Predicting co-contraction with an open source musculoskeletal shoulder model during dynamic and static tasks

Abstract

The shoulder is one of the most complex joints in the body. It has a large range of motion and has active, as well as passive, components to its stabilization. Many injuries occur every year due to overexertion and strain on the shoulder. Musculoskeletal models can be used as a proactive ergonomics tool for shoulder specific job task design, and to help prevent these injuries before they occur. The purpose of this thesis was to critically evaluate the performance of four optimization criteria (sum of squared activation, sum of cubed activation, sum of quartic activation, and entropy assisted) using the open source modeling platform OpenSIM. Experimental torque, kinematic, and EMG data were collected using ten participants for a variety of dynamic arm movements, and static arm postures, in different planes of action. The kinematic and torque data were processed and used as inputs to OpenSIM to calculate predicted muscle activations and joint reaction forces. Experimental EMG was cross correlated with the predicted muscle activity of 8 muscles, and RMSD was calculated between experimental and predicted muscle activity for evaluation. A co-contraction index was also used to assess the model’s ability to predict co-activation between muscle pairs. Overall, the sum of cubed activation and sum of quartic activation model predictions explained significantly more variance (38 ±2.5%, p<0.01) than the sum of squares and entropy models, when compared with experimental EMG. In conclusion, the type of optimization criterion chosen had an effect on the accuracy of the model predictions. Future research, in the development of optimization criterions for the shoulder, will create better model predictions of muscle forces and joint reaction forces, enabling musculoskeletal models to be more useful as a tool to the clinical and ergonomic populations.