Novel Neural Interfaces For Upper-Limb Motor Rehabilitation After Stroke

Abstract

Stroke is the third most common cause of death and the main cause of acquired adult disability in developed countries. The most common consequence of stroke is motor impairment, which becomes chronic in 56% of stroke survivors. However, reorganization of brain networks can occur in response to sensory input, expe- rience and learning. Although several post-stroke neurorehabilitation techniques have been investigated, there is no standardized therapy for severely impaired chronic stroke patients except for brain-machine interfaces (BMIs), which have shown positive results but still fail to elicit full motor function restoration. This work presents novel neural interfaces that aim to improve the existing rehabilita- tion therapies and to offer an alternative treatment to severely paralyzed stroke patients. First, we propose a novel myoelectric interface (MI) that is calibrated with electromyographic (EMG) data from the healthy limb, mirrored and used as a reference model for the paretic arm in order to reshape the pathological muscle synergy organization of stroke patients. A 4-session motor training with this mir- ror MI sufficed to induce motor learning in 10 healthy participants, suggesting that it might be a potential tool for the correction of maladaptive muscle activations and by extension, for the subsequent motor rehabilitation after stroke. Second, although significant positive results have been achieved with non-invasive BMIs based on electroencephalographic (EEG) activity, the functional motor recovery induced by such therapies still remains modest mainly due to poor decoding per- formance. Here, we explored the possibility of using novel algorithms to increase the performance of multi-class EEG-decoding of movements from the same limb, showing encouraging but still limited results. Finally, we propose integrating the novel mirror MI into a cortico-muscular hybrid BMI that combines brain and resid- ual muscle activity to increase decoding accuracy and hence, allow a more natural and dexterous control of the interface, facilitating neuroplasticity and motor re- covery. The system was validated in a healthy participant and a stroke patient, setting the premise for its application in a clinical setup.