Real-Time Myocontrol of a Human-Computer Interface by Paretic Muscles After Stroke

Biosignals from skeletal muscle have been used to control human-machine interface. Signals from paretic muscles of humans with stroke are distorted and highly variable. Here, we examine the stability of surface electromyography (sEMG) features from paretic hand muscles to enable continuous, real-time multicommand control of a human-computer interface (HCI). Subjects with long standing cortical strokes (>6 months, n = 12) and neurologically intact controls (n = 12) performed two wrist rotations (wrist extend and wrist flex) and two grips (power grip and tine grip) with the nondominant (controls) or paretic (stroke patients) hand. Data reduction analyses revealed a distinct pattern of coactivation across muscles for each gesture. These synergies were similar for control and stroke groups and stable across sessions. Results of offline experiments involving wrist rotation and hand grips confirmed that gestures performed in isolation or combination were recognized at greater than chance level in both groups (p < 0.01). In online experiments, HCI control was evaluated with a balloon shooter game. Users in both groups were able to control the direction and speed of a simulated bullet to a balloon target with greater than chance-level accuracy (p < 0.01). Taken together, these results demonstrate that sEMG synergy features from paretic hand muscles can be used to drive continuous, real-time multicommand control of an HCI.