AI gait predictions varied by joint and scoring method
A review found different leaders depending on how prediction quality was measured, with ankle estimates performing better than knee or hip estimates.
Based on the published abstract. The full paper may contain additional methods, results and limitations.
The 30-second takeaway
The review suggests AI can estimate lower-limb joint moments during typical walking, but no model family led on every measure. Deep neural networks performed best on the measure of model fit, while traditional machine learning led on normalized prediction error. These are prediction results, so they do not demonstrate improvements in walking or rehabilitation.
The abstract gives broad uncertainty intervals and no study follow-up durations. Its sample concerned typically developed gait, leaving performance in people with gait disorders unresolved.
What researchers found
Joint-moment prediction quality
Compared with knee and hip predictions, ankle estimates performed better on R-squared and normalized root mean square error; assessment timing was not reported.
What was being predicted
The models predicted a biomechanical measure called joint moment during walking. Researchers compared the algorithms, the signals supplied to them, and the joints they were asked to predict, including ankles, knees, and hips.
Why the scoring method matters
The reported leader changed between model fit and normalized prediction error. That makes a single overall winner difficult to name, and the authors call for further testing in people whose gait is affected by disease.
The original publication
Artificial intelligence in lower limb joint moment prediction during typically developed gait: A systematic review and multilevel random-effects meta-analysis.
Pan J, Gao Z, Zhou Z et al.
Gait Posture · 2026
- PubMed ID
- 41412059
- Record checked
AI-assisted research and writing. This explains one selected publication; it is not a complete review of everything known. Our approach.
One more question, understood.
Keep track of the research you’ve explored.