Robert F. Kirsch, Arthur T.C. Au, Purvee P. Parikh, Ana Maria AcostaDepartment of Biomedical Engineering, Case Western Reserve UniversityCleveland, Ohio 44106
Individuals with cervical spinal cord injuries (SCI) retain voluntary control of some shoulder muscles while others are paralyzed. Functional neuromuscular stimulation (FNS) of the paralyzed muscles can be used to restore their force-generating capacities. However, a control method to coordinate the actions of several paralyzed (but stimulated) muscles with the actions of the remaining voluntary muscles must be devised. We have investigated the potential use of electromyographic (EMG) recordings from voluntary shoulder muscles in individuals with C5-C6 spinal cord injury to automatically control the stimulation to paralyzed shoulder muscles in a task-appropriate manner. Experimentally, we have evaluated the ability of a time-delayed artificial neural network to predict shoulder and elbow motions using only EMG signals recorded from six shoulder and elbow muscles as inputs, both in able-bodied subjects and in subjects with tetraplegia arising from C5-C6 spinal cord injury. For both subject populations, all four joint angles (elbow flexion-extension and shoulder horizontal flexion-extension, elevation-depression, and internal-external rotation) and the corresponding joint velocities and accelerations, were accurately predicted during movements of widely different complexities performed at different speeds and with different hand loads. These results indicate that the EMG signals from shoulder and elbow muscles with voluntary contain a significant amount of information about arm movement kinematics that could be exploited to control shoulder muscle FNS, although exactly how to perform this control was not investigated. Thus, a second study using a model-based approach was performed to determine if the stimulation of several paralyzed muscles could be predicted directly from the activation patterns or muscles with retained voluntary control. A musculoskeletal model of the human shoulder and elbow was used in simulation to train an artificial neural network to automatically generate appropriate stimulation patterns for the "paralyzed" muscles based on "voluntary" muscle activations. Substantial shoulder strength was provided by adding just two muscles (pectoralis major and latissimus dorsi), and the needed activations of these "stimulated" muscles could be predicted with reasonable accuracy using just two muscles (trapezius and rhomboids) as control sources. Taken together, the results of these two studies indicate that EMG-based control of shoulder muscle FNS in individuals with C5-C6 level SCI should be feasible.