RESUMEN
PURPOSE: Individuals with walking impairment, such as those with cerebral palsy, often face challenges in leading physically active lives due to the high energy cost of movement. Assistive devices like powered exoskeletons aim to alleviate this burden and improve mobility. Traditionally, optimizing the effectiveness of such devices has relied on time-consuming laboratory-based measurements of energy expenditure, which may not be feasible for some patient populations. To address this, our study aimed to enhance the state-of-the-art predictive model for estimating steady-state metabolic rate from 2-min walking trials to include individuals with and without walking disabilities and for a variety of terrains and wearable device conditions. METHODS: Using over 200 walking trials collected from eight prior exoskeleton-related studies, we trained a simple linear machine learning model to predict metabolic power at steady state based on condition-specific factors, such as whether the trial was conducted on a treadmill (level or incline) or outdoors, as well as demographic information, such as the participant's weight or presence of walking impairment, and 2 minutes of metabolic data. RESULTS: We demonstrated the ability to predict steady-state metabolic rate to within an accuracy of 4.71 ± 2.7% on average across all walking conditions and patient populations, including with assistive devices and on different terrains. CONCLUSION: This work seeks to unlock the use of in-the-loop optimization of wearable assistive devices in individuals with limited walking capacity. A freely available MATLAB application allows other researchers to easily apply our model.
RESUMEN
The clinical efficacy of robotic rehabilitation interventions hinges on appropriate neuromuscular recruitment from the patient. The first purpose of this study was to evaluate the use of supervised machine learning techniques to predict neuromuscular recruitment of the ankle plantar flexors during walking with ankle exoskeleton resistance in individuals with cerebral palsy (CP). The second goal of this study was to utilize the predictive models of plantar flexor recruitment in the design of a personalized biofeedback framework intended to improve (i.e., increase) user engagement when walking with resistance. First, we developed and trained multilayer perceptrons (MLPs), a type of artificial neural network (ANN), utilizing features extracted exclusively from the exoskeleton's onboard sensors, and demonstrated 85-87% accuracy, on average, in predicting muscle recruitment from electromyography measurements. Next, our participants completed a gait training session while receiving audio-visual biofeedback of their personalized real-time planar flexor recruitment predictions from the online MLP. We found that adding biofeedback to resistance elevated plantar flexor recruitment by 24 16% compared to resistance alone. This study highlights the potential for online machine learning frameworks to improve the effectiveness and delivery of robotic rehabilitation systems in clinical populations.
RESUMEN
Background: Age-related deficits in plantar flexor muscle function during the push-off phase of walking likely contribute to the decline in mobility that affects many older adults. Isolated strengthening of the plantar flexor muscles has failed to improve push-off power or walking economy in this population. New mobility aids and/or functional training interventions may help slow or prevent ambulatory decline in the elderly. Objective: The overarching objective of this study was to explore the feasibility of using an untethered, dual-mode ankle exoskeleton for treating walking disability in the elderly; testing the device in assistance mode as a mobility aid to reduce energy consumption, and as a resistive gait training tool to facilitate functional recruitment of the plantar flexor muscles. Methods: We recruited 6 older adults between the ages of 68 to 83 years to evaluate the feasibility of the dual-mode exoskeleton across two visits. On the first visit, we quantified acute metabolic and neuromuscular adaption to ankle exoskeleton assistance during walking in older adults, and subsequently determined if higher baseline energy cost was related to an individual's potential to benefit from untethered assistance. On the second visit, we validated the potential for push-off phase ankle resistance combined with plantar pressure biofeedback to facilitate functional utilization of the ankle plantar flexors during walking. We also conducted a twelve-session ankle resistance training protocol with one pilot participant to explore the effects of gait training with wearable ankle resistance on mobility and plantar flexor strength. Results: Participants reached the lowest net metabolic power, soleus variance ratio, and soleus iEMG at 6.6 ± 1.6, 19.8 ± 1.6, and 5.8 ± 4.9 minutes, respectively, during the 30-minute exoskeleton assistance adaptation trial. Four of five participants exhibited a reduction (up to 19%) in metabolic power during walking with assistance relative to baseline, but there was no group-level change. Participants who had greater baseline metabolic power exhibited a greater reduction during walking with assistance. Walking with resistance increased stance-phase soleus iEMG by 18 - 186% and stance-phase average positive ankle power by 9 - 88% compared to baseline. Following ankle resistance gait training, the participant exhibited a 5% increase in self-selected walking speed, a 15% increase in fast walking speed, a 36% increase in 6-min-walk-test distance, and a 31% increase in plantar flexor strength compared to pre-intervention measurements. Conclusions: Our results suggest that dual-mode ankle exoskeletons appear highly applicable to treating plantar flexor dysfunction in the elderly, with assistance holding potential as a mobility aid and resistance holding potential as a functional gait training tool. We used an untethered design to maximize the relevance of this for informing the design of intervention studies that may take place at home and in the community to improve mobility and quality of life in older adults. Future studies with larger sample sizes are recommended to expand on the results of this feasibility investigation.