In the current study, we examined the modulation of the cortical mechanisms involved in unimanual and bimanual tasks in the presence of cervical tSCS. The choice of studying the M1 and S1 sensorimotor cortical regions was based on the pivotal role of these areas in modulating bimanual performance [39]. We found that beta cortical oscillations associated with left sensorimotor regions were significantly modulated by tSCS during the execution of both unimanual and bimanual common-goal movements, pointing to the increase in synchronous neural firing in M1 and S1 induced by tSCS. In the alpha band however, we observed ERS of sensorimotor cortical activity only during unimanual movement. Our finding demonstrated that there is no significant modulation of interhemispheric connectivity between left and right M1 and S1 when cervical tSCS was applied. Furthermore, our study revealed that cervical tSCS improved performance during the bimanual common-goal task as characterized by MT and RMSE, but had no effect on movement kinematics during the execution of bimanual dual-goal and unimanual tasks. To the best of our knowledge, this is the first study to investigate the neural correlates of three behaviorally distinct unimanual and bimanual tasks under the influence of cervical tSCS using EEG measures. Up until now, knowledge about the effect of tSCS on cortical networks underlying bimanual motor control as well as cortical neurophysiological mechanisms of tSCS was very limited.

A number of studies over the past few years demonstrated that tSCS may be effective in improving sensorimotor function after SCI [14, 23, 40]. These studies used metrics such as spinally-evoked potentials, MEPs [34], cervicomedullary evoked potentials (CMEP) [3], and H-reflexes [1, 37] to investigate the underlying mechanisms of this electrical stimulation neuromodulatory technique. Only one previous study provided information regarding the effect of tSCS on cortical oscillations captured by EEG. McGeady et al. reported that 10 min of cervical tSCS is not sufficient to produce significant modulation of sensorimotor brain rhythm, but this finding was not consistent among all participants and relied on the intensity of stimulation [32]. Participants who received the highest doses of stimulation had suppression of cortical activity (10% ERS), implying that stimulation intensity is a critical factor at cortical level. In line with this view, a crucial finding in our work was a significant suppression of sensorimotor cortical activity in some of the performed tasks. Nonetheless, there are two important differences between our procedure for electrode placement and determining stimulus intensity and the work by McGeady and colleagues. First, we placed two adhesive cathodic electrodes midline at C3–4 and C6–C7 [37], instead of single electrode at C5–6 as was the case in the McGeady et al. study [32]. Second, instead of subjectively setting the current intensity by asking the participants about their maximum tolerance level, we followed the procedure outlined in Benavides et al., in which stimulation intensity was determined based on the threshold that induces spinally-evoked potentials [3]. In addition, we used a stimulation frequency of 40 Hz instead of the 30Hz used in McGeady et al. [32].

This suppression of cortical activity is not surprising in this study. We previously reported unchanged MEPs in the presence of cervical tSCS with 10 kHz modulation [37]. However, a recent study determined that tSCS with 5 kHz carrier frequency facilitated the amplitude of CMEPs but did not modulate the amplitude of MEPs [3], and suggested that tSCS activates cortical inhibitory networks projecting to corticospinal neurons. Interestingly, the facilitation of MEPs in Benavides et al. only happened when the carrier frequency was removed from the stimulation waveform, suggesting that the carrier frequency contributed to the inhibitory mechanism [3]. This effect was further substantiated by an increase in the level of short-interval intracortical inhibition (SICI) only when tSCS was applied with 5 kHz carrier frequency [3]. It has been suggested that the modulation of SICI is mediated by intracortical GABA inhibitory networks [11]. Therefore, with the presence of 10 kHz carrier frequency in our study, it is rational to contemplate that a similar inhibitory intracortical mechanism is responsible for the suppression of motor and sensory cortical activity (i.e., ERS) found in this study.

Alternatively, the ERS may be a consequence of exposure to discomfort caused by stimulation [29, 38]. Participants in our study verbally reported a strong fluttering or vibration sensation at the cathodic sites. The amplitude of stimulation for each participant was also close to maximal tolerance; at this level the participants could not tolerate the stimulation for more than 3–4 min (the duration of the task). Maximal tolerance with tSCS applied laterally across the spinous process between lumbar L1 and L2 vertebrae was shown to be more than 50% lower than the stimulation level required to elicit spinally-evoked potentials [30]. The stimulation amplitude in our study was set similarly (i.e., at the level that induces spinally-evoked potentials), which would have caused experience of discomfort. Moreover, discomfort and painful sensations are associated with reduced ERD during movement [49]. Thus, discomfort experienced by the participants may have contributed to the suppressive effect on cortical activity observed in this study.

Parallels can be drawn from neuromuscular electrical stimulation (NMES) studies. Modulation of brain activation induced by NMES has been reported previously [45, 47]. For example, NMES of wrist extensor induced stimulation intensity dependent modulation of sensorimotor cortical activity, with above motor threshold intensities producing cortical facilitation and below motor threshold intensities causing cortical inhibition [24]. Importantly, motor threshold level in the NMES study is defined as the intensity that induces finger twitches, and at this level proprioceptive receptors as well as cutaneous mechanoreceptors are activated [4, 17]. With below motor threshold stimulation, however, only cutaneous mechanoreceptors are activated [24]. The procedure to determine the amplitude of stimulation ensured that all the participants in our study received stimulation at the level that elicits spinally-evoked potentials. At this level, posterior root afferents are recruited [2, 36]. Thus, we may conclude that tSCS through recruitment of posterior root afferents should produce the same facilitation of ERD observed in the NMES study; however, we found the opposite. This effect may be due to exposure to high-intensity stimulation and activation of intracortical inhibitory networks which could have interfered with the conduction of sensory information [3, 32].

We did not find significant modulation of interhemispheric connectivity in cortical sensorimotor regions. Our results suggest a trend towards increased beta band interhemispheric connectivity between left and right M1when tSCS was delivered relative to when tSCS was off across all movement conditions, but the opposite of this trend was seen between the left and right S1 (i.e., decrease of interhemispheric connectivity). No particular trend was observed in the alpha band interhemispheric connectivity results. A recent study suggested both a decrease (in areas associated with direct motor control) and an increase (in areas of motor planning) of functional connectivity in the presence of lumbar tSCS during tonic and rhythmic muscle contraction of the lower limbs [48]. However, this effect was only observed at the level of cortical sources, and was absent for the EEG electrode-based analysis [48]. Similarly, NMES has been shown to strengthen interhemispheric functional connectivity between cortical sensorimotor regions [18]. Modulation of interhemispheric inhibition may explain increased/decreased functional connectivity between sensorimotor regions [7, 18]. The current investigation is unable to identify specific neural pathways or regions responsible for changes in the level of interhemispheric connectivity. Future investigation is needed using measures such as fMRI-based functional connectivity and TMS-based IHI to further explore tSCS-induced modulation of interhemispheric connectivity. Moreover, as suggested by Steele et al. [48], connectivity analysis is more accurate when performed at cortical source levels as opposed to when sensor-based information is employed.

Stimulation applied to regions near EEG electrodes is considered a major source of artifacts in the data and complicates the interpretation of the results. To alleviate the effect of stimulation artifacts in the EEG recordings, two approaches including artifact removal and inter-stimulus data extraction have been suggested previously [28]. The limitation of these approaches is the exclusion of brain data in the analysis during stimulation. To overcome this challenge, a recent study suggested that EEG during tSCS “bares statistically similar characteristics to that of normal EEG” if the frequency band of interest does not overlap with stimulation frequency [32, 33]. Therefore, no artifact removal techniques were thought to be necessary, but notch filtering was recommended in the frequency domain. In this study, we followed the procedure outlined in [32], which only involves applying a band-pass filter between 0.1 and 200 Hz. Additionally, we suppressed the tSCS-induced contamination of EEG data in the time domain by replacing artifacts with an average of clean EEG signal. This additional step was necessary because we observed the signs of stimulation artifact spreading beyond its frequency to nearby frequency bands (i.e., alpha and beta) in the spectral power analysis. This was evident as an abnormal brief peak near 20Hz in Fig. 5D. However, we acknowledge that removing parts of the EEG signal (the artifacts) and replacing them with average values, effectively alters the original dynamics of the signal during those intervals. This means removing not just the artifact but also deprivation of results from genuine brain signals during the stimulation period. The artifact removal approach led to a substantial reduction in both alpha and beta band tSCS-induced sensorimotor ERS relative to when only band-pass filtering was applied. This reduction of ERS demonstrates that retaining the simulation artifact in the data comes at the cost of exaggerated ERS, and hence misinterpretation of the results. Additionally, when a period of an EEG signal is flattened, this introduces discontinuities in the signal, which can manifest as spectral leakage in the frequency domain. Spectral leakage can lead to power being spread across different frequencies, which might influence spectral power estimates, not just at 40 Hz but to a broad range of frequencies. This is also a valid concern even if we used total removal of artifacts instead of replacing them with an average. The effect of artifact removal affects the immediate nearby frequencies, but as we move farther away from 40 Hz, the effect becomes less pronounced. Therefore, the selection of 40 Hz for the stimulation frequency aimed to separate the stimulation frequency from the frequency bands of interest, such as the beta band. Finally, we applied a windowing function before spectral power computation to limit the spectral leakage.

In this study, pulse duration was set to 1ms but the stimulation artifact recorded by EEG persisted for approximately 7–11ms. In other words, ~ 28–45% of the EEG data were replaced with an average of the clean EEG when the artifact removal approach was used. Thus, these two artifact removal approaches create a trade-off between the possibility of inaccurate frequency domain results and data loss. Notably, the type of tSCS stimulator used affects the duration of the resulting artifact. Using two DS8R stimulators for the cathodes in this study resulted in a relatively narrower artifact in the EEG recordings because the stimulus pulses were delivered simultaneously by the two stimulators. Delivering the stimulus pulses sequentially as is the case with other tSCS stimulators would increase the duration of the artifact to nearly twice the duration seen in our data, rendering any analysis of EEG activity virtually impossible.

We speculate that there are two underlying reasons contributing to the discrepancy between our view of handling stimulation artifact and what was suggested in McGeady et al. [32, 33]. First, stimulation frequency was 30 Hz in the McGeady et al. study [32, 33] compared to 40 Hz in this study. This means that the inter-stimulus interval was wider in the previous study allowing for ~ 25% higher amount of clean and useful data for frequency-domain analyses. Second, delivering stimulation through two cervical electrodes may have caused the pronounced stimulation artifact in the time series EEG data, which led to having only ~ 55–72% clean EEG signals between successive stimulation pulses. It is perhaps the case that with lower stimulation frequencies, tSCS presents no threat to frequency domain analyses as suggested by McGeady et al. [33], but artifact removal is required at higher frequencies. Future research is necessary to explore whether EEG recordings are feasible with tSCS at different stimulation frequencies, especially to assess how artifacts affect spectral power in frequency bands of interest such as alpha, beta, and gamma, as well as interhemispheric connectivity. EEG activity contaminated with stimulation artifacts can provide misleading representations in the frequency domain [48]. Thus, we believe that artifact removal is necessary, at least for higher stimulation frequencies such as 40 Hz.

Importantly, our results suggest that tSCS improves MT and RMSE of bimanual common-goal movements in participants with no history of neural injury or disease, but is ineffective in improving bimanual dual-goal and unimanual tasks. It has been previously shown than tSCS primarily activates afferent fibers of the dorsal roots and dorsal horn of the spinal cord [2, 15]. Through monosynaptic and oligosynaptic connections from sensory afferents, spinal α-motoneurons are recruited [9, 20]. We speculate that there is an increase in the transmission of proprioceptive information that enhanced the performance of the bimanual common-goal task. Successful performance of common-goal reaching movements requires extensive coordination between the two arms and constant sharing of spatial location between the two arms [12]. A recent study [10] investigated the activation of proprioceptive fibers during cervical transcutaneous spinal cord stimulation. This computational study suggests preferential activation of both Aα and Aβ fibers compared to α-motor fibers. This finding shows the contribution of both proprioceptive and cutaneous input to tSCS-evoked potentials. While this does not directly translate to enhanced proprioceptive function, we speculate that preferential activation of proprioceptive fibers may have contributed to better performance of the bimanual common-goal task which relies on proprioception more than the other two tasks. Therefore, it is plausible that the increased proprioceptive input during cervical tSCS contributed to improved kinematics particularly during bimanual common-goal movements.

Although improvements in unimanual hand and arm function were previously reported with tSCS after SCI [23], the reason improvements in kinematic performance were not seen during the unimanual and dual-goal tasks in this study is likely because the participants were neurologically intact. Nonetheless, our current findings critically highlight the potential of tSCS in promoting recovery of bimanual movements after neurological conditions. If tSCS is capable of improving movement accuracy and movement time in participants with no history of neural injuries/diseases, it is possible that kinematic outcomes can be improved for participants with SCI or stroke. We posit that this behavioral improvement can be achieved through hybrid rehabilitation training that consists of bimanual coordination tasks and tSCS. Moreover, our findings highlight the importance of comprehensively and accurately assessing bimanual impairments and quantifying bimanual performance after SCI/stroke. Stroke survivors exhibit varying performance levels when engaged in different bimanual movements [26]. Elucidating what aspect of bimanual movements is primarily targeted by tSCS in participants with neurological conditions is a question for future studies.

If the kinematic performance of common-goal movements is improved with tSCS in people with SCI or stroke, analogous to what the present study found, our results could serve to inform the optimal bimanual rehabilitation training design. Our task design offers a precise and sensitive measure for kinematic analyses of arm function before and after rehabilitation training. At this time, it is difficult to make a conclusion about the link between cortical synchronization and potential behavioral improvement caused by tSCS when tested in a clinical population. The key is to track the changes in the level of cortical ERS (or ERD) during the course of a tSCS-based bimanual rehabilitation training paradigm and correlate it with behavioral improvements. Our study suggests sensorimotor cortical inhibition when tSCS is applied with a modulated waveform (i.e., 10 kHz waveform). A previous study suggested stronger corticospinal excitability after the application of tSCS for 20 min only when the kHz modulation was removed [3]. Another study found no significant changes in corticospinal excitability with stimulation at 40% of the posterior root reflex (PRR) threshold, but found modulation of MEPs at 60% and 80% threshold [31]. Taken together, our present work and these two previous studies demonstrate that cervical tSCS using a 10 kHz carrier frequency influences cortical activation and excitability. While it is not possible to draw definitive conclusions from the results of this study and previous work at this time, it seems that tSCS, when administered at a sufficient intensity level tailored to the specific stimulation paradigm, can increase corticospinal excitability. Moreover, it may be the case that a non-modulated tSCS waveform that does not cause cortical inhibition and leads to stronger corticospinal excitability is more beneficial for improving upper limb function, especially performance of bimanual common-goal movement, after neural injury.