Introduction
A recent study suggested that chronic stroke patients maintain the capacity to increase synchronization of neural activity between different brain regions as measured by EEG connectivity. These changes of functional connectivity in the motor cortex through neurofeedback correlate with improvements in motor performance [
1]. Previously, we demonstrated that a novel specific training, the turning-based treadmill program, yielded greater improvements in gait speed and temporal symmetry than regular treadmill training for people with chronic stroke [
2]. We presumed the turning-based treadmill training, which is a challenging and unfamiliar training task for chronic stroke patients, may facilitate brain reorganization and behavioral recovery [
3]. Thus, we sought to understand how such novel gait training promotes brain reorganization in this study.
An EEG-based method has the advantage of real-time recording during walking due to the relative ease of data acquisition. As indicated by the authors of the first study to use an EEG signal recorded during walking, the power increases within numerous frequency bands (3–150 Hz) in the sensorimotor cortex and is more pronounced during the end of the stance phase of walking [
4]. Source localization EEG analysis revealed the importance of the primary somatosensory, somatosensory association, primary motor and cingulate cortex in gait control [
5]. Focal lesions due to stroke may not only affect the functional connectivity of cortical areas [
6] but also impede the neural transmission of descending motor pathways [
7]. Based on spectral analysis, the direct relationship of cortical activities with peripheral movements is still unknown. Accordingly, an analysis of EEG-EMG coherence recorded during treadmill walking was done by Petersen et al. [
8], who demonstrated that cortical activity in the primary motor cortex within the gamma band (24–40 Hz) was transmitted via the corticospinal tract to the leg muscles during the swing phase of walking. In addition, a recent study confirmed the strong correlation between kinematic errors of the lower extremities and fronto-centroparietal connectivity during gait training and post-training in healthy subjects [
9]. However, it remains unknown how novel and challenging gait training shapes the cortico-cortical network and cortico-spinal network during walking in individuals with chronic stroke. Therefore, the aims of the current study were to explore the effects of the turning-based treadmill training, a novel gait training program, on cortico-cortical connectivity and corticomuscular connectivity and to investigate the relationship between connectivity changes and gait performance in chronic stroke patients.
Discussion
In this study, we demonstrated that EEG-EEG connectivity and EEG-EMG connectivity during walking can be enhanced more by a novel gait training program that uses a turning-based treadmill instead of a regular treadmill. Moreover, the improvement in gait symmetry, but not the gait speed, correlated with the modulations in the EEG-EEG and EEG-EMG connectivity over frontal-central-parietal areas of the brain.
Restoring walking ability, regardless of speed or symmetry, is an important goal of poststroke rehabilitation, but conventional approaches are not always successful at improving gait asymmetry [
24,
25]. In our previous and present study, we found that changes in walking speed and temporal symmetry were considered clinically significant as well as statistically significant following the turning-based treadmill training [
2,
22]. Interlimb asymmetry of the single limb support time stems from less time spent on the affected limb or more time spent on the unaffected limb. Less time on the affected limb may indicate poorer balance control of the affected limb, and more time on the unaffected limb may be due to difficulty advancing the affected leg [
23]. During turning-based treadmill training, the outer limb requires relatively greater activation of the ankle dorsiflexors to swing and greater activation of the ankle plantar flexors for propulsion [
26]. Alternately, the inner limb increases extensor muscle activity for stance and increases ankle dorsiflexors use for rapid swing [
26,
27]. When we reviewed the individual data on the single support time of the unaffected limb and affected limb, we found that normalization of temporal asymmetry in our participants was primarily due to a reduction in the single support time on the unaffected leg. Therefore, our turning-based treadmill training seems to have a stronger effect on advancing the affected leg than regular treadmill training. In addition, temporal gait symmetry is an important gait parameter that provides information on energy consumption [
19,
28]. The gait asymmetry was not improved after 12 sessions of regular treadmill training, which may have be due to task repetition reinforcement of asymmetry on the treadmill [
29]. However, the gait asymmetry was improved after turning-based treadmill training in the present study. We thus suggest that this novel training can be applied to improve affected leg advancement and possibly reduce energy consumption in stroke rehabilitation.
In this study, we further demonstrated the improvements in walking speed and gait symmetry accompanied by concurrent modulation in EEG-EEG and EEG-EMG connectivity. Moreover, the larger the changes in functional connectivity were, either in EEG-EEG or EEG-EMG connectivity, the better the recovery of gait symmetry. Therefore, such neuromuscular modulation or neural plasticity induced by training may, at least partially, explain the gait improvement. A previous study suggested that individuals with chronic stroke preserve the capacity to increase synchronization of neural activity between different brain regions as measured by EEG connectivity through neurofeedback training [
1]. It has been noted that a challenging, unfamiliar and more difficult task for stroke patients may increase the possibilities for brain activation to stimulate behavioral recovery [
3]. Turning-based walking requires more balance and limb coordination than walking in straight line [
30]. Therefore, walking on a turning-based treadmill seems to fulfill the features necessary for a task to induce brain activation and thus behavior recovery.
In this study, the EEG-EEG connectivity and EEG-EMG connectivity measured during treadmill walking was used to explore the possible mechanisms for reorganization of functional cortical network and functional coupling between cortical commands and subsequent muscle activation. Involvement of the motor cortex and corticospinal tract in the control of walking has been demonstrated in healthy subjects [
5,
8]. However, how novel gait training shapes brain activities and the descending pathway for chronic stroke survivors remains unclear. We revealed both cortico-cortical and cortico-muscular reorganization in the gamma band over the middle central area, middle frontal area, ipsilesional frontal, and contralesional central area can be further enhanced by challenging gait training, such as the turning-based treadmill training in this study. One possible role of coherent oscillation is to link and promote synchronous neural firing between different neuronal populations with different spatial distributions but that are functionally related [
31]. The coupling between EEG and EMG indicated that cortical control drives peripheral muscular activities through the corticospinal tract during walking. Increased connectivity in the gamma band after specific walking training is in line with previous results that showed that the peak EEG-EMG coherence frequency always shifted to higher frequency (25–40 Hz) from the beta-band during walking compared to those during static contraction [
8]. The gamma-band oscillations in the frontal-central areas play an important role in the execution of the complex goal-directed task which involved motor coordination, cognitive processes and sensorimotor integration [
32]. Therefore, the turning-based treadmill training, which includes specific training and requires complex integration of coordinated muscle activity and multiple sensory systems by the cortex, could result in better walking performance.
Assessment of EEG functional connectivity has been widely used to understand the correlation of brain activities in different cortical areas [
14]. The coherent neural oscillation in the beta band over the ipsilesional motor cortex was able to predict motor improvements in the first week after stroke [
33]. Loss of connectivity thus suggests a low capacity to integrate sensorimotor communication between distant brain regions in stroke survivors [
34]. Following 28 days of upper-extremity treatment increased the resting-state connectivity of the beta-band between the ipsilesional primary motor cortex and premotor cortex paralleling the motor gains [
34]. A recent study showed that 40-session gait training with powered exoskeletons could improve the strength of functional connectivity in the affected hemisphere [
35]. Consistent with above results, we provide evidence that this 12-session turning-based treadmill training for chronic stroke patients could induce synchronization of the cortical network around distinct brain areas together with gait improvement. Furthermore, this challenging turning-based training was demonstrated to be more effective than regular treadmill training with regard to brain modulation and gait performance.
Besides the involvement of neural relocation in the ipsilesional hemisphere, we observed the network contribution over the contralesional frontal-central-parietal area and between the contralesional central area and the lower-extremity muscle for gait recovery. The role of the contralesional motor area for treatment-induced poststroke brain reorganization is controversial. Decreased frontoparietal networking in the contralesional hemisphere has only been found in stroke patients who gained improved motor behavior after ankle robotics training [
36]. However, as some recent studies emphasized the role of the contralesional hemisphere during recover [
34], our results showed that activation in the contralesional hemisphere did not necessarily contribute to unfavorable recovery, especially for chronic stroke survivors. In addition, motor task selection and duration of training session may also account for such discrepancies.
In addition to the intrahemispheric interactions, increased interhemispheric connectivity after turning-based treadmill training was observed. To our knowledge, evidence of the interhemispheric connectivity induced by lower-extremity training is nonexistent. However, these findings were consistent with previous studies’ findings of the strong influence of the ipsilesional dorsal premotor cortex on its contralesional homologue, with improvements in behavioral performance after 3 weeks of upper limb rehabilitation therapy in stroke patients [
37]. Pellegrino et al. [
38] also demonstrated that the 12-week robotic treatment for chronic stroke patients improved hand control and modulated the interhemispheric connectivity in the high beta and gamma bands at rest and such changes correlated with hand control improvements. We reported the first evidence that interhemispheric connectivity during walking can be shaped by novel treadmill training. Considering the bilateral involvement of the lower-extremities during walking, neurophysiological mechanisms of walking regarding the interaction between the lesioned and contralesioned hemispheres in stroke patients warrants further study.
Interestingly, increased coherence in the alpha band was only present in the EEG-EMG connectivity in this study. It is well known that the coherence in the 8–12 Hz frequency over the motor area is most prominent during rest, and such coherence is suppressed (or called desynchronization) during voluntary movement [
39]. However, a recent study found the coherence between EEG and EMG of TA around this frequency band could be induced by peripheral nerve stimulation while performing static muscle contractions [
40]. Petersen et al. also reported a coherence peak at approximately 10 Hz during walking in healthy human subjects [
8]. Our present findings may reflect the possibility of treatment-induced reorganization from afferent inputs to the cortical network during walking [
2].
It is noteworthy that enhancements in gait symmetry, but not the gait speed, were related to the modulations in EEG-EEG and EEG-EMG connectivity over the frontal-central-parietal areas in the brain. The network changes may thus account for the associated gait recovery underlying the central and peripheral neuromuscular mechanisms. The temporal gait symmetry provides information on the differences of single support times between the 2 legs [
20]. As mentioned above, improvement of temporal gait symmetry may be explained by the successful combination of a forced compelled weight-bearing approach and intensive practice of the normal swing phase component. The control of a normal gait pattern demands that patients achieve more muscular control and whole-body coordination [
19]. Therefore, the restoration of gait pattern, rather than gait speed, is more likely to relate to the alteration of brain activities. Although speed is the most widely used measurement and can differentiate levels of gait disability, it is determined by multiple factors [
20]. Investigation of cortical contribution for gait speed in response to therapeutic approaches is needed in the future. Recently, numerous studies have started to design the EEG-based brain-machine interfaces (BMI) to decode the association between the angles of specific joints and neuroelectric cortical activity during gait training [
41,
42]. Since gait symmetry assists in understanding the underlying impairments and treatment-induced cortical changes, a measure of temporal gait symmetry should be included in quantitative gait analysis to better reflect this aspect in gait training with an EEG-based BMI.
This study has several limitations. The sample size is relatively small, and we only assessed limited gait parameters. A larger, randomized controlled clinical trial including comprehensive assessment of gait parameters (i.e., distinct aspects of symmetry or accelerometry signals in time, frequency and time-frequency domains [
43]) is needed to validate the reported brain characteristics of the treadmill training in this study. The improved gait speed and temporal symmetry maintained at 1-month follow-up according to our previous works [
2], however, whether the brain network-related gait recovery can be seen at follow-up is not known. Additionally, only ambulatory patients with chronic stroke were recruited. Therefore, our results may not extend to individuals with acute stroke or more severe motor deficits. In this study, we only provided 32-channel cortical data and one muscle signal to explore possible connectivity during walking. Future studies should use a combination of anatomical and functional neuroimaging techniques to determine the spatial patterns of brain reorganization after gait training. In addition, the task chosen for neurophysiological assessment was walking on the regular treadmill for both groups. Therefore, the assessed task might be more familiar to participants in the regular treadmill training group, but not to participants in the turning-based training group. Whether neural connectivity is influenced by task familiarity is not clear and needs further elucidation. The lack of a complete understanding of the cortical contribution to walking ability and walking recovery after a stroke is due to lacking of valid tools to thoroughly investigate the functional role of the cortex during walking in humans. As mentioned in the methods section, although we have done our best to exclude all motion artifacts and provide stable task-related EEG signals, we believe future studies using more sophisticated engineering solutions for eliciting artifacts could advance our understanding of the “real-time” brain activities during walking and boost development of effective intervention with a neural basis [
44,
45].