Projection neurons wire the brain over long distances and provide a network between different brain regions. In particular, both hemispheres are structurally connected by transcallosal projections and exhibit functional cross talks (Hofer and Frahm, 2006; Meyer et al., 1995); such interhemispheric interaction is essential for higher order cognitive and sensorimotor brain functions. An early argument for interhemispheric interaction was that dynamic interplay via the callosum not only allows for simple coordination of processing between the hemispheres but also has profound effects on attentional functioning (Banich, 1998). In the motor domain, neurons in the monkey primary motor cortex are firing bilaterally and motor signals from the two hemispheres interact during unimanual motor tasks (Ames and Churchland, 2019); in the human motor cortex, measuring these neural activities via electroencephalography (EEG) and paired-pulse transcranial magnetic stimulation (TMS) recently revealed that the two hemispheres act together and the related cortical oscillatory activity influences the inhibitory interhemispheric brain network (Picazio et al., 2014; Stefanou et al., 2018).

A critical issue is whether a specific interhemispheric activity can be volitionally controlled. Previous studies have focused on inhibitory interhemispheric sensorimotor network, evaluated by interhemispheric inhibition (IHI), from movement-related manner (Duque et al., 2007; Duque et al., 2005; Liang et al., 2014; Morishita et al., 2012; Murase et al., 2004; Nelson et al., 2009), and passive neuromodulation effects by non-invasive brain stimulation techniques such as transcranial direct current stimulation (tDCS) or repetitive TMS (rTMS) (Boddington and Reynolds, 2017; Gilio et al., 2003; Williams et al., 2010). Due to the experimental limitations related to these observational or open-loop paradigms, much less is known about the effects of changes in sensorimotor activity patterns in both hemispheres on IHI, and the possibility of manipulating IHI. Therefore, the closed-loop paradigms to manipulate bilateral sensorimotor activity patterns and IHI should be of great value for the understanding of human sensorimotor brain function (Figure 1A).

Figure 1

Download asset Open asset

As a recent neural manipulative tool, numerous studies have used a brain-computer interface (BCI)-based neurofeedback whereby participants learn to volitionally desynchronize and synchronize oscillatory sensorimotor rhythms (SMR-ERD/ERS) in the contralateral sensorimotor cortex (SM1) through visual and/or somatosensory feedback (Ang and Guan, 2017; Chaudhary et al., 2016; Ramos-Murguialday et al., 2013; Soekadar et al., 2015b). BCI-based neurofeedback is supported by the fact that the intensity of SMR-ERD in EEG represents states of high versus low excitability of not only the SM1 (Neuper et al., 2006; Neuper and Pfurtscheller, 2001; Pfurtscheller et al., 2006), but also the corticospinal descending pathway, as measured by the motor-evoked potential (MEP) amplitude (Takemi et al., 2015; Takemi et al., 2013). Because both hemispheres are structurally and functionally connected, it is likely that the balance of bilateral SMR-ERDs and transcallosal excitability states are linked, as indicated by common variation of conditioning MEP amplitude and IHI (Ferbert et al., 1992; Ghosh et al., 2013; Ni et al., 2009).

Here, we sought to investigate the association of IHI and the balance of SMR-ERDs to understand the inhibitory sensorimotor functions of interhemispheric interaction that may critically depend on the oscillatory brain activity in both hemispheres. Furthermore, if changes in the oscillatory brain activity influenced IHI, we tested whether it can be manipulated using a dual-coil paired-pulse TMS protocol (Daskalakis et al., 2002; Ferbert et al., 1992). To assess the association of IHI with ongoing oscillatory brain activity, we used a recently developed spatially bivariate BCI-based neurofeedback technique that allows volitional modulation of SMR amplitude in both hemispheres (Hayashi et al., 2021; Hayashi et al., 2020) and triggers the TMS pulses in real time at predetermined bilateral SMR amplitudes. Our system, therefore, overcomes previously unresolved experimental limitations of the ordinary observational or open-loop experimental paradigm. That is, our paradigm enables participants to control the bivariate sensorimotor excitability, and determines whether it is possible to manipulate the IHI magnitude.

Using the novel closed-loop bivariate (bi-hemispheric brain state-dependent) EEG-triggered dual-coil paired-pulse TMS system (bi-EEG-triggered dual-TMS), we evaluated the effects of different states of the targeted bidirectional up- or down-regulated one-sided hemisphere on effective inhibitory interhemispheric network expressed by IHI: (1) resting-state (REST), (2) during right finger motor imagery (MI) without visual feedback (NoFB), (3) high (HIGH), (4) middle (MID), and (5) low (LOW) excitability states of the ipsilateral SM1 (conditioning side) to the unilateral imagined hand movement while maintaining constant contralateral excitability during BCI-based neurofeedback (Figure 1B). Inspired by the findings that SMR-ERD was associated with corticospinal excitability amplitude (Takemi et al., 2015; Takemi et al., 2013), we hypothesized that the ipsilateral SMR-ERD to the imagined (right) hand is a potent up- or down-regulator of IHI from the ipsilateral (right) to the contralateral (left) hemisphere. This multimodal research provides strong evidence for the dynamic interplay between distinct regions underlying IHI through BCI-based neurofeedback, and may therefore form the basis for interhemispheric sensorimotor activity.

Values corresponding to each statistical figure are presented in each table (Supplementary file 1) for readability.

Modulation effects of bilateral SM1 at EEG level

We tested whether participants could learn volitional up- or down-regulation of the ipsilateral sensorimotor excitability to the imagined right hand while maintaining constant contralateral sensorimotor excitability at the EEG level using a spatially bivariate BCI-based neurofeedback. For the contralateral SMR-ERD, a one-way rmANOVA for the sessions (five levels: REST, NoFB, HIGH, MID, and LOW) revealed a significant difference (F_(4,102) = 4.81, p=0.014, and η²=0.16). Post-hoc two-tailed paired t-tests demonstrated no significant differences in the contralateral SMR-ERD between the neurofeedback sessions (HIGH-MID: Cohen’s d=0.13, p=1.00; HIGH-LOW: Cohen’s d=0.22, p=1.00; MID-LOW: Cohen’s d=0.11, p=1.00), whereas significant differences were found between REST and other sessions (all p<0.05; Figure 2A). In contrast, after showing significant differences in the ipsilateral SMR-ERD between sessions (F_(4,102) = 156.0, p<0.001, and η²=0.86), post-hoc two-tailed paired t-tests showed that ipsilateral SMR-ERD increased during the HIGH session (HIGH-MID: Cohen’s d=3.12, p<0.001; HIGH-LOW: Cohen’s d=6.13, p<0.001) and decreased during the LOW session (MID-LOW: Cohen’s d=4.44, p<0.001), revealing a significant bidirectional modulation compared to baseline sensorimotor endogenous activity (Figure 2B). The spatial patterns of SMR-ERD in each session are depicted in Figure 2C. We found that the SMR-ERDs were predominantly localized in bilateral parieto-temporal regions (C3 and C4 channels and their periphery), and strong ipsilateral SMR-ERD was observed in the HIGH session with constant contralateral SMR-ERD. Modulation effects of the contralateral and ipsilateral excitabilities (power attenuation) in the theta (4–7 Hz), low beta (14–20 Hz), high beta (21–30 Hz), and gamma (31–50 Hz) bands were shown in Figure 2—figure supplement 1 (see also Supplementary file 1–Table 2).

Figure 2 with 1 supplement see all

Download asset Open asset

IHI curves at rest

To validate IHI measurement under bi-EEG-triggered dual-TMS setup, IHI curves were obtained at rest (Figure 3A) in 20 out of 24 participants, with conditioning stimulus (CS) of varying intensity (five different intensities, 100–140% of resting motor threshold [RMT], in steps of 10% RMT). A one-way rmANOVA for intensities (six levels: 0 [single test stimulus: TS-only], 100, 110, 120, 130, and 140% RMT) revealed significant difference in intensity (F_(5,109) = 8.31, p<0.001, η²=0.28). Post-hoc two-tailed paired t-tests showed significant difference between TS-only and 100–140% of RMT (TS-only versus 100% RMT: p=0.039, TS-only versus 110–140% RMT: all p<0.001). There were no significant differences across CS intensities, while the size of MEP amplitude tended to be smaller for larger CS intensities. The IHI was approximately half of the maximum when the CS intensity was approximately 130% of RMT, which was compatible with a previous EEG-TMS experiment (Stefanou et al., 2018; Tsutsumi et al., 2012). In addition, the stimulus intensities of the left and right M1 to evoke MEP of 1 mV peak-to-peak amplitude from the relaxed right and left first dorsal interosseous (FDI) muscles (SI_1mV) were determined and used for the following dual-coil paired-pulse TMS setup. The average SI_1mV of TS and CS were 124 ± 11% RMT and 132 ± 13% RMT, respectively. We also confirmed that the SI_1mV of CS showed approximately 50% of the mean conditioned MEP over the mean unconditioned test MEP (Figure 3B).

Figure 3

Download asset Open asset

IHI manipulation via spatially bivariate BCI-based neurofeedback

Typical examples of MEP amplitude elicited by single TS (TS-only) and paired-pulse stimulation (CS + TS) of a representative participant are shown in Figure 4A. The manipulation range of IHI (i.e. the difference between HIGH and LOW sessions) was 32.6 ± 30.7% (Cohen’s d=1.50) in all participants excluding 2 participants with data corruption (i.e. 22 participants). A one-way rmANOVA of the sessions (five levels: REST, NoFB, HIGH, MID, and LOW) revealed significant differences in IHI magnitude (F_(4,101) = 6.85, p<0.001, η²=0.22). Across the three BCI-based neurofeedback sessions (i.e. HIGH, MID, and LOW sessions) for the comparison of IHI magnitude, post-hoc two-tailed paired t-tests showed significant difference between HIGH and MID sessions (Cohen’s d=0.94, p=0.025), and between HIGH and LOW sessions (Cohen’s d=1.36, p<0.001), but not between MID and LOW sessions (Cohen’s d=0.42, p=0.424; Figure 4B). In addition, we found disinhibition in NoFB session compared to REST session, although there was no statistical difference (Cohen’s d=0.42, p=0.279); this trend is consistent with the previous studies, which reported that in healthy participants, IHI targeting the moving index finger leads to disinhibition before the movement onset to produce voluntary movement (Duque et al., 2005; Murase et al., 2004) and during isometric contraction (Nelson et al., 2009). Importantly, rmANOVA for the MEP amplitudes induced by TS-only revealed no significant differences between all sessions (F_(4,101) = 2.44, p=0.104, η²=0.09), suggesting that participants could learn to volitionally regulate the ipsilateral sensorimotor excitability while maintaining constant contralateral sensorimotor excitability (Figure 4C). A linear mixed-effect model that considered the contralateral SMR-ERD and ipsilateral SMR-ERD as fixed effect in the statistical regression model, and participants as random effect revealed a significant effect of the ipsilateral SMR-ERD (p<0.001) on IHI magnitude, but no main effect of the contralateral SMR-ERD or contralateral SMR-ERD × ipsilateral SMR-ERD interaction (p=0.828 and p=0.058, respectively).

Figure 4 with 4 supplements see all

Download asset Open asset

To examine the IHI magnitude during neurofeedback, we further compared the IHI magnitude across sessions by calculating the percentage changes in IHI magnitude to baseline (i.e. NoFB session), and investigating the difference between REST, HIGH, MID, and LOW sessions. A one-way rmANOVA for sessions (four levels: REST, HIGH, MID, and LOW) revealed significant differences (F_(3,81) = 8.32, p<0.001, η²=0.24). Post-hoc two-tailed paired t-tests showed significant differences between HIGH and MID sessions (Cohen’s d=1.06, p=0.017), and between HIGH and LOW sessions (Cohen’s d=1.42, p<0.001), but not between MID and LOW sessions (Cohen’s d=0.45, p=0.734; Figure 4—figure supplement 1). The normalized results were compatible with the results without normalization presented in Figure 4.

To further examine the influence of spontaneous SMR fluctuations on IHI, non-triggered TMS trial (referred to as the failed trial) was delivered in random timing ranging from 5.5 to 6 s during the MI epoch. For example, data from HIGH session in Figure 4—figure supplement 2 were collected from MID and LOW sessions, with the target ranges of SMR-ERD in HIGH session (Figure 4—figure supplement 2). A two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and trials (two levels: triggered TMS trials and non-triggered TMS trials) revealed a significant main effect for sessions (F_{(2, 117)} = 5.08, p=0.008, and η²=0.08), but no main effect for trials (F_{(1, 117)} = 0.32, p=0.575, and η²<0.01) and for interaction (F_{(2, 117)} = 1.82, p=0.167, and η²=0.03). Thus, no significant difference in IHI between the volitional control of SMR-ERDs in the closed-loop environment differs from spontaneous SMR fluctuations was observed. Post-hoc two-tailed paired t-tests showed significant differences between the HIGH and MID sessions (Cohen’s d=0.94, p=0.017), and between HIGH and LOW sessions (Cohen’s d=1.36, p<0.001) in the triggered TMS trials. In contrast, in the non-triggered TMS trials, post-hoc two-tailed paired t-tests showed no significant differences between HIGH and MID sessions (Cohen’s d=0.20, p=1.00), and between HIGH and LOW sessions (Cohen’s d=0.28, p=1.00).

Although we revealed that modulating the excitability of the right hemisphere changes IHI magnitude from the right to the left hemisphere, the specificity of the IHI manipulation has not been proved yet. We presented additional results related to changes in corticospinal excitability as assessed by MEPs in the left FDI in response to the CS over the right hemisphere (Figure 4—figure supplement 3B). To investigate whether changes in IHI from the right to the left hemisphere and changes in corticospinal excitability in the right hemisphere are independent phenomena, a generalized linear model with the percentage changes of left MEP amplitude as a covariate of no interest was applied. This approach tests whether the effect of IHI manipulation survives even if the covariate is added. During the triggered TMS trials, a generalized linear model for sessions (three levels: HIGH, MID, and LOW) revealed a significant main effect for sessions (F_{(3, 60)} = 12.75, p<0.001, and η²=0.44). Post-hoc two-tailed paired t-tests in the increase in IHI showed significant differences between HIGH and MID sessions (Cohen’s d=1.06, p<0.001), and between HIGH and LOW sessions (Cohen’s d=1.42, p<0.001). During the non-triggered TMS trials, a generalized linear model for sessions (three levels: HIGH, MID, and LOW) revealed no main effect for sessions (F_{(3, 60)} = 1.13, p=0.319, and η²=0.10). Therefore, we successfully demonstrated that IHI changes greater than the variance explained by the CS effect and manipulation is not simply an epiphenomenon.

In the control muscle (abductor digiti minimi, ADM), such observed modulation was not driven. A one-way rmANOVA for sessions (five levels: REST, NoFB, HIGH, MID, and LOW) revealed significant differences (F_(4,101) = 2.51, p<0.048, and η²=0.12). Across the three BCI-based neurofeedback sessions (i.e. HIGH, MID, and LOW sessions) for the comparison of IHI magnitude, post-hoc two-tailed paired t-tests showed no significant difference between sessions (HIGH-MID: Cohen’s d=0.10, p=1.00; HIGH-LOW: Cohen’s d=0.05, p=1.00; MID-LOW: Cohen’s d=0.15, p=1.00), whereas significant difference was observed between REST and NoFB sessions (Cohen’s d=0.45, p=0.024; Figure 4—figure supplement 4A). Statistical analysis for TS-only revealed no significant differences in MEP amplitude between the sessions (F_(4,101) = 0.62, p=0.649, and η²=0.03; REST: 0.54 ± 0.51 mV, NoFB: 0.50 ± 0.39 mV, HIGH: 0.73 ± 0.61 mV, MID: 0.59 ± 0.50 mV, and LOW: 0.65 ± 0.51 mV; Figure 4—figure supplement 4B), indicating that IHI manipulation occurred only in the targeted muscle.

Associations between IHI magnitude and bilateral EEG patterns

To examine the association between IHI magnitude and bilateral EEG patterns, we performed within- and across-participant correlation analyses, respectively. In the within-participant correlation analysis between IHI magnitude and bilateral SMR-ERDs in each participant, 7 of the 22 participants showed a significant correlation between IHI magnitude and ipsilateral SMR-ERD (r = –0.307 ± 0.252 [mean ± SD]), but not between IHI magnitude and contralateral SMR-ERD in three neurofeedback sessions (r = –0.059 ± 0.298 [mean ± SD]). A chi-squared test was performed to determine the overall significance of the 22 individual correlations between IHI magnitude and contralateral or ipsilateral SMR-ERDs. A comparison of the proportions of significant results obtained from the ipsilateral (7 out of 22 participants) and contralateral (1 out of 22 participants) SMR-ERDs with the proportion expected due to chance (i.e. alpha level = 0.05) showed statistically significant differences (chi squared value = 5.50, p=0.019 [Fisher’s exact test: p=0.046]). The distributions of correlation coefficients across all participants were shown in Figure 5A.

Figure 5 with 1 supplement see all

Download asset Open asset

In the across-participant correlation between IHI magnitude and contralateral or ipsilateral SMR-ERDs, we performed a repeated measures correlation analysis (Bakdash and Marusich, 2017). We found a significant correlation in the ipsilateral SMR-ERD (r_rm=–0.436, p=0.004; Figure 5B), but not in the contralateral SMR-ERD (r_rm=0.008, p=0.960; Figure 5C).

To analyze the associations between IHI magnitude and bilateral neural network from another perspective, interhemispheric functional connectivity during MI was examined. Interhemispheric functional connectivity was quantified as a Network-intensity measure (Hayashi et al., 2020), which is calculated from the corrected imaginary part of coherence (ciCOH) (Ewald et al., 2012; Vukelić and Gharabaghi, 2015). For the Network-intensity_MI, a one-way rmANOVA for the sessions (five levels: REST, NoFB, HIGH, MID, and LOW) revealed significant differences (F_(4,102) = 3.04, p=0.020, and η²=0.10). Post-hoc two-tailed paired t-tests in all five sessions demonstrated significant differences between HIGH and MID sessions (Cohen’s d=0.83, p=0.033; Figure 5D). We found no significant difference between HIGH and LOW sessions (Cohen’s d=0.40, p=1.00; Figure 5D), suggesting that interhemispheric functional connectivity did not reflect the excitatory or inhibitory activity. Associations of IHI magnitude and contralateral or ipsilateral EEG patterns in the theta (4–7 Hz), low beta (14–20 Hz), high beta (21–30 Hz), and gamma (31–50 Hz) bands were shown in Figure 5—figure supplement 1 (see also Supplementary file 1–Table 10).

Individual characteristics associated with the manipulation capability of IHI

Finally, we investigated the neural characteristics associated with the manipulation capability of IHI calculated from the percentage change between HIGH and LOW sessions using correlation-based analysis (Figure 6A). In the relationships between the manipulation capability of IHI and IHI magnitude in REST session (IHI_rest), we found a significant correlation (r = –0.447, p=0.044), indicating that participants with greater IHI at rest were able to strongly manipulate the IHI (Figure 6B). For the relationships between resting-state effective inhibitory interhemispheric network assessed by IHI_rest and interhemispheric functional connectivity at the EEG level (Network-intensity_rest), participants with greater IHI_rest showed larger Network-intensity_rest (r=0.547, p=0.013; Figure 6C). Furthermore, we verified whether IHI_rest may be associated with intrinsic EEG profiles in NoFB sessions including bilateral SMR-ERDs. We found a significant correlation between IHI_rest and ipsilateral SMR-ERD during MI (r = –0.619, p=0.004), but not between IHI_rest and contralateral SMR-ERD during MI (r = –0.283, p=0.228).

Figure 6 with 1 supplement see all

Download asset Open asset

In addition to the alpha band, the beta band is also a well-established EEG signature of motor execution and imagery (Crone et al., 1998; Pfurtscheller, 2001). To verify whether resting-state functional connectivity in other frequency bands (i.e. outside of feedback frequency) was associated with IHI, an across-participant Pearson’s correlation was calculated. In the across-participant correlations between IHI_rest and Network-intensity_rest in the theta (4–7 Hz), low beta (14–20 Hz), high beta (21–30 Hz), and gamma (31–50 Hz) bands, we found significant correlation in the high beta (r=–0.618, p=0.004), but not in theta (r=–0303, p=0.194), low beta (r=–0.229, p=0.331), and gamma (r=–0.390, p=0.089) bands (Figure 6—figure supplement 1A). To attenuate the aberrant effect of values in some especially low Network-intensity_rest or high (CS + TS)/TS at rest (outliers) in the theta, low-beta, high-beta, and gamma bands, the jackknife correlation and the bias were calculated (theta: r=–0.207, bias = 2.02; low-beta: r=–0.121, bias = 3.65; high-beta: r=–0.528, bias = –4.09; gamma: r=–0.310, bias = 0.06). In addition, the correlation between IHI_rest and the contralateral or ipsilateral EEG patterns in frequency bands outside of a target alpha band were in Figure 6—figure supplement 1B and C, respectively.

In the present study, we aimed to determine whether it is possible to manipulate the IHI magnitude and uncover related neural activity behind IHI modulation, while controlling bilateral SMR-ERDs via spatially bivariate BCI-based neurofeedback. This was the first study to show IHI-state manipulation with a large dynamic range induced by volitional variation in ipsilateral sensorimotor excitability expressed by SMR-ERD. In addition, resting-state interhemispheric networks at the TMS (IHI_rest) and EEG (interhemispheric Network-intensity_rest) levels were associated with the manipulation capability of IHI magnitude.

Previous neurofeedback studies have demonstrated that it is possible to gain voluntary control over central nervous system activity without externally administered interventions, if appropriate neurofeedback is embedded in a reinforcement learning task, such as food rewards for animals (Engelhard et al., 2013; Fetz, 2013) and visually rewarding stimuli for humans (Thompson et al., 2009). However, conventional BCI-based neurofeedback of the SMR signal from one hemisphere does not guarantee spatial specific activation of the sensorimotor region in the targeted hemisphere (Buch et al., 2008; Caria et al., 2011; Soekadar et al., 2015a) because sensorimotor activities in the left and right hemispheres potentially influence one another, making IHI manipulation difficult. Using spatially bivariate BCI-based neurofeedback enables participants to volitionally increase or decrease (bidirectional) the ipsilateral sensorimotor excitability, while maintaining constant contralateral sensorimotor excitability. Ipsilateral SMR-ERD to the imagined hand (but not contralateral), reflecting IHI magnitude (Figure 2B) and their significant correlation (Figure 5C), were explained by previous studies (Haegens et al., 2011; Khademi et al., 2018; Kraus et al., 2016; Madsen et al., 2019; Naros et al., 2020; Ros et al., 2010; Sauseng et al., 2009; Takemi et al., 2013; Thies et al., 2018; Zarkowski et al., 2006). The in vivo cortical recordings in monkeys revealed that pericentral alpha power was inversely related with the normalized firing rate in the sensorimotor regions (Haegens et al., 2011). In humans, several studies using single-pulse TMS of the motor hand area combined with EEG tested how ongoing pericentral oscillatory activity impacts corticomotor excitability reflected by the MEP amplitude. These EEG-TMS studies found a negative relationship between pre-stimulus alpha power and MEP amplitude through both offline (Sauseng et al., 2009; Takemi et al., 2013; Zarkowski et al., 2006) and online (Madsen et al., 2019) approaches, whereas the opposite of that was also found (Thies et al., 2018). Furthermore, another EEG-TMS study using neurofeedback demonstrated that the endogenous suppression of alpha rhythms at rest can produce robust increase in MEP amplitude and decrease in short-interval intracortical inhibition (Ros et al., 2010). Similarly, in the beta band, some evidences were revealed that a negative relationship between beta power and MEP amplitudes (Khademi et al., 2018; Kraus et al., 2016; Naros et al., 2020). As for IHI through the transcallosal fiber, it is predominantly regulated through direct postsynaptic mechanisms in the apical dendritic shafts of pyramidal neurons and a specific cortical microcircuitry mediated by dendritic GABA_B receptors in inhibitory interneurons (Palmer et al., 2012). Additionally, consistent with this phenomenon, down-regulation of ipsilateral SMR-ERD, e.g., may influence the excitation of the ipsilateral transcallosal pyramidal neuron followed by disinhibition of the contralateral inhibitory interneurons during unilateral upper limb MI. Therefore, pronounced changes in cortical mechanism despite the absence of sensory input and constant MEP amplitude in the imagined right hand elicited by single TS suggest that the modulation of IHI magnitude from the ipsilateral (right) to the contralateral (left) hemisphere was at least partly of cortical, rather than spinal origin. These neural mechanisms can be explored further by using triple pulse procedures (Ni et al., 2011) in which (dis)inhibition can be directly measured as the modulation of short-interval intracortical inhibition.

The potential contribution of resting-state IHI magnitude and interhemispheric functional connectivity in the alpha and high beta bands to the current results of manipulation capability in IHI (Figure 6B and C and Figure 6—figure supplement 1A) can be speculated by considering the previous studies. Inter-regional communication is accompanied by synchronized oscillations in different brain regions (Fries, 2005; Varela et al., 2001), and this synchronization can be evaluated by functional connectivity. Because both hemispheres are structurally connected by transcallosal projection and exhibit functional cross talks (Hofer and Frahm, 2006; Meyer et al., 1995), the manipulation capability of IHI is likely associated with a structural connectivity as well as local oscillatory power entrainment. As for frequency band outside of neurofeedback target, it is well known that the intensity of SMR-ERD in the beta band reflects the sensorimotor cortical excitability and corticomuscular activation (Hussain et al., 2019; Schulz et al., 2014). This multimodal EEG-TMS approach proves that the bilateral alpha and beta activities observed in the current study served to modulate the functional interhemispheric interaction over motor cortices.

For the other modalities, e.g., real-time fMRI-based neurofeedback to a single region of interest (Sitaram et al., 2017; Weiskopf et al., 2004) or inter-regional functional connectivity (Liew et al., 2016; Pereira et al., 2019) can be used for volitional modulation of neuronal connectivity and could serve as a possible therapeutic tool for motor or cognitive training in diseases related to impaired inter-regional connectivity. However, even fMRI-based neurofeedback with high spatial resolution, based on blood oxygenation level-dependent signal, cannot easily distinguish between the excitatory and the inhibitory activities from each region (Moon et al., 2021). Additionally, neurofeedback using interhemispheric functional connectivity has the possibility to counteract the modulation of IHIs from the right to left hemisphere and from the left to right hemisphere. Our results showed no significant difference in Network-intensity during MI between HIGH and LOW sessions (Figure 5D), indicating that interhemispheric functional connectivity at the EEG level was not changed, whereas IHI was modulated. This may indicate that IHI and Network-intensity simply reflect the activity of different neural populations; interhemispheric projections act via surround/lateral inhibition in the sensory and motor cortices (Carson, 2020), while the synchrony at the EEG level reflects changes in the activity of local interactions between pyramidal neurons and interneurons in the thalamocortical loops (Pfurtscheller and Lopes da Silva, 1999). Furthermore, it is possible that Network-intensity reflects the degree of bidirectional synchrony between the two motor cortices, while IHI reflects one-way (right-to-left) inhibitory effects. Therefore, it is unclear whether fMRI-based neurofeedback using interhemispheric functional connectivity is linked to ipsilateral excitability, and whether it modulates IHI magnitude.

Interhemispheric activity is also passively modulated by externally administered interventions, e.g., tDCS or rTMS over the motor cortices (Gilio et al., 2003; Pena-Gomez et al., 2012; Williams et al., 2010). Although such tools have neuromodulation efficacy, their long-term sustained effects are often limited and they do not have spatial specificity due to the remote effects (Di Pino et al., 2014; Notturno et al., 2014; Weiskopf et al., 2004). Conversely, under the conscious selflearning environment, volitional control of MEP amplitudes is retained for at least 6 months without further training (Ruddy et al., 2018), which supports the future prediction of long-term efficacy of IHI manipulation. In addition, we successfully demonstrated that IHI changes greater than the variance explained by the CS effect. Based on our findings, further experiments may help to better understand the pathway specificity. To further confirm whether the participants learned pathway-specific IHI manipulation, the CS effect should be kept constant across the conditions by adjusting the CS intensity to evoke left MEP of 1 mV peak-to-peak amplitude. Furthermore, the effect size of IHI manipulation in the current study (Cohen’s d=1.50) was comparable with that of representative tDCS (Cohen’s d=1.55) (Williams et al., 2010) and superior to that of rTMS studies (Gilio et al., 2003) (Cohen’s d=0.80; note that it is a read from the graph). Two meta-analyses of clinical trials using rTMS and tDCS also indicated that the effect sizes were medium (0.4–0.6) (Adeyemo et al., 2012; Hsu et al., 2012). Based on a previous study of neurofeedback training combined with externally administered interventions (Ang et al., 2015), their combination may contribute to facilitating neural plasticity.

Our techniques and evidences are expected to be applied in various fields, e.g., in the context of neurorehabilitation. In stroke patients, the interhemispheric imbalance/competition model predicts the presence of asymmetry in the interhemispheric sensorimotor network, with excessive inhibition from the non-affected hemisphere limiting motor performance and maximal recovery (Duque et al., 2005; Murase et al., 2004). Therefore, it is believed that guiding inhibitory interhemispheric network to the appropriate pattern through both targeted up-conditioning in the affected hemisphere and down-conditioning in the non-affected hemisphere may contribute to reduced abnormal IHI and enhanced achievable functional recovery (Chieffo et al., 2013; Di Pino et al., 2014; Dong et al., 2006; Hummel and Cohen, 2006). Although the rehabilitation strategy of attempting to rebalance interhemispheric networks in order to improve motor recovery after stroke is controversial (Bundy et al., 2017; Carson, 2020; Xu et al., 2019), the current technique can be tailored either to up-conditioning the damaged hemisphere, down-conditioning the intact hemisphere, a combination of both, or vice versa depending on the patient’s specific states. Furthermore, a previous review advocated the conceptual framework and suggested that a fundamental role of IHI is to support the narrowing of excitatory focus by co-opting the capacities of the two cerebral hemispheres (Carson, 2020). This framework is consistent with the known associations between the structural integrity of callosal projections and the magnitude of the motor deficits after stroke (Auriat et al., 2015; Granziera et al., 2012; Koh et al., 2018). Thus, our spatially bivariate EEG-neurofeedback approach might be highly relevant to application perspective in the context of stroke recovery as well as a basic science perspective. Future work probing the residual ability to manipulate IHI in stroke patients is warranted.

Source data used to generate the figures are publicly available via Dryad Digital Repository, accessible here: Hayashi, Masaaki (2021), Spatially bivariate EEG-neurofeedback can manipulate interhemispheric rebalancing of M1 excitability, Dryad, Dataset, https://doi.org/10.5061/dryad.hhmgqnkj3 Scripts used for the neurofeedback experiment are available on GitHub (https://github.com/MasaakiHayashi/elife-neurofeedback-experiment, (copy archived at swh:1:rev:20fab186058b4f59577c1d4da31f9fff377a129e)).

1. 1. Hayashi M

   (2021) Dryad Digital Repository

   Data from Spatially bivariate EEG-neurofeedback can manipulate interhemispheric rebalancing of M1 excitability.

   https://doi.org/10.5061/dryad.hhmgqnkj3
2. 1. Hayashi M

   (2022) GitHub

   ID elife-neurofeedback-experiment. Spatially bivariate EEG-neurofeedback can manipulate interhemispheric inhibition.

   https://github.com/MasaakiHayashi/elife-neurofeedback-experiment

1. 1. Adeyemo BO
   2. Simis M
   3. Macea DD
   4. Fregni F

   (2012) Systematic review of parameters of stimulation, clinical trial design characteristics, and motor outcomes in non-invasive brain stimulation in stroke

   Frontiers in Psychiatry 3:88.

   https://doi.org/10.3389/fpsyt.2012.00088

   • PubMed
   • Google Scholar
2. 1. Ames KC
   2. Churchland MM

   (2019) Motor cortex signals for each arm are mixed across hemispheres and neurons yet partitioned within the population response

   eLife 8:e46159.

   https://doi.org/10.7554/eLife.46159

   • PubMed
   • Google Scholar
3. 1. Ang KK
   2. Guan C
   3. Phua KS
   4. Wang C
   5. Zhao L
   6. Teo WP
   7. Chen C
   8. Ng YS
   9. Chew E

   (2015) Facilitating effects of transcranial direct current stimulation on motor imagery brain-computer interface with robotic feedback for stroke rehabilitation

   Archives of Physical Medicine and Rehabilitation 96:S79–S87.

   https://doi.org/10.1016/j.apmr.2014.08.008

   • PubMed
   • Google Scholar
4. 1. Ang KK
   2. Guan C

   (2017) EEG-Based strategies to detect motor imagery for control and rehabilitation

   IEEE Transactions on Neural Systems and Rehabilitation Engineering 25:392–401.

   https://doi.org/10.1109/TNSRE.2016.2646763

   • PubMed
   • Google Scholar
5. 1. Arai N
   2. Müller-Dahlhaus F
   3. Murakami T
   4. Bliem B
   5. Lu MK
   6. Ugawa Y
   7. Ziemann U

   (2011) State-dependent and timing-dependent bidirectional associative plasticity in the human SMA-M1 network

   The Journal of Neuroscience 31:15376–15383.

   https://doi.org/10.1523/JNEUROSCI.2271-11.2011

   • PubMed
   • Google Scholar
6. 1. Auriat AM
   2. Borich MR
   3. Snow NJ
   4. Wadden KP
   5. Boyd LA

   (2015) Comparing a diffusion tensor and non-tensor approach to white matter fiber tractography in chronic stroke

   NeuroImage. Clinical 7:771–781.

   https://doi.org/10.1016/j.nicl.2015.03.007

   • PubMed
   • Google Scholar
7. 1. Bakdash JZ
   2. Marusich LR

   (2017) Repeated measures correlation

   Frontiers in Psychology 8:13.

   https://doi.org/10.3389/fpsyg.2017.00456

   • PubMed
   • Google Scholar
8. 1. Banich MT

   (1998) The missing link: the role of interhemispheric interaction in attentional processing

   Brain and Cognition 36:128–157.

   https://doi.org/10.1006/brcg.1997.0950

   • PubMed
   • Google Scholar
9. 1. Boddington LJ
   2. Reynolds JNJ

   (2017) Targeting interhemispheric inhibition with neuromodulation to enhance stroke rehabilitation

   Brain Stimulation 10:214–222.

   https://doi.org/10.1016/j.brs.2017.01.006

   • PubMed
   • Google Scholar
10. 1. Buch E
    2. Weber C
    3. Cohen LG
    4. Braun C
    5. Dimyan MA
    6. Ard T
    7. Mellinger J
    8. Caria A
    9. Soekadar S
    10. Fourkas A
    11. Birbaumer N

    (2008) Think to move: A neuromagnetic brain-computer interface (BCI) system for chronic stroke

    Stroke 39:910–917.

    https://doi.org/10.1161/STROKEAHA.107.505313

    • PubMed
    • Google Scholar
11. 1. Bundy DT
    2. Souders L
    3. Baranyai K
    4. Leonard L
    5. Schalk G
    6. Coker R
    7. Moran DW
    8. Huskey T
    9. Leuthardt EC

    (2017) Contralesional brain-computer interface control of a powered exoskeleton for motor recovery in chronic stroke survivors

    Stroke 48:1908–1915.

    https://doi.org/10.1161/STROKEAHA.116.016304

    • PubMed
    • Google Scholar
12. 1. Caria A
    2. Weber C
    3. Brötz D
    4. Ramos A
    5. Ticini LF
    6. Gharabaghi A
    7. Braun C
    8. Birbaumer N

    (2011) Chronic stroke recovery after combined BCI training and physiotherapy: A case report

    Psychophysiology 48:578–582.

    https://doi.org/10.1111/j.1469-8986.2010.01117.x

    • PubMed
    • Google Scholar
13. 1. Carson RG

    (2020) Inter-hemispheric inhibition sculpts the output of neural circuits by co-opting the two cerebral hemispheres

    The Journal of Physiology 598:4781–4802.

    https://doi.org/10.1113/JP279793

    • PubMed
    • Google Scholar
14. 1. Carter Leno V
    2. Tomlinson SB
    3. Chang SAA
    4. Naples AJ
    5. McPartland JC

    (2018) Resting-state alpha power is selectively associated with autistic traits reflecting behavioral rigidity

    Scientific Reports 8:7.

    https://doi.org/10.1038/s41598-018-30445-2

    • PubMed
    • Google Scholar
15. 1. Chaudhary U
    2. Birbaumer N
    3. Ramos-Murguialday A

    (2016) Brain-computer interfaces for communication and rehabilitation

    Nature Reviews. Neurology 12:513–525.

    https://doi.org/10.1038/nrneurol.2016.113

    • PubMed
    • Google Scholar
16. 1. Chieffo R
    2. Inuggi A
    3. Straffi L
    4. Coppi E
    5. Gonzalez-Rosa J
    6. Spagnolo F
    7. Poggi A
    8. Comi G
    9. Comola M
    10. Leocani L

    (2013) Mapping early changes of cortical motor output after subcortical stroke: A transcranial magnetic stimulation study

    Brain Stimulation 6:322–329.

    https://doi.org/10.1016/j.brs.2012.06.003

    • PubMed
    • Google Scholar
17. 1. Cohen J

    (1992) A power primer

    Psychological Bulletin 112:155–159.

    https://doi.org/10.1037//0033-2909.112.1.155

    • PubMed
    • Google Scholar
18. 1. Crone NE
    2. Miglioretti DL
    3. Gordon B
    4. Sieracki JM
    5. Wilson MT
    6. Uematsu S
    7. Lesser RP

    (1998)

    Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis I Alpha and beta event-related desynchronization

    Brain: A Journal of Neurology 121:2271–2299.

    • Google Scholar
19. 1. Daskalakis ZJ
    2. Christensen BK
    3. Fitzgerald PB
    4. Roshan L
    5. Chen R

    (2002) The mechanisms of interhemispheric inhibition in the human motor cortex

    The Journal of Physiology 543:317–326.

    https://doi.org/10.1113/jphysiol.2002.017673

    • PubMed
    • Google Scholar
20. 1. Devanne H
    2. Lavoie BA
    3. Capaday C

    (1997) Input-output properties and gain changes in the human corticospinal pathway

    Experimental Brain Research 114:329–338.

    https://doi.org/10.1007/pl00005641

    • PubMed
    • Google Scholar
21. 1. Di Pino G
    2. Pellegrino G
    3. Assenza G
    4. Capone F
    5. Ferreri F
    6. Formica D
    7. Ranieri F
    8. Tombini M
    9. Ziemann U
    10. Rothwell JC
    11. Di Lazzaro V

    (2014) Modulation of brain plasticity in stroke: A novel model for neurorehabilitation

    Nature Reviews. Neurology 10:597–608.

    https://doi.org/10.1038/nrneurol.2014.162

    • PubMed
    • Google Scholar
22. 1. Dong Y
    2. Dobkin BH
    3. Cen SY
    4. Wu AD
    5. Winstein CJ

    (2006) Motor cortex activation during treatment may predict therapeutic gains in paretic hand function after stroke

    Stroke 37:1552–1555.

    https://doi.org/10.1161/01.STR.0000221281.69373.4e

    • PubMed
    • Google Scholar
23. 1. Dunst B
    2. Benedek M
    3. Koschutnig K
    4. Jauk E
    5. Neubauer AC

    (2014) Sex differences in the IQ-white matter microstructure relationship: A DTI study

    Brain and Cognition 91:71–78.

    https://doi.org/10.1016/j.bandc.2014.08.006

    • PubMed
    • Google Scholar
24. 1. Duque J
    2. Hummel F
    3. Celnik P
    4. Murase N
    5. Mazzocchio R
    6. Cohen LG

    (2005) Transcallosal inhibition in chronic subcortical stroke

    NeuroImage 28:940–946.

    https://doi.org/10.1016/j.neuroimage.2005.06.033

    • PubMed
    • Google Scholar
25. 1. Duque J
    2. Murase N
    3. Celnik P
    4. Hummel F
    5. Harris-Love M
    6. Mazzocchio R
    7. Olivier E
    8. Cohen LG

    (2007) Intermanual Differences in movement-related interhemispheric inhibition

    Journal of Cognitive Neuroscience 19:204–213.

    https://doi.org/10.1162/jocn.2007.19.2.204

    • PubMed
    • Google Scholar
26. 1. Engelhard B
    2. Ozeri N
    3. Israel Z
    4. Bergman H
    5. Vaadia E

    (2013) Inducing γ oscillations and precise spike synchrony by operant conditioning via brain-machine interface

    Neuron 77:361–375.

    https://doi.org/10.1016/j.neuron.2012.11.015

    • PubMed
    • Google Scholar
27. 1. Ewald A
    2. Marzetti L
    3. Zappasodi F
    4. Meinecke FC
    5. Nolte G

    (2012) Estimating true brain connectivity from EEG/MEG data invariant to linear and static transformations in sensor space

    NeuroImage 60:476–488.

    https://doi.org/10.1016/j.neuroimage.2011.11.084

    • PubMed
    • Google Scholar
28. 1. Faul F
    2. Erdfelder E
    3. Buchner A
    4. Lang AG

    (2009) Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses

    Behavior Research Methods 41:1149–1160.

    https://doi.org/10.3758/BRM.41.4.1149

    • PubMed
    • Google Scholar
29. 1. Ferbert A
    2. Priori A
    3. Rothwell JC
    4. Day BL
    5. Colebatch JG
    6. Marsden CD

    (1992) Interhemispheric inhibition of the human motor cortex

    The Journal of Physiology 453:525–546.

    https://doi.org/10.1113/jphysiol.1992.sp019243

    • PubMed
    • Google Scholar
30. 1. Ferree TC
    2. Luu P
    3. Russell GS
    4. Tucker DM

    (2001) Scalp electrode impedance, infection risk, and EEG data quality

    Clinical Neurophysiology 112:536–544.

    https://doi.org/10.1016/s1388-2457(00)00533-2

    • PubMed
    • Google Scholar
31. 1. Fetz EE

    (2013) Volitional control of cortical oscillations and synchrony

    Neuron 77:216–218.

    https://doi.org/10.1016/j.neuron.2013.01.003

    • PubMed
    • Google Scholar
32. 1. Fries P

    (2005) A mechanism for cognitive dynamics: neuronal communication through neuronal coherence

    Trends in Cognitive Sciences 9:474–480.

    https://doi.org/10.1016/j.tics.2005.08.011

    • PubMed
    • Google Scholar
33. 1. Ghosh S
    2. Mehta AR
    3. Huang G
    4. Gunraj C
    5. Hoque T
    6. Saha U
    7. Ni Z
    8. Chen R

    (2013) Short- and long-latency interhemispheric inhibitions are additive in human motor cortex

    Journal of Neurophysiology 109:2955–2962.

    https://doi.org/10.1152/jn.00960.2012

    • PubMed
    • Google Scholar
34. 1. Gilio F
    2. Rizzo V
    3. Siebner HR
    4. Rothwell JC

    (2003) Effects on the right motor hand-area excitability produced by low-frequency rTMS over human contralateral homologous cortex

    The Journal of Physiology 551:563–573.

    https://doi.org/10.1113/jphysiol.2003.044313

    • PubMed
    • Google Scholar
35. 1. Granziera C
    2. Daducci A
    3. Meskaldji DE
    4. Roche A
    5. Maeder P
    6. Michel P
    7. Hadjikhani N
    8. Sorensen AG
    9. Frackowiak RS
    10. Thiran JP
    11. Meuli R
    12. Krueger G

    (2012) A new early and automated MRI-based predictor of motor improvement after stroke

    Neurology 79:39–46.

    https://doi.org/10.1212/WNL.0b013e31825f25e7

    • PubMed
    • Google Scholar
36. 1. Groppa S
    2. Oliviero A
    3. Eisen A
    4. Quartarone A
    5. Cohen LG
    6. Mall V
    7. Kaelin-Lang A
    8. Mima T
    9. Rossi S
    10. Thickbroom GW
    11. Rossini PM
    12. Ziemann U
    13. Valls-Solé J
    14. Siebner HR

    (2012) A practical guide to diagnostic transcranial magnetic stimulation: report of an IFCN committee

    Clinical Neurophysiology 123:858–882.

    https://doi.org/10.1016/j.clinph.2012.01.010

    • PubMed
    • Google Scholar
37. 1. Haegens S
    2. Nácher V
    3. Luna R
    4. Romo R
    5. Jensen O

    (2011) α-Oscillations in the monkey sensorimotor network influence discrimination performance by rhythmical inhibition of neuronal spiking

    PNAS 108:19377–19382.

    https://doi.org/10.1073/pnas.1117190108

    • PubMed
    • Google Scholar
38. 1. Hallett M

    (2007) Transcranial magnetic stimulation: A primer

    Neuron 55:187–199.

    https://doi.org/10.1016/j.neuron.2007.06.026

    • PubMed
    • Google Scholar
39. 1. Harris-Love ML
    2. Perez MA
    3. Chen R
    4. Cohen LG

    (2007) Interhemispheric inhibition in distal and proximal arm representations in the primary motor cortex

    Journal of Neurophysiology 97:2511–2515.

    https://doi.org/10.1152/jn.01331.2006

    • PubMed
    • Google Scholar
40. 1. Hayashi M
    2. Tsuchimoto S
    3. Mizuguchi N
    4. Miyatake M
    5. Kasuga S
    6. Ushiba J

    (2019) Two-stage regression of high-density scalp electroencephalograms visualizes force regulation signaling during muscle contraction

    Journal of Neural Engineering 16:14.

    https://doi.org/10.1088/1741-2552/ab221a

    • PubMed
    • Google Scholar
41. 1. Hayashi M
    2. Mizuguchi N
    3. Tsuchimoto S
    4. Ushiba J

    (2020) Neurofeedback of scalp bi-hemispheric EEG sensorimotor rhythm guides hemispheric activation of sensorimotor cortex in the targeted hemisphere

    NeuroImage 223:117298.

    https://doi.org/10.1016/j.neuroimage.2020.117298

    • PubMed
    • Google Scholar
42. 1. Hayashi M
    2. Mizuguchi N
    3. Tsuchimoto S
    4. Ushiba J

    (2021) Neurofeedback of scalp bi-hemispheric EEG sensorimotor rhythm guides hemispheric activation of sensorimotor cortex in the targeted hemisphere. Brain-Computer Interface Research: A State-of-the-Art Summary

    SpringerBriefs in Electrical and Computer Engineering 9:25–38.

    https://doi.org/10.1007/978-3-030-60460-8_3

    • Google Scholar
43. Software

    1. Hayashi M

    (2022) Masaaki Hayashi. 2022. elife-neurofeedback-experiment, version 20fab18

    GitHub.

    https://github.com/MasaakiHayashi/elife-neurofeedback-experiment
44. 1. He S
    2. Everest-Phillips C
    3. Clouter A
    4. Brown P
    5. Tan H

    (2020) Neurofeedback-linked suppression of cortical β bursts speeds up movement initiation in healthy motor control: a double-blind sham-controlled study

    The Journal of Neuroscience 40:4021–4032.

    https://doi.org/10.1523/JNEUROSCI.0208-20.2020

    • PubMed
    • Google Scholar
45. 1. Hofer S
    2. Frahm J

    (2006) Topography of the human corpus callosum revisited--comprehensive fiber tractography using diffusion tensor magnetic resonance imaging

    NeuroImage 32:989–994.

    https://doi.org/10.1016/j.neuroimage.2006.05.044

    • PubMed
    • Google Scholar
46. 1. Hsu WY
    2. Cheng CH
    3. Liao KK
    4. Lee IH
    5. Lin YY

    (2012) Effects of repetitive transcranial magnetic stimulation on motor functions in patients with stroke: A meta-analysis

    Stroke 43:1849–1857.

    https://doi.org/10.1161/STROKEAHA.111.649756

    • PubMed
    • Google Scholar
47. 1. Hummel FC
    2. Cohen LG

    (2006) Non-invasive brain stimulation: a new strategy to improve neurorehabilitation after stroke?

    The Lancet. Neurology 5:708–712.

    https://doi.org/10.1016/S1474-4422(06)70525-7

    • PubMed
    • Google Scholar
48. 1. Hussain SJ
    2. Claudino L
    3. Bönstrup M
    4. Norato G
    5. Cruciani G
    6. Thompson R
    7. Zrenner C
    8. Ziemann U
    9. Buch E
    10. Cohen LG

    (2019) Sensorimotor oscillatory phase-power interaction gates resting human corticospinal output

    Cerebral Cortex 29:3766–3777.

    https://doi.org/10.1093/cercor/bhy255

    • PubMed
    • Google Scholar
49. 1. Iwama S
    2. Tsuchimoto S
    3. Hayashi M
    4. Mizuguchi N
    5. Ushiba J

    (2020) Scalp electroencephalograms over ipsilateral sensorimotor cortex reflect contraction patterns of unilateral finger muscles

    NeuroImage 222:117249.

    https://doi.org/10.1016/j.neuroimage.2020.117249

    • PubMed
    • Google Scholar
50. 1. Khademi F
    2. Royter V
    3. Gharabaghi A

    (2018) Distinct beta-band oscillatory circuits underlie corticospinal gain modulation

    Cerebral Cortex 28:1502–1515.

    https://doi.org/10.1093/cercor/bhy016

    • PubMed
    • Google Scholar
51. 1. Kober SE
    2. Witte M
    3. Grinschgl S
    4. Neuper C
    5. Wood G

    (2018) Placebo hampers ability to self-regulate brain activity: A double-blind sham-controlled neurofeedback study

    NeuroImage 181:797–806.

    https://doi.org/10.1016/j.neuroimage.2018.07.025

    • PubMed
    • Google Scholar
52. 1. Koh CL
    2. Tang PF
    3. Chen HI
    4. Hsu YC
    5. Hsieh CL
    6. Tseng WYI

    (2018) Impaired callosal motor fiber integrity and upper extremity motor impairment are associated with stroke lesion location

    Neurorehabilitation and Neural Repair 32:602–612.

    https://doi.org/10.1177/1545968318779730

    • PubMed
    • Google Scholar
53. 1. Kraus D
    2. Naros G
    3. Bauer R
    4. Leão MT
    5. Ziemann U
    6. Gharabaghi A

    (2016) Brain-robot interface driven plasticity: Distributed modulation of corticospinal excitability

    NeuroImage 125:522–532.

    https://doi.org/10.1016/j.neuroimage.2015.09.074

    • PubMed
    • Google Scholar
54. 1. Liang N
    2. Funase K
    3. Takahashi M
    4. Matsukawa K
    5. Kasai T

    (2014) Unilateral imagined movement increases interhemispheric inhibition from the contralateral to ipsilateral motor cortex

    Experimental Brain Research 232:1823–1832.

    https://doi.org/10.1007/s00221-014-3874-4

    • PubMed
    • Google Scholar
55. 1. Liew S-L
    2. Rana M
    3. Cornelsen S
    4. Fortunato de Barros Filho M
    5. Birbaumer N
    6. Sitaram R
    7. Cohen LG
    8. Soekadar SR

    (2016) Improving motor corticothalamic communication after stroke using real-time fMRI connectivity-based neurofeedback

    Neurorehabilitation and Neural Repair 30:671–675.

    https://doi.org/10.1177/1545968315619699

    • PubMed
    • Google Scholar
56. 1. Madsen KH
    2. Karabanov AN
    3. Krohne LG
    4. Safeldt MG
    5. Tomasevic L
    6. Siebner HR

    (2019) No trace of phase: Corticomotor excitability is not tuned by phase of pericentral mu-rhythm

    Brain Stimulation 12:1261–1270.

    https://doi.org/10.1016/j.brs.2019.05.005

    • PubMed
    • Google Scholar
57. 1. McFarland DJ
    2. McCane LM
    3. David SV
    4. Wolpaw JR

    (1997) Spatial filter selection for EEG-based communication

    Electroencephalography and Clinical Neurophysiology 103:386–394.

    https://doi.org/10.1016/s0013-4694(97)00022-2

    • PubMed
    • Google Scholar
58. 1. Menzler K
    2. Belke M
    3. Wehrmann E
    4. Krakow K
    5. Lengler U
    6. Jansen A
    7. Hamer HM
    8. Oertel WH
    9. Rosenow F
    10. Knake S

    (2011) Men and women are different: diffusion tensor imaging reveals sexual dimorphism in the microstructure of the thalamus, corpus callosum and cingulum

    NeuroImage 54:2557–2562.

    https://doi.org/10.1016/j.neuroimage.2010.11.029

    • PubMed
    • Google Scholar
59. 1. Meyer BU
    2. Röricht S
    3. von Einsiedel HG
    4. Kruggel F
    5. Weindl A

    (1995) Inhibitory and excitatory interhemispheric transfers between motor cortical areas in normal humans and patients with abnormalities of the corpus callosum

    Brain: A Journal of Neurology 118:429–440.

    https://doi.org/10.1093/brain/118.2.429

    • Google Scholar
60. 1. Moher D
    2. Schulz KF
    3. Altman D

    (2001) The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials

    JAMA 285:1987–1991.

    https://doi.org/10.1001/jama.285.15.1987

    • PubMed
    • Google Scholar
61. 1. Moon HS
    2. Jiang H
    3. Vo TT
    4. Jung WB
    5. Vazquez AL
    6. Kim SG

    (2021) Contribution of excitatory and inhibitory neuronal activity to bold fMRI

    Cerebral Cortex 31:4053–4067.

    https://doi.org/10.1093/cercor/bhab068

    • PubMed
    • Google Scholar
62. 1. Morishita T
    2. Uehara K
    3. Funase K

    (2012) Changes in interhemispheric inhibition from active to resting primary motor cortex during a fine-motor manipulation task

    Journal of Neurophysiology 107:3086–3094.

    https://doi.org/10.1152/jn.00888.2011

    • PubMed
    • Google Scholar
63. 1. Murase N
    2. Duque J
    3. Mazzocchio R
    4. Cohen LG

    (2004) Influence of interhemispheric interactions on motor function in chronic stroke

    Annals of Neurology 55:400–409.

    https://doi.org/10.1002/ana.10848

    • PubMed
    • Google Scholar
64. 1. Naros G
    2. Lehnertz T
    3. Leão MT
    4. Ziemann U
    5. Gharabaghi A

    (2020) Brain state-dependent gain modulation of corticospinal output in the active motor system

    Cerebral Cortex 30:371–381.

    https://doi.org/10.1093/cercor/bhz093

    • PubMed
    • Google Scholar
65. 1. Nelson AJ
    2. Hoque T
    3. Gunraj C
    4. Ni Z
    5. Chen R

    (2009) Bi-directional interhemispheric inhibition during unimanual sustained contractions

    BMC Neuroscience 10:31.

    https://doi.org/10.1186/1471-2202-10-31

    • PubMed
    • Google Scholar
66. 1. Neuper C
    2. Pfurtscheller G

    (2001) Event-related dynamics of cortical rhythms: frequency-specific features and functional correlates

    International Journal of Psychophysiology 43:41–58.

    https://doi.org/10.1016/s0167-8760(01)00178-7

    • PubMed
    • Google Scholar
67. 1. Neuper C
    2. Scherer R
    3. Reiner M
    4. Pfurtscheller G

    (2005) Imagery of motor actions: differential effects of kinesthetic and visual-motor mode of imagery in single-trial EEG

    Brain Research. Cognitive Brain Research 25:668–677.

    https://doi.org/10.1016/j.cogbrainres.2005.08.014

    • PubMed
    • Google Scholar
68. 1. Neuper C
    2. Wörtz M
    3. Pfurtscheller G

    (2006) ERD/ERS patterns reflecting sensorimotor activation and deactivation

    Progress in Brain Research 159:211–222.

    https://doi.org/10.1016/S0079-6123(06)59014-4

    • PubMed
    • Google Scholar
69. 1. Ni Z
    2. Gunraj C
    3. Nelson AJ
    4. Yeh IJ
    5. Castillo G
    6. Hoque T
    7. Chen R

    (2009) Two phases of interhemispheric inhibition between motor related cortical areas and the primary motor cortex in human

    Cerebral Cortex 19:1654–1665.

    https://doi.org/10.1093/cercor/bhn201

    • PubMed
    • Google Scholar
70. 1. Ni Z
    2. Müller-Dahlhaus F
    3. Chen R
    4. Ziemann U

    (2011) Triple-pulse TMS to study interactions between neural circuits in human cortex

    Brain Stimulation 4:281–293.

    https://doi.org/10.1016/j.brs.2011.01.002

    • PubMed
    • Google Scholar
71. 1. Notturno F
    2. Marzetti L
    3. Pizzella V
    4. Uncini A
    5. Zappasodi F

    (2014) Local and remote effects of transcranial direct current stimulation on the electrical activity of the motor cortical network

    Human Brain Mapping 35:2220–2232.

    https://doi.org/10.1002/hbm.22322

    • PubMed
    • Google Scholar
72. 1. Oldfield RC

    (1971) The assessment and analysis of handedness: the Edinburgh inventory

    Neuropsychologia 9:97–113.

    https://doi.org/10.1016/0028-3932(71)90067-4

    • PubMed
    • Google Scholar
73. 1. Palmer LM
    2. Schulz JM
    3. Murphy SC
    4. Ledergerber D
    5. Murayama M
    6. Larkum ME

    (2012) The cellular basis of GABA(B)-mediated interhemispheric inhibition

    Science 335:989–993.

    https://doi.org/10.1126/science.1217276

    • PubMed
    • Google Scholar
74. 1. Pena-Gomez C
    2. Sala-Lonch R
    3. Junqué C
    4. Clemente IC
    5. Vidal D
    6. Bargalló N
    7. Falcón C
    8. Valls-Solé J
    9. Pascual-Leone Á
    10. Bartrés-Faz D

    (2012) Modulation of large-scale brain networks by transcranial direct current stimulation evidenced by resting-state functional MRI

    Brain Stimulation 5:252–263.

    https://doi.org/10.1016/j.brs.2011.08.006

    • PubMed
    • Google Scholar
75. 1. Pereira J
    2. Direito B
    3. Sayal A
    4. Ferreira C
    5. Castelo-Branco M

    (2019) Self-modulation of premotor cortex interhemispheric connectivity in a real-time functional magnetic resonance imaging neurofeedback study using an adaptive approach

    Brain Connectivity 9:662–672.

    https://doi.org/10.1089/brain.2019.0697

    • PubMed
    • Google Scholar
76. 1. Pfurtscheller G
    2. Neuper C

    (1997) Motor imagery activates primary sensorimotor area in humans

    Neuroscience Letters 239:65–68.

    https://doi.org/10.1016/s0304-3940(97)00889-6

    • PubMed
    • Google Scholar
77. 1. Pfurtscheller G
    2. Lopes da Silva FH

    (1999) Event-related EEG/MEG synchronization and desynchronization: basic principles

    Clinical Neurophysiology 110:1842–1857.

    https://doi.org/10.1016/s1388-2457(99)00141-8

    • PubMed
    • Google Scholar
78. 1. Pfurtscheller G

    (2001) Functional brain imaging based on ERD/ERS

    Vision Research 41:1257–1260.

    https://doi.org/10.1016/s0042-6989(00)00235-2

    • PubMed
    • Google Scholar
79. 1. Pfurtscheller G
    2. Brunner C
    3. Schlögl A
    4. Lopes da Silva FH

    (2006) Mu rhythm (de)synchronization and EEG single-trial classification of different motor imagery tasks

    NeuroImage 31:153–159.

    https://doi.org/10.1016/j.neuroimage.2005.12.003

    • PubMed
    • Google Scholar
80. 1. Picazio S
    2. Veniero D
    3. Ponzo V
    4. Caltagirone C
    5. Gross J
    6. Thut G
    7. Koch G

    (2014) Prefrontal control over motor cortex cycles at beta frequency during movement inhibition

    Current Biology 24:2940–2945.

    https://doi.org/10.1016/j.cub.2014.10.043

    • PubMed
    • Google Scholar
81. 1. Ramos-Murguialday A
    2. Broetz D
    3. Rea M
    4. Läer L
    5. Yilmaz O
    6. Brasil FL
    7. Liberati G
    8. Curado MR
    9. Garcia-Cossio E
    10. Vyziotis A
    11. Cho W
    12. Agostini M
    13. Soares E
    14. Soekadar S
    15. Caria A
    16. Cohen LG
    17. Birbaumer N

    (2013) Brain-machine interface in chronic stroke rehabilitation: A controlled study

    Annals of Neurology 74:100–108.

    https://doi.org/10.1002/ana.23879

    • PubMed
    • Google Scholar
82. 1. Robertson MM
    2. Furlong S
    3. Voytek B
    4. Donoghue T
    5. Boettiger CA
    6. Sheridan MA

    (2019) EEG power spectral slope differs by ADHD status and stimulant medication exposure in early childhood

    Journal of Neurophysiology 122:2427–2437.

    https://doi.org/10.1152/jn.00388.2019

    • PubMed
    • Google Scholar
83. 1. Ros T
    2. Munneke MAM
    3. Ruge D
    4. Gruzelier JH
    5. Rothwell JC

    (2010) Endogenous control of waking brain rhythms induces neuroplasticity in humans

    The European Journal of Neuroscience 31:770–778.

    https://doi.org/10.1111/j.1460-9568.2010.07100.x

    • PubMed
    • Google Scholar
84. 1. Ros T
    2. Enriquez-Geppert S
    3. Zotev V
    4. Young KD
    5. Wood G
    6. Whitfield-Gabrieli S
    7. Wan F
    8. Vuilleumier P
    9. Vialatte F
    10. Van De Ville D
    11. Todder D
    12. Surmeli T
    13. Sulzer JS
    14. Strehl U
    15. Sterman MB
    16. Steiner NJ
    17. Sorger B
    18. Soekadar SR
    19. Sitaram R
    20. Sherlin LH
    21. Schönenberg M
    22. Scharnowski F
    23. Schabus M
    24. Rubia K
    25. Rosa A
    26. Reiner M
    27. Pineda JA
    28. Paret C
    29. Ossadtchi A
    30. Nicholson AA
    31. Nan W
    32. Minguez J
    33. Micoulaud-Franchi JA
    34. Mehler DMA
    35. Lührs M
    36. Lubar J
    37. Lotte F
    38. Linden DEJ
    39. Lewis-Peacock JA
    40. Lebedev MA
    41. Lanius RA
    42. Kübler A
    43. Kranczioch C
    44. Koush Y
    45. Konicar L
    46. Kohl SH
    47. Kober SE
    48. Klados MA
    49. Jeunet C
    50. Janssen TWP
    51. Huster RJ
    52. Hoedlmoser K
    53. Hirshberg LM
    54. Heunis S
    55. Hendler T
    56. Hampson M
    57. Guggisberg AG
    58. Guggenberger R
    59. Gruzelier JH
    60. Göbel RW
    61. Gninenko N
    62. Gharabaghi A
    63. Frewen P
    64. Fovet T
    65. Fernández T
    66. Escolano C
    67. Ehlis AC
    68. Drechsler R
    69. Christopher deCharms R
    70. Debener S
    71. De Ridder D
    72. Davelaar EJ
    73. Congedo M
    74. Cavazza M
    75. Breteler MHM
    76. Brandeis D
    77. Bodurka J
    78. Birbaumer N
    79. Bazanova OM
    80. Barth B
    81. Bamidis PD
    82. Auer T
    83. Arns M
    84. Thibault RT

    (2020) Consensus on the reporting and experimental design of clinical and cognitive-behavioural neurofeedback studies (CRED-nf checklist

    Brain: A Journal of Neurology 143:1674–1685.

    https://doi.org/10.1093/brain/awaa009

    • PubMed
    • Google Scholar
85. 1. Rossi S
    2. Hallett M
    3. Rossini PM
    4. Pascual-Leone A
    5. Safety of TMS Consensus Group

    (2009) Safety, ethical considerations, and application guidelines for the use of transcranial magnetic stimulation in clinical practice and research

    Clinical Neurophysiology 120:2008–2039.

    https://doi.org/10.1016/j.clinph.2009.08.016

    • PubMed
    • Google Scholar
86. 1. Rossini PM
    2. Barker AT
    3. Berardelli A
    4. Caramia MD
    5. Caruso G
    6. Cracco RQ
    7. Dimitrijević MR
    8. Hallett M
    9. Katayama Y
    10. Lücking CH
    11. Noordhout AL
    12. Marsden CD
    13. Murray NMF
    14. Rothwell JC
    15. Swash M
    16. Tomberg C

    (1994) Non-invasive electrical and magnetic stimulation of the brain, spinal cord and roots: basic principles and procedures for routine clinical application

    Electroencephalography and Clinical Neurophysiology 91:79–92.

    https://doi.org/10.1016/0013-4694(94)90029-9

    • PubMed
    • Google Scholar
87. 1. Rossini PM
    2. Burke D
    3. Chen R
    4. Cohen LG
    5. Daskalakis Z
    6. Di Iorio R
    7. Di Lazzaro V
    8. Ferreri F
    9. Fitzgerald PB
    10. George MS
    11. Hallett M
    12. Lefaucheur JP
    13. Langguth B
    14. Matsumoto H
    15. Miniussi C
    16. Nitsche MA
    17. Pascual-Leone A
    18. Paulus W
    19. Rossi S
    20. Rothwell JC
    21. Siebner HR
    22. Ugawa Y
    23. Walsh V
    24. Ziemann U

    (2015) Non-invasive electrical and magnetic stimulation of the brain, spinal cord, roots and peripheral nerves: Basic principles and procedures for routine clinical and research application. An updated report from an I.F.C.N. Committee

    Clinical Neurophysiology 126:1071–1107.

    https://doi.org/10.1016/j.clinph.2015.02.001

    • PubMed
    • Google Scholar
88. 1. Ruddy K
    2. Balsters J
    3. Mantini D
    4. Liu Q
    5. Kassraian-Fard P
    6. Enz N
    7. Mihelj E
    8. Subhash Chander B
    9. Soekadar SR
    10. Wenderoth N

    (2018) Neural activity related to volitional regulation of cortical excitability

    eLife 7:e40843.

    https://doi.org/10.7554/eLife.40843

    • PubMed
    • Google Scholar
89. 1. Safeldt MG
    2. Tomasevic L
    3. Karabanov A
    4. Siebner H
    5. Madsen KH

    (2017) Towards brain-state dependent transcranial magnetic stimulation: Targeting the phase of oscillatory neocortical activity with singe-pulse TMS

    Brain Stimulation 10:449–450.

    https://doi.org/10.1016/j.brs.2017.01.323

    • Google Scholar
90. Book

    1. Sanei S
    2. Chambers JA

    (2013) Introduction to EEGEEG Signal Processing

    John Wiley & Sons, Ltd.

    https://doi.org/10.1002/9780470511923.ch1

    • Google Scholar
91. 1. Sauseng P
    2. Klimesch W
    3. Gerloff C
    4. Hummel FC

    (2009) Spontaneous locally restricted EEG alpha activity determines cortical excitability in the motor cortex

    Neuropsychologia 47:284–288.

    https://doi.org/10.1016/j.neuropsychologia.2008.07.021

    • PubMed
    • Google Scholar
92. 1. Schulz H
    2. Ubelacker T
    3. Keil J
    4. Müller N
    5. Weisz N

    (2014) Now I am ready-now I am not: The influence of pre-TMS oscillations and corticomuscular coherence on motor-evoked potentials

    Cerebral Cortex 24:1708–1719.

    https://doi.org/10.1093/cercor/bht024

    • PubMed
    • Google Scholar
93. 1. Sitaram R
    2. Ros T
    3. Stoeckel L
    4. Haller S
    5. Scharnowski F
    6. Lewis-Peacock J
    7. Weiskopf N
    8. Blefari ML
    9. Rana M
    10. Oblak E
    11. Birbaumer N
    12. Sulzer J

    (2017) Closed-loop brain training: the science of neurofeedback

    Nature Reviews. Neuroscience 18:86–100.

    https://doi.org/10.1038/nrn.2016.164

    • PubMed
    • Google Scholar
94. 1. Soekadar SR
    2. Birbaumer N
    3. Slutzky MW
    4. Cohen LG

    (2015a) Brain-machine interfaces in neurorehabilitation of stroke

    Neurobiology of Disease 83:172–179.

    https://doi.org/10.1016/j.nbd.2014.11.025

    • PubMed
    • Google Scholar
95. 1. Soekadar SR
    2. Witkowski M
    3. Birbaumer N
    4. Cohen LG

    (2015b) Enhancing hebbian learning to control brain oscillatory activity

    Cerebral Cortex 25:2409–2415.

    https://doi.org/10.1093/cercor/bhu043

    • PubMed
    • Google Scholar
96. 1. Stefanou MI
    2. Desideri D
    3. Belardinelli P
    4. Zrenner C
    5. Ziemann U

    (2018) Phase synchronicity of μ-rhythm determines efficacy of interhemispheric communication between human motor cortices

    The Journal of Neuroscience 38:10525–10534.

    https://doi.org/10.1523/JNEUROSCI.1470-18.2018

    • PubMed
    • Google Scholar
97. 1. Takemi M
    2. Masakado Y
    3. Liu M
    4. Ushiba J

    (2013) Event-related desynchronization reflects downregulation of intracortical inhibition in human primary motor cortex

    Journal of Neurophysiology 110:1158–1166.

    https://doi.org/10.1152/jn.01092.2012

    • PubMed
    • Google Scholar
98. 1. Takemi M
    2. Masakado Y
    3. Liu M
    4. Ushiba J

    (2015) Sensorimotor event-related desynchronization represents the excitability of human spinal motoneurons

    Neuroscience 297:58–67.

    https://doi.org/10.1016/j.neuroscience.2015.03.045

    • PubMed
    • Google Scholar
99. 1. Takemi M
    2. Maeda T
    3. Masakado Y
    4. Siebner HR
    5. Ushiba J

    (2018) Muscle-selective disinhibition of corticomotor representations using a motor imagery-based brain-computer interface

    NeuroImage 183:597–605.

    https://doi.org/10.1016/j.neuroimage.2018.08.070

    • PubMed
    • Google Scholar
100. 1. Thies M
     2. Zrenner C
     3. Ziemann U
     4. Bergmann TO

     (2018) Sensorimotor mu-alpha power is positively related to corticospinal excitability

     Brain Stimulation 11:1119–1122.

     https://doi.org/10.1016/j.brs.2018.06.006

     • PubMed
     • Google Scholar
101. 1. Thompson AK
     2. Chen XY
     3. Wolpaw JR

     (2009) Acquisition of a simple motor skill: task-dependent adaptation plus long-term change in the human soleus H-reflex

     The Journal of Neuroscience 29:5784–5792.

     https://doi.org/10.1523/JNEUROSCI.4326-08.2009

     • PubMed
     • Google Scholar
102. 1. Tsuchimoto S
     2. Shibusawa S
     3. Iwama S
     4. Hayashi M
     5. Okuyama K
     6. Mizuguchi N
     7. Kato K
     8. Ushiba J

     (2021) Use of common average reference and large-Laplacian spatial-filters enhances EEG signal-to-noise ratios in intrinsic sensorimotor activity

     Journal of Neuroscience Methods 353:109089.

     https://doi.org/10.1016/j.jneumeth.2021.109089

     • PubMed
     • Google Scholar
103. 1. Tsutsumi R
     2. Shirota Y
     3. Ohminami S
     4. Terao Y
     5. Ugawa Y
     6. Hanajima R

     (2012) Conditioning intensity-dependent interaction between short-latency interhemispheric inhibition and short-latency afferent inhibition

     Journal of Neurophysiology 108:1130–1137.

     https://doi.org/10.1152/jn.00300.2012

     • PubMed
     • Google Scholar
104. 1. Varela F
     2. Lachaux JP
     3. Rodriguez E
     4. Martinerie J

     (2001) The brainweb: phase synchronization and large-scale integration

     Nature Reviews. Neuroscience 2:229–239.

     https://doi.org/10.1038/35067550

     • PubMed
     • Google Scholar
105. 1. Vukelić M
     2. Gharabaghi A

     (2015) Oscillatory entrainment of the motor cortical network during motor imagery is modulated by the feedback modality

     NeuroImage 111:1–11.

     https://doi.org/10.1016/j.neuroimage.2015.01.058

     • PubMed
     • Google Scholar
106. 1. Waters S
     2. Wiestler T
     3. Diedrichsen J

     (2017) Cooperation not competition: Bihemispheric tdcs and fmri show role for ipsilateral hemisphere in motor learning

     The Journal of Neuroscience 37:7500–7512.

     https://doi.org/10.1523/JNEUROSCI.3414-16.2017

     • PubMed
     • Google Scholar
107. 1. Weiskopf N
     2. Mathiak K
     3. Bock SW
     4. Scharnowski F
     5. Veit R
     6. Grodd W
     7. Goebel R
     8. Birbaumer N

     (2004) Principles of a brain-computer interface (BCI) based on real-time functional magnetic resonance imaging (fMRI

     IEEE Transactions on Bio-Medical Engineering 51:966–970.

     https://doi.org/10.1109/TBME.2004.827063

     • PubMed
     • Google Scholar
108. 1. Williams JA
     2. Pascual-Leone A
     3. Fregni F

     (2010) Interhemispheric modulation induced by cortical stimulation and motor training

     Physical Therapy 90:398–410.

     https://doi.org/10.2522/ptj.20090075

     • PubMed
     • Google Scholar
109. 1. Xu J
     2. Branscheidt M
     3. Schambra H
     4. Steiner L
     5. Widmer M
     6. Diedrichsen J
     7. Goldsmith J
     8. Lindquist M
     9. Kitago T
     10. Luft AR
     11. Krakauer JW
     12. Celnik PA
     13. SMARTS Study Group

     (2019) Rethinking interhemispheric imbalance as a target for stroke neurorehabilitation

     Annals of Neurology 85:502–513.

     https://doi.org/10.1002/ana.25452

     • PubMed
     • Google Scholar
110. 1. Yuan H
     2. Liu T
     3. Szarkowski R
     4. Rios C
     5. Ashe J
     6. He B

     (2010) Negative covariation between task-related responses in alpha/beta-band activity and BOLD in human sensorimotor cortex: an EEG and fMRI study of motor imagery and movements

     NeuroImage 49:2596–2606.

     https://doi.org/10.1016/j.neuroimage.2009.10.028

     • PubMed
     • Google Scholar
111. 1. Zarkowski P
     2. Shin CJ
     3. Dang T
     4. Russo J
     5. Avery D

     (2006) EEG and the variance of motor evoked potential amplitude

     Clinical EEG and Neuroscience 37:247–251.

     https://doi.org/10.1177/155005940603700316

     • PubMed
     • Google Scholar
112. 1. Zrenner C
     2. Desideri D
     3. Belardinelli P
     4. Ziemann U

     (2018) Real-time EEG-defined excitability states determine efficacy of TMS-induced plasticity in human motor cortex

     Brain Stimulation 11:374–389.

     https://doi.org/10.1016/j.brs.2017.11.016

     • PubMed
     • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Spatially bivariate EEG-neurofeedback can manipulate interhemispheric rebalancing of M1 excitability" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Nicole Wenderoth as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by and Richard Ivry as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Tomas Ros (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) The main claim of the authors that the spatially bivariate EEG-neurofeedback lead to changes in interhemispheric inhibition is currently not sufficiently supported by the experimental evidence because (i) the authors did not demonstrate that IHI could be modulated bidirectionally, (ii) the training did not lead to a corresponding change in interhemispheric coherence and, therefore, no conclusions with regard to interhemispheric "rebalancing" can be drawn, and (iii) it is unclear whether the effects were mainly driven by excitability changes in the right hemisphere such that changes in IHI are merely an epiphenomenon. It will be important to address these concerns by presenting control analyses and/or additional data.

2) There is no significant difference in Network-intensity (better: interhemispheric coherence) during MI between HIGH and LOW sessions; the provided explanations in the discussion are highly speculative. The most obvious explanation is that interhemispheric coherence was simply not changed by MI and ERD feedback. How can this lack of an effect be reconciled with the main claim of the authors?

3) The paired-pulse TMS effects could result from modulating the excitability of the right motor cortex. It is important to show that IHI changed over and above the CS effect.

4) Bidirectional control of IHI would need to be significantly demonstrated above and below the resting baseline. Although the study shows bidirectional control of SMR power (in the ipsilateral hemisphere), this is strictly not the case for IHI (which shows only significant increases relative to rest).

5) The result section is difficult to understand because context information is often missing since the detailed Methods are reported at the end. Please revise the results to improve accessibility for the reader.

6) Some statements like "suggesting that the volitional control of SMR-ERDs in the closed-loop environment differs from spontaneous SMR fluctuations" require a direct statistical comparison

7) Revise correlation analyses to properly account for within versus between-subject variability

8) It remains an open question, whether the more sophisticated neurofeedback approach (of modulating both hemispheres separately) was indeed necessary to achieve the findings of this work.

9) The dynamic range of SMR modulation occurs mainly in the ipsilateral hemisphere (in terms of bidirectional control). The contralateral hemisphere consistently shows increased SMR power relative to rest, with no bidirectional modulation. Please explain why this was the case and discuss potential implications

10) Why was the gender ratio of the subjects (1:11, F:M) so highly skewed? This should be explained, and potential limitations should be discussed

11) The discussion of the neurophysiological mechanisms mediating IHI would benefit from more depth and a better embedding into the literature.

Reviewer #1 (Recommendations for the authors):

The manuscript reports the interesting phenomenon that selectively modulating the excitability of the right hemisphere changes interhemispheric inhibition from the right to the left hemisphere. While this is interesting, I'm wondering how the interhemispheric result relates to changes in corticospinal excitability as assessed via MEPs in response to the conditioning stimulus.

I understood that the CS has been chosen such that TS MEPs are reduced by approximately 50%. Was that done while participants were at rest or while they performed the MI task /BCI control?

If the CS intensity was determined at rest, did the authors analyse the CS MEPs and particularly how they change from IHI at rest to IHI during the MI/BCI control? Does IHI change independently of MEP modulation in response to the CS? This would be an important analysis also for the shuffled non-triggered trials since it might reveal whether corticospinal excitability and IHI can be controlled independently.

Please note that there has been some debate in the field whether or not the CS effect has to be kept constant across conditions. Thus, if there are differences in CS MEP amplitudes between the sessions, the authors might want to address this methodological issue in the discussion.

The result section is difficult to understand because context information is often missing since the detailed Methods are reported at the end. Please revise the results to improve accessibility for the reader. In that regard, I would recommend swapping the order in the result section and presenting first the EEG states and then the associated IHI measures.

What is the rationale for using right finger MI as the no-FB condition? Please explain

Line 257ff: there seems to be a mistake: «one-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) revealed significant differences F(2,44) = 1.43, p < 0.248,…» should probably read « revealed no significant differences».

Line 260 ff there seems to be a mistake: Although the IHI magnitude in the LOW session was larger than that in the HIGH and LOW sessions.… replace second LOW by MID.

I liked the control analysis of the authors shown in suppl. Figure 3-2, However, the conclusion that the LOW,MID,HIGH data shown in Figure 3B is different from the data shown in suppl. Figure 3-2 requires a direct statistical comparison.

Figure 4C: I'm not convinced that presenting the laterality index adds any information to Figure 4 A and B. In fact, I found the caption confusing since it seems to suggest that (i) the negative LI for LOW results from contralateral activity which is inconsistent with the topoplot that shows high activity in the ipsilateral hemisphere; and (ii) the positive LI for HIGH results from ipsilateral activity which is also inconsistent with the topo which shows desynchronisation in the ipsilateral hemisphere. Given the shown in panels A and B, the LI measure seems inappropriate since there are virtually no activity changes in the contralateral hemisphere while the ipsilateral hemisphere shows either synchronized or desynchronized SMR activity.

Associations between IHI magnitude and bilateral EEG patterns: The authors might want to consider calculating a repeated measures correlation because it does not seem meaningful to pool data across the LOW, MID, HIGH sessions.

Discussion

Line 470ff: The first sentence of the discussion follows a reversed logic. Participants did not consciously self-regulate IHI but they modulated their EEG pattern based on the closed-loop spatially bivariate BCI-based neurofeedback. Changes in IHI were the consequence of self-regulating the EEG activity pattern. Please revise throughout the discussion.

The discussion of the neurophysiological mechanisms mediating IHI could be a bit more in-depth. The review of Carson RG. J Physiol. 2020 Nov;598(21):4781-4802 might help.

The authors argue that "This multimodal EEG-TMS approach proves that the bilateral α and β activities observed in the current study served to regulate the inhibition and facilitation of inhibitory functional coupling of interhemispheric interaction over motor cortices.» This argument would be strengthened if the interhemispheric effects were independent of the general excitability of the ipsilateral hemisphere as reflected in CS MEP amplitudes. Only then, the authors can argue that their intervention specifically modulated interhemispheric circuits.

Lines 538-544: Explanatory theoretical model: I cannot follow the argument of the authors. Is it possible that IHI and "network intensity" simply reflect the activity of different neural populations? IHI effects are rather specific and I doubt that this could be picked up via general EEG markers.

Line 561ff This seems to be consistent «Furthermore, the effect size of IHI manipulation in the current study (Cohen's d = 1.50) was approximately 1-2-fold higher than that of representative tDCS (Cohen's d = 1.55)…».

Discussion of IHI in stroke: As stated by the authors, it is somewhat controversial whether IHI is a useful therapy target early after stroke. However, their EEG-neurofeedback approach allows to selectively up- or down-regulate one hemisphere which might be highly relevant in the context of stroke recovery both from a basic science and application perspective. It might be worthwhile to expand this aspect of the discussion.

Reviewer #2 (Recommendations for the authors):

Although the Introduction is comprehensive in terms of the mechanistic relation between EEG and TMS variables, it does not sufficiently include the "clinical" significance of modifying IHI e.g. in terms of rehabilitation potential. It would be nice to include a few references in clinical populations where IHI and/or SMR are impacted, or where IHI/SMR have been used as prognostic markers of motor recovery. I would also include a sentence or two in the Abstract underlining the potential clinical significance of manipulating IHI.

In terms of related studies combining the TMS/EEG/BCI trio, the work by Ros et al. 2010 was one of the first to demonstrate that MEP measures of cortical excitability and inhibition (MEP, SICI) can be modulated using closed-loop feedback of oscillations from the motor cortex.

Ros, T., Munneke, M.A.M., Ruge, D., Gruzelier, J.H. and Rothwell, J.C. (2010), Endogenous control of waking brain rhythms induces neuroplasticity in humans. European Journal of Neuroscience, 31: 770-778. https://doi.org/10.1111/j.1460-9568.2010.07100.x

It is not clear what behavioural and neural variable the authors are referring to in the first Results section starting with. "We confirmed the difficulties of each session" (line 171).

Line 171: "The mean value of the sum of the triggered trials". Mean and sum have different definitions, do you mean 'the mean value of all the triggered trials'?

There is no reference to what the abbreviation 'TS' stands for (I am assuming it is "test stimulus")

Line 219: The manipulation range of IHI (i.e., the difference between HIGH and LOW sessions) was 32.6. However elsewhere the means (Line 222/ 223) are separately indicated as HIGH: 62.0 and LOW: 96.9. Their difference (96.9 – 62.0) is 34.9.

Line 263: "suggesting that the volitional control of SMR-ERDs in the closed-loop environment differs from spontaneous SMR fluctuations". This depends on how one defines the word "differs". In this case, the comparison is technically qualitative as the contrast is between a significant result (volitional control, p <0.05) and a non-significant one (spontaneous fluctuation, p > 0.05). However, a more direct comparison would involve comparing the two effects directly (volitional vs spontaneous) by statistically testing for a significant difference in effect sizes (e.g. using a two-way rmANOVA).

Taking the effect size of the HIGH-vs-LOW contrast as an example, for volitional control this is d = 1.42, while for spontaneous fluctuation it is d = 0.49, which yields a differential d = 0.93 (1.42 -0.49). This is likely statistically significant but it will depend on the error term, hence the need for a formal two-way rmANOVA analysis with an interaction effect. This should be done for all the contrast analyses in the paper (including SMR power), instead of using the one-way ANOVA.

The dynamic range of SMR modulation occurs mainly in the ipsilateral hemisphere (in terms of bidirectional control). The contralateral hemisphere consistently shows increased SMR power relative to rest, with no bidirectional modulation. How do the authors explain this ? Could this be due to how the neurofeedback was implemented for each hemisphere (thresholds/rewards)? Right hand dominance? If the latter, where there any differential effects in the subgroup of left hand dominant participants?

Line 580: "the current technique can be tailored either to upregulate the damaged hemisphere, down-regulate the intact hemisphere…" and Line 585: "In conclusion, we presented an innovative approach to voluntarily and bidirectionally manipulate the state of IHI"

Critically, bidirectional control would need to be significantly demonstrated above and below the resting baseline. Although the study shows bidirectional control of SMR power (in the ipsilateral hemisphere), this is strictly not the case for IHI (which shows only significant increases relative to rest). Hence, the term "bidirectional" should not be used in the context of the IHI, but rather the phrase "control of interhemispheric disinhibition". Hence, the authors could state that the current results should motivate future studies investigating whether upregulation of IHI in the damaged hemisphere is possible above and beyond spontaneous resting values.

Reviewer #3 (Recommendations for the authors):

This is a well-conducted study building and expanding on the pioneering work of this group (Takemi et al. 2013, 2015). The novel contribution of this study is that self-regulated ERD is not modulating the TMS effects on the periphery (MEP) only, but also the TMS effects on the opposite brain hemisphere. This has been probed by IHI, but may probably have also been shown by TEP.

All other interpretations with regard to "rebalancing", "pathway-specificity", "interhemispheric connectivity", and "challenging the concept of neurorehabilitation" etc. are far fetching and should be toned down. Such statements would have necessitated (i) either direct neurofeedback of interhemispheric coherence, (ii) or demonstration of a direct link between the self-regulated ERD and interhemispheric coherence, and (iii) the application in patient populations.

Therefore, it is advisable to better link this study to the large body of literature in the IHI field, and discuss where this work adds additional insights. It is surprising, for example, that relevant previous literature is not referred to and discussed: Liang et al. (2014) have shown that imagined movements change IHI, Nelson et al. (2009) demonstrated bi-directional IHI, and Morishita et al. (2011, 2012) contributed with regard to ipsilateral inhibitory circuits and IHI. This work should be cited.

The weaknesses listed above should also be acknowledged as limitations in the discussion.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Spatially bivariate EEG-neurofeedback can manipulate interhemispheric inhibition" for further consideration by eLife. Your revised article has been evaluated by Richard Ivry (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

Based on the new analyses, it is unclear whether the original finding that interhemispheric inhibition can be modulated by a new EEG neurofeedback approach does still hold. Reviewer #1 points out an alternative statistical approach that might strengthen the original argument of the authors while controlling for changes in the excitability of the modulated hemisphere.

Reviewer #3 points out that it is still unclear whether bihemispheric EEG is superior to unihemispheric EEG. The authors might want to address this comment by adding an in-depth discussion of how their new neurofeedback method compares with existing methods with regard to modulating interhemispheric interactions.

Reviewer #1 (Recommendations for the authors):

The authors provided very thorough and extensive revisions and they have addressed most of my comments. It is particularly reassuring to see the additional analyses of corticomotor excitability as probed by the CS.

However, I'm still not completely convinced by the author's statistical approach.

If I understood the approach correctly, % change in IHI and % change of CS MEP amplitude were entered into one rmANOVA model and the authors tested for a condition x index interaction effect (which did not reach significance). This is a rather strict comparison. A more traditional approach would be to add % CS change as a covariate of no interest and test whether the IHI effect survives even if the covariate is added. If that is the case, the authors have demonstrated that IHI changes over and above the variance explained by the CS effect which would support their initial hypothesis that IHI can be modulated via their EEG-BCI and is not simply an epiphenomenon. No other statistical analyses would be required. However, the discussion would need to be modified!

The authors also analysed changes in CS MEP amplitude across conditions. CS MEPs did not significantly differ across the different EEG-BCI conditions. However, the absence of a significant effect does not allow the authors to conclude that there is "no effect". This would require a Bayesian approach which would test whether the data provides evidence in favour of H0. This could be another argument that IHI changes over and above CS changes. However, I don't think that these additional analyses would be needed if the co-variance analysis of IHI would reach significance. Maybe the authors might want to consult a stats expert for this analysis.

Reviewer #3 (Recommendations for the authors):

I am satisfied with the author's responses and revisions, which are extensive and much more consistent with the collected evidence from the experiment.

Ultimately, I am ambivalent (i.e. neither for nor against) on recommending this paper for publication in eLife based on its scientific significance. On a positive note, the authors clearly demonstrate that it is possible to modulate IHI (albeit unidirectionally) using bivariate neurofeedback, and this is an original and scientifically novel contribution. On the other hand, from the experiments conducted, it is not clear whether the bivariate neurofeedback approach is critical for the observed findings since there is no comparison to the conventional BCI-based neurofeedback that exploits SMR-ERD from one hemisphere. Hence the results show evidence of "sufficient" conditions to reduce inter-hemispheric inhibition (IHI) but do not provide evidence of "necessary" conditions i.e. whether uni-hemispheric neurofeedback would lead to similar effects but with a much more parsimonious experimental setup that does not involve controlling EEG from both hemispheres.

I, therefore, leave it to the senior editor for the ultimate decision regarding publication, taking also into account the other reviewers' comments.

Essential revisions:

1) The main claim of the authors that the spatially bivariate EEG-neurofeedback lead to changes in interhemispheric inhibition is currently not sufficiently supported by the experimental evidence because (i) the authors did not demonstrate that IHI could be modulated bidirectionally, (ii) the training did not lead to a corresponding change in interhemispheric coherence and, therefore, no conclusions with regard to interhemispheric "rebalancing" can be drawn, and (iii) it is unclear whether the effects were mainly driven by excitability changes in the right hemisphere such that changes in IHI are merely an epiphenomenon. It will be important to address these concerns by presenting control analyses and/or additional data.

Overall, we agree with these opinions. We addressed all concerns corresponding as follows: (i) the response to No.4 below; (ii) the response to No.2 below; (iii) the response to No.3 below.

2) There is no significant difference in Network-intensity (better: interhemispheric coherence) during MI between HIGH and LOW sessions; the provided explanations in the discussion are highly speculative. The most obvious explanation is that interhemispheric coherence was simply not changed by MI and ERD feedback. How can this lack of an effect be reconciled with the main claim of the authors?

Our results presented a spatially bivariate EEG-based neurofeedback approach that enables healthy participants to independently self-regulate excitability of the left versus right hemispheres resulting in IHI changes. However, interhemispheric coherence was not associated with ipsilateral excitability (Figure 5D). We would like to emphasize that this exploratory analysis only provides a limited minor result.

We think that our results are reasonable since it is possible that interhemispheric coherence did not reflect inhibitory modulation via neurofeedback; IHI and Network-intensity may reflect the activity of different neural populations. Our explanatory theoretical model (Lines 538-544 in the previous manuscript) has not been fully clarified based on the experimental results, so we have removed the relevant text and simplified the discussion. It is unclear whether fMRI-based neurofeedback using interhemispheric functional connectivity is linked to ipsilateral excitability, and whether it modulates IHI magnitude.

In addition, the interhemispheric coherence was calculated with one-to-one electrodes, while the Network-intensity is an expanded index for evaluation between multiple electrodes. In this study, Network-intensity was adopted to evaluate coherence involving not only C3 (or C4) but also peripheral multiple electrodes. Previous studies described that Network-intensity was calculated from the corrected imaginary part of coherence (ciCOH) (Ewald et al., 2012; Hayashi et al., 2020; Vukelić and Gharabaghi, 2015). Interhemispheric Network-intensity was computed as follows (written in “Connectivity analysis”): ${\displaystyle \text{Network intensity}\left(f\right)=\sum _{cont=1}^7\sum _{ipsi=1}^7{\text{ciCOH}}_{cont,ipsi}(f)}$ where Network-intensity is computed between left and right hemisphere, ∑∑ciCOH is the sum of significant ciCOH values for interhemispheric interaction, cont denotes the 7 channels of interest (i.e., C3 and its neighboring 6 channels), ipsi denotes the 7 channels of interest in the opposite hemisphere (i.e., C4 and its neighboring 6 channels), and f indicates the frequency of interest (e.g., the predefined α frequencies).

We have revised the related parts of the manuscript.

[Lines 858-860] “There was a significant difference in interhemispheric functional connectivity between HIGH and MID sessions, but not between HIGH and LOW sessions.”

[Lines 1142-1157] “However, even fMRI-based neurofeedback with high spatial resolution, based on blood oxygenation level-dependent signal, cannot easily distinguish between the excitatory and the inhibitory activities from each region (Moon et al., 2021). Additionally, neurofeedback using interhemispheric functional connectivity has the possibility to counteract the modulation of IHIs from the right to left hemisphere and from the left to right hemisphere. Our results showed no significant difference in Network-intensity during MI between HIGH and LOW sessions (Figure 5D), indicating that interhemispheric functional connectivity at the EEG level was not changed, whereas IHI was modulated. This may indicate that IHI and Network-intensity simply reflect the activity of different neural populations; interhemispheric projections act via surround/lateral inhibition in the sensory and motor cortices (Carson, 2020), while the synchrony at the EEG level reflects changes in the activity of local interactions between pyramidal neurons and interneurons in the thalamocortical loops (Pfurtscheller and Lopes da Silva, 1999). Furthermore, it is possible that Network-intensity reflects the degree of bidirectional synchrony between the two motor cortices, while IHI in this study reflects one-way (right-to-left) inhibitory effects.”

[Lines 1157-1159] “Therefore, it is unclear whether fMRI-based neurofeedback using interhemispheric functional connectivity is linked to ipsilateral excitability, and whether it modulates IHI magnitude.”

3) The paired-pulse TMS effects could result from modulating the excitability of the right motor cortex. It is important to show that IHI changed over and above the CS effect.

We appreciate your valuable comments. We have added additional results related to changes in corticospinal excitability as assessed by MEPs in the left FDI in response to the conditioning stimulus over the right hemisphere (Figure 4—figure supplements 3B). A direct statistical comparison between increase in IHI and left MEP amplitude is shown in Figure 4—figure supplements 3C. To compare the two indices, the percent changes based on the values in NoFB session were calculated. The stimulus intensity of the CS during the experiment was set such that the CS evoked MEP of peak-to-peak amplitude from the left FDIs (SI_1mV) “at rest” (Lines 1482-1484, “The stimulus intensities of the left and right M1 to evoke MEP …”). In addition, we analyzed the non-triggered trials to show the CS effect (Figure 4—figure supplements 3D).

Based on the additional results, the MEP amplitudes in the left FDI elicited by CS did not vary across sessions, indicating that the modulation of the ipsilateral SMR-ERD caused no changes in corticospinal excitability in the contralateral side. Although we successfully revealed that modulating the excitability of the right hemisphere changes IHI from the right to the left hemisphere, a significant interaction between IHI modulation and changes in the left MEP amplitude was not observed. Thus, the specificity of the IHI manipulation could not be fully demonstrated. To confirm whether the participants learned pathway-specific IHI manipulation, the CS effect should have been kept constant across the conditions by adjusting the CS intensity to evoke left MEP of 1 mV peak-to-peak amplitude. Because such an experiment was not possible in the present study due to the time limitation (the experimental time would need to be doubled to 5–6 h to perform such an experiment), further studies are warranted.

We have toned down our results and conclusions, such as the use of the term "pathway-specific" manipulation, since our experiment did not directly demonstrate whether changes in interhemispheric inhibition and corticospinal excitability are independent phenomena. We have revised the manuscript to include these additional analyses, figures, and limitations.

[Lines 555-599] “Although we revealed that modulating the excitability of the right hemisphere changes IHI magnitude from the right to the left hemisphere, the specificity of the IHI manipulation was not proved yet. We presented additional results related to changes in corticospinal excitability as assessed by MEPs in the left FDI in response to the CS over the right hemisphere (Figure 4—figure supplements 3B). A one-way rmANOVA for sessions (five levels: REST, NoFB, HIGH, MID, and LOW) revealed no significant differences in the left MEP amplitude between all sessions (F_(4,101) = 0.39, p = 0.813, η² = 0.02). Furthermore, in order to investigate whether changes in IHI from the right to the left hemisphere and changes in corticospinal excitability in the right hemisphere are independent phenomena, a direct statistical comparison between the normalized IHI and the normalized left MEP was applied. During the triggered TMS trials, a two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and indices (two levels: increase in IHI and left MEP amplitude) revealed a significant main effect for sessions (F_{(2, 122)} = 3.96, p = 0.022, η² = 0.06), but no main effect for indices (F_{(1, 122)} = 1.11, p = 0.294, η² = 0.01) and for interaction (F_{(2, 122)} = 1.65, p = 0.195, η² = 0.03). A post-hoc two-tailed paired t-test in the increase in IHI showed significant differences between HIGH and MID sessions (Cohen's d = 1.06, p = 0.011), and between HIGH and LOW sessions (Cohen's d = 1.42, p < 0.001). In contrast, a post-hoc two-tailed paired t-test in the left MEP amplitude showed no significant differences between HIGH and LOW sessions (Cohen's d = 0.16, p = 1.00). During the nontriggered TMS trials, a two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and indices (two levels: increase in IHI and eft MEP amplitude) revealed no main effect for sessions (F_{(2, 122)} = 0.95, p = 0.390, η² = 0.02), for indices (F_{(1, 122)} = 0.81, p = 0.369, η² = 0.01) and for interaction (F_{(2, 122)} = 0.49, p = 0.614, η² = 0.01). Therefore, we successfully revealed that modulating the excitability of the right hemisphere changes IHI from the right to the left hemisphere, but a significant interaction between IHI modulation and changes in the left MEP amplitude was not observed.”

[Lines 1276-1317] “Our study has several limitations that need to be considered when interpreting the results. … Additionally, we successfully revealed that modulating the excitability of the right hemisphere changes IHI from the right to the left hemisphere, but the specificity of the IHI manipulation could not be fully demonstrated (Figure 4—figure supplements 3B and C). … To confirm whether changes in interhemispheric inhibition and corticospinal excitability are independent phenomena, the CS effect should have been kept constant across the conditions by adjusting the CS intensity to evoke left MEP of 1 mV peak-to-peak amplitude. Because such an experiment was not possible in the present study due to time limitation (the experimental time would need to be doubled to 5–6 h to perform such an experiment), future studies will be needed.”

[Lines 1853-1860] “The statistical differences between corticospinal excitability as assessed by MEPs in the left FDI in response to the CS over the right hemisphere was tested by a one-way rmANOVA for sessions (five levels: REST, NoFB, HIGH, MID, and LOW). To investigate whether changes in interhemispheric inhibition and corticospinal excitability are independent phenomena, a direct statistical comparison using a two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and indices (two levels: increase in IHI and eft MEP amplitude) was applied in the triggered and non-triggered TMS trials.”

4) Bidirectional control of IHI would need to be significantly demonstrated above and below the resting baseline. Although the study shows bidirectional control of SMR power (in the ipsilateral hemisphere), this is strictly not the case for IHI (which shows only significant increases relative to rest).

Although we successfully demonstrated the up- and down-regulate the ipsilateral SMR-ERD (i.e., bidirectional changes) at EEG-level, the bidirectional changes in IHI from the right to the left hemisphere were not statistically shown since there is no significant difference in MID and LOW sessions (Figure 4B). Therefore, we have revised the expression of the “bidirectional” modulation in IHI and added the future experiment in the Limitation.

[Lines 1277-1282] “Although we successfully demonstrated the bidirectional changes in the ipsilateral SMR-ERD, the bidirectional changes in IHI from the right to the left hemisphere were not statistically significant since there was no significant difference between the MID and LOW sessions (Figure 4B). The current results should prompt future studies to ascertain whether down-regulation of IHI (i.e., disinhibition) during MI and up-regulation of IHI at rest in the non-dominant hemisphere (intact hemisphere) are possible in a similar setup.”

[Lines 1350-1355] “In conclusion, we presented an innovative approach to manipulate the state of IHI, by directly and bidirectionally modulating SMR-ERDs in a spatially bivariate BCI-based neurofeedback paradigm.”

5) The result section is difficult to understand because context information is often missing since the detailed Methods are reported at the end. Please revise the results to improve accessibility for the reader.

According to this suggestion, we have revised throughout the manuscript and swapped the order in the results to improve accessibility for the reader.

e.g., explanations of “triggered trials” and “waiting time” [Lines 247-251], stimulus intensities during TMS experiment [Lines 351-356], explanations of each five sessions including REST and NoFB sessions [Lines 218-220], and how to evaluate interhemispheric functional connectivity [Lines 804-808].

In addition, values corresponding to each statistical figure are presented in each table (Supplementary File 1) for readability.

6) Some statements like "suggesting that the volitional control of SMR-ERDs in the closed-loop environment differs from spontaneous SMR fluctuations" require a direct statistical comparison

We have revised to the correct sentences as below.

In addition, we performed a direct statistical comparison and replaced the result in Figure 4—figure supplements 2.

[Lines 455-554] “A two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and trials (two levels: triggered TMS trials and non-triggered TMS trials) revealed a significant main effect for sessions (F_{(2, 117)} = 5.08, p = 0.008, η² = 0.08), but no main effect for trials (F_{(1, 117)} = 0.32, p = 0.575, η² < 0.01) and for interaction (F_{(2, 117)} = 1.82, p = 0.167, η² = 0.03). Thus, no significant difference in IHI between the volitional control of SMR-ERDs in the closed-loop environment differs from spontaneous SMR fluctuations was observed. A post-hoc two-tailed paired t-test showed significant differences between HIGH and MID sessions (Cohen's d = 0.94, p = 0.017), and between HIGH and LOW sessions (Cohen's d = 1.36, p < 0.001) in the triggered TMS trials. In contrast, in the non-triggered TMS trials, a posthoc two-tailed paired t-test showed no significant differences between HIGH and MID sessions (Cohen's d = 0.20, p = 1.00), and between HIGH and LOW sessions (Cohen's d = 0.28, p = 1.00).”

[Lines 1850-1853] “To investigate the influence of spontaneous SMR fluctuations on IHI, a two-way rmANOVA for sessions (HIGH, MID, and LOW) and trials (triggered TMS trial and non-triggered TMS trial) as the withinparticipant factors.”

7) Revise correlation analyses to properly account for within versus between-subject variability

In accordance with your comments regarding the across-participant corrections between IHI magnitude and bilateral EEG patterns, we performed a repeated measures correlation since it was not reasonable to combine different sessions of different participants. Simple correlation is often applied to non-independent observations or aggregated data. A previous study argued that it may produce biased, specious results due to violation of independence and/or differing patterns (across-participants vs. within-participants) (Bakdash and Marusich, 2017).

We also calculated within-participant corrections between IHI magnitude and bilateral EEG patterns in accordance with the suggestion by Reviewer 3. We have added the overall significance of the 22 individual correlations using a chi-squared test and the distribution of correlation coefficients across all participants. In the analysis, data from HIGH, MID, and LOW sessions were aggregated as independent observations. We have integrated the above suggestions and modified the text as follows.

[Lines 1772-1776] “In the across-participant corrections between contralateral or ipsilateral SMR-ERDs and IHI magnitude, we performed a repeated measures correlation. A previous study argued that it may produce biased, specious results due to violation of independence and/or differing patterns across-participants versus withinparticipants (Bakdash and Marusich, 2017). “

[Lines 717-803] “In the across-participant correlation between IHI magnitude and contralateral or ipsilateral SMR-ERDs, we performed a repeated measures correlation analysis (Bakdash and Marusich, 2017). We found a significant correlation in the ipsilateral SMR-ERD (r_rm = 0.436, p = 0.004; Figure 5B), but not in the contralateral SMR-ERD (r_rm = 0.008, p = 0.960; Figure 5C).”

8) It remains an open question, whether the more sophisticated neurofeedback approach (of modulating both hemispheres separately) was indeed necessary to achieve the findings of this work.

Spatially bivariate BCI-based neurofeedback enables us to modulate the ipsilateral SMR-ERD bidirectionally while maintaining constant contralateral SMR-ERD. It provided new findings that the ipsilateral SMR-ERD influenced the IHI magnitude. As you mentioned, it remains an open question, whether the conventional BCI-based neurofeedback that exploits SMR-ERD from one hemisphere is substitutable to achieve the findings of this work. In this context, in case the contralateral SMR-ERD was up- or down-regulated instead of maintained at a moderate level, it is unclear whether the ipsilateral SMR-ERD was modulated bidirectionally and whether the IHI changed with it. The study performed a total of 11 sessions, including REST and NoFB, and the combinations of SMR-ERDs in both hemispheres (3 conditions vs. 3 conditions) may be worth evaluating in the future.

We have added future directions as follows:

[Lines 1329-1339] “Moreover, in case the contralateral SMR-ERD was up- or down-regulated instead of maintained at a moderate level, it is unclear whether the ipsilateral SMR-ERD was modulated bidirectionally and whether the IHI changed with it. The study performed a total of 11 sessions, including REST and NoFB, and the combinations of SMR-ERDs in both hemispheres (3 conditions vs. 3 conditions) may be worth evaluating in the future. In addition, it is not clear whether the bivariate neurofeedback approach indeed contributed to the observed findings since there is no comparison to the conventional BCI-based neurofeedback that exploits SMR-ERD from one hemisphere. A future study is required to investigate whether the conventional univariate neurofeedback that exploits SMR-ERD from one hemisphere achieves bidirectional modulation of the ipsilateral SMR-ERD and accompanying IHI magnitude.”

9) The dynamic range of SMR modulation occurs mainly in the ipsilateral hemisphere (in terms of bidirectional control). The contralateral hemisphere consistently shows increased SMR power relative to rest, with no bidirectional modulation. Please explain why this was the case and discuss potential implications

We aimed to modulate the ipsilateral hemisphere bidirectionally while maintaining constant the contralateral excitability using our spatially bivariate BCI-based neurofeedback. To evaluate the effect of SMR modulation in the ipsilateral hemisphere, participants always aimed to maintain constant contralateral SMR-ERD. As written in the manuscript ("Spatially bivariate BCI-based neurofeedback"), visual feedback was provided on a computer screen in the form of cursor movements in a two-dimensional coordinate, in which x- and y-axis corresponded to the degree of the ipsilateral and contralateral SMR-ERD, respectively. For example, in HIGH session, participants were instructed to move the cursor toward the middle right (x > 0, y = 0) in the two-dimensional coordinate, the position showed a strong SMR-ERD in the ipsilateral hemisphere and moderate SMR-ERD in the contralateral hemisphere. On the contrary, in LOW session, participants were instructed to move the cursor toward the middle left (x < 0, y = 0) in the two-dimensional coordinate, the position showed a low SMR-ERD (or strong SMR-ERS) in the ipsilateral hemisphere and moderate SMR-ERD in the contralateral hemisphere. Therefore, spatially bivariate BCI-based neurofeedback enables us to modulate the ipsilateral SMR-ERD bidirectionally while maintaining constant contralateral SMR-ERD in the three sessions (i.e., HIGH, MID and LOW). In fact, Figures 2A and B showed the intended results. To help readers’ understanding, we prepared a sample video as Video 1.

The above is the answer to your point, but we notice that we need to describe the experimental limitation regarding (1) handedness and (2) the other patterns of the contralateral excitability. In this study, all participants were right-handed (Laterality Quotient: 72.2 ± 30.9%), as assessed by the Edinburgh Inventory (Oldfield, 1971), to eliminate bias due to the dominance. If a left-handed participant performs the same experiment, the modulation effect via BCI-based neurofeedback may be different. Moreover, in case the contralateral SMR-ERD was up- or downregulated instead of maintained at a moderate level, it is unclear whether the ipsilateral SMR-ERD was modulated bidirectionally and whether the IHI changed with it. The study performed a total of 11 sessions, including REST and NoFB, and the combinations of SMR-ERDs in both hemispheres (3 conditions vs. 3 conditions) may be worth evaluating in the future.

[Line 1325-1334] “For the limitations of the volitional control of SMR-ERD during unilateral MI, all participants were right-handed (Laterality Quotient: 72.2 ± 30.9%), as assessed by the Edinburgh Inventory (Oldfield, 1971), to eliminate bias due to the dominance. If a left-handed participant performs the same experiment, the modulation effect via BCI-based neurofeedback may be different. Moreover, in case the contralateral SMR-ERD was up- or downregulated instead of maintained at a moderate level, it is unclear whether the ipsilateral SMR-ERD was modulated bidirectionally and whether the IHI changed with it. The study performed a total of 11 sessions, including REST and NoFB, and the combinations of SMR-ERDs in both hemispheres (3 conditions vs. 3 conditions) may be worth evaluating in the future.”

10) Why was the gender ratio of the subjects (1:11, F:M) so highly skewed? This should be explained, and potential limitations should be discussed

The study participants were freely recruited from the university without any criteria for sex; there was no experimental plan to equalize the sex ratio. Because there were few female participants in the study, it is unclear whether our findings can be generalized to females. Previous studies revealed that the brain microstructure, including corpus callosum, differed between men and women (Dunst et al., 2014; Menzler et al., 2011). It is unknown whether these sex differences influenced our IHI study and additional studies are required to determine this.

We added this limitation as follows:

[Lines 1341-1347] “Finally, the sex ratio in this study was the consequence of recruiting freely from the university; there was no experimental plan to equalize the sex ratio. Because there were few female participants in the study, it is unclear whether our findings can be generalized to females. Previous studies revealed that the brain microstructure, including corpus callosum, differed between men and women (Dunst et al., 2014; Menzler et al., 2011). It is unknown whether these sex differences influenced our IHI study and additional studies are required to determine this.”

11) The discussion of the neurophysiological mechanisms mediating IHI would benefit from more depth and a better embedding into the literature.

We have added the information related to the neurophysiological mechanisms mediating IHI. In addition, the review by Carson provided new insight into reassessing the inter-hemispheric competition model. The alternative conceptual framework, which states that IHI acts to sculpt the output of neural circuits by co-opting the two cerebral hemispheres, is shown to be consistent with the known associations between the structural integrity of callosal projections and the magnitude of the motor deficits after stroke. This may emphasize the benefit of our neurofeedback approach in the context of neurorehabilitation.

With have added several additional references and revised the discussion as follows:

[Lines 1150-1155] “This may indicate that IHI and Network-intensity simply reflect the activity of different neural populations; interhemispheric projections act via surround/lateral inhibition in the sensory and motor cortices (Carson, 2020), while the synchrony at the EEG level reflects changes in the activity of local interactions between pyramidal neurons and interneurons in the thalamocortical loops (Pfurtscheller and Lopes da Silva, 1999).”

[Lines 1265-1270] “Furthermore, a previous review advocated the conceptual framework and suggested that a fundamental role of IHI is to support the narrowing of excitatory focus by co-opting the capacities of the two cerebral hemispheres (Carson, 2020). This framework is consistent with the known associations between the structural integrity of callosal projections and the magnitude of the motor deficits after stroke (Auriat et al., 2015; Granziera et al., 2012; Koh et al., 2018).”

[Lines 1307-1311] “Previous studies suggested that the IHI magnitude and the corticomotor excitability were modulated independently (Morishita et al., 2011, 2012). Although it is difficult to directly compare these results due to the significantly different motor tasks and experimental settings, the assertion that the projections from the transcallosal and the corticospinal pathways are distinct supports our findings.”

Reviewer #1 (Recommendations for the authors):

The manuscript reports the interesting phenomenon that selectively modulating the excitability of the right hemisphere changes interhemispheric inhibition from the right to the left hemisphere. While this is interesting, I'm wondering how the interhemispheric result relates to changes in corticospinal excitability as assessed via MEPs in response to the conditioning stimulus.

I understood that the CS has been chosen such that TS MEPs are reduced by approximately 50%. Was that done while participants were at rest or while they performed the MI task /BCI control?

If the CS intensity was determined at rest, did the authors analyse the CS MEPs and particularly how they change from IHI at rest to IHI during the MI/BCI control? Does IHI change independently of MEP modulation in response to the CS? This would be an important analysis also for the shuffled non-triggered trials since it might reveal whether corticospinal excitability and IHI can be controlled independently.

Please note that there has been some debate in the field whether or not the CS effect has to be kept constant across conditions. Thus, if there are differences in CS MEP amplitudes between the sessions, the authors might want to address this methodological issue in the discussion.

We appreciate your valuable comments. We have added additional results related to changes in corticospinal excitability as assessed by MEPs in the left FDI in response to the conditioning stimulus over the right hemisphere (Figure 4—figure supplements 3B). A direct statistical comparison between increase in IHI and left MEP amplitude is shown in Figure 4—figure supplements 3C. To compare the two indices, the percent changes based on the values in NoFB session were calculated. The stimulus intensity of the CS during the experiment was set such that the CS evoked MEP of peak-to-peak amplitude from the left FDIs (SI_1mV) “at rest” (Lines 1482-1484, “The stimulus intensities of the left and right M1 to evoke MEP …”). In addition, we analyzed the non-triggered trials to show the CS effect (Figure 4—figure supplements 3D).

Based on the additional results, the MEP amplitudes in the left FDI elicited by CS did not vary across sessions, indicating that the modulation of the ipsilateral SMR-ERD caused no changes in corticospinal excitability in the contralateral side. Although we successfully revealed that modulating the excitability of the right hemisphere changes IHI from the right to the left hemisphere, a significant interaction between IHI modulation and changes in the left MEP amplitude was not observed. Thus, the specificity of the IHI manipulation could not be fully demonstrated. To confirm whether the participants learned pathway-specific IHI manipulation, the CS effect should have been kept constant across the conditions by adjusting the CS intensity to evoke left MEP of 1 mV peak-to-peak amplitude. Because such an experiment was not possible in the present study due to the time limitation (the experimental time would need to be doubled to 5–6 h to perform such an experiment), further studies are warranted.

We have toned down our results and conclusions, such as the use of the term "pathway-specific" manipulation, since our experiment did not directly demonstrate whether changes in interhemispheric inhibition and corticospinal excitability are independent phenomena. We have revised the manuscript to include these additional analyses, figures, and limitations.

[Lines 555-599] “Although we revealed that modulating the excitability of the right hemisphere changes IHI magnitude from the right to the left hemisphere, the specificity of the IHI manipulation was not proved yet. We presented additional results related to changes in corticospinal excitability as assessed by MEPs in the left FDI in response to the CS over the right hemisphere (Figure 4—figure supplements 3B). A one-way rmANOVA for sessions (five levels: REST, NoFB, HIGH, MID, and LOW) revealed no significant differences in the left MEP amplitude between all sessions (F_(4,101) = 0.39, p = 0.813, η² = 0.02). Furthermore, in order to investigate whether changes in IHI from the right to the left hemisphere and changes in corticospinal excitability in the right hemisphere are independent phenomena, a direct statistical comparison between the normalized IHI and the normalized left MEP was applied. During the triggered TMS trials, a two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and indices (two levels: increase in IHI and left MEP amplitude) revealed a significant main effect for sessions (F_{(2, 122)} = 3.96, p = 0.022, η² = 0.06), but no main effect for indices (F_{(1, 122)} = 1.11, p = 0.294, η² = 0.01) and for interaction (F_{(2, 122)} = 1.65, p = 0.195, η² = 0.03). A post-hoc two-tailed paired t-test in the increase in IHI showed significant differences between HIGH and MID sessions (Cohen's d = 1.06, p = 0.011), and between HIGH and LOW sessions (Cohen's d = 1.42, p < 0.001). In contrast, a post-hoc two-tailed paired t-test in the left MEP amplitude showed no significant differences between HIGH and LOW sessions (Cohen's d = 0.16, p = 1.00). During the nontriggered TMS trials, a two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and indices (two levels: increase in IHI and eft MEP amplitude) revealed no main effect for sessions (F_{(2, 122)} = 0.95, p = 0.390, η² = 0.02), for indices (F_{(1, 122)} = 0.81, p = 0.369, η² = 0.01) and for interaction (F_{(2, 122)} = 0.49, p = 0.614, η² = 0.01).”

Therefore, we successfully revealed that modulating the excitability of the right hemisphere changes IHI from the right to the left hemisphere, but a significant interaction between IHI modulation and changes in the left MEP amplitude was not observed.

[Lines 1276-1317] “Our study has several limitations that need to be considered when interpreting the results. … Additionally, we successfully revealed that modulating the excitability of the right hemisphere changes IHI from the right to the left hemisphere, but the specificity of the IHI manipulation could not be fully demonstrated (Figure 4—figure supplements 3B and C). … To confirm whether changes in interhemispheric inhibition and corticospinal excitability are independent phenomena, the CS effect should have been kept constant across the conditions by adjusting the CS intensity to evoke left MEP of 1 mV peak-to-peak amplitude. Because such an experiment was not possible in the present study due to time limitation (the experimental time would need to be doubled to 5–6 h to perform such an experiment), future studies will be needed.”

[Lines 1853-1860] “The statistical differences between corticospinal excitability as assessed by MEPs in the left FDI in response to the CS over the right hemisphere was tested by a one-way rmANOVA for sessions (five levels: REST, NoFB, HIGH, MID, and LOW). To investigate whether changes in interhemispheric inhibition and corticospinal excitability are independent phenomena, a direct statistical comparison using a two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and indices (two levels: increase in IHI and eft MEP amplitude) was applied in the triggered and non-triggered TMS trials.”

The result section is difficult to understand because context information is often missing since the detailed Methods are reported at the end. Please revise the results to improve accessibility for the reader. In that regard, I would recommend swapping the order in the result section and presenting first the EEG states and then the associated IHI measures.

According to your suggestion, we have revised throughout the manuscript and swapped the order in the results to improve accessibility for the reader. e.g., explanations of “triggered trials” and “waiting time” [Lines 247-251], stimulus intensities during TMS experiment [Lines 351-356], explanations of each five sessions including REST and NoFB sessions [Lines 218-220], and how to evaluate interhemispheric functional connectivity [Lines 804-808].

In addition, values corresponding to each statistical figure are presented in each table (Supplementary File 1) for readability.

What is the rationale for using right finger MI as the no-FB condition? Please explain

The NoFB session served as a baseline control to examine changes in IHI magnitude during MI without any influence of visual input and volitional control. In the three neurofeedback HIGH, MID, and LOW sessions, participants received visual feedback during right finger MI and learned to modulate the sensorimotor activity patterns using their own volition. In other words, these three sessions are influenced by visual input and volitional control, in addition to performing MI (Takemi et al., 2018).

We added explanation of the rationale of NoFB session in the Results section as follows:

[Lines 215-220] “In the three neurofeedback sessions (i.e., HIGH, MID, and LOW), participants received visual feedback during right finger MI and tried to modulate the sensorimotor activity patterns. The NoFB session served as a baseline to examine the IHI magnitude without visual stimuli and volitional control. The REST session also served as a control to determine the intrinsic IHI magnitude at rest.”

Line 257ff: there seems to be a mistake: «one-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) revealed

significant differences F(2,44) = 1.43, p < 0.248,…» should probably read « revealed no significant differences».

Corrected. [Lines 451-554]

Line 260 ff there seems to be a mistake: Although the IHI magnitude in the LOW session was larger than that in the HIGH and LOW sessions.… replace second LOW by MID.

I liked the control analysis of the authors shown in suppl. Figure 3-2, However, the conclusion that the LOW,MID,HIGH data shown in Figure 3B is different from the data shown in suppl. Figure 3-2 requires a direct statistical comparison.

We have revised to the correct sentences as below.

In addition, we performed a direct statistical comparison and replaced the result in Figure 4—figure supplements 2

[Lines 455-554] “A two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and trials (two levels: triggered TMS trials and non-triggered TMS trials) revealed a significant main effect for sessions (F_{(2, 117)} = 5.08, p = 0.008, η² = 0.08), but no main effect for trials (F_{(1, 117)} = 0.32, p = 0.575, η² < 0.01) and for interaction (F_{(2, 117)} = 1.82, p = 0.167, η² = 0.03). Thus, no significant difference in IHI between the volitional control of SMR-ERDs in the closed-loop environment differs from spontaneous SMR fluctuations was observed. A post-hoc two-tailed paired t-test showed significant differences between HIGH and MID sessions (Cohen's d = 0.94, p = 0.017), and between HIGH and LOW sessions (Cohen's d = 1.36, p < 0.001) in the triggered TMS trials. In contrast, in the non-triggered TMS trials, a posthoc two-tailed paired t-test showed no significant differences between HIGH and MID sessions (Cohen's d = 0.20, p = 1.00), and between HIGH and LOW sessions (Cohen's d = 0.28, p = 1.00).”

[Lines 1850-1853] “To investigate the influence of spontaneous SMR fluctuations on IHI, a two-way rmANOVA for sessions (HIGH, MID, and LOW) and trials (triggered TMS trial and non-triggered TMS trial) as the withinparticipant factors.”

Figure 4C: I'm not convinced that presenting the laterality index adds any information to Figure 4 A and B. In fact, I found the caption confusing since it seems to suggest that (i) the negative LI for LOW results from contralateral activity which is inconsistent with the topoplot that shows high activity in the ipsilateral hemisphere; and (ii) the positive LI for HIGH results from ipsilateral activity which is also inconsistent with the topo which shows desynchronisation in the ipsilateral hemisphere. Given the shown in panels A and B, the LI measure seems inappropriate since there are virtually no activity changes in the contralateral hemisphere while the ipsilateral hemisphere shows either synchronized or desynchronized SMR activity.

We agree that the laterality index (LI) does not provide additional important information to Figures 2A and B. To avoid confusion for the readers, the results and descriptions related to the LI have been deleted from the manuscript.

Associations between IHI magnitude and bilateral EEG patterns: The authors might want to consider calculating a repeated measures correlation because it does not seem meaningful to pool data across the LOW, MID, HIGH sessions.

In accordance with your comments regarding the across-participant corrections between IHI magnitude and bilateral EEG patterns, we performed a repeated measures correlation since it was not reasonable to combine different sessions of different participants. Simple correlation is often applied to non-independent observations or aggregated data. A previous study argued that it may produce biased, specious results due to violation of independence and/or differing patterns (across-participants vs. within-participants) (Bakdash and Marusich, 2017).

We also calculated within-participant corrections between IHI magnitude and bilateral EEG patterns in accordance with the suggestion by Reviewer 3. We have added the overall significance of the 22 individual correlations using a chi-squared test and the distribution of correlation coefficients across all participants. In the analysis, data from HIGH, MID, and LOW sessions were aggregated as independent observations. We have integrated the above suggestions and modified the text as follows.

[Lines 1772-1776] “In the across-participant corrections between contralateral or ipsilateral SMR-ERDs and IHI magnitude, we performed a repeated measures correlation. A previous study argued that it may produce biased, specious results due to violation of independence and/or differing patterns (across-participants vs. withinparticipants) (Bakdash and Marusich, 2017).”

[Lines 717-803] “In the across-participant correlation between IHI magnitude and contralateral or ipsilateral SMR-ERDs, we performed a repeated measures correlation analysis (Bakdash and Marusich, 2017). We found a significant correlation in the ipsilateral SMR-ERD (r_rm = 0.436, p = 0.004; Figure 5B), but not in the contralateral SMR-ERD (r_rm = 0.008, p = 0.960; Figure 5C)”

Discussion

Line 470ff: The first sentence of the discussion follows a reversed logic. Participants did not consciously self-regulate IHI but they modulated their EEG pattern based on the closed-loop spatially bivariate BCI-based neurofeedback. Changes in IHI were the consequence of self-regulating the EEG activity pattern. Please revise throughout the discussion.

We have revised the explanations throughout the Discussion. [Lines 1024-1046 (“In the present study, we aimed…”), Lines 1350-1355 (“In conclusion, we presented an innovative approach to…”)]. Similarly, the claim was simply used “manipulate” or “manipulation” instead of “consciously/voluntary self-manipulate” or “consciously/voluntary selfmanipulation”.

The discussion of the neurophysiological mechanisms mediating IHI could be a bit more in-depth. The review of Carson RG. J Physiol. 2020 Nov;598(21):4781-4802 might help.

We appreciate for suggesting the study by Carson. We have added the information related to the neurophysiological mechanisms mediating IHI. In addition, the review by Carson provided new insight into reassessing the interhemispheric competition model. The alternative conceptual framework, which states that IHI acts to sculpt the output of neural circuits by co-opting the two cerebral hemispheres, is shown to be consistent with the known associations between the structural integrity of callosal projections and the magnitude of the motor deficits after stroke. This may emphasize the benefit of our neurofeedback approach in the context of neurorehabilitation.

With have added several additional references and revised the discussion as follows:

[Lines 1150-1155] “This may indicate that IHI and Network-intensity simply reflect the activity of different neural populations; interhemispheric projections act via surround/lateral inhibition in the sensory and motor cortices (Carson, 2020), while the synchrony at the EEG level reflects changes in the activity of local interactions between pyramidal neurons and interneurons in the thalamocortical loops (Pfurtscheller and Lopes da Silva, 1999).”

[Lines 1265-1270] “Furthermore, a previous review advocated the conceptual framework and suggested that a fundamental role of IHI is to support the narrowing of excitatory focus by co-opting the capacities of the two cerebral hemispheres (Carson, 2020). This framework is consistent with the known associations between the structural integrity of callosal projections and the magnitude of the motor deficits after stroke (Auriat et al., 2015; Granziera et al., 2012; Koh et al., 2018).”

[Lines 1307-1311] “Previous studies suggested that the IHI magnitude and the corticomotor excitability were modulated independently (Morishita et al., 2011, 2012). Although it is difficult to directly compare these results due to the significantly different motor tasks and experimental settings, the assertion that the projections from the transcallosal and the corticospinal pathways are distinct supports our findings.”

The authors argue that "This multimodal EEG-TMS approach proves that the bilateral α and β activities observed in the current study served to regulate the inhibition and facilitation of inhibitory functional coupling of interhemispheric interaction over motor cortices.» This argument would be strengthened if the interhemispheric effects were independent of the general excitability of the ipsilateral hemisphere as reflected in CS MEP amplitudes. Only then, the authors can argue that their intervention specifically modulated interhemispheric circuits.

As stated in to our response to the first comment, our results were not adequate to confirm that the changes in IHI and changes in the corticospinal excitability were independent of each other. Therefore, we have deleted the sentence. In addition, the methodological limitations of the study have been revised, as stated in our response to the first comment.

Lines 538-544: Explanatory theoretical model: I cannot follow the argument of the authors. Is it possible that IHI and "network intensity" simply reflect the activity of different neural populations? IHI effects are rather specific and I doubt that this could be picked up via general EEG markers.

We agree with your concern that different neural populations might be involved in IHI and Network-intensity. In addition, Network-intensity reflects the bidirectional synchrony between the two hemispheres, while IHI reflects oneway (right-to-left) inhibitory effects. Therefore, we have deleted the sentences to simplify the Discussion section.

[Lines 1142-1157] “However, even fMRI-based neurofeedback with high spatial resolution, based on blood oxygenation level-dependent signal, cannot easily distinguish between the excitatory and the inhibitory activities from each region (Moon et al., 2021). Additionally, neurofeedback using interhemispheric functional connectivity has the possibility to counteract the modulation of IHIs from the right to left hemisphere and from the left to right hemisphere. Our results showed no significant difference in Network-intensity during MI between HIGH and LOW sessions (Figure 5D), indicating that interhemispheric functional connectivity at the EEG level was not changed, whereas IHI was modulated. This may indicate that IHI and Network-intensity simply reflect the activity of different neural populations; interhemispheric projections act via surround/lateral inhibition in the sensory and motor cortices (Carson, 2020), while the synchrony at the EEG level reflects changes in the activity of local interactions between pyramidal neurons and interneurons in the thalamocortical loops (Pfurtscheller and Lopes da Silva, 1999). Furthermore, it is possible that Network-intensity reflects the degree of bidirectional synchrony between the two motor cortices, while IHI in this study reflects one-way (right-to-left) inhibitory effects.”

[Lines 1157-1159] “Therefore, it is unclear whether fMRI-based neurofeedback using interhemispheric functional connectivity is linked to ipsilateral excitability, and whether it modulates IHI magnitude.”

Line 561ff This seems to be consistent «Furthermore, the effect size of IHI manipulation in the current study (Cohen's d = 1.50) was approximately 1-2-fold higher than that of representative tDCS (Cohen's d = 1.55)…».

We have revised this sentence as follows:

[Lines 1168-1171] “Furthermore, the effect size of IHI manipulation in the current study (Cohen's d = 1.50) was comparable with that of representative tDCS (Cohen's d = 1.55) (Williams et al., 2010) and superior to that of rTMS studies (Gilio et al., 2003) (Cohen's d = 0.80; note that it is a read from the graph).”

Discussion of IHI in stroke: As stated by the authors, it is somewhat controversial whether IHI is a useful therapy target early after stroke. However, their EEG-neurofeedback approach allows to selectively up- or down-regulate one hemisphere which might be highly relevant in the context of stroke recovery both from a basic science and application perspective. It might be worthwhile to expand this aspect of the discussion.

We thank your encouragement. Following the original sentences, we complemented the pointed aspects.

[Lines 1270-1272] “Thus, our spatially bivariate EEG-neurofeedback approach might be highly relevant to application perspective in the context of stroke recovery as well as a basic science perspective.”

Reviewer #2 (Recommendations for the authors):

Although the Introduction is comprehensive in terms of the mechanistic relation between EEG and TMS variables, it does not sufficiently include the "clinical" significance of modifying IHI e.g. in terms of rehabilitation potential. It would be nice to include a few references in clinical populations where IHI and/or SMR are impacted, or where IHI/SMR have been used as prognostic markers of motor recovery. I would also include a sentence or two in the Abstract underlining the potential clinical significance of manipulating IHI.

During the initial review of this manuscript, we received the following comment from a senior editor: “the introduction focused on the relevance of the work for stroke rehabilitation seemed a stretch and, in some sense, to weaken the work given that you do not include any behavioral outcome measures, and that the rebalancing argument has proven somewhat contentious”.

Because Reviewer 1 also provided a similar suggestion, we added the clinical significance of our results. Thus, we addressed the rehabilitation issue in the Discussion section and focused on the core question of our work and relevant literature in the Introduction section.

We have revised the manuscript to state the importance of our findings in the context of neurorehabilitation as follows:

[Lines 1252-1260] “In stroke patients, the interhemispheric imbalance/competition model predicts the presence of asymmetry in the interhemispheric sensorimotor network, with excessive inhibition from the non-affected hemisphere limiting motor performance and maximal recovery (Duque et al., 2005; Murase et al., 2004). Therefore, guiding effective inhibitory interhemispheric network represented by IHI to the appropriate pattern through both targeted up-conditioning in the affected hemisphere and down-conditioning in the non-affected hemisphere may contribute to reduced abnormal IHI and enhanced achievable functional recovery (Chieffo et al., 2013; Di Pino et al., 2014; Dong et al., 2006; Hummel and Cohen, 2006).”

In terms of related studies combining the TMS/EEG/BCI trio, the work by Ros et al. 2010 was one of the first to demonstrate that MEP measures of cortical excitability and inhibition (MEP, SICI) can be modulated using closed-loop feedback of oscillations from the motor cortex.

Ros, T., Munneke, M.A.M., Ruge, D., Gruzelier, J.H. and Rothwell, J.C. (2010), Endogenous control of waking brain rhythms induces neuroplasticity in humans. European Journal of Neuroscience, 31: 770-778. https://doi.org/10.1111/j.1460-9568.2010.07100.x

Thank you for introducing the previous study. We cited the recommended literature.

[Lines 1056-1073] “Using spatially bivariate BCI-based neurofeedback enables participants to volitionally increase or decrease (bidirectional) the ipsilateral sensorimotor excitability, while maintaining constant contralateral sensorimotor excitability. Ipsilateral SMR-ERD (but not contralateral), reflecting IHI magnitude (Figure 2B) and their significant correlation (Figure 5C), were compatible with previous studies (Haegens et al., 2011; Madsen et al., 2019; Ros et al., 2010; Sauseng et al., 2009; Takemi et al., 2013; Thies et al., 2018; Zarkowski et al., 2006). The … Furthermore, another EEG-TMS study using neurofeedback demonstrated that the endogenous suppression of α rhythms at rest can produce robust increase in MEP amplitude and decrease in short-interval intracortical inhibition (Ros et al., 2010).”

It is not clear what behavioural and neural variable the authors are referring to in the first Results section starting with. "We confirmed the difficulties of each session" (line 171).

We have added the explanations in the Results (the same thing was pointed out by Reviewer 1).

[Lines 247-251] “To quantify the difficulties in each neurofeedback session, two indices were examined: (1) the number of triggered trials represents the BCI performance (i.e., success trials) since TMS was only triggered when the EEG signal exceeded the predetermined SMR-ERD threshold; (2) mean waiting time for a triggered event from task onset, which yields a range of 0–5 s (triggered TMS timing).”

Line 171: "The mean value of the sum of the triggered trials". Mean and sum have different definitions, do you mean 'the mean value of all the triggered trials'?

We have corrected the relevant part.

[Lines 251-255] “The mean values of all triggered trials (± 1 SD) of the neurofeedback sessions were 14.2 ± 4.7 trials and post-hoc paired t-tests following a one-way repeated-measures ANOVA (rmANOVA) revealed no significant difference between the three sessions (all p > 0.05; HIGH: 12.0 ± 5.0 trials, MID: 16.1 ± 3.8 trials, LOW: 15.9 ± 5.4 trials).”

There is no reference to what the abbreviation 'TS' stands for (I am assuming it is "test stimulus")

We have already written the abbreviation in Line 342. We also added it in the “abbreviation” section. [Line 58]

Line 219: The manipulation range of IHI (i.e., the difference between HIGH and LOW sessions) was 32.6. However elsewhere the means (Line 222/ 223) are separately indicated as HIGH: 62.0 and LOW: 96.9. Their difference (96.9 – 62.0) is 34.9.

Due to a few corrupted data in HIGH session (Lines 1401-1403), the manipulation range was not calculated in two participants, so the two values were different. The MEP amplitude in LOW session excluding two participants was 94.6%. Thus, the difference between HIGH and LOW sessions was 32.6% which was consistent with written information. To avoid the confusion of readers, we have corrected the sentence as follows:

[Lines 388-391] “The manipulation range of IHI (i.e., the difference between HIGH and LOW sessions) was 32.6 ± 30.7% (Cohen's d = 1.50) in all participants excluding 2 participants with data corruption (i.e., 22 participants).”

Line 263: "suggesting that the volitional control of SMR-ERDs in the closed-loop environment differs from spontaneous SMR fluctuations". This depends on how one defines the word "differs". In this case, the comparison is technically qualitative as the contrast is between a significant result (volitional control, p <0.05) and a non-significant one (spontaneous fluctuation, p > 0.05). However, a more direct comparison would involve comparing the two effects directly (volitional vs spontaneous) by statistically testing for a significant difference in effect sizes (e.g. using a two-way rmANOVA).

Taking the effect size of the HIGH-vs-LOW contrast as an example, for volitional control this is d = 1.42, while for spontaneous fluctuation it is d = 0.49, which yields a differential d = 0.93 (1.42 -0.49). This is likely statistically significant but it will depend on the error term, hence the need for a formal two-way rmANOVA analysis with an interaction effect. This should be done for all the contrast analyses in the paper (including SMR power), instead of using the one-way ANOVA.

We appreciate your valuable comment. A direct statistical comparison was shown in Figure 4—figure supplements 2. Again, the "non-triggered TMS trials" were collections of the trials when SMR-ERD reached an equivalent threshold during the other two sessions. Therefore, we think that it is an understandable result that there is no significant interaction. We have revised the related parts in the manuscript as below.

Similarly, a two-way rmANOVA was applied when a direct statistical comparison is required (Figure 4—figure supplements 3C, D).

[Lines 455-554] “A two-way rmANOVA for sessions (three levels: HIGH, MID, and LOW) and trials (two levels: triggered TMS trial and non-triggered TMS trials) revealed a significant main effect for sessions (F_{(2, 117)} = 5.08, p = 0.008, η² = 0.08), but no main effect for trials (F_{(1, 117)} = 0.32, p = 0.575, η² < 0.01) and for interaction (F_{(2, 117)} = 1.82, p = 0.167, η² = 0.03). A post-hoc two-tailed paired t-test showed significant differences between HIGH and LOW sessions (Cohen's d = 0.71, p = 0.006). Thus, no significant difference in IHI between the volitional control of SMR-ERDs in the closedloop environment differs from spontaneous SMR fluctuations was observed. A post-hoc two-tailed paired t-test showed significant differences between HIGH and MID sessions (Cohen's d = 0.94, p = 0.017), and between HIGH and LOW sessions (Cohen's d = 1.36, p < 0.001) in the triggered TMS trials. In contrast, in the non-triggered TMS trials, a post-hoc two-tailed paired t-test showed no significant differences between HIGH and MID sessions (Cohen's d = 0.20, p = 1.00), and between HIGH and LOW sessions (Cohen's d = 0.28, p = 1.00).”

[Lines 1850-1853] “To investigate the influence of spontaneous SMR fluctuations on IHI, a two-way rmANOVA for sessions (HIGH, MID, and LOW) and trials (triggered TMS trial and non-triggered TMS trial) as the withinparticipant factors.”

The dynamic range of SMR modulation occurs mainly in the ipsilateral hemisphere (in terms of bidirectional control). The contralateral hemisphere consistently shows increased SMR power relative to rest, with no bidirectional modulation. How do the authors explain this ? Could this be due to how the neurofeedback was implemented for each hemisphere (thresholds/rewards)? Right hand dominance? If the latter, where there any differential effects in the subgroup of left hand dominant participants?

We aimed to modulate the ipsilateral hemisphere bidirectionally while maintaining constant the contralateral excitability using our spatially bivariate BCI-based neurofeedback. To evaluate the effect of SMR modulation in the ipsilateral hemisphere, participants always aimed to maintain constant contralateral SMR-ERD. As written in the manuscript ("Spatially bivariate BCI-based neurofeedback"), visual feedback was provided on a computer screen in the form of cursor movements in a two-dimensional coordinate, in which x- and y-axis corresponded to the degree of the ipsilateral and contralateral SMR-ERD, respectively. For example, in HIGH session, participants were instructed to move the cursor toward the middle right (x > 0, y = 0) in the two-dimensional coordinate, the position showed a strong SMR-ERD in the ipsilateral hemisphere and moderate SMR-ERD in the contralateral hemisphere. On the contrary, in LOW session, participants were instructed to move the cursor toward the middle left (x < 0, y = 0) in the two-dimensional coordinate, the position showed a low SMR-ERD (or strong SMR-ERS) in the ipsilateral hemisphere and moderate SMR-ERD in the contralateral hemisphere. Therefore, spatially bivariate BCI-based neurofeedback enables us to modulate the ipsilateral SMR-ERD bidirectionally while maintaining constant contralateral SMR-ERD in the three sessions (i.e., HIGH, MID and LOW). In fact, Figures 2A and B showed the intended results. To help readers’ understanding, we prepared a sample video as Video 1.

The above is the answer to your point, but we notice that we need to describe the experimental limitation regarding (1) handedness and (2) the other patterns of the contralateral excitability. In this study, all participants were right-handed (Laterality Quotient: 72.2 ± 30.9%), as assessed by the Edinburgh Inventory (Oldfield, 1971), to eliminate bias due to the dominance. If a left-handed participant performs the same experiment, the modulation effect via BCI-based neurofeedback may be different. Moreover, in case the contralateral SMR-ERD was up- or downregulated instead of maintained at a moderate level, it is unclear whether the ipsilateral SMR-ERD was modulated bidirectionally and whether the IHI changed with it. The study performed a total of 11 sessions, including REST and NoFB, and the combinations of SMR-ERDs in both hemispheres (3 conditions vs. 3 conditions) may be worth evaluating in the future.

[Line 1325-1334] “For the limitations of the volitional control of SMR-ERD during unilateral MI, all participants were right-handed (Laterality Quotient: 72.2 ± 30.9%), as assessed by the Edinburgh Inventory (Oldfield, 1971), to eliminate bias due to the dominance. If a left-handed participant performs the same experiment, the modulation effect via BCI-based neurofeedback may be different. Moreover, in case the contralateral SMR-ERD was up- or downregulated instead of maintained at a moderate level, it is unclear whether the ipsilateral SMR-ERD was modulated bidirectionally and whether the IHI changed with it. The study performed a total of 11 sessions, including REST and NoFB, and the combinations of SMR-ERDs in both hemispheres (3 conditions vs. 3 conditions) may be worth evaluating in the future.”

Line 580: "the current technique can be tailored either to upregulate the damaged hemisphere, down-regulate the intact hemisphere…" and Line 585: "In conclusion, we presented an innovative approach to voluntarily and bidirectionally manipulate the state of IHI"

Critically, bidirectional control would need to be significantly demonstrated above and below the resting baseline. Although the study shows bidirectional control of SMR power (in the ipsilateral hemisphere), this is strictly not the case for IHI (which shows only significant increases relative to rest). Hence, the term "bidirectional" should not be used in the context of the IHI, but rather the phrase "control of interhemispheric disinhibition". Hence, the authors could state that the current results should motivate future studies investigating whether upregulation of IHI in the damaged hemisphere is possible above and beyond spontaneous resting values.

We thank your important comments. Although we successfully demonstrated the up- and down-regulate the ipsilateral SMR-ERD (i.e., bidirectional changes) at EEG-level, the bidirectional changes in IHI from the right to the left hemisphere were not statistically shown since there is no significant difference in MID and LOW sessions (Figure 4B). Therefore, we have revised the expression of the “bidirectional” modulation in IHI and added the future experiment in the Limitation.

[Lines 1277-1282] “Although we successfully demonstrated the bidirectional changes in the ipsilateral SMR-ERD, the bidirectional changes in IHI from the right to the left hemisphere were not statistically significant since there was no significant difference between the MID and LOW sessions (Figure 4B). The current results should prompt future studies to ascertain whether down-regulation of IHI (i.e., disinhibition) during MI and up-regulation of IHI at rest in the non-dominant hemisphere (intact hemisphere) are possible in a similar setup.”

[Lines 1350-1355] “In conclusion, we presented an innovative approach to manipulate the state of IHI, by directly and bidirectionally modulating SMR-ERDs in a spatially bivariate BCI-based neurofeedback paradigm.”

Reviewer #3 (Recommendations for the authors):

This is a well-conducted study building and expanding on the pioneering work of this group (Takemi et al. 2013, 2015). The novel contribution of this study is that self-regulated ERD is not modulating the TMS effects on the periphery (MEP) only, but also the TMS effects on the opposite brain hemisphere. This has been probed by IHI, but may probably have also been shown by TEP.

All other interpretations with regard to "rebalancing", "pathway-specificity", "interhemispheric connectivity", and "challenging the concept of neurorehabilitation" etc. are far fetching and should be toned down. Such statements would have necessitated (i) either direct neurofeedback of interhemispheric coherence, (ii) or demonstration of a direct link between the self-regulated ERD and interhemispheric coherence, and (iii) the application in patient populations.

Therefore, it is advisable to better link this study to the large body of literature in the IHI field, and discuss where this work adds additional insights. It is surprising, for example, that relevant previous literature is not referred to and discussed: Liang et al. (2014) have shown that imagined movements change IHI, Nelson et al. (2009) demonstrated bi-directional IHI, and Morishita et al. (2011, 2012) contributed with regard to ipsilateral inhibitory circuits and IHI. This work should be cited.

Thank you for your valuable comments and for introducing relevant previous studies. Although we agree that the parts pointed out need to be toned down, we would like to emphasize that we successfully revealed that modulating the excitability of the right hemisphere changes IHI from the right to the left hemisphere via BCI-based neurofeedback. This is fundamentally different from observational studies investigating movement-related IHI (Duque et al., 2007, 2005; Liang et al., 2014; Morishita et al., 2012; Murase et al., 2004; Nelson et al., 2009).

In addition, previous studies suggested that the IHI magnitude and the corticomotor excitability were modulated independently (Morishita et al., 2011, 2012). Although it is difficult to directly compare these results due to the significantly different motor tasks and experimental settings, the assertion that the projections from the transcallosal and the corticospinal pathways are distinct supports our findings (Figure 4—figure supplements 3B and C).

Furthermore, we discussed the neurophysiological mechanisms mediating IHI based on, for example, this literature (Carson, 2020). Regarding the results of interhemispheric coherence, we answered the next question.

We paid attention to the expression, and revised the manuscript as follows:

[Line 2] “Spatially bivariate EEG-neurofeedback can manipulate interhemispheric inhibition”

[Lines 1051-1056] “However, conventional BCI-based neurofeedback of the SMR signal from one hemisphere does not guarantee spatial specific activation of the sensorimotor activity in the targeted hemisphere (Buch et al., 2008; Caria et al., 2011; Soekadar et al., 2015a) because sensorimotor activities in the left and right hemispheres potentially influence one another, making IHI manipulation difficult.”

[Lines 45-47] “Our results show an interesting approach for neuromodulation, which might provide new treatment opportunities, for example, in patients suffering from a stroke. “

[Lines 1165-1168] “Conversely, under the conscious self-learning environment, volitional control of MEP amplitudes is retained for at least 6 months without further training (Ruddy et al., 2018), which supports the future prediction of long-term efficacy of IHI manipulation.”

[Lines 96-100] “A critical issue is whether a specific interhemispheric activity can be volitionally controlled. Previous studies have focused on inhibitory interhemispheric sensorimotor network, represented by interhemispheric inhibition (IHI), from movement-related manner (Duque et al., 2007, 2005; Liang et al., 2014; Murase et al., 2004; Nelson et al., 2009),”

[Lines 398-431] “In addition, we found disinhibition in NoFB session compared to REST session, although there was no statistical difference (Cohen's d = 0.42, p = 0.279); this trend is consistent with the previous studies, which reported that in healthy participants, IHI targeting the moving index finger leads to disinhibition before the movement onset to produce voluntary movement (Duque et al., 2005; Murase et al., 2004) and during isometric contraction (Nelson et al., 2009).”

[Lines 1307-1311] “Previous studies suggested that the IHI magnitude and the corticomotor excitability were modulated independently (Morishita et al., 2011, 2012). Although it is difficult to directly compare these results due to the significantly different motor tasks and experimental settings, the assertion that the projections from the transcallosal and the corticospinal pathways are distinct supports our findings.”

[Lines 1150-1155] “This may indicate that IHI and Network-intensity simply reflect the activity of different neural populations; interhemispheric projections act via surround/lateral inhibition in the sensory and motor cortices (Carson, 2020), while the synchrony at the EEG level reflects changes in the activity of local interactions between pyramidal neurons and interneurons in the thalamocortical loops (Pfurtscheller and Lopes da Silva, 1999).”

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Reviewer #1 (Recommendations for the authors):

The authors provided very thorough and extensive revisions and they have addressed most of my comments. It is particularly reassuring to see the additional analyses of corticomotor excitability as probed by the CS.

However, I'm still not completely convinced by the author's statistical approach.

If I understood the approach correctly, % change in IHI and % change of CS MEP amplitude were entered into one rmANOVA model and the authors tested for a condition x index interaction effect (which did not reach significance). This is a rather strict comparison. A more traditional approach would be to add % CS change as a covariate of no interest and test whether the IHI effect survives even if the covariate is added. If that is the case, the authors have demonstrated that IHI changes over and above the variance explained by the CS effect which would support their initial hypothesis that IHI can be modulated via their EEG-BCI and is not simply an epiphenomenon. No other statistical analyses would be required. However, the discussion would need to be modified!

The authors also analysed changes in CS MEP amplitude across conditions. CS MEPs did not significantly differ across the different EEG-BCI conditions. However, the absence of a significant effect does not allow the authors to conclude that there is "no effect". This would require a Bayesian approach which would test whether the data provides evidence in favour of H0. This could be another argument that IHI changes over and above CS changes. However, I don't think that these additional analyses would be needed if the co-variance analysis of IHI would reach significance. Maybe the authors might want to consult a stats expert for this analysis.

We appreciate the suggestion to use an alternative statistical approach. According to the suggestion, we tested whether the effect of IHI manipulation remains significant after a covariate is added. As a result, we confirmed that there was a significant difference in IHI magnitude between sessions (p < 0.001) using a generalized linear model with the percent changes of left MEP amplitude as a covariate of no interest (Figure 4—figure supplements 3C). Therefore, we did not analyze the left MEP amplitude, and only displayed the results (Figure 4—figure supplements 3B). Based on this, we successfully demonstrated that IHI changes greater than the variance explained by the CS effect and manipulation is not simply an epiphenomenon. In addition, we analyzed the non-triggered trials to investigate the CS effect (Figure 4—figure supplements 3D). During the non-trigged trials, no significant difference was observed in IHI magnitude (p = 0.319).

Based on the findings that IHI can be manipulated via BCI-based neurofeedback, further experiments may help to better understand the pathway specificity. In future studies, to further confirm whether the participants learned pathway-specific IHI manipulation, the CS effect should be kept constant across the conditions by adjusting the CS intensity to evoke left MEP of 1 mV peak-to-peak amplitude.

We have revised the manuscript to include these additional analyses, figures, and discussion.

[Lines 337-354] “Although we revealed that modulating the excitability of the right hemisphere changes IHI magnitude from the right to the left hemisphere, the specificity of the IHI manipulation was not proved yet. We presented additional results related to changes in corticospinal excitability as assessed by MEPs in the left FDI in response to the CS over the right hemisphere (Figure 4—figure supplements 3B).

To investigate whether changes in IHI from the right to the left hemisphere and changes in corticospinal excitability in the right hemisphere are independent phenomena, a generalized linear model with the percent changes of left MEP amplitude as a covariate of no interest was applied. This approach tests whether the effect of IHI manipulation survives even if the covariate is added. During the triggered TMS trials, a generalized linear model for sessions (three levels: HIGH, MID, and LOW) revealed a significant main effect for sessions (F_{(3, 60)} = 12.75, p < 0.001, η² = 0.44). Post-hoc two-tailed paired t-test in the increase in IHI showed significant differences between HIGH and MID sessions (Cohen's d = 1.06, p < 0.001), and between HIGH and LOW sessions (Cohen's d = 1.42, p < 0.001). During the non-triggered TMS trials, a generalized linear model for sessions (three levels: HIGH, MID, and LOW) revealed no main effect for sessions (F_{(3, 60)} = 1.13, p = 0.319, η² = 0.10). Therefore, we successfully demonstrated that IHI changes greater than the variance explained by the CS effect and manipulation is not simply an epiphenomenon.”

[Lines 647-652] “In addition, we successfully demonstrated that IHI changes greater than the variance explained by the CS effect. Based on our findings, further experiments may help to better understand the pathway specificity. To further confirm whether the participants learned pathway-specific IHI manipulation, the CS effect should be kept constant across the conditions by adjusting the CS intensity to evoke left MEP of 1 mV peak-to-peak amplitude.”

[Lines 1122-1125] “To investigate whether changes in IHI and corticospinal excitability in response to the CS over the right hemisphere are independent phenomena, a generalized linear model with the percent changes of left MEP amplitude as a covariate of no interest was conducted for the triggered and non-triggered TMS trials.”

Reviewer #3 (Recommendations for the authors):

I am satisfied with the author's responses and revisions, which are extensive and much more consistent with the collected evidence from the experiment.

Ultimately, I am ambivalent (i.e. neither for nor against) on recommending this paper for publication in eLife based on its scientific significance. On a positive note, the authors clearly demonstrate that it is possible to modulate IHI (albeit unidirectionally) using bivariate neurofeedback, and this is an original and scientifically novel contribution. On the other hand, from the experiments conducted, it is not clear whether the bivariate neurofeedback approach is critical for the observed findings since there is no comparison to the conventional BCI-based neurofeedback that exploits SMR-ERD from one hemisphere. Hence the results show evidence of "sufficient" conditions to reduce inter-hemispheric inhibition (IHI) but do not provide evidence of "necessary" conditions i.e. whether uni-hemispheric neurofeedback would lead to similar effects but with a much more parsimonious experimental setup that does not involve controlling EEG from both hemispheres.

I, therefore, leave it to the senior editor for the ultimate decision regarding publication, taking also into account the other reviewers' comments.

We have added an in-depth comparison of the new bivariate and univariate neurofeedback methods.

First, this study did not perform to show the necessary and sufficient condition that IHI can be manipulated only via bivariate BCI-based neurofeedback. However, we emphasize that our results are still significant because this was the first study to show IHI-state manipulation induced by volitional variation in ipsilateral sensorimotor excitability to the imagined hand, which has not been achieved with the conventional univariate BCI-based neurofeedback paradigm.

A direct comparison between bivariate and univariate (i.e., ipsilateral hemisphere to the imagined hand) BCI-based neurofeedback might help solidifying the necessity of bivariate BCI-based neurofeedback for targeted IHI manipulation, but it was not feasible to incorporate experiments in the present study due to experimental time limitation (the experimental time would need to be doubled to 5–6 h to perform such an experiment, violating ethical concerns). A future study should be performed to directly compare the effect size of the IHI manipulation between bivariate and univariate BCI-based neurofeedback in a within-participant cross-over design with adjusting experimental time by omitting some experimental conditions. Again, the present study focused to show the feasibility of bivariate BCI-based neurofeedback for targeted IHI manipulation, and required 2.5–3 h of the experimental protocol for proving this.

When conducted a future experiment for direct comparison of bivariate and univariate BCI-based neurofeedback, it is highly probable that the necessary conditions will be satisfied for the following reasons.

Our findings indicated that unilateral (not bilateral) up- or down-regulation of the sensorimotor excitability manipulates the IHI magnitude (Figure 2C and Figure 4B), which is consistent with previous studies (Carson, 2020; Duque et al., 2005; Murase et al., 2004). However, conventional methods did not guarantee unilateral specific sensorimotor activation (Buch et al., 2008; Caria et al., 2011; Soekadar et al., 2015). In addition, SMR-ERD was observed focused over the scalp sensorimotor areas, mainly bilaterally (e.g., Pichiorri et al., 2011; Vidaurre et al., 2019). Therefore, the effect of IHI modulation may be minimal because the bilateral up-regulation of the sensorimotor excitability via univariate BCI-based neurofeedback may inhibit the other hemispheres. We did not investigate whether the conventional univariate neurofeedback that exploits SMR-ERD from one hemisphere achieves bidirectional modulation of the target-hemisphere-dependent SMR-ERD and the accompanying IHI magnitude, but it is assumed that the sensorimotor excitability in the opposite hemisphere would be also up-regulated through univariate neurofeedback, leading to a weaker range of IHI manipulation.

Based on the aforementioned factors, we presumed that it is difficult to produce a similar effect as this study via univariate BCI-based neurofeedback. Even if IHI-state manipulation is equally achieved via univariate BCI-based neurofeedback, our scientific significance remains established. Direct comparison would be executed beyond this work.

According to the reviewer’s suggestion, we have revised the discussion and limitations in the manuscript as follows:

[Lines 709-723] “In addition, it is not clear from previous experiments whether the bivariate neurofeedback approach is superior to the conventional BCI-based neurofeedback that exploits SMR-ERD from one hemisphere. Since both hemispheres are connected by intrinsic transcallosal projections and exhibit functional crosstalk (Arai et al., 2011; Hofer and Frahm, 2006; Meyer et al., 1995; Waters et al., 2017), it is expected that the sensorimotor excitability in the opposite hemisphere would also be up-regulated through univariate (i.e., ipsilateral hemisphere to the imagined hand) BCI-based neurofeedback. Consistent with the findings that the unilateral up- or down-regulation of the sensorimotor excitability manipulates the IHI magnitude, such sensorimotor activation without spatial specificity may inhibit the other hemisphere, leading to weaker range of IHI manipulation. Thus, we presumed that it is difficult to produce a similar effect as this study via univariate BCI-based neurofeedback. A future study is required to directly compare the effect size of IHI manipulation between bivariate and univariate BCI-based neurofeedback in a within-participant cross-over design with adjusting experimental time by omitting certain experimental conditions.”

References

Arai N, Müller-Dahlhaus F, Murakami T, Bliem B, Lu M-K, Ugawa Y, Ziemann U. State-dependent and timingdependent bidirectional associative plasticity in the human SMA-M1 network. J Neurosci 31(43), 15376–15383, 2011.

Buch E, Weber C, Cohen LG, Braun C, Dimyan MA, Ard T, Mellinger J, Caria A, Soekadar S, Fourkas A, Birbaumer N. Think to move: A neuromagnetic brain-computer interface (BCI) system for chronic stroke. Stroke 39(3), 910–917, 2008.

Caria A, Weber C, Brötz D, Ramos A, Ticini LF, Gharabaghi A, Braun C, Birbaumer N. Chronic stroke recovery after combined BCI training and physiotherapy: A case report. Psychophysiology 48(4), 578–582, 2011.

Carson RG. Inter-hemispheric inhibition sculpts the output of neural circuits by co-opting the two cerebral hemispheres. The Journal of Physiology 598(21), 4781–4802, 2020.

Duque J, Hummel F, Celnik P, Murase N, Mazzocchio R, Cohen LG. Transcallosal inhibition in chronic subcortical stroke. Neuroimage 28(4), 940–946, 2005.

Hofer S, Frahm J. Topography of the human corpus callosum revisited: Comprehensive fiber tractography using diffusion tensor magnetic resonance imaging. NeuroImage 32(3), 989–994, 2006.

Meyer B-U, Röricht S, von Einsiedel HG, Kruggel F, Weindl A. Inhibitory and excitatory interhemispheric transfers between motor cortical areas in normal humans and patients with abnormalities of the corpus callosum. Brain 118(2), 429–440, 1995.

Murase N, Duque J, Mazzocchio R, Cohen LG. Influence of interhemispheric interactions on motor function in chronic stroke. Ann Neurol 55(3), 400–409, 2004.

Pichiorri F, Fallani FDV, Cincotti F, Babiloni F, Molinari M, Kleih SC, Neuper C, Kübler A, Mattia D. Sensorimotor rhythm-based brain–computer interface training: The impact on motor cortical responsiveness. J Neural Eng 8(025020), 9, 2011.

Soekadar SR, Birbaumer N, Slutzky MW, Cohen LG. Brain–machine interfaces in neurorehabilitation of stroke.

Neurobiology of Disease 83(1), 172–179, 2015.

Vidaurre C, Ramos Murguialday A, Haufe S, Gómez M, Müller K-R, Nikulin VV. Enhancing sensorimotor BCI performance with assistive afferent activity: An online evaluation. NeuroImage 199375–386, 2019.

Waters S, Wiestler T, Diedrichsen J. Cooperation not competition: Bihemispheric tDCS and fMRI show role for ipsilateral hemisphere in motor learning. J Neurosci 37(31), 7500–7512, 2017.

Author details

1. Masaaki Hayashi

   Graduate School of Science and Technology, Keio University, Kanagawa, Japan

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing

   Competing interests

   is employed by LIFESCAPES Inc

   "This ORCID iD identifies the author of this article:" 0000-0002-7104-6765

2. Kohei Okuyama

   Department of Rehabilitation Medicine, School of Medicine, Keio University, Tokyo, Japan

   Contribution

   Data curation, Validation, Investigation, Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

3. Nobuaki Mizuguchi

   Research Organization of Science and Technology, Ritsumeikan University, Shiga, Japan

   Contribution

   Validation, Investigation, Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

4. Ryotaro Hirose

   Graduate School of Science and Technology, Keio University, Kanagawa, Japan

   Contribution

   Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

5. Taisuke Okamoto

   Graduate School of Science and Technology, Keio University, Kanagawa, Japan

   Contribution

   Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

6. Michiyuki Kawakami

   Department of Rehabilitation Medicine, School of Medicine, Keio University, Tokyo, Japan

   Contribution

   Resources, Validation, Writing – review and editing

   Competing interests

   No competing interests declared

7. Junichi Ushiba

   Faculty of Science and Technology, Keio University, Kanagawa, Japan

   Contribution

   Conceptualization, Resources, Supervision, Funding acquisition, Methodology, Writing – review and editing

   For correspondence

   ushiba@bio.keio.ac.jp

   Competing interests

   is a founder and the Representative Director of the University Startup Company, LIFESCAPES Inc. involved in the research, development, and sales of rehabilitation devices including brain-computer interfaces. He receives a salary from Connect Inc., and holds shares in Connect Inc. This company does not have any relationships with the device or setup used in the current study

   "This ORCID iD identifies the author of this article:" 0000-0003-1161-983X

• 1,173

  Page views

• 332

  Downloads

• 4

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.