Predicting lower body joint moments and electromyography signals using ground reaction forces during walking and running: An artificial neural network approach.

The source data came from healthy adults walking or running, so it does not show how the model would work in patients with motion disorders. The model was checked by splitting the same source data into training, validation and testing groups, not by testing new patients or different real-world settings. EMG signals were not normalized to maximum voluntary contraction, so muscle activation levels cannot be compared across subjects. The original musculoskeletal model error was not reported, so the joint-moment targets may include unknown error. Resampling and normalizing each stride to 100 points may smooth real differences between people and steps. The paper reports model-fit accuracy, not patient outcomes, treatment effects, or clinical usefulness.

The authors declared no conflicts of interest. The article metadata lists NIH and U.S. Government non-PHS research support.

Do not read the high model accuracy as proof that ground reaction force can replace motion capture or EMG in patient care. The models were built only from healthy adults in controlled walking and running studies, and the paper says error rates may be higher in patients with motion disorders.