An Evaluation of Optimization Techniques for the Prediction of Muscle Activation Patterns During Isometric Tasks

Abstract

The purpose of this study was to critically evaluate the modeling potential of proposed optimization cost functions for predicting muscle forces during isometric loading. Models of the muscles about the elbow (eleven muscles) and wrist (five muscles) were constructed. The models accounted for muscle moment arms, physiological cross-sectional area, specific tension, and percent fiber type. Five nonlinear optimization cost functions, a representative sample of those proposed to date, were analyzed: minimizing the sums of muscle force2, stress2, stress3, (normalized force)2, and minimizing fatigue. Several different protocols were implemented, including elbow models which balanced combinations of flexion-extension, supination-pronation, and varus-valgus loads. Theoretical predictions were compared with EMG data of muscle activation changes as a function of load direction and muscle coactivation relationships. Results indicate a strong dependence of muscle coordination predictions on the number of degrees of freedom balanced. The choice of cost function had little influence on the results. The cost functions examined were not able to reliably estimate muscle activation as a function of load direction. Furthermore, specific synergic relationships between muscle pairs could not be accurately represented. An error analysis indicated that the discrepancies between predicted values and actual values could not be explained by errors in physiological measurements, as the differences between these two were relatively insensitive to changes in the anatomical parameters. In short, no particular cost function was found to adequately represent actual muscle activity at the elbow, although predictions at the wrist were more favorable due to differences in the degrees of freedom at the joints.