Rapid energy expenditure estimation for assisted and inclined loaded walking

Background Estimating energy expenditure with indirect calorimetry requires expensive equipment and provides slow and noisy measurements. Rapid estimates using wearable sensors would enable techniques like optimizing assistive devices outside a lab. Existing methods correlate data from wearable sensors to measured energy expenditure without evaluating the accuracy of the estimated energy expenditure for activity conditions or subjects not included in the correlation process. Our goal is to assess data-driven models that are capable of rapidly estimating energy expenditure for new conditions and subjects. Methods We developed models that estimated energy expenditure from two datasets during walking conditions with (1) ankle exoskeleton assistance and (2) various loads and inclines. The estimation was portable and rapid, using input features that are possible to measure with wearable sensors and restricting the input data length to a single gait cycle or four second interval. The performance of the models was evaluated for three use cases. The first case estimated energy expenditure during walking conditions for subjects with some subject specific training data available. The second case estimated all conditions in the dataset for a new subject not included in the training data. The third case estimated new conditions for a new subject. The models also ordered the magnitude of energy expenditure across all conditions for a new subject. Results The average errors in energy expenditure estimation during assisted walking conditions were 4.4%, 8.0%, and 8.1% for the three use cases, respectively. The average errors in energy expenditure estimation during inclined and loaded walking conditions were 6.1%, 9.7%, and 11.7% for the three use cases. The models ordered the magnitude of energy expenditure with a maximum and average percentage of correctly ordered conditions of 56% and 43% for assisted walking and 85% and 55% for incline and loaded walking. Conclusions Our data-driven models determined the accuracy of energy expenditure estimation for three use cases. For experiments where the accuracy of a data-driven model is sufficient, standard indirect calorimetry can be replaced. The energy expenditure ordering could aid in selecting optimal assistance conditions. The models, code, and datasets are provided for reproduction and extension of our results.