Training wrist extensor function and detecting unwanted movement strategies in an EMG-controlled visuomotor task

Stroke patients often suffer from severe upper limb paresis. Rehabilitation treatment typically targets motor impairments as early as possible, however, muscular contractions, particularly in the wrist and fingers, are often too weak to produce overt movements, making the initial phase of rehabilitation training difficult. Here we propose a new training tool whereby electromyographic (EMG) activity is measured in the wrist extensors and serves as a proxy of voluntary corticomotor drive. We used the Myo armband to develop a proportional EMG controller which allowed volunteers to perform a simple visuomotor task by modulating wrist extensor activity. In this preliminary study six healthy participants practiced the task for one session (144 trials), which resulted in a significant reduction of the average trial time required to move and hold a cursor in different target zones (p < 0.001, ANOVA), indicating skill learning. Additionally, we implemented an EMG based classifier to distinguish between the desired movement strategy and unwanted alternatives. Validation of the classifier indicated that accuracy for detecting rest, wrist extension and unwanted strategies was 92.5 + 6.9% (M + SD) across all participants. When performing the motor task the classification algorithm flagged 4.3 + 3.5% of the trials as 'unwanted strategies', even in healthy subjects. We also report initial feedback from a survey submitted to two chronic stroke patients to inquire about feasibility and acceptance of the general setup by patients.