I. Pirogov Russian National Research Medical University of the Ministry of Health of the Russian Federation. The inclusion criteria were: signed voluntary informed consent; age 18–80 years; any gender; righthandedness confirmed by the Edinburgh Handedness Inventory (score above 40). The exclusion criteria were: disagreement to participate in the study; intake of medications that affect the central nervous system at the time of the study; acute conditions or exacerbations of chronic diseases; chronic pain syndrome and disabling conditions (including amputations); serious vision problems that prevent seeing the reflection of the limb in the mirror; skin diseases in the head area; a desire to drop out of the study; deterioration of the health condition during the study.

The study included 30 healthy volunteers, 12 male and 18 female, median age 21 [20.0; 23.0] years, all right-handed. No participants dropped out of the study; the data describes the entire sample (n = 30).

The procedure

The participants were randomized into two groups by the active arm, right or left. The active arm was the one placed in front of the mirror and moved or subjected to sensory stimulation (fig. 1A). The active left arm group included 18 people, the active right arm group 12 people.

The experiments were conducted in an electrically shielded and soundproof chamber (Neuroiconica Assistive, Russia, model EK-1). The participants wore a cap with 21 EEG electrodes, 10 NIRS sources and 10 NIRS detectors. A conductive gel was applied under each electrode. The procedure included three blocks: 1) synchronous movement of two hands without a mirror; 2) imitation of mirror therapy with a motor paradigm; 3) imitation of mirror therapy with sensory stimulation. After the second and third blocks, the participants were offered to fill out questionnaires reflecting subjective parameters of MI.

For the first block of the experiment, the participants performed a bimanual synchronous movement "fist-edge-palm" (fig. 1B) 20 times every 10 seconds, prompted by an audio signal.

For the second block, a mirror was placed before the participant so that it reflected the active arm, while the other hand was behind the mirror, relaxed. The participant performed the "fist-edge-palm" movement with only the active arm 20 times every 10 seconds, prompted by an audio signal; the gaze of the participant was focused on the reflection of the moving hand in the mirror.

The third block involved sensory stimulation of the participant's active arm: the first tool was a blunt needle, used from the middle of the forearm to the tip of the middle finger, the second tool — a brush, driven from the middle of the forearm to the tip of the middle finger in constant contact with the arm. The intensity of stimulation with a blunt needle was pre-calibrated by the experimenter to provide a comfortable tactile sensation without pain or discomfort. Each stimulation included alternating sequential use of the needle and the brush for a total of 20 times. The applications of the tools were 10 seconds apart from each other. The second arm was behind the mirror, relaxed; the participant watched the reflection of the stimulated arm in the mirror.

Hereinafter, the terms "contralateral" and "ipsilateral" are used in relation to the active arm when referring to the lateralization of sources of activity in the brain. For the bimanual movement stage, the active arm is that which performs movements in the second and third blocks of the experiment.

Assessment of the subjective parameters of the mirror illusion

After the second and third blocks, the participants assessed the following parameters of MI on a 10-point scale:

• vividness of the illusion: "I had a feeling of movement in the hand reflected in the mirror" (second block) or "I had a feeling of touch in the reflected hand" (third block);
• ownership: "I felt that the hand in the mirror was a part of my body, and not just a reflection of another hand" (second and third blocks);
• agency: "It seemed to me that I could directly control the movements of my hand in the mirror" (second block).

Additionally, the participants mentioned when the MI began to manifest: immediately or in the first seconds of the stimulation, or in the first, second, third or fourth quarter of the procedure.

After completing three blocks of the experiment, the participants assessed the intensity of the emotional reaction to MI in five domains: surprise, interest, delight, anxiety, and calmness (absence of a pronounced emotional reaction). They rated the intensity of each emotion on a scale from 0 (completely absent) to 10 (maximum intensity).

The questionnaire was developed specifically for this study.

Registration of brain activity signals

NeoRec Cap 21 (Medical Computer Systems, Russia) was used to record EEG; the setup included 21 leads: Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, C7, C8, P7, P8, Cp5, Cp6, Fz, Cz, Pz. The reference lead was Afz. The sampling rate was 250 Hz. The signal was not filtered during recording. We made EEG recordings for all 30 participants, one recording per block.

Photon Cap C20 NIRS spectroscope device (CortiVision, Poland) with 10 sources and 10 detectors was used to register the metabolic activity of the brain. Of all the source–detector pairs, we selected 28 channels corresponding to positions C5h, Fcc3, Ccp3, C3h, Ccp1, Fcc1, C1h, Fccz, Fc1h, Ccpz, C2h, Ccp2, Cp2h, Fcc2, C4h, Fcc4, Ccp4, C6h, Cp1h, Cppz,

P1h, P2h, Ppoz, Fc2h, Ffcz, F2h, F1h, Affz of the 10-5 system (fig. 2). The NIRS sampling rate was about 5 Hz, reduced to this value by linear interpolation of the signal based on the timestamps sent by the device. There were NIRS recordings of 29 participants available, as one recording was lost.

Both devices transmitted data via the Lab Streaming Layer protocol using software provided by the manufacturers, NeoRec (version 1.6.19830, Medical Computer Systems, Russia) and CortiView (v1.11.5, CortiVision, Poland). The data were synchronized and recorded using a proprietary PsychoPy script (v2022.1.1, Open Science Tools Ltd, UK).

Recording hand movements or sensory stimulation

The participant's hand movements and the experimenter's actions were recorded with a CW 870FHD video camera (CBR, China), resolution: 640 × 480, horizontal FOV:75°, frame rate: 10 fps. The script that synchronously recorded the NIRS and EEG data assigned numbers of the frames to the saved data samples. The videos were analyzed frame-by-frame. In blocks with bimanual movements or mirrored movements, we distinguished the beginning of movement, complete clenching into a fist, full opening of the palm, touching the table with the palm, complete cessation of movements. In blocks with stimulation — the beginning of the experimenter's hand movement, the first touch with a hard object, the end of the last touch with a hard object, the beginning of the brush touch, the removal of the brush from the hand, the end of the experimenter's hand movement. The EEG and NIRS recordings were marked up in selected frames: the first data sample corresponding to the marked frame was taken as the beginning of the event.

Processing of brain activity signals

We used the MATLAB 2023b environment to process EEG and NIRS data. A bidirectional notch filter suppressing 50 Hz interference was used for EEG filtering; it was followed by a unidirectional FIR filter with a bandwidth of 0.5–40 Hz. The delay in bandpass filtering was compensated by sample shifts. To isolate artifacts, we analyzed 200 ms of the signal epoch: those with more than 1% of samples that exceeded ±200 microvolts on any of the channels were discarded. Only 3 out of 60 recordings had 0.5–0.8% of such epochs, and in the remaining recordings this figure was significantly smaller. The recordings of each participant were combined after filtering, and then the resulting signal was decomposed into independent components with the help of a proprietary implementation of the AMICA algorithm [26]. Based on the criteria described earlier [27], we selected components reflecting activity in the primary sensorimotor cortex of left and right hemispheres (mu rhythm sources), an additional motor region, and in the posterior parietal cortex/preclinium. If such components could not be identified, we used the LCMV beamforming method based on the averaged topographic maps of the components of the needed type, and thus searched for the activity.

For the selected components, we calculated the degree of desynchronization of rhythmic activity in the alpha range (8–13 Hz) during movements and stimulation: applied filtering using a unidirectional FIR filter with shift compensation, then the Hilbert transformation, and calculation of absolute values. The degree of desynchronization was calculated as follows:

ERD = (A_rest ‒ A_task)/(A_rest + A_task),

where Arest is the average amplitude in the window (–6, –2) before the start of arm movement in blocks 1 and 2, and before the start of movement of the experimenter's hand in block 3; Atask is the average amplitude of the signal in the interval from the beginning of movement to its end for blocks 1 and 2, or the interval from the first to the last touch with a hard or soft object for block 3.

The NIRS signals, recorded as radiation intensities at two wavelengths, were converted into concentrations of oxy-(HbO) and deoxyhemoglobin (HbR). A filter with a bandwidth of 0.002–0.09 Hz was applied to the resulting data. In each session, we isolated 10-second segments with hemodynamic responses; the beginning was start of the movement for blocks 1 and 2, and first touch for block 3. The signal value th was average within the 5 seconds of the recording preceding the considered segment was subtracted therefrom. The value of the hemodynamic response was calculated as the response value averaged over all segments and over the four NIRS channels of the respective region. We considered four regions: those corresponding to the primary sensorimotor cortex of the left (NIRS channels in positions C5h, Fcc3, Ccp3, C3h) and the right hemispheres (C4h, Fcc4, Ccp4, C6h), sources in the additional motor region (Ffcz, F2h, F1h, Affz), and those in the posterior parietal cortex/preclinium (Cppz, P1h, P2h, Ppoz).

Statistical analysis

For statistical analysis, we used the STATISTICA 6.0 program (StatSoft, USA). The sample distribution normalcy hypothesis was checked with the help of the Shapiro–Wilk test. Spearman's rank correlation coefficient (r_s) was used for correlation analysis, and the Wilcoxon test for comparison of related samples. The level of significance was set at p < 0.05. Quantitative data are given as median and interquartile ranges (25^th and 75^th percentiles).

RESULTS

The degree of desynchronization of the mu rhythm

We managed to establish EEG laterality only for the mu rhythm sources, i.e. those localized in primary sensorimotor areas. In cases of bimanual movement, a comparable degree of desynchronization of the mu rhythm was observed over both hemispheres. A similar symmetrical desynchronization of the mu rhythm was recorded for the unimanual movement in front of the mirror both over the contralateral and ipsilateral hemispheres relative to the active limb (tab. 1). In contrast, sensory stimulation of the active arm caused significantly more pronounced desynchronization of the mu rhythm over the contralateral hemisphere (tab. 1).

Hemodynamic response

In all experimental blocks, we registered a comparable intensity of the hemodynamic response in both hemispheres of the brain (tab. 2). This pattern was evident in the concentration of both HbO and HbR. Tactile stimulation caused atypical changes in hemodynamic parameters: the concentration of HbO decreased during active stimulation, and the concentration of HbR decreased to a lesser extent or increased (tab. 2).

Correlation of quantitative characteristics of EEG and NIRS signals

The intensity of desynchronization of rhythmic activity in the alpha range also correlated with changes in HbO concentration (r_s = 0.527, p = 0.003) in the block involving unimanual movement in front of the mirror, as registered in the posterior parietal cortex. No further statistically significant correlations between the magnitude of desynchronization and hemodynamic response measures were observed for identical experimental conditions.

Psychometric characteristics of the mirror illusion

The subjective intensity of the mirror illusion was comparable between hand movement and sensory stimulation conditions in terms of illusion vividness and sense of ownership (tab. 3). The majority of participants noted the onset of MI during the first half of the corresponding block: 53% in block 2 and 79% in block 3 (tab. 3).

The analysis of correlations between the subjective parameters of MI and the degree of desynchronization or a change in the concentration of HbO or HbR associated with unimanual movements in front of the mirror revealed only a weak correlation of the vividness of the illusion with the degree of desynchronization of the ipsilateral mu rhythm (r_s = 0.370, p = 0.044) and a change in the concentration of HbO in the area of the supplementary motor cortex (r_s = 0.425, p = 0.022), as well as a weak correlation of the sense of agency with a change in the concentration of HbO (r_s = 0.392, p = 0.036) and HbR (r_s = –0.413, p = 0.026) in the area of the supplementary motor cortex. This task yielded no other significant correlations.

As for the sensory stimulation of the arm in front of the mirror, we registered no significant correlations between the subjective parameters of MI and the degree of desynchronization of the mu rhythm (both contra- and ipsilateral).

Most of the subjective parameters of MI significantly positively correlated with the intensity of the emotional reaction to the illusion (tab. 4).

For the unimanual movement in front of the mirror, the degree of interest in the emerging illusory sensations negatively correlated with the degree of desynchronization of the mu rhythm in the contralateral hemisphere (r_s = –0.471, p = 0.009) and in the area of the supplementary motor cortex (r_s = –0.419, p = 0.021), as well as with changes in the concentration of HbO in the area of the supplementary motor cortex (r_s = –0.444, p = 0.016). We registered no other statistically significant correlations between the vividness of emotional reactions and quantitative indicators of the EEG or NIRS response.

DISCUSSION

The simultaneous recording of EEG and NIRS signals during perception of mirror visual feedback revealed that a unimanual movement reflected in the mirror and seen by a healthy individual activated both hemispheres of the brain, same as bimanual movement, as evidenced by electrophysiological and hemodynamic indicators. For the tactile stimulation, EEG registered activation of only the contralateral hemisphere relative to the stimulated arm, while NIRS showed comparable hemodynamic changes in both hemispheres, but atypical in the direction of changes of HbO and HbR concentrations. Most indicators of electrophysiological and hemodynamic response did not correlate with each other.

Other studies have shown bilateral activation of the brain associated with arm movement and visual mirror feedback using either EEG [17, 18] or NIRS [24, 25]. An unimanual movement in front of a mirror was accompanied (to a greater extent) by activation of the additional motor [24, 25], upper or lower parietal [18, 24], premotor [18, 25] and primary somatomotor [18] cortex of the ipsilateral hemisphere in relation to the active arm. These areas include, but are not limited to, the structures of the mirror neuron system. One of the explanations for the MI mechanism is the blockade of inhibitory processes in the ipsilateral (i.e., contralateral to the mirror image) frontal and parietal associative sensorimotor areas of the cortex, which leads to the disinhibition of these areas and the development of the illusion of movement of a stationary limb relaxed behind the mirror [18]. Perhaps this is due to a mirror feedbackinduced shift in the activation of cortical structures in the direction of the ipsilateral hemisphere, which helps to reduce the interhemispheric imbalance in unimanual movement [5].

EEG investigation of changes associated with tactile stimulation with a mirror [28] has shown stronger activation on the contralateral side of the stimulated arm, which is consistent with our data. Another finding was the activation of the secondary sensorimotor cortex and areas of the mirror neuron system in the ipsilatreal hemisphere, which justified the expectation of changes in the amplitude of the mu rhythm on the ipsilateral side depending on the intensity of the illusion; however, no such changes have been detected. The observed changes in HbO and HbR concentrations may reflect the so-called "sensorimotor paradox": it is impossible to clearly determine the direction of the hemodynamic response to tactile stimulation due to the complexity of the pattern of simultaneous activation and deactivation of various neural populations [29]. This phenomenon poses a serious methodological problem for the use of NIRS in the context of investigation of brain activity within the framework of the mirror paradigm, especially taking into account the influence of stimulation parameters and the involvement of multiple sensory modalities [30]. It can be assumed that simpler stimulation protocols may be beneficial to further research and explanation of atypical changes in hemodynamic parameters during sensory stimulation with mirror feedback (exclusive use of either a hard item or a soft object).

In this study, we revealed no correlation between the majority of quantitative EEG and NIRS indicators for identical experimental conditions. The correlation between the degree of rhythm desynchronization in the alpha range and an increase in HbO concentration was observed only in the area of the posterior parietal cortex in the context of unimanual movement in front of the mirror. The available literature does not offer other works involving simultaneous registration of electrophysiological and hemodynamic parameters of brain activity during experiments with mirror visual feedback. Further analysis of additional indicators, such as lateralization or connectivity indices, is required. Overall, it can be concluded that in a comprehensive study of the neurophysiological mechanisms of MI, EEG and NIRS can complement but not replace each other.

Individual subjective parameters of MI weakly correlated with the degree of desynchronization of the ipsilateral mu and the activation of the additional motor cortex. However, most of them did not correlate with electrophysiological and hemodynamic parameters of the cerebral cortex activity but showed a significant correlation with the intensity of the emotional reaction to the illusion. The dissociation between objective neurophysiological indicators and subjective experiences of illusion may indicate a multilevel or network-based formation of the illusory experience rather than the activity of isolated cortical regions [18]. The emotional reaction to the illusion can serve as an integrative indicator reflecting not only the perceptual aspects but also the personal characteristics of the participants, their willingness to accept the illusory experience and their general emotional reactivity.

The revealed bilateral activation of primary sensorimotor areas associated with unimanual movement in front of the mirror has a potential clinical significance in the context of optimization of mirror therapy protocols. These findings support the rationale for applying the motor paradigm of mirror therapy in patients with stroke hemiparesis, as it provides activation of motor structures in the ipsilateral hemisphere, which may facilitate restoration of interhemispheric functional balance. The revealed close relationship between the subjective parameters of the mirror illusion and the emotional reaction suggests that the positive emotional response of the patient in the first sessions can be a predictor of the effectiveness of therapy, reflecting the degree of involvement of the patient in the rehabilitation process. The demonstrated possibility of combining EEG and NIRS opens up prospects for the development of hybrid neurobiological control technologies. This enables prediction of treatment efficacy and individualization of rehabilitation programs based on neurophysiological biomarkers, including for patients with contraindications to conventional mirror therapy, such as bilateral amputation with phantom limb pain syndrome.

This study has the following limitations: the participants had no technical means to record the time of the MI onset; there were no control block that would involve unimanual movement without a mirror; the questionnaire collecting perception of MI was a simplified one. However, these limitations do not affect the main conclusions of the study. It should also be noted that in this experiment, it is impossible to completely separate the emotional reaction to MI and the reaction caused by the novelty of an unusual sensory experience: there was only one assessment, the measurements were not taken repeatedly. Some of the recorded emotional states (surprise, interest) may partially reflect an orientation reflex to a new perceptual experience. The differentiation of these components may be the subject of future longitudinal studies.

CONCLUSIONS

Thus, arm movement in front of the mirror is accompanied by bilateral activation of the primary sensorimotor areas of both hemispheres. The correlations between EEG and NIRS indicators were found only in the posterior parietal cortex, which means that EEG and NIRS complement but not replace each other in the study of MI. The subjective characteristics of the illusion correlated with the emotional reaction, and only some of them weakly correlated with neurophysiological indicators, which corresponds to the concept of a multi-level network organization of the mechanisms behind the illusion. In subsequent studies, it is advisable to jointly register EEG and NIRS indicators during mirror therapy in patients with post-amputation phantom pain or post-stroke hemiparesis, as well as to evaluate the dynamics of the identified MI neuromarkers. The data obtained in such studies can be compared with the norm and taken into account when developing EEG and NIRS-backed neurobiocontrol technologies aimed at correcting these pathologies.