Predicting reactive stepping in response to perturbations by using a classification approach.
Amber R Emmens, Edwin H F van Asseldonk, Vera Prinsen and 1 others
PMID 32616066WHAT IT FOUND
Computer models trained on body-motion and foot-pressure data from ten healthy adults could tell whether a person would step after a forward push, but predictions were weak until the moment just before a step began.
This is a lab method, not a patient test.
Key findings
01Neural networks were the best-performing tested classifiers for predicting whether a perturbation response would be a step.
02Conventional stability boundary, extrapolated center of mass, and center-of-mass time-to-boundary methods performed worse than the trained classifiers.
03Prediction performance improved as the observation window lengthened, and early predictions before the stepping leg began to unload were less accurate.
STILL TO COME
How it was doneWhat they found
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
Only ten healthy young adults with no neurological, musculoskeletal or other medical impairments were studied, so it does not show what would happen in patients with balance problems. The perturbations were delivered in a laboratory with a harness, taped feet, crossed arms, fixed gaze and a counting task, which may not match everyday balance responses. Only forward perturbations were analyzed, so backward perturbations and other perturbation types are not covered. Training, validation and test data came from the same subjects and the same experimental setting, so the classifiers may not generalize to new people or different perturbations. The classifiers were optimized for correct classification, not for early step detection, and early step detection was poor, so many stepping responses could be missed before step initiation. Foot marker clusters often failed, so body center-of-mass calculations were made without foot data, with the effect assumed negligible. Hyperparameters were not systematically optimized.
The easy way to misread this
Do not treat this as a clinical fall-prediction tool. It was developed in ten healthy young adults in a lab, and early predictions before the stepping leg began to unload were poor, so it does not show that patients can be assessed this way.