Intramuscular Coherence of the Lower Flexor Muscles During Robotic Ankle-assisted Gait

1 Background: A close-fitting assisted walking device (RE-Gait) designed to assist ankle 2 movements might be a novel approach for acquiring the forefoot rocker function in the gait cycle. 3 The purpose of the present study was to investigate the effects of using RE-Gait by evaluating the 4 intramuscular coherence (IMC) of the two parts of the tibialis anterior muscles (TA) in the initial, 5 mid, and terminal swing phase, which could indicate whether the common synaptic drive of motor 6 neurons was populated. 7 Methods: Seventeen healthy volunteers walked on a treadmill at a comfortable speed before, 8 during, and immediately after 15-minute RE-Gait intervention (pre / RG / post). RE-Gait 9 supported plantar flexion at toe lift-off in the terminal stance phase and dorsiflexion in the initial 10 swing phase. Electromyograms of the right lower leg and gait parameters were analyzed for each 11 session. 12 Results: After RE-Gait intervention, the step length was significantly increased. IMC of the two 13 parts of the TA muscles in the beta frequency band in the initial swing phase was significantly 14 enhanced during RE-Gait intervention compared with pre session. In addition, IMCs in the beta 15 and low-gamma frequency bands were significantly correlated with the enhancement ratio of the 16 step length. 17 Conclusions: These results suggest that robotic ankle planter flexion and dorsiflexion assistance


Introduction
Walking in humans depends on the integrated hierarchical activities of supraspinal and spinal neural controls, including the primary motor cortex and corticospinal tract [1][2][3][4][5][6].The corticospinal inputs are critical for foot flexor muscle activities in the initial swing phase of the gait cycle, while they have little effect on the extensor muscles [1,3,4].A recent study demonstrated that gaitcombined rhythmic brain stimulation over the primary motor cortex in the swing phase could alter neuroplasticity in the corticospinal pathways and enhance gait performance [7].The required corticospinal inputs are, however, impaired in patients with central nervous system disorders [8,9].
From the pre-swing to the initial swing phase of the gait, the forefoot rocker function plays a crucial role in smooth energy utilization, generating plantar flexion torque in the pre-swing phase and switching dorsiflexion movement.Smooth switching is essential for efficient walking; however, we cannot provide enough assistance for the forefoot rocker function using manual support alone.To address this, robotic walking rehabilitation devices have been developed.A close-fitting assisted walking device (RE-Gait; Space Bio-Laboratories Co., Ltd, Hiroshima, Japan) designed to assist ankle dorsiflexion and plantar flexion movements [10] might be an effective approach for supporting the forefoot rocker function.Our previous studies revealed that short-term use of RE-Gait enhanced step length [10,11] and modulated spinal cord excitability [11].Continuous use of RE-Gait might be effective for re-learning of gait with smooth switching.However, there is currently insufficient neurophysiological evidence to support this effect.
In addition to background electromyographic (EMG) recordings of the lower legs, intramuscular coherence (IMC) analysis of the paired electromyographic recordings has previously been utilized to evaluate corticospinal function during walking [1,[12][13][14][15][16][17].IMC in the alpha, beta, and lowgamma frequency bands are thought to reflect coupling of motor unit activities that originate from common cortical sources.Therefore, IMC from the two parts of the tibialis anterior (TA) muscles can indicate whether the common synaptic drive of TA motor neurons is populated in the swing phase of the gait cycle.
In clinical studies, coupling of motor unit activities in the 10-to 20-Hz frequency band of the two parts of the TA muscles was severely reduced in patients with spinal cord injuries (SCI), suggesting that motor unit synchrony during walking depends on an intact supraspinal drive to the spinal cord [18].Similar findings were reported in the paralyzed lower legs of hemiplegic patients [16].Children with cerebral palsy showed reductions in beta and low-gamma band coherence [19,20], but this improved after four weeks of gait training [20] and also reflected improved maturation of ankle control [21].These findings support the view that a loss of IMC during walking is associated with a deficit in motor drive from the corticospinal tract.
The aim of the present study was to investigate the effects of RE-Gait by evaluating corticospinal function in the swing phase of the gait using background EMG and IMC analysis of 1 the two parts of the TA muscles. 2

Participants
Seventeen healthy volunteers (age 25.7±6.3years, 8 women) participated in the present study.
None of the participants had a history of neurological or orthopedic disorders.This study was approved in advance by the Ethical Committee for Epidemiology of Hiroshima University in accordance with the Declaration of Helsinki, and written consent was obtained from all participants.

RE-Gait
RE-Gait is a close-fitting assisted walking device shaped like an ankle-foot orthosis.It weighs 1 kg, and is easy to wear (Fig. 1A).It decodes the gait cycle by using two pressure sensors located at the toe and heel, and a geared motor assists ankle movement, which can control angular velocity in each gait cycle at the preferred timing.In the present study, the participants wore RE-Gait on their right leg, and the parameter of the assist was set to aid their forefoot rocker function; plantar flexion torque was applied at toe lift-off in the pre-swing phase (50-60% of the gait cycle), and dorsiflexion torque was applied in the initial swing phase (60-80% of the gait cycle) (Fig. 1B).
The length of the gait cycle was measured before the experiment, and was set individually.

Experimental protocols
EMG of the right lower leg was recorded before, during, and immediately after a 15-minute RE-Gait intervention (pre / RG / post).For each session, participants were asked to walk on a treadmill at their comfortable speed (2.0-3.3 km/h), which was measured before the experiment.
The speed was constant across the three sessions.The pre and post sessions lasted five minutes, with the first two minutes serving as a warm-up.The last three minutes of each session were used for analysis.
EMG was recorded by single differential surface electrodes (FAD-SEMG1: 4Assist, Inc, Tokyo, Japan) during walking.The electrodes were placed on the proximal and distal ends of the tibialis anterior muscles (TAp, TAd; 8-10 cm distance between the two sets of electrodes).The EMG signals were amplified (x100) with an EMG amplifier system (FAD-ABOX8, 4Assist, Inc), and digitized at 1000 Hz using a Power Lab system (PowerLab8/35, AD Instruments, Australia).To identify the gait cycle, foot sensors (FA-DL-250: 4Assist, Inc) were located on the heel and forefoot of the right foot.
The movement of the right leg was also recorded from the sagittal plane using a digital video camera during each session (frame rate=30 fps).

Analysis
The EMG data were band-pass filtered at 10-200 Hz and rectified for off-line analysis with LabChart v.8.1.12(AD Instruments).We defined the gait cycle by identifying two consecutive heel contacts, and time-normalized and rectified the EMG signals into 100 data points for each gait cycle.The EMG signals were averaged over the gait cycle for each muscle and each participant.Epochs with motion artifacts were excluded from averaging.The averaged EMG data were divided into seven phases as follows: initial loading (0-12% of the gait cycle), mid stance (12-30%), terminal stance (30-50%), pre swing (50-62%), initial swing (62-75%), mid swing (75-87%), and terminal swing phase (87-100%), in accordance with some previous studies [22,23].
The correlations between the EMG signals were assessed with IMC functions [14].As rectified EMG signals have been shown to enhance test-retest reliability [17], we used the rectified EMG signals of the TAp and TAd muscles.Because remarkable IMC was previously detected in the initial and terminal swing phase [13], we analyzed IMC by dividing the swing phase into initial, mid, and terminal parts.The frequency-domain analyses were performed based on previous studies [6, 12-18, 20, 21] .The IMC between the two rectified EMG signals (x and y) at frequency (λ) was defined as: where f xx(λ) and fyy(λ) represent values of auto spectra, and f xy(λ) represents values of cross spectra, calculated with a discrete Fourier transformation.The IMC function (|Rxy(λ)| 2 ) provides normative measures of linear association on a scale from 0 to 1 in the frequency domain, where 1 indicates a perfect linear correlation.Meaningful coherence was defined when the values exceeded a confidence limit (CL) with a probability of 95% (α=0.05).CL was calculated as follows: where N represents the number of analyzed segments.
In the time domain analysis, estimates of cumulant density function were used to characterize the correlation as follows: qxy(u)=fxy(λ) e iλu dλ where qxy(u) is the cumulant density function defined as the inverse Fourier transform of the cross spectrum.The dominant feature is that the cumulant density reflects the presence of common synaptic input.
To evaluate the coherence variables, the areas between the coherence spectrum and the CLs of the alpha (10-15 Hz), beta (15-30 Hz), and low-gamma (30-45 Hz) frequency bands in the initial, mid, and terminal swing phase were examined.The IMC analyses were performed using Neurospec 2.0 (http://www.neurospec.org/) and MATLAB R2019b (MathWorks, Inc) software.
The gait parameters were assessed from the step length and trailing limb angle (TLA).The right and left step lengths were obtained from as the anterior-posterior distance between both heels at the moment of initial contact phase.TLA was defined as the angle between the vertical axis and the vector joining the greater trochanter with the fifth metatarsal head at the moment of toe-off phase.They were analyzed from 10 gait cycles in each session by using Image J software (https://imagej.nih.gov/ij/).
In addition, the enhancement of step length on the right leg after RE-Gait intervention (Δstep) was calculated respectively as follows: Δstep=step length (post) -step length (pre) / step length (post) + step length (pre) The step length was compared using a two-way repeated ANOVA (3 sessions × right and left legs), and TLA was compared using a one-way repeated ANOVA.Each EMG activity and the areas of IMC in the alpha, beta, and low-gamma frequency bands were compared using one-way repeated ANOVA among the three sessions (pre, RG, and post).Post-hoc analyses were carried out where appropriate to compare differences in the sessions using paired t-tests with Bonferroni's correction.To determine the association between the behavioral and neurophysiological data, the correlation between Δstep and the area of IMC in each frequency band in RG session was also calculated.The level of significance was set at p<0.05 for each analysis.All statistical analyses were carried out using the JMP software (version 15, IBM, New York, U.S.A.).

Results
All participants could perform the intervention without stumbling or feeling strong fatigue.
A two-way repeated ANOVA (3 sessions (pre, RG, post) × 2 legs (right and left)) revealed a significant difference in the step length among the sessions (F(2, 80)=16.56,p<0.001), whereas no significant differences were found between legs (F(1, 80)=2.93,p=0.09) or in the interaction between the two factors (F(2, 80)=0.85,p=0.43).Post-hoc tests revealed that the step length of the RG session was significantly longer than that of the pre session for both legs (p<0.01), and that of the post session was significantly longer than that of the pre session for the right leg (p<0.05) (Fig. 2A).This tendency was in agreement with our previous study, in which RG intervention was effective for the enhancement of step length [10,11].
Significant difference was also revealed in TLA (F(2, 32)=13.62,p<0.001).Post-hoc test showed that of the RG session was significantly increased compared to the pre (p<0.01) and post (p<0.05)session, and that of the post session was increased than the pre session (p<0.05) (Fig. 2B).
The pooled EMG activities of the right TA muscles are shown in Fig. 3.The EMG patterns were not different among the three sessions, but the RG session showed unique features during the swing phase.TA activities tended to be higher in the initial swing phase, but lower in the terminal swing phase and initial loading phases.The TAd activities were significantly increased compared with the post session (p<0.05).In contrast, the pre and post session EMG activities were not different.
IMC between the two parts of the TA muscles were analyzed to investigate the effects of the cortico-spinal function in the swing phase.The averaged IMC from all participants in the initial, mid, and terminal swing phase are shown in Fig. 4.Each IMC was obtained from restricted frequency bands in all participants without high coherence across the frequency bands, which means that crosstalk did not affect the data [18,24].IMC for frequency ranges of 10-40 Hz were enhanced in the initial swing phase in the RG session.The areas of IMC in the beta frequency band were significantly different in the initial swing phase (alpha: F(2,32)=2.22,p=0.12, beta: F(2,32)=3.40,p<0.05, low-gamma: F(2,32)=0.45,p=0.64).Post hoc test revealed that the RG session significantly enhanced the IMC in the beta band compared with the pre session (p<0.05).
Comparing the behavioral and neurophysiological data, a significant correlation was found between the enhancement ratio of the right step length after RE-Gait intervention (Δstep) and IMC in the beta and low-gamma frequency bands in the RG session (alpha: r=0.41, p=0.11, beta: 1 r=0.56,p=0.018, low-gamma: r=0.54, p=0.025) (Fig. 5).. 2

Discussion
This is the first study to assess EMG activities and IMC of the two parts of the TA muscles during a robot-supported gait using an ankle assistive device, RE-Gait.We revealed that IMC in the beta frequency band of the two parts of the TA muscles in the initial swing phase was significantly enhanced by using RE-Gait.In addition, the amplitudes of IMC in the beta and lowgamma frequency bands during RE-Gait intervention were correlated with the enhancement ratio of the step length.These results suggest that robotic ankle planter flexion and dorsiflexion assistance in the pre-and initial swing phase would be effective for learning an adaptively modified gait by activating the corticospinal tracts.
After 15-minute RE-Gait intervention, step lengths on both sides were significantly extended.This finding was consistent with our previous studies [10,11].We consider this might be due to better toe clearance after the intervention resulting from smooth energy utilization by generating plantar-flexion and dorsiflexion torque.It was supported by the enhancement of TLA during and after the intervention.TLA was mainly contributed to the increase in propulsion force [25], and correlated with step length and walking speed [26].A previous study suggested that gait retraining in the chronic phase of stroke recovery were related to the propulsive force on paretic limb in the neuromotor recovery [27].By inducing the plantar flexion torque in the pre-swing phase, the potential energy for lifting the heel was generated, and subsequent dorsiflexion torque in the initial swing phase might support smooth replacement with the kinetic energy.
A characteristic EMG pattern was observed during RE-Gait intervention.TA activation tended be smaller in the terminal swing and initial loading phases, but was obviously enhanced in the initial swing phase.Several previous studies have suggested that robot-assisted walking is associated with lower levels of muscle activities [22,28], and increasing body weight support and guidance force decreased the activity of specific muscles during robot assisted walking [29], which means that the robotic devices could facilitate movement control of the legs by reducing muscle loads.The decrement in the TA activities related to initial heel strike in the present study could reflect a similar effect of robotic support.That is, the ankle dorsiflexion support and heel rocker support from the exoskeleton could produce a clear heel strike with a small TA muscle load from eccentric contraction.
In contrast, TA activity in the initial swing phase was enhanced in the RG session.Notably, IMC in the beta frequency band of the two parts of the TA muscles was significantly increased compared with the pre session.
IMC analysis is a helpful tool to evaluate the corticospinal pathway in walking [6, 12-18, 20, 21].The test-retest reliability was recently confirmed, indicating that beta-band IMC is a reliable method to assess corticospinal control of gait in young subjects [30].In addition, IMC was higher in young subjects than older subjects [31].Therefore, the frequency-dependent IMC changes of young participants in the present study would be reliable.Some previous studies have reported a linkage between the IMC and behavioral deficits.One study demonstrated a positive correlation between the IMC of the two parts of the TA muscles and highest toe elevation in SCI patients [12].Another showed a significant correlation between gait speed and IMC on the paretic side in hemiplegic patients [15].Lower step-to-step variability in the swing phase was positively correlated with IMC in healthy children [32].These studies suggest that IMC would reflect the corticospinal control of ankle movement and ambulatory function.
During visual guided walking with visual feedback, IMC was significantly enhanced compared with normal treadmill walking [31,33].In split-belt treadmill training, IMC was increased during the initial adaptation stage.Interestingly, a significant correlation was found between IMC in the initial adaptation stage and the double support symmetry ratio, implying that IMC might be a predictor for locomotor adaptation [34].In the present study, we also observed an IMC increment in the initial swing phase during RE-Gait intervention, which indicates that the participants required corticospinal activity to control ankle lifting even with the assistance of RE-Gait.In other words, the aid can be a useful intervention for learning adaptively modified walking.This possibility was supported by the significant correlation between the area IMC in the beta and lowgamma frequency band and the enhancement ratio of the step length after the intervention.In addition, significant changes of IMC were observed in the initial swing phase but not in the mid and terminal phases, which matched the timing of the assistance.
A significant IMC increment during RE-Gait intervention was found in the beta frequency band, and a significant correlation between the behavioral and neurophysiological results was found in the beta and low-gamma frequency bands.These frequency-dependent IMC changes might be explained by the neural origin of the IMC.Beta band oscillatory activation of the sensorimotor cortex (SMC) plays a key role in motor control [35], and beta band IMC is thought to originate from the SMC [36,37].The low-gamma band also indicates corticospinal activity, which is clearly observed during voluntary muscle contractions [38].On the other hand, the neural origin of the alpha band oscillatory activity is debated; it is considered to reflect a different mechanism from the beta and low-gamma bands, because cortico-muscular coherence is absent in the alpha band [32,39,40].IMC was found in the restricted lower-frequency band in patients after complete SCI, suggesting that the origin is likely to be spinal [41].Subcortical activities including the cerebellum may contribute to alpha oscillatory coherence [37].
The present study had several limitations.First, the study was conducted on healthy participants.
Compared with healthy participants, patients with central nervous system disorders have lower IMCs associated with deficits in motor drive during their gait [12,15,16].Although only a small number of studies have been reported, gait training increased the IMC in children with cerebral palsy [20].Further studies are needed to investigate the effects of RE-Gait intervention for patients.
Second, our study did not directly measure corticospinal function (e.g., magnetic evoked potentials, cortico-muscular coherence) in parallel with IMC.IMC indicates not only the corticospinal drive, but also signals from other cortical areas including inhibitory signals [42].
Because not only the cortico-spinal tracts are involved in gait control [43], but the effects on other signals should also be considered in future studies.

Conclusion
Robotic ankle planter flexion and dorsiflexion assistance in the pre-and initial swing phase would be effective for learning adaptively modified walking by activating corticospinal tracts.RE-Gait will be a useful tool for re-learning of gait with smooth switching with appropriate forefoot rocker function.

Fig. 2
Fig.2 Gait parameters.(A) The averaged data of the step length of the right leg (filled circles) and

Fig. 3
Fig.3 EMG activities.(A) Pooled EMG activities of the right leg during the gait cycle.The black

Fig. 4 Fig. 5
Fig.4 IMC of the paired TA muscles.(A-C) Pooled IMC of the paired TA muscles in frequency

Figures Figure 1 RE
Figures