Uncertainty Modeling and Propagation in Musculoskeletal Modeling

Abstract

Biomechanical input data of in silico models are subject to uncertainties due to subject variability, experimental protocol and computing technique. Traditional perturbation analysis showed important drawbacks such as unobvious definition of the true range of value for the sensitivity analysis. In this present study, we used a novel framework to model the uncertainties of thigh mass property as well as to quantify their impact on the thigh muscle force estimation. A simplified patient specific musculoskeletal model (3 segments, 2 joints and 8 hip flexor muscles) of a post-polio residual paralysis subject was developed. Knowledge-based fusion pbox was used to model the uncertainties of the thigh mass property. Then, a Monte Carlo simulation was performed to quantify their impact on the thigh muscle force estimation through forward dynamics simulation. The global range of value of the rectus femoris force is from 2327.59 ± 39.32 N to 3353.16 ± 383.8 N at the peak level. The global range of value of the gracilis force is from 143.53 ± 2.35 N to 159.27 ± 8 N at the peak level. Cumulative probability functions of these ranges were presented and discussed. Our study suggested that under input data uncertainties, the musculoskeletal simulation results needs to be determined within a global range of values. Consequently, the clinical use of such global range will make the decision making more reliable. Thus, our study could be used as a guideline for such a purpose.