Skip to main content
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Apr 1.
Published in final edited form as: Ann Biomed Eng. 2020 Jan 16;48(4):1309–1321. doi: 10.1007/s10439-020-02454-8

Adaptive Ankle Resistance from a Wearable Robotic Device to Improve Muscle Recruitment in Cerebral Palsy.

Benjamin C Conner 1, Jason Luque 2, Zachary F Lerner 1,2
PMCID: PMC7096247  NIHMSID: NIHMS1550090  PMID: 31950309

Abstract

Individuals with cerebral palsy can have weak and poorly coordinated ankle plantar flexor muscles that contribute to inefficient walking patterns. Previous studies attempting to improve plantar flexor function have had inconsistent effects on mobility, likely due to a lack of task-specificity. The goal of this study was to develop, validate, and test the feasibility and neuromuscular response of a novel wearable adaptive resistance platform to increase activity of the plantar flexors during the propulsive phase of gait. We recruited eight individuals with spastic cerebral palsy to walk with adaptive plantar flexor resistance provided from an untethered exoskeleton. The resistance system and protocol was safe and feasible for all of our participants. Controller validation demonstrated our ability to provide resistance that proportionally- and instantaneously-adapted to the biological ankle moment (R = 0.92±0.04). Following acclimation to resistance (0.16 ± 0.02 Nm/kg), more-affected limbs exhibited a 45±35% increase in plantar flexor activity (p = 0.02), a 26±24% decrease in dorsiflexor activity (p < 0.05), and a 46±25% decrease in co-contraction (tibialis anterior and soleus) (p = 0.02) during the stance phase. This adaptive resistance system warrants further investigation for use in a longitudinal intervention study.

Keywords: gait, task-specificity, co-contraction, training, soleus, untethered

Introduction

The ankle plantar flexors are a key component of efficient gait, accounting for approximately 50% of propulsive forces during walking 16. In addition to propulsion, the plantar flexors extend the knee joint during midstance for improved posture, and play an indispensable role in the exchange between the potential and kinetic energy of the center of mass, contributing to an inverted pendulum movement40. Reduced function of the ankle plantar flexor muscles is likely a significant contributor to the gait dysfunction in a majority of children with spastic cerebral palsy (CP), a neuromuscular disorder arising from injury to the brain before cerebral development is complete 32. Children with spastic CP have significantly weaker isometric strength of the lower extremities than their typically developing peers 44, and muscle weakness in children with CP is significantly correlated to walking ability and gross motor function, with the strongest correlation existing with plantar flexion strength 14. This is in agreement with muscle weakness being more pronounced distally in children with CP 5. Additionally, strength to weight ratio in children with CP is predictive of independent ambulatory ability 10. This relationship is supported by the findings of a feasibility study on powered plantar flexion assistance using an ankle exoskeleton for children with CP, which resulted in improved gait kinematics and reduced the energy cost of walking25, which is typically 2 to 3 times higher in this population 34.

Several interventions are typically implemented to ameliorate gait dysfunction in children with CP. Some children receive orthopedic surgery to addresses lever arm dysfunction caused by the secondary bone deformities due to abnormal loading forces 15. However, these procedures do not address muscle weakness, which is an equally important component in the moments produced at joints for movement. Physical therapy (PT) is central to the maintenance of function in CP, but there is limited evidence on the effectiveness of various PT interventions for improving mobility in this population. Strength training, specifically, was found to have moderate evidence of no effect on walking speed or stride length, and inconclusive evidence regarding improvements in gross motor function 1,31. In a randomly controlled trial of 51 children with CP, 12 weeks of progressive strength training did not lead to improved mobility 37. There is an apparent disconnect between the contribution of plantar flexion weakness to gait dysfunction and the effectiveness of strength training for improving gait in CP.

One possible explanation for this may be the specificity of training principle of motor learning that is applied in therapeutic interventions, which suggests that to improve a motor task, that specific task needs to be practiced within a functional context 45. Previous studies 37,38,41 looking at the effect of strength training on walking ability in children with CP have been limited by resistance exercises that strengthen the lower limbs through an isolated movement. While this serves to increase strength, it does not train the motor control that would allow an individual to effectively utilize this strength during gait. Research suggests that improving neuromuscular function requires repetitive volitional engagement during functional tasks 36.

There have also been studies investigating the use of task-specific gait training. In particular, focus has been on partial body weight supported treadmill training via devices such as the Lokomat, with observed improvements in gross motor function, walking speed, and endurance3,4,12,29,39. One potential weakness of this task-specific training is the relative unloading of body weight, which serves as an important afferent signal for modifying the amplitude of muscle activation2,27,35; prolonged unloading over time may result in muscle atrophy and declines in neuromotor performance of the plantar flexors7. In addition, the majority of robot-assisted treadmill training for individuals with CP have used a tethered system (e.g., Lokomat19), which does not offer the possibility for training overground or in more ecologically valid contexts, such as at home. Another promising modality for task-specific training in individuals with CP is the use of biofeedback, where features such as gamification and virtual reality provide cues for improved walking. Studies on biofeedback training have been generally positive6,8,13, yet more research is needed to determine if biofeedback alone is sufficient for transferring motor learning to real world performance28.

There is a growing understanding that to improve walking ability, individuals with CP require both increased strength and improved selective neuromuscular control. Coupled with our knowledge on the important role of the ankle plantar flexor muscles during walking, this suggests that effective gait rehabilitation interventions should be designed to provide ankle plantar flexor resistance that is perfectly timed during walking to elicit the desired neuromuscular firing pattern that reinforces appropriate function, leading to improved strength, coordination, and gait. This is supported by the finding that increasing engagement of the soleus muscle while walking can lead to improved gait kinematics 20. Additionally, to promote neuroplasticity and motor learning, active neuromuscular engagement has been shown to be superior to passive engagement for improved performance and cortical reorganization 26. This motivates the need for controlling a functional resistance intervention so that it is responsive to user input to maximize active engagement. We theorize that adaptive plantar flexor resistance provided during the propulsive phase of walking could lead to both increased strength and improved neuromuscular control following a suitable training intervention. However, we are not aware of any studies that have demonstrated a joint-level wearable intervention capable of eliciting improved neuromuscular engagement of a limb with neurological-deficiency during a functional locomotor task.

The first goal of this study was to develop and validate a control scheme for a wearable exoskeleton capable of providing adaptive resistance proportional to the biological ankle moment during walking for individuals with spastic CP. The second goal of this study was to test the feasibility of walking with plantar flexor resistance for individuals with spastic CP, and assess the neuromuscular response to resisted walking for preliminary clinical validation of this functional training concept. We hypothesized that walking with resistance would increase more-affected-limb, stance-phase plantar flexor muscle activity. In addition, we hypothesized that with our proportional resistance control scheme, the increase in stance-phase plantar flexor activity would not be matched by an increase in stance-phase dorsiflexor activity, leading to overall reduced ankle co-contraction. To test these hypotheses, we recruited eight total individuals with spastic CP to walk with our adaptive ankle resistance platform.

Materials and Methods

A. Exoskeleton Platform for Wearable Resistance

We used a previously developed untethered ankle exoskeleton suitable for use with individuals with CP as the platform for providing wearable plantar flexor resistance 17. Briefly, this lightweight ankle exoskeleton, with customizable foot inserts and calf cuffs for individualized fit (Figure 2), provides bilateral ankle torque using high-performance DC motors (EC-4pole, Maxon) powered by an onboard battery (910 mAh E-flite 6S) to actuate a pulley aligned with each ankle joint in the sagittal plane via a Bowden cable transmission. Torque sensors (TRT-500, Transducer Techniques) at the ankle joint provide feedback for the motor torque controller. The exoskeleton’s control module contains a custom printed circuit board, wireless communication, a microcontroller, motor drivers, and signal processors. The exoskeleton is controlled wirelessly via a custom MATLAB graphical user interface (GUI). The microcontroller implements a finite state machine and resistance control algorithm, and streams experimental data to the GUI.

Figure 2. Experimental setup for controller and neuromuscular validation:

Figure 2.

(A) Participant outfitted with a custom ankle exoskeleton and reflective markers on an instrumented treadmill for collection of 3D kinematics and kinetics while walking with resistance for validation of the adaptive resistance control scheme (subjects V1 & V2). (B) Participant outfitted with a custom ankle exoskeleton and bilateral soleus and tibialis anterior surface EMG sensors for assessment of the neuromuscular response to walking with resistance (subjects 1 – 6). (C) Ankle exoskeleton components worn by all participants.

We developed a purpose-built exoskeleton resistance controller based on two primary design criteria. The first criterion was task-specificity: the controller must facilitate increased neuromuscular firing of the plantar flexor muscles during the portion of the gait cycle when they function to propel the body forward (i.e. push-off). The second criterion was user engagement: the controller should be responsive and provide a mechanical cue in response to changing user input (i.e. as someone pushes harder, more resistance is provided). To meet these goals, we implemented a proportional joint-moment control scheme designed to provide adaptive resistance for a user to actively engage with during the stance phase of walking (Figure 1) 17. Embedded foot sensors (FlexiForce A201, Tekscan) placed under the ball of the foot were used to estimate the biological ankle moment in real-time. Plantar flexor resistance from the exoskeleton during walking was provided by multiplying a nominal torque setpoint (e.g. 0.15 Nm/kg) by the relative estimated ankle joint moment established from the embedded foot sensors and a calibration trial. The relative estimated ankle joint moment was calculated by normalizing the instantaneous estimated ankle moment during training by the peak estimated ankle moment measured during a calibration trial. Therefore, if the real-time estimated ankle moment reached the estimated peak from the calibration trial, the relative estimated ankle joint moment would be equal to 1, and 0.15 Nm/kg of resistance torque would be applied. If the relative estimated ankle joint moment was not equal to 1, the resistance torque would be scaled accordingly. The torque setpoint was modified to prescribe an appropriate level of resistance for each user across visits.

Figure 1. Theoretical concept of the adaptive functional resistance intervention:

Figure 1.

1) Embedded sensors are used to estimate the biological ankle moment in real-time; 2) Plantar flexor resistance is provided proportional to each individual user’s estimated ankle moment in real-time; as a result, resistance is controlled to be responsive to user input; 3) Adaptive resistance requires the user to increase their volitional engagement of the plantar flexors and acts as a mechanical cue for the neuromuscular system to recruit additional motor units to improve “push off” during late stance.

B. Participants

Eight individuals diagnosed with spastic CP were recruited for this study to validate the control scheme and clinically assess the neuromuscular response of walking with the functional resistance intervention. This study was approved by the Northern Arizona University Institutional Review Board (#986744), and informed written consent was received from adult participants, or from a parent or legal guardian for minor participants. An additional consent was obtained for publication of any identifying videos or images. Controller validation (Goal 1), requiring access to an instrumented treadmill and motion capture system, was completed at the Northern Arizona University Human Performance Laboratory (Flagstaff, AZ). The feasibility and neuromuscular response protocol (Goal 2), requiring electromyography measurement and increased access to potential participants, was completed at the Northern Arizona University – Phoenix Biomedical Campus (Phoenix, AZ).

Inclusion criteria for participation in this study included the ability to walk continuously for at least six minutes with or without support, age 9 to 21 years, Gross Motor Function Classification System (GMFCS) Levels I – II, and the ability to follow simple directions. Exclusion criteria included orthopedic surgery within the past six months or any conditions that would prevent safe participation.

We recruited eight individuals with CP for participation in this study (Table 1): ages 9 – 17 years old, seven males and one female, and GMFCS I (n = 4) and II (n = 4). The first two participants (Subjects V1 & V2) completed a one visit protocol for the validation of the adaptive resistance control scheme. The other six participants (Subjects 1 – 6) completed a 4–5 visit protocol for the neuromuscular validation of the resistance intervention.

Table 1.

Participant characteristics and walking protocol information

Subject Gender Age (yrs) Height (cm) Body mass (kg) GMFCS levela Treadmill supportb More-affected limb Gait typec Gait speed (m/s) Resistance level (Nm/kg) MVC (Kg-f)d
R L
V1 M 9 137.0 30.7 I None left Mild crouch and apparent equinus 0.75 0.15 N/A
V2 M 13 151.0 44.0 I None left Crouch 1.00 0.15 N/A
1 M 17 173.5 54.4 II Supported Right Crouch 0.58 0.13 14 28
2 M 14 143.0 42.0 II Supported Right Crouch 0.45 0.15 9 13
3 M 16 157.0 42.5 II Supported Right Winters type II (late-stance hyperextension) 0.40 0.16 27 14
4 M 15 160.5 59.0 I None Right Crouch 0.67 0.17 21 29
5 F 16 160.5 68.5 I None left Poor eccentric control of pre-tibial group* 0.72 0.15 32 19
6 M 12 131.5 30.2 II None Right Crouch 0.49 0.20 13 16
a

GMFCS: Gross Motor Function Classification System

b

Treadmill support: indicates whether or not a participant held on while walking on the treadmill; supported means the participant held with at least one hand, while none means the participant did not hold on. This was consistent between conditions.

c

Gait type: as identified by a licensed physical therapist, with ‘crouch’ gait defined by those parameters set by Gage et al16 and ‘Winters type II’ defined by those parameters set by Winters et al46. Note: all subjects were characterized by a weakened push-off on their affected side.

*

Poor eccentric control of pre-tibial group: Subject 5’s gait type did not fit a specific criteria as defined by Gage or Winters, but instead was characterized by poor control of the ankle on heel strike and push-off.

d

MVC: maximum voluntary contraction of the ankle plantar flexors, average of three trials for each side.

C. Experimental evaluation

Goal 1: Validation of the adaptive resistance control scheme

Participants were first evaluated by a licensed physical therapist for physical characteristics, GMFCS level, and gait type (Table 1). To validate that the ankle resistance torque was provided proportional to the instantaneous biological ankle moment, we measured the correlation between estimated biological ankle moment and time-synced exoskeleton torque. For this controller validation, two participants (Subjects V1 – V2) walked with resistance (torque magnitude set to 0.15 Nm/kg) at a self-selected speed on a split-belt in-ground instrumented treadmill (Bertec; 960 Hz). Participants were outfitted with reflective markers on the lower extremity, pelvis, and trunk. Three-dimensional marker trajectories were recorded using a motion capture system (Vicon, 10 cameras; 120 Hz; Figure 1) (for marker set description 25). Participants began the trial with a 60 – 120 second calibration period while walking at their self-selected speed on the treadmill. Next, the torque setpoint of 0.15 Nm/kg of resistance was applied, and exoskeleton torque, estimated biological ankle moment, three-dimensional marker trajectories, and ground reaction forces were recorded over ten consecutive gait cycle. Using the recorded marker trajectories, a previously reported OpenSim full-body musculoskeletal gait model24 was scaled to the anthropometrics of each participant. The model had 18 primary body segments and 21 degrees of freedom; a ball-and-socket joint at the third lumbar vertebra, 3 translations and 3 rotations at the pelvis, a ball-and-socket joint at each hip, tibiofemoral joints with coupled translation and rotation, and revolute ankle and subtalar joints. Joint kinematics and kinetics were calculated in OpenSim via inverse kinematics and inverse dynamics, respectively 11. Kinematic and kinetic data were filtered using a 4th order zero-lag Butterworth filter at 6 Hz and 12 Hz, respectively.

Goal 2: Feasibility of plantar flexor resistance and assessment of neuromuscular response

Participants were evaluated by a licensed physical therapist on their first visit for physical characteristics, GMFCS level, gait type, and plantar flexor strength (Table 1), and outfitted with a custom ankle exoskeleton. Plantar flexor strength was tested using a hand-held dynamometer placed under the pad of the foot with the participant lying supine. Participants were instructed to push as hard as possible for three seconds, and the average of three trials was recorded. Participants then spent four to five “acclimation” visits walking at a self-selected speed on a treadmill with exoskeleton ankle resistance, as we anticipated a learning effect with this new training modality. The magnitude of resistance was initially set at 0.075 Nm/kg and increased over subsequent visits based on user performance (i.e. reaching the prescribed torque setpoint), rate of perceived exertion, and self-identified soreness before and after a visit. Soreness was self-identified at the end of each visit on the following scale: none, mild, moderate, severe, and very severe. Participants walked for 20 minutes with resistance on each acclimation visit. Seated rest breaks were provided based on participant preference. While walking with resistance, participants were advised to focus on their more-affected side, or limb with greatest impairment, as determined by asking the subject and/or parent or guardian, and confirmed by the physical therapist. Additionally, to distinguish the more-affected limb from the less-affected (or unaffected) limb, maximum voluntary contractions (MVC) of the plantar flexors of each limb was tested (Table 1). Differences in ankle torque generated between limbs aligned with the self-reported more-affected/less-affected limbs for all subjects but Subject 3, who was non-verbal (his parent believed that he did not understand the MVC directions and as a result, limb dominance was based on self-report & therapist assessment).

Frequent verbal instruction was given to the participants while walking with resistance: variations of “push against the resistance”. Output from the control assembly provided the research team with real-time feedback on engagement with the exoskeleton (i.e. how well participants were pushing against the resistance during stance phase), and the level of resistance was decreased or increased according to performance and rate of perceived exertion. Subject 3 was unable to receive or acknowledge the instruction to push against resistance because of verbal communication deficits. In an attempt to address this shortcoming, we provided this participant with our real-time feedback of his plantar flexor engagement.

Assessment of more-affected limb neuromuscular response to resisted plantar flexion was completed following the visit when the research team assessed appropriate acclimation and engagement with the device, as indicated by reaching the prescribed resistance setpoint on ≥ 50% of gait cycles and subjective feedback from participants that they could feel the resistance when walking. Participants’ lower extremities were bilaterally outfitted with surface electromyography (EMG) sensors (Noraxon, Scottsdale, AZ; 1500 Hz) on the soleus (SOL) and tibialis anterior (TA). To clinically assess the intervention, we isolated the effects of adaptive resistance by having participants walk with the exoskeleton on the treadmill at a self-selected speed under two conditions in the following order: 1) No Resistance: wearing the exoskeleton while it maintained zero torque at the ankle joint – motors actively compensated for system dynamics and friction so the joint rotated freely, and 2) Resistance: wearing the exoskeleton as it provided plantar flexor resistance via the proportional control scheme (Figure 1). Walking speed was not changed between conditions. Consistent with the acclimation visits, participants walked for a total of 20 minutes with two to three breaks depending on participant preference, and the last five minute block was used for data collection (i.e. the first 15 minutes served as warm-up and practice). EMG activity was recorded for 30 seconds during each condition within this last five minute block. Embedded foot sensors detected gait events.

D. Data processing & statistical analysis

All data were analyzed in MATLAB (v2019a, The MathWorks, Inc., Natick, MA). The distribution of all variables was found to be normal using a Kolmogrov-Smirnov test with small-sample Lilliefors correction18. Significance level for this feasibility study was set at α = 0.05 without correction for multiple comparisons.

Goal 1: Validation of the adaptive resistance control scheme

To validate the adaptive resistance control scheme, we quantified the Pearson’s product moment correlation coefficient (R) over ten gait cycles between the biological ankle moment (normalized by peak moment within a trial) and exoskeleton torque (normalized by the prescribed level of assistance). With coefficients of 0.7 – 0.9 indicating a high correlation 30, our goal was to fall within this range, which would indicate that the controller was responding directly to user input and ankle demand. We also quantified the root mean square error (RMSE) between the normalized peak biological ankle moment and normalized peak exoskeleton resistance over the ten gait cycles.

Goal 2: Feasibility of plantar flexor resistance and assessment of neuromuscular response

EMG signals were de-meaned, band pass filtered (4th order butterworth, 15 – 380 Hz band-pass cutoff), rectified, and low-pass filtered (4th order butterworth, 7 Hz low-pass cutoff), and then down-sampled to align with the gait events recorded by the force sensitive resistors. Muscle activity over the 30 second trial was then segmented by gait events. We computed the average EMG amplitude of each muscle during stance phase. Average amplitudes were then normalized to the average peak amplitude from the No Resistance trials. Additionally, co-contraction during push-off phase was calculated as the normalized ratio of TA activity (antagonist) to SOL activity (agonist) for each stance phase cycle. Outcomes were grouped by more-affected and less-affected limb. To assess the relationship between level of resistance and magnitude of neuromuscular response to resistance in the more-affected limb, we quantified the Pearson’s product moment correlation coefficient (R) between prescribed level of resistance and percent change in average muscle activation levels and co-contraction between No Resistance and Resistance conditions30. Paired t-tests were used to compare average muscle activation levels and co-contraction between No Resistance and Resistance conditions for both more-affected and less-affected limbs. Cohen’s d (d) was used to calculate effect size, where 0.2 was considered a small effect, 0.5 a medium effect, and 0.8 a large effect 21. Significance level for this feasibility study was set at α = 0.05 without correction for multiple comparisons.

Results

Goal 1: Validation of the adaptive resistance control scheme (Subjects V1 – V2)

There was a very high correlation between the normalized biological ankle moment and rectified normalized exoskeleton torque (R = 0.92 ± 0.04) for both Subjects V1 and V2 (see Figure 3). In addition, the RMSE between peak biological ankle moment and peak exoskeleton resistance torque over ten gait cycles was 6.3 ± 0.4%.

Figure 3. Validation of adaptive ankle resistance control:

Figure 3.

A) recorded ankle moments and resistance torque values from Subjects V1 and V2, with very high correlation (R = 0.92 ± 0.04) between biological ankle moment (blue) and magnitude of torque resistance by the exoskeleton (red). The resistance torque curve was rectified (i.e. plotted in the same direction as the biological moment) to assist with visualizing proportionality; B) Normalized biological ankle moment plotted against normalized exoskeleton resistance torque for Subjects V1 (closed circles) and V2 (open circles) over ten consecutive gait cycles.

Goal 2: Feasibility of plantar flexor resistance and assessment of neuromuscular response (Subjects 1 – 6)

All participants (Subjects 1 – 6) tolerated increased level of resistance beyond the initial setpoint of 0.075 Nm/kg. Following acclimation, the average tolerated resistance was 0.16 ± 0.02 Nm/kg, which ranged from 0.13 – 0.20 Nm/kg (Table 1). Self-identified soreness levels ranged from none – moderate throughout the visits. In the more-affected limb, there was a moderate correlation between prescribed level of resistance and increase in SOL activity (R = 0.69), and high correlations between prescribed level of resistance and decrease in TA activity and co-contraction (R = −0.84 and −0.73, respectively). We observed changes in more-affected limb muscle activity and co-contraction (Figure 4AC). For the more-affected limb, SOL activity increased 45 ± 35% (p = 0.02, d = 1.33), TA activity decreased 26 ± 24% (p < 0.05, d = −1.07), and co-contraction level decreased 46 ± 25% (p = 0.02, d = −1.42). A representative SOL activity curve (Subject 1) can be seen in Figure 5. Individual EMG plots of more-affected limb stance-phase SOL and TA activity are reported in supplementary material. No significant difference was seen in the less-affected limb SOL (p = 0.24, d = 0.55) or TA (p = 0.38, d = 0.39) activity, or less-affected limb co-contraction level (p = 0.55, d = 0.26).

Figure 4. Neuromuscular response:

Figure 4.

more-affected limb average stance phase soleus (A) and tibialis anterior (B) muscle activity, and co-contraction level (C) for No Resistance and Resistance conditions for Subjects 1 – 6 while walking at a self-selected speed on the treadmill. Muscle activity was normalized to the peak stance phase activity of the No Resistance condition and averaged over the 30 second walking trial. Co-contraction was calculated as the ratio of averaged, normalized tibialis anterior to soleus muscle activity for each stance phase of gait. *p < 0.05

Figure 5. Representative neuromuscular response during resisted gait:

Figure 5.

Subject 1 displayed increased (25%) average soleus activity from No Resistance to Resistance condition during stance phase.

6. Discussion

We successfully completed our first goal of validating that the adaptive resistance controller applied resistance proportional to the biological ankle moment during walking. For the two participants who completed the validation protocol of this study, there was a very high correlation between biological ankle moment and the magnitude of exoskeleton resistance (R = 0.92 ± 0.04, Figure 3) and low RMSE between peak biological ankle moment and peak resistance torque (6.3 ± 0.4%). This result agrees with our previous study on proportional control for ankle plantar flexor assistance 17.

With regards to Goal 2, the findings from this study indicate that gait training with plantar flexor resistance delivered via an untethered exoskeleton is feasible for individuals with spastic CP, GMFCS levels I and II. All individuals tolerated walking with resistance, and were able to maintain their baseline self-selected walking speed when resistance was applied. For the six participants who underwent the neuromuscular response protocol, a similar level of resistance (0.16 ± 0.02 Nm/kg) was attained after four to five sessions of 20-minute resisted walking trials. Previous studies have found that children with CP have peak biological ankle moments of around 1 Nm/kg while walking 9,25, meaning the average tolerated resistance after acclimation represented approximately 16% of the biological plantar flexor ankle moment. Participants’ perceived soreness level after each training session ranged from none – moderate. Importantly, no subjects selected severe or extremely severe after training. Four to five acclimation visits seemed to be appropriate for the anticipated “learning” that is required with this modality, as it allowed all six participants to reach a resistance level and degree of engagement that resulted in moderate soreness. While it is plausible that further acclimation visits would result in an even greater neuromuscular response, this must be balanced with the expected training effect of this intervention.

The results of this study support our hypothesis that proportional ankle resistance during gait in individuals with CP would increase plantar flexor muscle activity. Specifically, soleus activity significantly increased only for the more-affected limb. On the less-affected side, soleus activity increased for three of the six subjects. This may be explained by the direction to focus on the more-affected side. Additional research is needed to assess how the user’s focus on an individual limb affects their neuromuscular response to the intervention. An important consideration from these results is the level of resistance and magnitude of response. As an initial, feasibility study, level of soreness was closely monitored and resistance levels were increased conservatively. It is reasonable to assume that participants could train at a higher level of resistance than prescribed in this study, and this would likely be associated with an even greater neuromuscular response. This assumption is supported by the finding that that there was a moderate correlation between prescribed level of resistance and increase in soleus activity (R = 0.69), and high correlations between prescribed level of resistance and decrease in tibialis anterior activity and co-contraction (R = −0.84 and −0.73, respectively). These moderate to high correlations between resistance level and neuromuscular response also support the unique effect of adaptive resistance compared to verbal coaching alone.

Similar to the findings here, a previous study 20 found that a downward pelvic pull was also able to improve activation of the soleus, as well as coordination of the gastrocnemius muscle in children with CP. This improvement in neuromuscular activation led to an improved walking pattern, as the gastrocnemius could more effectively extend the knee during mid-stance and propel the center of mass forward at toe off. With the finding that our exoskeleton resistance protocol can also increase plantar flexor activity during stance phase, training with our adaptive ankle resistance device may lead to improved walking kinematics similar to what was observed from the downward pelvic pull system. It is important to note, however, that our study did not look directly at gastrocnemius activation profiles as this study did20, so this potential effect on kinematics should be interpreted with caution. In addition, we had the novel finding of decreased antagonist activity, as the previous study did not measure the response of the tibialis anterior.

A pleasantly surprising outcome of this study was the finding that stance-phase TA activity of the more-affected limb was significantly decreased during walking with resistance. Combined with the significant increase in SOL activity of the more-affected limb, this resulted in a substantial decrease in co-contraction during stance phase. While we anticipated that SOL activity would increase relative to TA activity resulting in a decrease in co-contraction, we did not anticipate that there would be a significant decrease in TA activity. On the less-affected side, changes in TA activity and co-contraction level was less consistent across subjects. Our significant finding of decreased co-contraction of the more-affected limb may be the most compelling feature of this proposed intervention. Children with spastic CP can present with increased co-contraction between the ankle plantar flexors and dorsiflexors 22, which likely plays a role in the higher metabolic cost of walking seen for this population 42. The mechanism behind this reduction in co-contraction, characterized by both an increase in agonist activity and decrease in antagonist activity, is not immediately clear. One possible explanation is that resisted plantarflexion while walking results in increased or improved reciprocal inhibition 23. With contraction of the soleus muscle and subsequent plantarflexion, there is stretch of the tibialis anterior. Stretch of muscle spindles within the tibialis anterior will lead to excitation of motoneurons and antagonist contraction. With typical, voluntary movements, however, the excitation of the stretched antagonist muscle is blocked by reciprocal inhibition via Ia inhibitory interneurons. It has also been shown that there is supraspinal involvement within reciprocal inhibition, with antagonist muscles decreasing activation right before an agonist muscle contracts. Reciprocal inhibition is impaired in CP, and this impairment is a common explanation for the co-contraction seen in this population. It is possible, given decreased TA activity despite increased SOL activity, that our protocol had an effect on reciprocal inhibition.

The key principle of adaptive ankle resistance with a wearable robotic platform compared to previous gait training interventions for individuals with CP is the ability to address both task-specificity and active engagement. One potential benefit of adaptive ankle resistance compared to other task-specific training protocols is the resulting increase in neuromuscular engagement while walking, which can help to drive motor learning. The ability to monitor a user’s instantaneous biological ankle moment in conjunction with the resistance torque applied via the wearable platform offers a unique opportunity to quantify and provide feedback of the user’s engagement in real-time while walking.

The successful validation of this wearable functional resistance exoskeleton system provides a unique opportunity to design future training interventions that match closely with the ASAP (Accelerated Skill Acquisition Program) framework45. The ASAP framework has eight principles for an evidence-based intervention for learning or relearning of motor skills to optimally affect neural plasticity: 1) Challenging and meaningful practice: With a controllable level of resistance, we are able to prescribe an appropriate level of difficulty; 2) Addresses important changeable impairments: plantar flexion weakness is a clear contributor to gait dysfunction in CP; 3) Overload and specificity: individuals are able to walk with resistance for repetitive muscle activation, and this practice occurs during the functional task of walking with no artificial breakdown of the movement; 4) Preserves natural task: we are able to provide resistance during walking; 5) Avoids artificial task breakdown: despite the focus on plantar flexion, participants still incorporate the whole task of walking; 6) Ensures active engagement: walking with resistance necessitates active engagement with the exoskeleton; 7) Balances immediate and future needs: the untethered aspect of the device allows for the potential of training in more ecologically valid contexts, such as at school or the playground; 8) Task-specific self-confidence: the ability to “level up” on resistance can serve as a motivator or confidence boost for users. 45

We are confident in the experimental design and results of this study, but there are notable limitations. First, as a pilot study focused on the potential to improve neuromuscular activation of the more-affected limb, we asked participants to only focus on one side while walking and did not assess gait kinematics or kinetics for Goal 2. It is possible that the same results could have been achieved without this direction, in addition to similar improvements on the less-affected side. It will also be critical for future investigations to assess the effects of this intervention on gait kinematics and kinetics. Second, it is possible that participants could increase plantar flexor activity simply with the instruction to “push” during the propulsion phase. However, this does not seem likely, as our data showed moderate to high positive correlations between neuromuscular response and resistance level, indicating the need for a resistive, afferent signal to recruit additional motor units. Additionally, resistance is necessary for modifying training intensity, which is an important factor in stimulating muscular adaptations 43 and promoting neuroplasticity 33. Third, future research should investigate the effects of plantar flexor resistance on gastrocnemius activity, which was not feasible in this study due to the location of the exoskeleton calf cuff. Finally, we had a limited sample size and our participants were all GMFCS I – II, meaning the proposed training has only been validated for ambulatory individuals with spastic CP. While this does not allow us to generalize our results to all individuals with CP, the findings of this pilot and feasibility study serve as an important first step for the development of this potentially transformative intervention. It is also important to note that this study serves as an initial investigation into the feasibility and neuromuscular response of wearable adaptive resistance. Any effects of this intervention, and changes in the neuromuscular response once training has finished, are a topic for future studies.

In conclusion, this study has shown the validity and feasibility of a novel training system and protocol for increasing plantar flexor activity within the functional context of walking for individuals with spastic CP. Proportional resistance of plantar flexion during the propulsive phase of gait leads to increased agonist muscle activity and decreased antagonist muscle activity of the more-affected limb in individuals with spastic CP, with the combined effect of reducing co-contraction. Future studies should explore the use of this wearable adaptive resistance platform in a longitudinal training intervention to improve neuromuscular function and gait in individuals with CP.

Supplementary Material

10439_2020_2454_MOESM1_ESM

Acknowledgements

The authors would like to thank Nushka Remec, P.T., Emily Frank, R.N., Elizabeth Orum, and Jennifer Lawson for their assistance with data collection and processing.

Research reported in this publication was supported in part by the Eunice Kennedy Shriver National Institute Of Child Health & Human Development of the National Institutes of Health under Award Number R03HD094583. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This work was also supported in part by The University of Arizona College of Medicine - Phoenix MD/PhD Program.

Abbreviations:

CP

cerebral palsy

PT

physical therapy

GUI

graphical user interface

GMFCS

Gross Motor Function Classification System

MVC

maximum voluntary contraction

EMG

electromyography

SOL

soleus

TA

tibialis anterior

R

correlation coefficient

d

Cohen’s d

Footnotes

Conflicts of Interest

ZFL is a named inventor on a pending utility patent application that describes the exoskeleton utilized in the study. ZFL is a co-founder of a company seeking to commercialize the device.

References

  • 1.Anttila H, Autti-Rämö I, Suoranta J, Mäkelä M, and Malmivaara A. Effectiveness of physical therapy interventions for children with cerebral palsy: A systematic review. BMC Pediatr. 8:14, 2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Bastiaanse CM, Duysens J, and Dietz V. Modulation of cutaneous reflexes by load receptor input during human walking. Exp. Brain Res. 135:189–198, 2000. [DOI] [PubMed] [Google Scholar]
  • 3.Bayón C, Lerma S, Ramírez O, Serrano JI, Del Castillo MD, Raya R, Belda-Lois JM, Martínez I, and Rocon E. Locomotor training through a novel robotic platform for gait rehabilitation in pediatric population: short report. J. Neuroeng. Rehabil. 13:98, 2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Borggraefe I, Schaefer JS, Klaiber M, Dabrowski E, Ammann-Reiffer C, Knecht B, Berweck S, Heinen F, and Meyer-Heim A. Robotic-assisted treadmill therapy improves walking and standing performance in children and adolescents with cerebral palsy. Eur. J. Paediatr. Neurol. 14:496–502, 2010. [DOI] [PubMed] [Google Scholar]
  • 5.Brown JK, Rodda J, Walsh EG, and Wright GW. Neurophysiology of lower‐limb function in hemiplegic children. Dev. Med. Child Neurol. 33:1037–1047, 2008. [DOI] [PubMed] [Google Scholar]
  • 6.Burdea GC, Cioi D, Kale A, Janes WE, Ross SA, and Engsberg JR. Robotics and gaming to improve ankle strength, motor control, and function in children with cerebral palsy - A case study series. IEEE Trans. Neural Syst. Rehabil. Eng. 21:165–173, 2013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Clark BC, Pierce JR, Manini TM, and Ploutz-Snyder LL. Effect of prolonged unweighting of human skeletal muscle on neuromotor force control. Eur. J. Appl. Physiol. 100:53–62, 2007. [DOI] [PubMed] [Google Scholar]
  • 8.Colborne GR, Wright FV, and Naumann S. Feedback of triceps surae EMG in gait of children with cerebral palsy: A controlled study. Arch. Phys. Med. Rehabil. 75:40–45, 1994. [PubMed] [Google Scholar]
  • 9.Dallmeijer AJ, Baker R, Dodd KJ, and Taylor NF. Association between isometric muscle strength and gait joint kinetics in adolescents and young adults with cerebral palsy. Gait Posture 33:326–332, 2011. [DOI] [PubMed] [Google Scholar]
  • 10.Davids JR, Oeffinger DJ, Bagley AM, Sison-Williamson M, and Gorton G. Relationship of Strength, Weight, Age, and Function in Ambulatory Children With Cerebral Palsy. J. Pediatr. Orthop. 35:523–529, 2015. [DOI] [PubMed] [Google Scholar]
  • 11.Delp SL, Anderson FC, Arnold AS, Loan P, Habib A, John CT, Guendelman E, and Thelen DG. OpenSim: Open-Source Software to Create and Analyze Dynamic Simulations of Movement. IEEE Trans. Biomed. Eng. 54:1940–1950, 2007. [DOI] [PubMed] [Google Scholar]
  • 12.Drużbicki M, Rusek W, Snela S, Dudek J, Szczepanik M, Zak E, Durmala J, Czernuszenko A, Bonikowski M, and Sobota G. Functional effects of robotic-assisted locomotor treadmill thearapy in children with cerebral palsy. J. Rehabil. Med. 45:358–363, 2013. [DOI] [PubMed] [Google Scholar]
  • 13.Dursun E, Dursun N, and Alican D. Effects of biofeedback treatment on gait in children with cerebral palsy. Disabil. Rehabil. 26:116–120, 2004. [DOI] [PubMed] [Google Scholar]
  • 14.Eek MN, and Beckung E. Walking ability is related to muscle strength in children with cerebral palsy. Gait Posture 28:366–371, 2008. [DOI] [PubMed] [Google Scholar]
  • 15.Gage JR, and Novacheck TF. An update on the treatment of gait problems in cerebral palsy. J. Pediatr. Orthop. B 10:265–74, 2001. [PubMed] [Google Scholar]
  • 16.Gage JR, Schwartz MH, Koop SE, and Novacheck TF. The identification and treatment of gait problems in cerebral palsy. London: Mac Keith Press, 2009. at <https://books.google.com/books?hl=en&lr=&id=PiiDMzb551sC&oi=fnd&pg=PR9&dq=gage+crouch+gait&ots=76vHwpHDa7&sig=naMsA5d6DLFRVvQPigbPZ3zN8U#v=onepage&q=gagecrouchgait&f=false> [Google Scholar]
  • 17.Gasparri GM, Luque J, and Lerner ZF. Proportional Joint-Moment Control for Instantaneously Adaptive Ankle Exoskeleton Assistance. IEEE Trans. Neural Syst. Rehabil. Eng. 27:751–759, 2019. [DOI] [PubMed] [Google Scholar]
  • 18.Ghasemi A, and Zahediasl S. Normality tests for statistical analysis: a guide for non-statisticians. Int. J. Endocrinol. Metab. 10:486–9, 2012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Jezernik S, Colombo G, Keller T, Frueh H, and Morari M. Robotic Orthosis Lokomat: A Rehabilitation and Research Tool. Neuromodulation Technol. Neural Interface 6:108–115, 2003. [DOI] [PubMed] [Google Scholar]
  • 20.Kang J, Martelli D, Vashista V, Martinez-Hernandez I, Kim H, and Agrawal SK. Robot-driven downward pelvic pull to improve crouch gait in children with cerebral palsy. Sci. Robot. 2:eaan2634, 2017. [DOI] [PubMed] [Google Scholar]
  • 21.Lachenbruch PA, and Cohen J. Statistical Power Analysis for the Behavioral Sciences (2nd ed.). J. Am. Stat. Assoc., 1989.doi: 10.2307/2290095 [DOI] [Google Scholar]
  • 22.Leonard CT, Hirschfeld H, and Forssberg H. The development of independent walking in children with cerebral palsy. Dev. Med. Child Neurol. 33:567–577, 2008. [DOI] [PubMed] [Google Scholar]
  • 23.Leonard CT, Moritani T, Hirschfeld H, and Forssberg H. Deficits in reciprocal inhibition of children with cerebral palsy as revealed by H reflex testing. Dev. Med. Child Neurol. 32:974–984, 2008. [DOI] [PubMed] [Google Scholar]
  • 24.Lerner ZF, DeMers MS, Delp SL, and Browning RC. How tibiofemoral alignment and contact locations affect predictions of medial and lateral tibiofemoral contact forces. J. Biomech. 48:644–650, 2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Lerner ZF, Harvey TA, and Lawson JL. A Battery-Powered Ankle Exoskeleton Improves Gait Mechanics in a Feasibility Study of Individuals with Cerebral Palsy. Ann. Biomed. Eng., 2019.doi: 10.1007/s10439-019-02237-w [DOI] [PubMed] [Google Scholar]
  • 26.Lotze M, Braun C, Birbaumer N, Anders S, and Cohen LG. Motor learning elicited by voluntary drive. Brain 126:866–872, 2003. [DOI] [PubMed] [Google Scholar]
  • 27.McGowan CP, Neptune RR, Clark DJ, and Kautz SA. Modular control of human walking: Adaptations to altered mechanical demands. J. Biomech. 43:412–419, 2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.de Mello Monteiro CB, Massetti T, da Silva TD, van der Kamp J, de Abreu LC, Leone C, and Savelsbergh GJP. Transfer of motor learning from virtual to natural environments in individuals with cerebral palsy. Res. Dev. Disabil. 35:2430–2437, 2014. [DOI] [PubMed] [Google Scholar]
  • 29.Meyer-Heim A, Ammann-Reiffer C, Schmartz A, Schafer J, Sennhauser FH, Heinen F, Knecht B, Dabrowski E, and Borggraefe I. Improvement of walking abilities after robotic-assisted locomotion training in children with cerebral palsy. Arch. Dis. Child. 94:615–620, 2009. [DOI] [PubMed] [Google Scholar]
  • 30.Mukaka MM Statistics corner: A guide to appropriate use of correlation coefficient in medical research. Malawi Med. J. 24:69–71, 2012. [PMC free article] [PubMed] [Google Scholar]
  • 31.Novak I, Mcintyre S, Morgan C, Campbell L, Dark L, Morton N, Stumbles E, Wilson S-A, and Goldsmith S. A systematic review of interventions for children with cerebral palsy: state of the evidence. Dev. Med. Child Neurol. 55:885–910, 2013. [DOI] [PubMed] [Google Scholar]
  • 32.Olney SJ, MacPhail HA, Hedden DM, and Boyce WF. Work and Power in Hemiplegic Cerebral Palsy Gait. Phys. Ther. 70:431–438, 1990. [DOI] [PubMed] [Google Scholar]
  • 33.Reid LB, Rose SE, and Boyd RN. Rehabilitation and neuroplasticity in children with unilateral cerebral palsy. Nat. Rev. Neurol. 11:390–400, 2015. [DOI] [PubMed] [Google Scholar]
  • 34.Rose J, Gamble JG, Burgos A, Medeiros J, and Haskell WL. Energy expenditure index of walking for normal children and for children with cerebral palsy. Dev. Med. Child Neurol. 32:333–340, 1990. [DOI] [PubMed] [Google Scholar]
  • 35.Rossignol S, Dubuc R, and Gossard J-P. Dynamic Sensorimotor Interactions in Locomotion. Physiol. Rev. 86:89–154, 2006. [DOI] [PubMed] [Google Scholar]
  • 36.Schmidt RA, and Lee TD. Motor Control and Learning: A Behavioral Emphasis. In: Human Kinetics; Champaign, IL: 2011. [Google Scholar]
  • 37.Scholtes VA, Becher JG, Comuth A, Dekkers H, Van Dijk L, and Dallmeijer AJ. Effectiveness of functional progressive resistance exercise strength training on muscle strength and mobility in children with cerebral palsy: a randomized controlled trial. Dev. Med. Child Neurol. 52:e107–e113, 2010. [DOI] [PubMed] [Google Scholar]
  • 38.Scholtes VA, Becher JG, Janssen-Potten YJ, Dekkers H, Smallenbroek L, and Dallmeijer AJ. Effectiveness of functional progressive resistance exercise training on walking ability in children with cerebral palsy: A randomized controlled trial. Res. Dev. Disabil. 33:181–188, 2012. [DOI] [PubMed] [Google Scholar]
  • 39.Schroeder AS, Homburg M, Warken B, Auffermann H, Koerte I, Berweck S, Jahn K, Heinen F, and Borggraefe I. Prospective controlled cohort study to evaluate changes of function, activity and participation in patients with bilateral spastic cerebral palsy after Robot-enhanced repetitive treadmill therapy. Eur. J. Paediatr. Neurol. 18:502–510, 2014. [DOI] [PubMed] [Google Scholar]
  • 40.Sutherland DH, Cooper L, and Daniel D. The role of the ankle plantar flexors in normal walking. J. Bone Joint Surg. Am. 62:354–63, 1980. [PubMed] [Google Scholar]
  • 41.Taylor NF, Dodd KJ, Baker RJ, Willoughby K, Thomason P, and Graham HK. Progressive resistance training and mobility-related function in young people with cerebral palsy: a randomized controlled trial. Dev. Med. Child Neurol. 55:806–812, 2013. [DOI] [PubMed] [Google Scholar]
  • 42.Unnithan VB, Dowling JJ, Frost G, and Bar-Or O. Role of cocontraction in the O2 cost of walking in children with cerebral palsy. Med. Sci. Sports Exerc. 28:1498–504, 1996. [DOI] [PubMed] [Google Scholar]
  • 43.Verschuren O, Ada L, Maltais DB, Gorter JW, Scianni A, and Ketelaar M. Muscle Strengthening in Children and Adolescents With Spastic Cerebral Palsy: Considerations for Future Resistance Training Protocols. Phys. Ther. 91:1130–1139, 2011. [DOI] [PubMed] [Google Scholar]
  • 44.Wiley ME, and Damiano DL. Lower-extremity strength profiles in spastic cerebral palsy. Dev. Med. Child Neurol. 40:100–7, 1998. [DOI] [PubMed] [Google Scholar]
  • 45.Winstein C, Lewthwaite R, Blanton SR, Wolf LB, and Wishart L. Infusing motor learning research into neurorehabilitation practice: a historical perspective with case exemplar from the accelerated skill acquisition program. J. Neurol. Phys. Ther. 38:190–200, 2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Winters TF, Gage JR, and Hicks R. Gait patterns in spastic hemiplegia in children and young adults. J. Bone Joint Surg. Am. 69:437–41, 1987. [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

10439_2020_2454_MOESM1_ESM

RESOURCES