Maximizing the Information Transfer Rate of a Myoelectric Classification System for Individuals With Spinal Cord Injury

Spinal cord injury (SCI) has a devastating impact on the lives of both those with the injury and those surrounding them. Myoelectric control systems have the potential to help control complex assistive technologies and enhance independence for the users, but do not take into consideration the diversity of neurological impairment observed following SCI. The objectives of this study were: 1) demonstrate that optimizing the design parameters of a myoelectric control system (number of gestures, gesture rate) for each participant can increase the information transfer rate (ITR), and 2) characterize the relationship between the optimal design parameters and the pattern of impairment. The ITR was used to capture trade-offs between increasing the number of possible gestures and gesture rate with the resulting drops in classification accuracy. Ten uninjured and ten participants with SCI were recruited. Using an 8-channel myoelectric control system, a series of trials were performed where the ITR was determined for different combinations of gesture numbers and gesture rates. A significant improvement in ITR was observed after optimization to the individual's level of injury in the SCI group ( $25.8\pm 10.9$ to $31.8\pm 8.0$ bits/min, p =0.002). Significant correlations were observed between the optimal gesture number and multiple metrics of impairment. These results demonstrate that the ITR of a myoelectric classification system can be increased by taking into account the presence of neurological impairment after SCI.