A. Subjects

Eighteen subjects (21.0 ± 2.8 years) participated in this study. Before obtaining their written informed consent, we provided the subjects with a detailed explanation of the study procedure. Subjects with an irregularity in their walking pattern, such as bow-leggedness, were excluded as this may interfere with our motion sensor’s ability to capture data. The subjects were informed of the physical fatigue they would experience during the experiment and were accustomed to walking on a treadmill. All protocols (065-1617-EXP) were approved by the Institutional Review Board of California Baptist University to ensure that the study adhered to national and international guidelines for research on humans. Additionally, all the researchers gained CITI certifications in Biomedical Researchers and Responsible conduct of research as an additional safeguard in adhering to regulations and protocols of working with subjects in a research environment. The recruitment phase for this study commenced on 02/05/2023 and ended on 08/15/2023.

B. Experimental setup

All experiments were performed on a treadmill (Woodway USA, Waukesha, WI) equipped with supporting handrails to ensure the safety of subjects. All subjects were given an approximate 5-minute preparatory period to adjust to walking on the treadmill. Two motion capturing markers were attached to the back of the subjects’ shoes and tracked by a camera-based motion-capturing system (Optotrak 3D Investigator, Norther Digital Inc. Canada). This system located the positions of the markers and sent the data to a PC in real-time using a program developed with LabVIEW (National Instruments Corp., TX) to measure the subjects’ spatiotemporal gait parameters such as step length, stride period, and stance duration. In this study, the position of the most forward marker was designated as the heel strike position, while the most backward marker position was identified as the toe-off position. We determined the stride period by measuring the interval from one heel-strike to the next, and the stance duration by measuring the interval from a heel-strike to the subsequent toe-off. It is important to note that the actual instance of toe-off occurs slightly after the marker reaches its most backward position. Nonetheless, we adhered to these measurement criteria consistently, prioritizing the assessment of symmetry between the two legs.

In all trials, subjects walked on a treadmill while being hooked up to a commercial electrical stimulator (MOTIONSTIM8, Krauth+Timermann, German), controllable through a computer program. Two 2” x 2” square electrodes, placed approximately 10 cm apart on the calf and the lower ankle, were used to trigger ankle movement by contracting the calf muscle (Fig 1A). Stimulation locations for the calf group included the sural nerve, gastrocnemius, peroneus, or soleus. The electrical stimulator was battery-powered, and a current isolator was used in conjunction with it for added safety. The stimulation threshold of each subject was determined by the magnitude of the current that induced slightly noticeable ankle plantar flexion. The preferred amplitude ranged from 20 mA– 30 mA during the trials. After documenting subjects’ preferred stimulation amplitude and walking speed, an experimental trial ensued where their preferred treadmill speed and amplitude threshold were employed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1.

(A) Experimental setup showing a subject’s trial with electrodes placed on the lower legs. (B) Each 16-minute trial consisted of a 4-minute baseline, a 7-minute perturbation, and a 5-minute post-perturbation period. For the perturbation period, the offset time was gradually adjusted to 0, -20, -40, -60, -80, and -100 ms from the initial τ. (C) During the initial FES perturbation period, electrical pulses were delivered to both legs at the same frequency as the initial stride, but the timings of the two pulses were separated by time τ, which signifies the initial time separation between the left toe-off and the subsequent right toe-off. Gradually, the timing interval between the two pulses was decreased by Δτ.

https://doi.org/10.1371/journal.pone.0312285.g001

C. Experimental protocol

Experiments were conducted with the following objectives: a) to primarily investigate the impacts of the FES perturbation on spatiotemporal gait symmetry among healthy subjects and b) to explore how the effects varied between the subjects’ explicit and implicit actions during training. Accordingly, all subjects were required to participate in two distinct trials separated by several days: a trial with implicit condition and a trial with explicit condition. The implicit trials were always completed prior to the explicit trials. For the implicit trial, which involved unconscious reaction to FES, subjects were instructed to walk normally and comfortably through the trial. The subjects received no specific instructions, aside from being informed about the periodic electrical stimulation that would occur during the trials. Conversely, the explicit trial required deliberate adjustments, with subjects consciously entraining their gait to the timing of electrical stimulation. Specifically, the subjects were asked to explicitly synchronize the toe-off phase of their gait for both legs with the stimulation.

Each 16-minute trial consisted of three stages: a 4-min baseline period, a 7-min perturbation period, and a 5-min post-perturbation period (Fig 1B). During the baseline period, we measured and calculated the average time (τ) between the left-toe-off and the subsequent right-toe-off phases, as well as the average stride duration from each gait cycle. During the perturbation period, we administered electrical pulses (0.12-second duration, 200 μs biphasic pulses at 90 Hz, 20–25 mA intensity) to both the right and left legs, at the same period of the baseline stride duration. In other words, periodic electrical stimulation was applied to each leg independently, but the stimulations were not delivered simultaneously to both legs; there was a time interval of τ between the stimulations. The time interval between the two electrical pulses delivered to both legs was gradually decreased by the offset time Δτ (Fig 1C). During the initial 60 seconds of the perturbation period, the electrical pulses were applied with zero offset to acclimate subjects to the stimulation. After that, the offset time (Δτ) was gradually adjusted to -20 ms, -40 ms, and -60 ms for 60 seconds each, and -80 ms and -100 ms for 90 seconds, respectively (a 100 ms offset time was approximately 7.5% of the average gait cycle duration in this study). In essence, the separation time between two pulses delivered to both the right and left legs was reduced by Δτ. This manipulation was intended to make the right-toe-off phase occur slightly earlier than in the typical gait cycle, thus inducing gait asymmetry during the perturbation period. Following the perturbation period, electrical stimulation was stopped, and subjects continued to walk for the remaining 5 minutes of the post-perturbation period.

D. Data analysis

The primary measures analyzed in this study were step length and stance duration time. The step length was determined as the distance between the marker on each leg at the heel strike of the leading leg. Specifically, right step length refers to the distance measured at the right heel strike, while left step length refers to the distance measured at the left heel strike. The gait cycle typically consists of two phases: stance and swing. The stance duration was calculated as the percentage of time the foot remained in contact with the ground during the gait cycle (stride time). Right stance duration represents the percentage of time the right foot was in contact with the ground, while left stance duration represents the percentage of time the left foot was in contact with the ground. To evaluate the symmetry of step length and stance duration, a ratio (%) between the measurements of the right and left legs was computed for each gait cycle by using the following formula: 100×(right leg measure–left leg measure)/(0.5×(right leg measure + left leg measure)). For example, a positive step symmetry ratio indicates that the right step length is longer than the left step length, while a negative symmetry ratio indicates that the right step length is shorter than the left step length. In addition, double limb support time is defined by the time in which both feet are in contact with the ground. There are two periods of double support per gait cycle: right double support and left double support. We define right double support as occurring at the end of the right limb’s stance (i.e., the time from left heel strike to right toe-off) and left double support at the end of the left limb’s stance (i.e., the time from right heel strike contact to left toe-off).

For data analysis, the mean and standard deviation (SD) of the symmetry ratios were calculated in successive 30-second intervals to examine the changes in symmetry during the perturbation and post-perturbation periods. The group means and SDs of the step length symmetry and stance duration symmetry were then calculated across all the subjects. To minimize between-subject variability in gait symmetry, the baseline symmetry was used as a reference. The baseline symmetry, which represented the symmetry value measured from the baseline period, was subtracted from the measured symmetry data obtained during all perturbation and post-perturbation periods (including the baseline itself) for each subject.

The primary objective of this study was to investigate the presence of spatiotemporal gait asymmetry induced by the FES perturbation during treadmill walking. We conducted statistical analysis for both the 7-minute perturbation period and the subsequent 5-minute post-perturbation period separately. Paired t-tests were first utilized to determine significant changes in step length and stance duration symmetries, compared to the baseline. These tests allowed us to identify the specific perturbation levels (the offset time) that exhibited significant changes in symmetry. Notably, each perturbation level in our trials lasted either 60 or 90 seconds, and we used the symmetry value obtained from the last 30-second window of each perturbation level. For the post-perturbation period, we assessed only three 30-second epochs (early post-perturbation, middle post-perturbation, late post-perturbation). Additionally, a one-way repeated measures ANOVA was performed to determine how increasing perturbation levels affected changes in step length and stance duration symmetries. All statistical tests were performed at a significance level of p ≤ 0.05, and the p-values are presented in the main text unless depicted in the figures.

Effects of FES-induced perturbations on temporal gait asymmetry adaptation (explicit trial)

In this study, we arranged the implicit trials to always precede the explicit trials to avoid any possible influence or recollection of subjects’ explicit actions on their implicit trials. Here, we first analyzed the outcomes of the explicit trials to comprehend how subjects adapted their gait patterns in response to the FES perturbations. We focused on two adaptive gait parameters: step length and the percentage time in the stance phase. Step length is a spatial gait parameter while percent time in the stance phase (referred to as stance duration) is a temporal measure. During explicit trials, as subjects attempted to synchronize their right and left toe-off phase with the timing of the administered electrical pulse timings, their step length and stance duration changed differently between the right and left leg (Fig 2). As the interval between the right electrical pulse and left electrical pulse decreased (targeting a shorter gap between the left toe-off and the subsequent right toe-off), we observed a decrease in the left stance relative to the right (Fig 2B bottom). Correspondingly, the left swing time became longer than the right swing time (not shown). This resulted in a positive directional change in stance duration symmetry (Fig 3B).

Initially, we thought that the right stance duration would decrease due to the altered timing interval between the two periodic electrical pulses, which was intended to slightly advance the phase of right-toe off. Contrary to our expectations, we observed that the subjects tended to adjust both their left and right gait steps in a complementary manner in response to the perturbation. This compensatory behavior resulted in reduced double limb support time for both legs (Fig 5A), accompanied by a slight decrease in the left stance duration (a corresponding subtle increase in the right stance duration) more than in the opposite leg. The precise mechanism behind why and how subjects adjusted their gait steps in this specific way is still being examined. Fig 3 shows overall changes in gait symmetry relative to the baseline level for the explicit trial (where subjects consciously attempted to entrain their gait with the perturbation). The stance duration symmetry tended toward a positive increase with higher perturbation levels (offset time). Overall, an increase in stance duration asymmetry was observed with higher perturbation levels, except the initial period when the perturbation started. The unexpectedly higher stance duration asymmetry at the lowest perturbation level may be due to subjects initially struggling to align their gait phase with the timing of the perturbations. The adapted stance duration asymmetry returned to symmetry once the perturbation was removed. However, subtle positive asymmetry persisted for up to 5 minutes in the post-perturbation period. These findings demonstrated that the FES perturbation procedure can modify subjects’ temporal gait pattern without necessarily utilizing mechanical devices such as a split-belt treadmill (SBT), which requires physical interaction with the subjects’ walking. Another important observation is the retention trend in the adapted temporal parameter. The positive directional stance duration asymmetry persisted slightly during the post-perturbation period, with only the last epoch showing statistical significance. While this may not provide conclusive evidence, this observation differs from studies using split-belt treadmills. In those studies, stance duration was rapidly changed during adaptation and both legs’ stance time rapidly returned to baseline levels, as did stance time asymmetry [11]. This disparity may be due to differing environmental influences between mechanical-based and the FES-based perturbation approaches. FES perturbation did not involve speed asymmetry, while SBT includes physical speed changes between the right and left legs. Several combinations of other factors, such as psychological and biomechanical factors or afferent sensory signal actions, may influence adaptive behavior differently.

Although stride times for both legs were largely equivalent across all periods, the percentage time in stance, swing, and double limb support changed during the perturbation period. At baseline, stance, swing, and double support times were symmetric for both legs (Fig 4). During the perturbation period, the percent time of right toe-off and right heel-strike (start and end times of the right swing) were slightly advanced, leading to a slight leftward shift of the right swing and consequently right stance time as well relative to those of the left (Fig 4A). This change in right swing and stance times remained consistent throughout all perturbation levels. The shifting right swing and stance times decreased both right and left double support times compared to the baseline (Fig 5A). It was reported that SBT walking induces asymmetry in double limb support time, which is known to proportionally change with the split-belt speed ratio [29]. Although this study observed significant decrease in double limb support times for both legs (Fig 5A), it is important to highlight that no statistically significance differences were found in the double limb support time symmetry during the perturbation period, in contrast to what is typically seen in SBT walking. This implies that physical speed differences between the right and left legs predominantly determine double limb support symmetry, which was not the case with our FES perturbation approach.

Effects of FES-induced perturbations on spatial gait asymmetry adaptation (explicit trial)

Under the explicit condition, the FES perturbation procedure used in this study was primarily oriented toward altering stance and swing symmetry (temporal gait parameter) by entraining toe-off phase with stimulation timing. We anticipated that changes in temporal gait symmetry would inherently influence spatial gait symmetry, such as step length symmetry. However, the change in spatial gait pattern was not as significant as the observed temporal changes. A slight increase in the right step length was noticed at higher perturbation levels (Fig 2B top), but it was not statistically significant. Moreover, despite slight positive directional changes in step length symmetry during the perturbation period (Fig 3A), the large variance inhibited the determination of statistical significance. In summary, the response of step length changes to FES perturbation varied substantially across individuals. This implies that changes in temporal gait patterns are not necessarily tied to changes in spatial gait patterns. Essentially, accommodating temporal gait patterns does not obligate a modification of the spatial gait pattern, or vice versa. Other studies utilizing SBT demonstrated that interlimb coordination can be independently controlled and adjusted in the adult human nervous system without necessarily changing various aspects of intralimb coordination [29, 30]. This suggests that different gait parameters are maintained and adapted through separate processes. Our study’s results are consistent with the prior studies’ findings, indicating that temporal and spatial controls for symmetric gait can be adapted independently.

Effects of FES perturbations under implicit condition

Effective interventions must facilitate efficient motor adaptation and ensure persistent changes (retention) in newly learned motor patterns. Various researchers have proposed that the retention benefit of learning relies on the unconscious or automatic mode within the learning process [19–21]. Therefore, we wanted to understand how FES perturbation affects gait asymmetry changes under implicit conditions. In the implicit trial, subjects walked on a treadmill while receiving electrical pulses but did not consciously alter their gait patterns. We observed a substantial variation in how subjects spontaneously compensated their gait in response to FES perturbation. When examining changes in step length and stance duration, we noticed a slightly reduced left stance duration compared to the right during the perturbation period (Fig 6A bottom), resulting in somewhat positive directional changes in stance duration symmetry (Fig 6B bottom). The degree of asymmetry was significantly smaller than that induced by the subjects’ conscious actions (explicit trial). In the post-perturbation period, there were no signs of aftereffects of the asymmetric stance duration. Regarding the spatial gait pattern changes under the implicit condition, due to large variances across the subjects, we found no significant changes in step length or step length symmetry (Fig 6B top). Additionally, no substantial changes were observed in stance and double support time relative to either leg (Figs 5B and 6).

During implicit trials, we observed that some subjects were unwittingly subjected to perturbation, exhibiting little to no gait entrainment. Others showed substantial entrainment, but most showed only sporadic entrainment throughout the implicit trials. In our earlier study which utilized FES on only one leg, approximately 55% of subjects showed some degree of entrainment [16]. However, the current study showed a lower proportion of subjects showing some degree of entrainment, which could be because FES perturbation was applied to both legs in this study, as opposed to one. It appears more challenging for both legs to spontaneously display entrainment to electrical pulses on each leg than to show entrainment to electrical pulses on a single leg.

We speculate that using a higher stimulation amplitude could have yielded more significant results. However, increasing the amplitude might have more robustly stimulated the muscles and led to greater ankle torque, which was not our intent as a method of inducing gait asymmetry. We maintained minimal stimulation amplitude throughout the implicit trials, as we aimed to understand the extent to which sensory afferent signals caused by FES influence symmetric gait adaptation. If the stimulation intensity needs to be increased to induce significant symmetric gait adaptation spontaneously, this inevitably results in uncomfortable tactile sensations and necessitates incorporation of electrically evoked joint torque as an additional factor in gait adaptation. Another reason for the lack of distinct results under the implicit condition could be that even when entrainment occurred spontaneously in walking, it did not necessarily synchronize with toe-off phases. Some subjects showed entrainment to their heel-strike phases. Consequently, the outcomes exhibited by subjects varied, leading to greater variations in spatiotemporal gait symmetry changes. Under implicit conditions, an experimental protocol designed to accommodate individual differences would likely yield more significant results. Therefore, this suggests the need for further investigation in this area.

Limitations and further direction

The current study represents a pioneering effort to evaluate a novel FES-based perturbation procedure to induce gait asymmetry adaptation during treadmill walking. There are, however, several limitations to consider. First, the stimulation amplitude utilized in this study was confined to a minimum threshold; therefore, the results might be different if varied stimulation intensities were used. For instance, as the stimulation amplitude increases, stronger joint torque follows, which may contribute to inducing asymmetric gait adaptations differently through FES-evoked sensory signals involved in gait control via neural circuits in the spinal cord. Second, the maximum offset time for the perturbation used in this study was capped at 100 ms, which is roughly equivalent to 7.5% of the average gait cycle duration. Should the perturbation become larger, entrainment is unlikely to occur during treadmill walking due to the constraints on walking speed, which is fixed on a treadmill, putting boundaries on gait entrainment. If experiments can be conducted in an over-ground walking setting, the perturbation size could be adjusted more freely, making the effects of FES perturbation more visible. Additionally, due to variations in walking speed and gait styles among the subjects, the range of the offset time (Δτ) did not correspond to the exact percentage time of the gait cycle duration. If gait favorably entrains to a particular percentage of the gait cycle duration, then the offset time should be adjusted based on an exact percentage time of each individual’s gait cycle duration, rather than using an absolute time for more significant results. Third, this study did not explore how subjects’ inherent baseline inherent asymmetry tendencies may affect the directional changes in gait symmetries in response to FES. Although the baseline asymmetry’s statistical significance is negligible, it would be valuable to investigate this further. Another limitation is the short post-FES perturbation observation period, lasting only 5 minutes. To explore the long-term effect, an extended post-perturbation period should be examined to accurately assess any retention effects. These limitations highlight the necessity for further research.