Opensim upper-extremity modelling: subject-specific scaling- and validation tools

Abstract

Muscle-driven simulations performed with a musculoskeletal model of the human upper-extremity need to include closed-loop kinematics to capture the limb's full mobility and model the actions of thoracoscapular muscles. On the OpenSim platform, a single upper-extremity model with closed-loop kinematics is available: the thoracoscapular shoulder model (TSM), which only includes muscle-elements crossing the shoulder-girdle, omitting upper-extremity musculature partially. OpenSim's native algorithms for subject-specific scaling of geometric- and muscle-length parameters function insufficiently for upper-extremity models, whilst inertial- and muscle-strength parameters are rarely scaled. Subject-specific models are impossible to validate in vivo, whilst indirect validation methods are limited. The goal of this study is to enable future OpenSim users to perform muscle-driven simulations with adequately scaled, subject-specific upper-extremity models. This study develops: (1) A generic model with the closed-loop kinematic structure of the TSM and all upper-extremity musculature; the thoracoscapular Delft shoulder and elbow model (TDSEM). (2) Subject-specific scaling tools for the TSM and TDSEM, performing geometric- and muscle-length parameter scaling based on optimization procedures, and inertial- and muscle-strength scaling based on a total muscle-volume estimation. (3) A method to estimate maximum isometric directional force at an end-effector with subject-specific models to indirectly validate the maximum force-generating capacity of these models. Geometric scaling accuracy is mainly limited by the accuracy of marker-data used. Both the TSM and TDSEM reached the desired RMS marker-error of ≤2cm and an average segmental-length error of ≤5%. Muscle-length parameter optimization fit is comparable to, or better than, studies employing similar algorithms. Scaled muscle-strength-, and inertial parameters are not validated, but the accuracy of the muscle-volume estimates they are derived from was known beforehand. The indirect validation method is not able to make estimations within ±10% of measured maximum forces. The method greatly overestimates measured values, regardless of the subject modelled or model used. Including upper-extremity musculature in the TDSEM results in improved muscle-driven analyses compared to the TSM. This can be improved further by improving the accuracy of muscle-elements attached to the radius, including ligaments in the model, and using a constraint to maintain glenohumeral stability. The presented scaling methods are recommended over OpenSim's native methods when scaling the TSM or TDSEM. The indirect validation method must be improved before it can be used to inform (in)validating conclusions about subject-specific models. For this, the effect of a constrained torso, task-specific practice, and inclusion of trained subjects on measurements must be evaluated and subject pose must be recorded. When combined with the stability constraint, estimation accuracy will likely improve greatly.