Neuromusculoskeletal Models to Simulate Rehabilitation of the Injured Ankle

Abstract

EMG-driven neuromusculoskeletal models have been developed by many investigators to estimate joint moments and muscle forces during human movements. We have been exploring how to use these models as a platform for studying how to alter human movement by changing muscle activation patterns. Muscle activity could be changed as part of physical therapy or athletic training, or through functional electrical stimulation. Hence, simulating changes in muscle activation provides a useful tool in rehabilitation following injury or impairment. PURPOSE: An ankle joint moment observed during healthy, normal gait was used to determine if other (theoretical) muscle activation patterns could be used to simulate a therapeutic intervention whereby unwanted muscles are re-trained using therapy (such as biofeedback) and new muscles are trained to produce gait changes. METHODS: A forward dynamics EMG-driven model was used to estimate forces in four muscles that cross the ankle. Gait analysis was used to estimate ankle joint moments using inverse dynamics, which was compared to model estimates of joint moments, and model parameters were adjusted to minimize errors. Subjects were studied during a simple walking task. An altered joint moment profile was then artificially created and EMGs were then adjusted in the model to determine if a set of EMGs could be found to produce the new joint moments. The “new EMGs” were based on minimizing changes to the existing EMGs. RESULTS: The EMG-driven neuromusculoskeletal model predicted the ankle joint moment well. A root mean square difference (RMS) of 3.68 Nm was found between the model-estimated joint moments and the inverse dynamic joint moments. When the muscle activations were changed to generate the new joint moment profile, the error was similar. Examination of the new muscle forces demonstrate that while some of the muscles required high levels of additional activation, a solution could be found that satisfied the constraint criteria and was within physiological limits. CONCLUSIONS: This demonstrates a novel yet practical way neuromusculoskeletal models can be used to study rehabilitation. This could be used as a part of muscle re-training using biofeedback and/or muscle stimulation to reduce undesired forces in specific muscles to yield an improved gait. Supported by NIH grant R01 HD38582.