EMG-Informed Neuromusculoskeletal Simulations Increase the Accuracy of the Estimation of Knee Joint Contact Forces During Sub-optimal Level Walking.

Purpose: Personalized musculoskeletal models are crucial to get insights into the mechanisms underpinning neuromusculoskeletal disorders and have the potential to support clinicians in the daily management and evaluation of patients. However, their use is still limited due to the lack of validation studies, which hinders people's trust in these technologies. The current study aims to assess the predictive accuracy of two common approaches to estimate knee joint contact forces, when employing musculoskeletal models.

Methods: Subject-specific musculoskeletal models were developed for four elderly subjects, exploiting the freely accessible Knee Grand Challenge datasets, and used to perform biomechanical simulations of level walking to estimate knee joint contact forces. The classical static optimization and EMG-assisted approaches were implemented to resolve the muscle redundancy problem. Their estimates were compared, in terms of predictive accuracy, against the experimental recordings from an instrumented knee implant and against one another. Spatiotemporal differences were identified through Statistical Parametrical Mapping, to complement traditional similarity metrics (R², RMSE, 95th percentile, and the maximal error).

Results: Both methods allowed to estimate the experimental knee joint contact forces experienced during walking with a high level of accuracy (R² > 0.82, RMSE < 0.56 BW). The EMG-assisted approach further enabled to highlight subject-specific features that were not captured otherwise, such as a prolonged or anticipated muscle-co-contraction.

Conclusion: While the static optimization approach provides reasonable estimates for subjects exhibiting typical gait, the EMG-assisted approach should be preferred and employed when studying clinical populations or patients exhibiting abnormal walking patterns.