Towards the development of a hybrid neural-machine interface for volitional control of a powered lower limb prosthesis

Currently available prosthetic devices lack the inherent inability to interpret the user's intent during locomotion. Volitional control of lower-limb prostheses has been demonstrated using electromyography (EMG)-based controllers, in which activation of the muscles of the lower-limb can be harnessed to decipher user intent. However, although robust during continuous locomotion, EMG-based systems (which rely only on peripheral nervous signaling) lack the ability to predict locomotion mode transitions (e.g., level ground to stairs, stairs to level ground) with advanced notice. In this study, we report on the electrocortical dynamics of human walking across different unconstrained walking conditions (i.e., level ground (LW), ramp ascent (RA), and stair ascent (SA)) using non-invasive electroencephalography (EEG). Additionally, the feasibility of detecting user locomotion modes (i.e., level-ground walking vs stair ascent) from EEG signals is investigated in able-bodied individuals. Subjects were instrumented for full-body mobile brain-body imaging (EEG, EMG, motion capture) while walking on an experimental gait course involving locomotion on level-ground, stairs, and ramps. A systematic EEG processing method was implemented to reduce artifacts (e.g., eye blinks/movements, muscle artifacts) and localize distinct brain sources within the EEG signal. Our results provide evidence that electrocortical amplitude modulations are correlated with the level of difficulty during locomotion tasks. Furthermore, we observe that activations in the occipital lobe appear to contain information relevant to identifying terrain transitions within the brain. This work is a step toward the development of a multimodal neural-machine Interface (NMI) that utilizes neural signaling from the brain (EEG) and muscles (EMG) signals for natural control of power prosthetic legs. This research is partly supported by NSF award IIS-1302339 and NIH award 1F99NS105210-01