Analysis of Brain and Muscle co-activation in Mobile Brain/Body Imaging settings

This paper analyzes the involvement of the cerebral area during robotic assisted gait training on a treadmill (Lokomat) in the two cases of passive and active locomotion. Imaging of the brain activity in such a natural environment is allowed by Mobile Brain/Body Imaging (MoBI) approach with high-density electroencephalography (EEG), which we used to characterize the differences between conditions. A key feature for this research is the pre-processing of the brain signals, which are recorded in a noisy and artefactual environment. For this reason, we analyzed the efficiency of Artifact Subspace Reconstruction (ASR) as cleaning method. We employed EEG, electromyography (EMG) and force-measuring treadmill to record brain and body signals from 5 healthy subjects walking with Lokomat. First, the choosing of the setting of the parameters in ASR is defined based on analysis over 11 healthy subjects, who are asked to walk on a steady-speed treadmill without any robotic orthosis. The effectiveness of ASR is evaluated on the results, mainly in terms of residual variance (rv), it has over the channels and the Independent Components (ICs), taking care of the artifactual and the brain-like ICs. Then, using an approach based on combination of Reliable Independent Component Analysis (RELICA), source localization and functional connectivity, we were able to compute time-warped Event Related Spectral Perturbation (ERSP) and information flow measures on five significative IC. It has been shown that ASR has great effect on the artifactual ICs like eye movement and blinking and neck movement in terms of variance and power removed, even though also brain ICs are affected. ASR, after an optimal choose of parameters, can be a very powerful tool in terms of reliability of the ICA decomposition and of reduced meaningfulness of the artifactual ICs related to the whole signal. From the analysis of the brain activation, cortical involvement during locomotion has been proven looking like the behaviour found in other researches (Gwin and Wagner), showing intra-stride changes in spectral power. The cortical connectivity evaluated from the information flow between different brain sources shows much more flows in the active locomotion (without guiding force) compared to passive one from 0% to 50% of the gait cycle. Future applications of results of this kind bring to an improvement in rehabilitation techniques. In addition, it can be a further step toward the application of brain-computer interfaces for gait training and brain monitoring.