Anticipatory and reactive responses to underfoot perturbations during gait in healthy adults and individuals with a recent mild traumatic brain injury

Abstract

Background:

Following mild traumatic brain injury individuals often exhibit quantifiable gait deficits over flat surfaces, but little is known about how they control gait over complex surfaces. Such complex surfaces require precise neuromotor control to anticipate and react to small disturbances in walking surfaces, and mild traumatic brain injury-related balance deficits may adversely affect these gait adjustments.

Methods:

This study investigates anticipatory and reactive gait adjustments for expected and unexpected underfoot perturbations in healthy adults (n = 5) and individuals with mild traumatic brain injury (n = 5). Participants completed walking trials with random unexpected or expected underfoot perturbations from a mechanized shoe and inertial measurement units collected kinematic data from the feet and sternum. Linear mixed-effects models assessed the effects of segment, group, and their interaction on standardized difference of accelerations between perturbation and non-perturbation trials.

Findings:

Both groups demonstrated similar gait strategies when perturbations were unexpected. During late swing phase before expected perturbations, persons with mild traumatic brain injury exhibited greater lateral acceleration of their perturbed foot and less lateral movement of their trunk compared with unperturbed gait. Control participants exhibited less lateral foot acceleration and no difference in mediolateral trunk acceleration compared with unperturbed gait during the same period. A significant group*segment interaction (p < 0.001) during this part of the gait cycle suggests the groups adopted different anticipatory strategies for the perturbation.

Interpretation:

Individuals with mild traumatic brain injury may be adopting cautious strategies for expected perturbations due to persistent neuromechanical deficits stemming from their injury.

1. Introduction

Mild traumatic brain injuries (mTBI) are associated with quantifiable balance deficits during a variety of balance and gait tasks (Fino et al., 2018), including static balance (Guskiewicz et al., 2001; King et al., 2017, 2014; Powers et al., 2014), gait initiation (Buckley et al., 2017), dual-task level-walking (Büttner et al., 2020) and walking with obstacle crossing (Catena et al., 2009, 2007). Despite the variety of tasks, most of the research on gait after mTBI examines walking over flat surfaces that may not represent the complexity of walking surfaces in daily living (Fino et al., 2018). Therefore, the full consequences of mTBI on gait and balance during daily living remain unclear.

Compared to level ground, complex, uneven terrain imposes additional balance challenges that can increase the demand for predictive and reactive control of gait. Healthy persons use predictive gait adjustments, such as altering foot placement or center of mass (CoM) position, to minimize the potential destabilizing effects of uneven terrain (Bruijn and Van Dieën, 2018). Foot placement is the primary balance control mechanism to accommodate for known upcoming perturbations that cause mediolateral shifts of the CoM. However, changing foot placement may not always be an option in the presence of environmental or temporal constraints. In such environments, ankle and hip torque compensate for the constraints on foot placement, with ankle torque effectively shifting the center of pressure (CoP) and hip torque shifting the position of the CoM (Bruijn and Van Dieën, 2018; Reimann et al., 2018). Tactics for recovering balance following a perturbation are similar to anticipatory adjustments, but balance recovery employs foot torque first during double support (Reimann et al., 2018), with changes in foot placement for correcting gait trajectory occurring over subsequent steps.

Following mTBI, cognitive inefficiencies are also common (Broglio and Puetz, 2008; Lempke et al., 2020) and may interfere with predictive and reactive control of gait over uneven terrain. Accurately predicting and reacting to uneven terrain creates a greater cognitive stress than when walking over flat surfaces (Nordin et al., 2019) and can be difficult for neurologically impaired populations with mobility deficits (Ellmers and Young, 2019). Following mTBI, many individuals struggle with tasks that require attention switching, such as dual-task gait (Fino et al., 2018), and also have poorer simple and choice reaction times (Lempke et al., 2020). Complex environments demand accurate distribution of attentional resources and quick reactions for identifying and managing upcoming perturbations; mTBI-induced neural inefficiencies at any point, from identifying environmental threats to executing movement, may interfere with safe or efficient ambulation in these environments. The confluence of balance and cognitive problems following mTBI suggest such individuals may struggle to make quick and effective adjustments during common complex gait tasks.

The purpose of our study was to compare preparatory and reactive adjustments to underfoot perturbations delivered by a mechanized shoe in healthy adults and adults with recent mTBI. We hypothesized that participants with mTBI would differ in their response to unexpected perturbations compared to healthy control subjects. Further, we expected that healthy participants, but not mTBI participants, would exhibit an anticipatory postural response when given warning of an upcoming perturbation.

2. Methods

2.1. Participants

Healthy controls and individuals with a recent mTBI were recruited from the local community and from the University of Utah outpatient rehabilitation facility. All participants provided informed written consent and participated in this IRB-approved study. Five healthy [4F, mean (SD) age = 23.8 (4.7) years] and five recently concussed [2F, mean (SD) age = 24.6 (5.2) years] adults participated in this study. All mTBI participants were a minimum of two weeks removed from their injury [mean (SD) = 27.4 (12.8) days] and still reporting residual symptoms. Participants with mTBI were required to have a diagnosis from a licensed medical professional such as a physician. Exclusion criteria included: (1) a history of mTBI in the previous year for healthy control participants, (2) a history of neurological or behavioral pathologies that could explain balance deficits, (3) a history of musculoskeletal history that could impact balance within the past year, (4) a history of multiple ankle sprains, (5) any reconstructive surgery of the lower extremities, (6) and any active medications that could contribute to balance deficits.

2.2. Procedure

Participants completed a structured interview to identify concussion history using the Ohio State TBI Identification Method and the Post-Concussion Symptom Scale (PCSS) to record self-reported symptoms before any balance assessment. Next, participants were outfitted with a pair of custom mechanized shoes containing a small aluminum block just proximal to the 5^th metatarsophalangeal joint of the left foot. The perturbation was controlled by a micro servo motor and recessed into the sole of the shoe so that it would not interfere with normal gait. A force-sensitive resistor was housed in the sole of the shoe to register steps. During specified tasks, the motor would rotate the aluminum block outward randomly between the third and sixth stride of the trial and elicit approximately 12° of ankle eversion unloaded and ~5° of eversion when fully loaded (due to compression and deformation of the shoe). The window between the third and sixth stride was chosen to remove the influence of gait initiation and termination from the data.

While wearing the shoes, participants completed several walking trials along a 7.5-m walkway including walking with and without a perturbation (Figure 1). Participants experienced perturbations under two conditions; unexpected and expected. Unexpected perturbation trials consisted of a perturbation without advanced warning. Expected perturbation trials consisted of a loud auditory warning tone delivered one stride prior to the perturbation, coincident with the push-off phase of the perturbed foot, to warn the participants of the upcoming underfoot disturbance. At the beginning of the session, participants acclimated to walking in the shoes by walking at their comfortable pace for two minutes. Next, participants were introduced to the perturbation and the warning tone with an introductory unexpected and expected perturbation trial. Once familiarized, participants completed two blocks of 16 trials, with each block corresponding to either unexpected perturbations or expected perturbations. In each block, four trials contained a single perturbation; the order of the trials were randomized within each block. Following the completion of the blocked trials, two final follow up trial were completed with an unexpected and expected perturbation, yielding 12 total perturbations (six unexpected, six expected) per subject. The initial introductory trial was excluded from all analysis for each condition, for a total of 5 perturbation trials per condition per subject. All perturbations were delivered to the left foot.

Figure 1:

Sequence of walking trials. Participants completed a baseline two-minute single task walk and were subsequently introduced to the underfoot perturbation. After completing the introductory trials, participants completed blocks of randomized trials featuring unexpected or expected perturbations, with the order of the block randomized. After the randomized blocks, two follow up perturbation trials and a final two-minute single task walk were completed.

Wireless inertial measurement units (IMUs) (Opal v2, APDM Inc., Portland OR, USA) were placed over both feet, the lumbar region (approximately L3), trunk (over the sternum), and wrists to collect accelerations, angular velocities, and magnetic field potentials during each trial at a frequency of 128 Hz. Perturbation times were annotated with a button trigger to synchronize all analysis with the perturbations for analysis.

2.3. Data Processing

Data processing and statistical analysis was conducted with custom code in MATLAB (ver. R2020b, The Mathworks Inc., Natick MA, USA). Raw acceleration signals were transformed from a sensor-based local frame to a body-fixed frame initially aligned with the global reference frame using a still period at the start of each trial. Gait events (e.g., heel contact and toe-off) were identified following the methods of Cain et al. (Cain et al., 2017) using local maxima in high-frequency power of the foot acceleration and angular velocity signals. Next, body-fixed acceleration data from all sensors were filtered using a 10 Hz, 4^th order recursive low-pass Butterworth filter. To capture potential changes in lateral foot motion (i.e., foot placement strategy) and in lateral trunk motion (i.e., hip torque strategy), the mediolateral accelerations of the sternum and left foot (perturbed foot) IMUs for three strides of each trial were extracted for further analysis. For perturbation trials, the three strides were: 1) the pre-perturbation stride, 2) the perturbation stride, and 3) the post-perturbation stride for each trial. For non-perturbation trials, three consecutive steady-state gait strides were used. The three non-perturbation strides were chosen to cover the same walking area as perturbation strides and remove any differences from gait initiation and termination, respectively.

To compare perturbation trials to non-perturbation trials within each subject, each stride was time normalized to 200 samples using spline interpolation, and the time normalized data from the perturbation trials were compared to time-normalized strides from five pseudo-randomly sampled non-perturbation trials within the same block using 1-dimensional statistical parametric mapping (1dSPM). The time series of t-scores from the within-subject 1dSPM comparisons produced a standardized difference between perturbation and non-perturbation trials for each condition for each subject (Patakay, spm1d version M.0.4.7). These t-scores served as the primary outcome variable for analysis in this study, where larger magnitude t-scores indicated greater differences between the perturbation and non-perturbation trials. The difference between group t-scores is reported as the standardized difference in figures 3 and 4.

Figure 3:

Mean standardized difference in mediolateral acceleration between unexpected and non-perturbation trials. Mean differences were examined for the three strides surrounding the perturbation. Standardized differences are reported for the trunk (A) and left foot (C) body segments. A score of zero indicates there was no difference in accelerations at that time point between perturbation and non-perturbation trials. Six epochs (dashed boxes) within the three-stride segment were selected for further analysis to assess difference between healthy (black) and mTBI (red) participants at different phases of the gait cycle. Linear mixed effects models (B) were fit for the mean standardized difference within each epoch to examine the presence of significant differences by group, segment, and their interaction. No significant group or interaction effects were identified when comparing unexpected and non-perturbation trials.

Figure 4:

Mean standardized difference in mediolateral acceleration between expected and non-perturbation trials. Comparison between perturbation and non-perturbation trials were done for the three strides occurring before, during and after the perturbation. Standardized difference is reported for the trunk (A) and left foot (C) segments. Six epochs (dashed boxes) within the three-stride segment were selected for further analysis to assess difference between healthy (black) and mTBI (red) participants at different phases of the gait cycle. Linear mixed effects models (B) were fit for the mean standardized difference within each epoch to examine the presence of significant effects of group, segment, and their interaction. A significant group*segment interaction was found in epoch 1 where participants would have been preparing for foot placement of the perturbation step.

Additionally, we calculated step time, terminal double support phase, and the angle of the perturbation. Step time and terminal double support phase were calculated using toe-off and heel contact gait event times for the six bilateral steps surrounding the perturbation. Step time is the time elapsed from heel contact of one step to heel contact of the contralateral foot. Terminal double support phase is the percentage of a stride occurring between contralateral heel strike and ipsilateral toe-off. Perturbation angle was calculated using the accelerometer to determine the inclination of the foot during stance. To calculate the perturbation angle, an Euler angle rotation sequence was applied to raw acceleration data, providing pitch and roll angle changes throughout each trial (Pedley, 2013). To confirm the size of the perturbation, the pitch and roll angle during stance phase of each pre-perturbation step was subtracted from the pitch and roll angle of the stance phase of each perturbation step. Stance periods were identified for each step using acceleration and angular velocity thresholds established by Rebula et al. (Rebula et al., 2013). This process was repeated for each perturbation trial to attain a mean perturbation angle for each condition per subject.

2.4. Statistical Analysis

Six consecutive quarter-stride epochs surrounding the perturbation, corresponding to critical phases of movement, were selected for statistical comparison. The first epoch encompassed the quarter-stride prior to the heel strike of the perturbation step during which time the soon-to-be-perturbed limb was in swing and participants could alter foot placement prior to the perturbation. The following epochs assess any differences that occur during heel strike and loading with the perturbation (epoch 2), mid-stance phase with the perturbation and heel strike of the contralateral limb (epoch 3) push off of the perturbed limb and loading of the contralateral limb (epoch 4), swing phase of the perturbed limb immediately after the perturbation (epoch 5), and heel strike and loading of the first recovery step of the perturbed limb (epoch 6). The mean t-score over the course of each epoch was calculated for each group and segment. To test for group differences in the mean t-score for each epoch, separate linear mixed effect models were fit with fixed effects for group, segment, and their interaction, and random intercepts by subject to account for repeated measures. In each model, the fixed effect of group, and the group*segment interaction, indicating different balance control strategies by group, were the effects of interest for this study. Normality was confirmed by checking the residuals of each model. A Bonferroni correction was used to account for multiple comparisons and the significance threshold was adjusted to α = 0.004

To compare the magnitude of the perturbation, as indicated by foot angle during stance, linear mixed models were fit to the mean perturbation angle using fixed effects of group, condition, and their interaction. Main effects and interactions were assessed using a 0.05 significance level. The fixed effect of group, and the group*condition interaction were the effects of interest.

3. Results

3.1. Spatiotemporal Results

Descriptive results of step times and terminal double support phases are given in Table 1. Descriptively, control and mTBI participants showed similar step times throughout all conditions. Terminal double support phase was largely similar between groups for each condition with the exception of the first step following the perturbation, where mTBI participants were only in terminal double support for 9.54% of the gait cycle.

Table 1:

Spatiotemporal characteristics during the right and left steps surrounding the perturbation. PTB+0 is the left step during which participants experienced heel strike with the perturbation present.

Step Time (s)
Condition	PTB − 2	PTB − 1	PTB + 0	PTB + 1	PTB + 2	PTB + 3
Normal
Control	0.56 (0.04)	0.56 (0.03)	0.56 (0.03)	0.57 (0.03)	0.56 (0.03)	0.57 (0.04)
mTBI	0.57 (0.03)	0.58 (0.03)	0.57 (0.03)	0.57 (0.03)	0.57 (0.03)	0.59 (0.03)
Unexpected
Control	0.57 (0.03)	0.57 (0.04)	0.56 (0.04)	0.57 (0.03)	0.57 (0.03)	0.58 (0.04)
mTBI	0.57 (0.03)	0.58 (0.03)	0.57 (0.04)	0.58 (0.03)	0.58 (0.03)	0.59 (0.04)
Expected
Control	0.56 (0.04)	0.56 (0.03)	0.57 (0.03)	0.57 (0.03)	0.57 (0.04)	0.58 (0.04)
mTBI	0.57 (0.03)	0.58 (0.02)	0.58 (0.03)	0.57 (0.02)	0.58 (0.03)	0.59 (0.03)
Terminal Double Support (% Gait Cycle)
Normal
Control	10.1 (1.1)	10.3 (0.8)	10.3 (1.2)	10.4 (1.0)	10.5 (1.4)	10.9 (1.4)
mTBI	9.9 (1.2)	10.0 (1.0)	9.9 (1.1)	9.9 (1.1)	10.0 (1.1)	10.5 (1.1)
Unexpected
Control	10.6 (0.3)	10.3 (0.7)	10.1 (0.9)	10.4 (1.2)	10.6 (1.1)	10.9 (1.6)
mTBI	10.1 (1.5)	9.9 (1.5)	9.7 (1.4)	10.0 (1.3)	10.1 (1.4)	10.9 (1.4)
Expected
Control	10.4 (1.1)	10.4 (0.8)	10.7 (1.1)	10.3 (1.2)	10.8 (1.4)	10.9 (2.2)
mTBI	10.0 (1.1)	10.1 (0.9)	10.1 (1.2)	9.5 (1.0)	10.0 (1.4)	10.9 (1.7)

3.2. Unexpected Condition

There were no significant interactions or main effects for group for any of the stride windows examined. A significant main effect of segment was observed for the epochs containing mid-stance and toe-off respectively (Figure 3, epochs 3 and 4). During these epochs, greater lateral accelerations of the left foot (i.e., perturbed foot) were measured relative to normal non-perturbed strides.

Expected Condition

There was a significant group*segment interaction and a significant main effect for segment during the swing phase before the perturbation (Figure 4, epoch 1). During this epoch, mTBI subjects exhibited greater lateral foot acceleration and less lateral trunk acceleration during the expected perturbation trials compared with unperturbed gait in the same block. During the same epoch, control subjects exhibited less lateral foot acceleration during the expected perturbation trials compared with unperturbed gait and exhibited no differences in mediolateral trunk acceleration relative to unperturbed gait (Table 2, epoch 1).

Table 2:

Results of the linear mixed-effects models for each epoch. Bolded p-values were significant at an alpha level of 0.05.

	EPOCH 1			EPOCH 2			EPOCH 3			EPOCH 4			EPOCH 5			EPOCH 6
Unexpected	β	SE	P	β	SE	P	β	SE	P	β	SE	P	β	SE	P	β	SE	P
Segment	−0.31	0.34	0.379	0.22	0.27	0.419	−8.73	0.92	<0.001	−6.82	0.94	<0.001	−1.05	0.48	0.043	−0.23	0.42	0.600
Group	−0.27	0.45	0.551	−0.34	0.42	0.428	−0.84	0.92	0.373	0.14	1.21	0.905	−0.04	0.62	0.949	0.10	0.42	0.825
Segment*Group	0.19	0.48	0.694	−0.44	0.38	0.263	1.14	1.30	0.394	−0.42	1.34	0.760	0.18	0.67	0.791	0.36	0.60	0.551
Expected
Segment	1.44	0.38	0.002	0.40	0.38	0.311	−6.92	1.47	<0.001	−4.98	1.18	<0.001	−0.55	0.55	0.339	−1.21	0.62	0.068
Group	1.07	0.42	0.022	0.69	0.43	0.127	2.99	1.47	0.059	2.29	1.18	0.070	0.83	0.65	0.216	0.14	0.69	0.843
Segment*Group	−2.92	0.54	<0.001	−0.78	0.53	0.161	−3.16	2.08	0.148	−3.48	1.67	0.053	−1.35	0.78	0.105	0.11	0.88	0.897

No significant interactions or main effects were observed during epoch 2. Similar to the unexpected condition, there was a significant effect for segment during the windows containing mid-stance (β = −6.92, SE = 1.47, p < 0.001) and toe-off (β = −4.98, SE = 1.18, p < 0.001) of the perturbation stride (Figure 4, epochs 3 and 4). This effect throughout mid-stance and toe-off, indicated a large standardized difference of the measured mediolateral acceleration of the foot between normal and perturbed steps, due to the presence of the underfoot disturbance, but no difference in the mediolateral acceleration of the trunk. Neither the swing phase nor the heel strike following the perturbation showed any significant interactions or main effects.

3.3. Perturbation Angle

There was a significant group*condition interaction for perturbation angle (roll) induced by the perturbation (β = 1.34, SE = 0.56, p < 0.019). Mean perturbation angle was similar between groups during unexpected perturbation trial [control mean angle (SD) = 4.99° (1.78°); mTBI mean angle (SD) = 5.21° (1.87°)]. During expected perturbation trials, control participants exhibited a similar mean perturbation angle to unexpected trials [control mean angle (SD) = 4.91° (1.56°)], while mTBI participants exhibited larger perturbation angles [mTBI mean angle (SD) = 6.73° (2.17°)] in the expected condition.

4. Discussion

This study investigated the effect of small unexpected and expected underfoot perturbations on gait in healthy adults and adults with recent mTBI. We hypothesized that individuals with mTBI, in comparison with healthy control subjects, would differ in their kinematic response when recovering from perturbations, and that they would exhibit smaller or insufficient anticipatory movements to prepare for upcoming expected perturbations. Contrary to our hypothesis, healthy control and mTBI participants exhibited no differences in mediolateral acceleration profiles at the trunk or perturbed limb when responding to unexpected or expected perturbations. However, when participants were given warning of a perturbation one stride before it occurred, the two groups appeared to utilize different anticipatory balance strategies, supporting our second hypothesis. Specifically, participants with mTBI limited lateral trunk acceleration and increased the lateral acceleration of their swing foot. Conversely, control participants did not change trunk motion and decreased the lateral acceleration of the swing foot. The results support our hypothesis that the control group would better prepare for expected perturbations.

The strategy adopted by the mTBI group in preparation for expected perturbations is similar to cautious preparatory strategies commonly observed in other clinical populations. For example, older adults walking over complex terrain typically adopt a larger step width and limit mediolateral trunk motion to increase stability in the frontal plane. Such strategies in older adults have been observed in environments that feature unpredictable terrain with limited visual information and multiple disturbances of different quality (Thies et al., 2005). Considering the small size of the perturbation, the consistent dimensions of the perturbation, and the frequency at which it occurred, the strategy we observed in mTBI participants appears overly cautious and perhaps unnecessary. When stepping with an underfoot perturbation that causes ankle inversion or eversion, the underfoot CoP shifts towards the perturbed side of the foot (Nnodim et al., 2013). Control subjects adopted a strategy that suggests a more medial foot placement prior to expected perturbation. Shifting foot placement medially may have helped control subjects compensate for the lateral shift of the CoP during the perturbation and preserving the preferred distance between the CoP and CoM. Given that healthy control participants were able to encode a specific response, it is possible that mTBI participants struggled to perceive the size or nature of the perturbation. If so, a lack of detailed knowledge about the perturbation would hinder the formation and updating of a safe and precise motor plan and could explain why the mTBI group adopted a generalized cautious strategy for the perturbation. While speculative, our results may also indicate that the mTBI group struggled to quickly interpret specific motor commands, namely, the warning tone prior to expected perturbations. Reaction time deficits are common following mTBI and have been observed as far as 60 days after injury and despite the resolution of clinical symptoms (Lempke et al., 2020). Recent studies have also shown that individuals with recent mTBI struggle to incorporate task-relevant sensory information during tests of functional reaction time (Lynall et al., 2018). Between the warning tone and the perturbation in this study, participants were afforded approximately 600-750 milliseconds between hearing the warning tone and the onset of the perturbation during stance. During this time, the participant was required to hear the tone, interpret the meaning, and select an appropriate balance strategy before the perturbation. Research of online adjustments during human gait suggests that adjustments can happen in under 200 milliseconds (Weerdesteyn et al., 2004; Zhang et al., 2020) and also that response latencies to obstacles worsen with aging, possibly due to cognitive processing deficits (Young and Hollands, 2012). While mTBI is notably heterogeneous with unclear, and at times conflicting, neuroanatomical origins, a large group of neurocognitive deficits, including slower reaction time and poor sensory integration, are common to mTBI (Broglio and Puetz, 2008; Lempke et al., 2020). Considering these deficits, it is possible that the cautious gait strategy was adopted by the mTBI group to compensate for an inability to develop a precise motor response in a short period of time.

Notably, the mTBI group exhibited significantly larger perturbation angles during expected trials compared to unexpected trials. The larger perturbation angle during expected trials may have been caused by a more lateral foot placement, as suggested by increased lateral acceleration during late swing. We suspect that the lateral foot placement shifted the distribution of plantar pressure medially and thereby decreased the deformation of the shoe around the perturbation flap. The unloaded perturbation angle was approximately 12 degrees but deformation of the sole resulted in the loaded perturbation being approximately 5 degrees. During expected trials, the mTBI group exhibited ankle roll around 6.5 degrees. By adopting a strategy that suggests a more lateral foot placement, the mTBI group likely made the gait disturbance larger than had they walked over it normally as in the unexpected condition.

This study should be interpreted within the context of its limitations. Principally, this data relied on a small sample size and acceleration data from IMUs. While we attempted to recruit a larger sample size, the servo motor and perturbation mechanism failed, and we were unable to source additional motors as they had been discontinued by the manufacturer. A new configuration, while possible, may have created small variation in the physical perturbation, and for that reason, we opted not to combine data from subsequent iterations of the perturbations with the data presented here. Additionally, it is difficult to draw conclusions regarding specific movement strategies without position data. Considering this, our results should be taken with some caution as the foot is not a fixed structure on the limb and acceleration differences could have been due to foot movement independent of foot placement. Lastly, constraints in the testing environment prevented the interpretation of gait corrections that may have occurred after the first recovery step that have been observed in other studies utilizing a similar perturbation paradigm (Kim et al., 2013). Future projects should focus on investigating anticipatory and reactionary movement to underfoot perturbations in mTBI with positional marker data in unconstrained walking environments that allow for postural analysis of multiple steps after the disturbance occurs. Future work should also aim to incorporate more types of perturbations. Incorporation of bilateral perturbations, medial perturbations, or perturbations at different parts of the foot will help answer more questions about how individuals with mTBI handle predictable and unpredictable underfoot disturbances during gait.

5. Conclusion

Individuals with recent mTBI demonstrated cautious anticipatory foot and trunk motion prior to a small, expected perturbation. The cautious balance strategy, not observed in controls, likely reflects either an inability to quickly prepare for the perturbation or an issue interpreting underfoot disturbances. Given these possibilities, future research should further examine relationships between mTBI and online control of movement while walking over varied terrain.

Figure 2:

(A) The mechanized shoe. Perturbations were delivered via a small aluminum block once per perturbation trial and caused approximately 5 degrees of ankle eversion when fully loaded. (B) A sample filtered acceleration trace for the three left foot strides surrounding the perturbation, with the middle stride (red foot) corresponding to the stride with the perturbation during stance. (C) A sample filtered acceleration trace for three consecutive strides from a trial without a perturbation.

Highlights.

• Persons with and without mild traumatic brain injury received gait perturbations

• Perturbations were similar to stepping on a pebble

• Both groups responded similarly to unexpected disturbances

• Healthy adults adopted a precise strategy for expected perturbations

• The injured group used a cautious strategy for expected perturbations

Abbreviations

mTBI

Mild Traumatic Brain Injury

CoM

Center of Mass

CoP

Center of Pressure

IMU

Inertial Measurement Unit

1dSPM

1-dimensional Statistical Parametric Mapping