Rapid energy expenditure estimation for ankle assisted and inclined loaded walking.
Patrick Slade, Rachel Troutman, Mykel J Kochenderfer and 2 others
PMID 31171003WHAT IT FOUND
Lab models using lower-limb muscle and foot-force signals estimated walking energy cost more closely for conditions similar to the training data than for new subjects, and ranked effort better across loads and inclines than across exoskeleton assistance settings.
They are not yet clinical tools.
Key findings
01For assisted walking, neural network estimates had 4.4% error for a new condition, and approximately 8% error when the subject was new or both subject and condition were new.
02For inclined loaded walking, neural network errors were 6.1% for a new condition, 9.7% for a new subject, and 11.7% for a new subject and new condition.
03The neural network models correctly ordered energy expenditure across conditions 87% of the time for inclined loaded walking and 61% for assisted walking.
STILL TO COME
How it was doneWhat they foundWhat it means for PTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The models were tested only on small lab datasets, not on patients or in clinical settings. Two of eight subjects in the assisted walking dataset were excluded because their metabolic rates were far from the group mean. The sensors used were immobile lab systems; the authors say wearable sensors could add noise and would need separate validation. The models did not estimate absolute energy expenditure as accurately as indirect calorimetry. Extreme conditions, such as exoskeleton settings with the most work, were harder to estimate, with roughly three times the error in one test. The validation set was included in cross-validation because datasets were small, which can make performance look better.
Declared interests
The work was funded by the National Science Foundation, the Department of Mechanical Engineering at Stanford University, AI Grant, the National Institutes of Health, and the National Center for Simulation in Rehabilitation Research. The supplied text does not include an author conflict-of-interest statement.
The easy way to misread this
Do not use these models as a clinical energy-expenditure monitor for patients. They were validated on small lab datasets with immobile sensors, and the authors state they do not estimate absolute energy expenditure as accurately as indirect calorimetry.