The effect of reducing the number of EMG channel inputs on loading and stiffness estimates from an EMG-driven model of the spine

Abstract

Electromyography (EMG)-driven models of the spine routinely require between ten and 14 EMG channels to estimate joint load and stiffness variables. This study was designed to determine the sensitivity of common EMG-driven model outputs to the removal of individual EMG channels, and to test two adapted models driven from eight channels. A total of 11 male participants performed a variety of static exertions designed to resist either an applied trunk flexion or right side trunk lateral bend moment. In this study, 14 channels of EMG were recorded and used to drive a biomechanical model of the spine to predict L4-L5 joint load and stiffness values. The model was subsequently re-run after the removal of individual pairs of bilateral EMG channels, and again with eight-channel models in which the rectus abdominus, latissimus dorsi and multifidus EMG-channels were eliminated. Results showed that the eight-channel model provided estimates for the majority of output variables that did not differ substantially from the 14-channel model, except in instances in which muscle force output was ramped to resist flexion moments. Estimates of the output variables were, in general, improved when multifidus fascicles were re-added to the model and driven from the lumbar erector spinae EMG sites.