A non-invasive brain-machine interface via independent control of individual motor units

Abstract

Brain-machine interfaces (BMIs) have the potential to augment human functions and restore independence in people with disabilities, yet a compromise between non-invasiveness and performance limits their relevance. Here, we demonstrate a BMI controlled by individual motor units non-invasively recorded from the biceps brachii. Through real-time auditory and visual neurofeedback of motor unit activity, 8 participants learned to skillfully and independently control three motor units in order to complete a two-dimensional center-out task, with marked improvements in control over 6 days of training. Concomitantly, dimensionality of the motor unit population increased significantly relative to naturalistic behaviors, largely violating recruitment orders displayed during stereotyped, isometric muscle contractions. Finally, participants demonstrated the potential of a motor unit BMI to power general applications by navigating a virtual keyboard in a spelling task, achieving performances comparable to spelling-tailored non-invasive BMIs that leverage less flexible control strategies to improve performance. These results highlight a largely unexplored level of flexibility of the sensorimotor system and show that this can be exploited to create a versatile, skillfully-controllable non-invasive BMI that has great potential to both provide translational benefit and augment motor functions.