The role of somatosensory feedback for brain-machine interfaces applications

Brain-machine interfaces (BMI) based on motor imagery (MI) have emerged as a promising approach to enhance motor skills and restore motor functions. However, the efficacy and efficiency of BMI systems remain limited. The current lack of usability can be explained by the fact that significant efforts have been dedicated to improve decoding efficiency and accuracy, but BMI studies have generally ignored the user-training component of BMI operation. It has been suggested that somatosensory feedback would be more suitable than standard visual feedback to train subjects to control a BMI. In this thesis, a novel feedback modality has been explored to improve BMI usability, namely sensory-threshold neuromuscular electrical stimulation (St-NMES). St-NMES delivers transcutaneous electrical stimulation that depolarizes sensory and motor axons without eliciting any muscular contraction. In order to assess the effect of this new feedback modality on BMI skill learning this thesis is composed of four experiments. In a first experiment, the effect of St-NMES on MI performance was investigated. Twelve healthy subjects participated in a cross-over design experiment comparing St-NMES with visual feedback. Offline analyses showed that St-NMES not only enhanced MI brain patterns, but also improved classification accuracy. Importantly, St-NMES alone did not induce detectable artefacts. In a second experiment, physiological impact of online BMI training on corticospinal tract (CST) plasticity was studied according to the feedback modality â either St-NMES or visual feedback. Ten healthy participants were enrolled in a cross-over design experiment testing both BMI systems. Results showed that BMI based on St-NMES significantly enhanced CST excitability compared to BMI based on visual feedback. Moreover, BMI system based on St-NMES was significantly more robust and accurate over days. A third experiment further explored the parallelism between BMI learning based on St-NMES feedback and natural motor learning, putting particular attention on the underlying physiology of the process. Apart from analyzing the evolution of BMI performance, we also examined changes in CST excitability and modulation of intracortical inhibition in the early learning phase (after one BMI session) as well as later learning stage (after 2 weeks training). Ten healthy participants were trained to control a BMI based on St-NMES feedback. Results showed that subjects improved their BMI control with practice, what might be explained by the adaptation of the central nervous system over time. Finally, the last experiment explored the feasibility of BMI-St-NMES for upper limb rehabilitation after stroke. A chronic stroke patient with a severe motor disability was trained with BMI-St-NMES over 3 weeks. After training, upper-limb motor function improved, reaching clinical relevance. Based on our previous observations, we believe that BMI-St-NMES training enhanced CST projections leading to motor recove