Uncertainty analysis and sensitivity of scapulothoracic joint angles to kinematic model parameters

The purposes of this study were to determine the influence of kinematic model parameter variability on scapulothoracic angle estimates, and to define which parameters of the kinematic model have the largest effect on scapulothoracic angle estimates. Nominal subject-specific kinematic models of nine participants were implemented. Fifteen parameters of the nominal models relative to the clavicle length, ellipsoid, sternoclavicular and acromioclavicular joint centers, and contact point location were altered from - 1 to 1 cm. Then, scapulothoracic angles were computed during four movements using multibody kinematic optimizations for nominal and altered models. The percentage of scapulothoracic angle variance explained by each parameter of the kinematic model was computed using Effective Algorithm for Computing Global Sensitivity Indices. When altering simultaneously the 15 parameters of the kinematic model, scapulothoracic angles varied up to 50°. For all movements and degrees of freedom, the clavicle length significantly explained the largest part of scapulothoracic angle variance (up to 25%, p < 0.01). In conclusion, kinematic model parameters need to be estimated accurately to avoid any bias in scapulothoracic angle estimates especially in a clinical context. The present sensitivity analysis may also be used as a benchmark for future works focusing on improving shoulder kinematic models. The curves represent mean scapulothoracic angles computed with the nominal model and their variability when kinematic model parameters are altered. The colormap graphs represent the percentage of scapulothoracic angle variance explained by each parameter of the kinematic model.