A neuromechanics-based powered ankle exoskeleton to assist walking post-stroke: a feasibility study

Abstract

Background: In persons post-stroke, diminished ankle joint function can contribute to inadequate gait propulsion. To target paretic ankle impairments, we developed a neuromechanics-based powered ankle exoskeleton. Specifically, this exoskeleton supplies plantarflexion assistance that is proportional to the user's paretic soleus electromyography (EMG) amplitude only during a phase of gait when the stance limb is subjected to an anteriorly directed ground reaction force (GRF). The purpose of this feasibility study was to examine the short-term effects of the powered ankle exoskeleton on the mechanics and energetics of gait.

Methods: Five subjects with stroke walked with a powered ankle exoskeleton on the paretic limb for three 5 minute sessions. We analyzed the peak paretic ankle plantarflexion moment, paretic ankle positive work, symmetry of GRF propulsion impulse, and net metabolic power.

Results: The exoskeleton increased the paretic plantarflexion moment by 16% during the powered walking trials relative to unassisted walking condition (p < .05). Despite this enhanced paretic ankle moment, there was no significant increase in paretic ankle positive work, or changes in any other mechanical variables with the powered assistance. The exoskeleton assistance appeared to reduce the net metabolic power gradually with each 5 minute repetition, though no statistical significance was found. In three of the subjects, the paretic soleus activation during the propulsion phase of stance was reduced during the powered assistance compared to unassisted walking (35% reduction in the integrated EMG amplitude during the third powered session).

Conclusions: This feasibility study demonstrated that the exoskeleton can enhance paretic ankle moment. Future studies with greater sample size and prolonged sessions are warranted to evaluate the effects of the powered ankle exoskeleton on overall gait outcomes in persons post-stroke.

Figures

Figure 1

Illustration of the proportional myoelectric propulsion (PMP) powered exoskeleton. The soleus electromyography (EMG) and anterior-posterior ground reaction force (GRF) from an instrumented treadmill were collected in real-time to control the magnitude and timing of exoskeleton actuation. The proportional myoelectric propulsion (PMP) controller supplies plantarflexion moment proportional to the soleus EMG activity only during a phase of gait when the stance limb is subjected to an anteriorly directed ground reaction force. The red highlighted region denotes the duration in which the exoskeleton is activated.

Figure 2

Paretic limb data (three steps) from a representative subject with and without the powered exoskeleton. During the powered walking (POWx3), the exoskeleton control signal was generated with magnitude proportional to the paretic soleus EMG (blue) only when the anterior-posterior GRF (black) was greater than 0 (region highlighted in red). Positive GRF denotes anterior (i.e., propulsive) force. During this phase, the exoskeleton supplied plantarflexion moment during late stance (red), contributing to the increased total ankle moment (gray) relative to the NoEXO condition. We note that there is a delay between the onset of the control signal and the onset of the exoskeleton moment (lag of approximately 83 ms).

Figure 3

Ankle joint mechanics (averaged over 5 subjects). Sagittal plane data (time-normalized to 101 data points across gait cycle) of paretic and non-paretic ankle mechanics (angle, moment, power) were analyzed from the last minute of each condition (NoEXO – black; UnPOW – red; POWx3 – blue). For clarity, data from POWx1 and POWx2 are not shown. The two vertical lines define the propulsion phase of stance (i.e., onset of propulsion and toe-off). During POW conditions, the exoskeleton generated plantarflexion moment during late stance (shown in dotted blue), and contributed to the increased total paretic moment (16% increase during POW relative to NoEXO, p < 0.05).

Figure 4

Paretic ankle positive work (averaged over 5 subjects). The paretic ankle joint positive work (J kg−1) across conditions of NoEXO (black), UnPOW (red), and three repetitions of POW (blue) were analyzed from the last minute of each condition. The exoskeleton’s contributions to the total positive work during POW are denoted in white. There was no statistically significant effect of the exoskeleton on the paretic ankle joint positive work (p =0.58). The error bars represent ± 1.0 standard deviation.

Figure 5

Anterior-posterior ground reaction force and percent paretic propulsion (averaged over 5 subjects). Anterior-posterior GRF data (time-normalized to 101 data points across gait cycle) of paretic and non-paretic lower extremities were analyzed from the last minute of each condition (NoEXO – black; UnPOW – red; POWx3 – blue). For clarity, time-series data from POWx1 and POWx2 are not shown. The two vertical lines define the propulsion phase of stance (i.e., onset of propulsion and toe-off). The percent paretic propulsion (described by Bowden et al. [44]) signifies the symmetry of the propulsion impulse (less than 50% indicates greater reliance on the non-paretic limb for propulsion). There was no statistically significant effect of the exoskeleton on the percent paretic propulsion (p = 0.81).

Figure 6

Whole-body net metabolic power (averaged over 5 subjects). Whole-body net metabolic power (W kg-1) across conditions of NoEXO (black), UnPOW (red), and three repetitions of POW (blue) were analyzed. Although there was no statistically significant effect of the exoskeleton on net metabolic power (p = 0.21), there was a tendency for a gradual reduction of metabolic cost with each bout of the powered walking conditions. The percent change values are expressed relative to the NoEXO condition. The error bars represent ± 1.0 standard deviation.

Figure 7

Linear-enveloped EMG and magnitude of time-integrated EMG during the propulsion phase (averaged over 3 subjects). EMG signals of SOL and TA (from paretic and non-paretic limbs) were analyzed during the last minute of each condition. Linear-enveloped EMG data (time-normalized to 101 data points across gait cycle) from NoEXO (black), UnPOW (red), and POWx3 (blue) are shown (but linear-enveloped EMG data for POWx1 and POWx2 are not for clarity). The two vertical lines define the propulsion phase of stance (i.e., onset of propulsion and toe-off). The magnitude of integrated EMG (iEMG) during the propulsion phase showed reduced paretic SOL activity (i.e., muscle that controlled the exoskeleton assistance) during all three POW conditions relative to NoEXO. The error bars represent ± 1.0 standard deviation. We note that two subjects’ EMG data were omitted due to technical difficulties, and thus we did not perform statistical analysis on EMG data because of the small sample size.