CBER Day 2007

PREDICTING MUSCLE FORCES AND JOINT MOMENTS USING SINGLE JOINT AND MULTI JOINT EMG-DRIVEN MODELS

Daniel N. Bassett, Qi Shao, Kurt T. Manal, Thomas S. Buchanan
Center for Biomedical Engineering Research, University of Delaware, Newark, DE

INTRODUCTION: Single-joint models may be adequate for some applications; however, it may be more appropriate to use a multi-joint model when studying complex motions. The present study investigates biarticular muscles in EMG-driven models accounting for their contributions to both joints they span.

METHODS: Six subjects performed normal walking, hopping, and hop-and-stop tasks while EMG, ground reaction forces, and motion data were collected. Three hybrid EMG-driven models were developed: single ankle, single knee, and multi-joint of the ankle and knee. An optimization algorithm was used to calibrate the forward dynamic Hill-type models by using the inverse dynamic joint moment as a benchmark.

RESULTS AND DISCUSSION: Normal walking comparison between forward dynamics and inverse dynamics joint moments at the ankle gave R2 values of 0.97 and 0.96 and RMS-error of 18.6% and 19.8%; whereas at the knee the R2 values were 0.80 and 0.79 and RMS-error of 20.9% and 23.1% for single-joint and multi-joint models respectively. New task predictions displayed the versatility of the calibrations for hopping ankle and hop-and-stop knee predictions which performed very similarly to walking, and compared to normal walking have similar kinematics and muscle activations. Muscle force predictions showed small variations between single and multi-joint models for the quadriceps, hamstrings, or the dorsiflexor. However, as expected, the gastrocnemii muscle forces varied significantly between the two types of models. Furthermore, a correlation was noted between the magnitude of the late stance knee flexion moment and relative magnitude of the gastrocnemii forces.

CONCLUSION: The three models perfomed very similarly for all subjects and all tasks; however, significant differences were found in the gastrocnemii force predictions. Implying single-joint models of the ankle should account for kinetics of the knee to replicate the presumably more realistic multi-joint force predictions.