Are subject-specific musculoskeletal models robust to the uncertainties in parameter identification?

Area of Science:

• Biomechanics
• Musculoskeletal Modeling
• Computational Biology

Background:

• Subject-specific musculoskeletal modeling aids in studying musculoskeletal disorders by incorporating personalized anatomy.
• Uncertainties in parameter identification during model creation can affect prediction accuracy, but their impact is not fully understood.

Purpose of the Study:

• To analyze the sensitivity of subject-specific model predictions (joint angles, moments, muscle/joint contact forces) during walking to uncertainties in body landmark positions, maximum muscle tension, and musculotendon geometry.
• To evaluate the effect of parameter uncertainties on musculoskeletal model predictions.

Main Methods:

• An MRI-based musculoskeletal model of the lower limbs (7-segment, 10-DOF, 84 musculotendon units) was created.
• A Monte Carlo probabilistic analysis was performed with 500 perturbed models to assess sensitivity.
• Freely available software was used for model creation, integration with OpenSim, and probabilistic simulations.

Main Results:

• Input variable uncertainties had a moderate effect on model predictions.
• Muscle and joint contact forces exhibited maximum standard deviations of 0.3 times body weight and ranges of 2.1 times body weight.
• Output variables showed significant correlations with only a few input variables (up to 7 out of 312) across the gait cycle.

Conclusions:

• Subject-specific models demonstrated moderate sensitivity to parameter identification, suggesting they are not markedly sensitive.
• Researchers must be aware of model precision relative to the intended application, as force predictions can be affected by uncertainties.
• While low probability, force prediction uncertainty can be of the same order of magnitude as the predicted value.