Effect of an exoskeleton-assisted training on brain and motor learning

Exoskeleton gait rehabilitation represents a developing area of research. Nevertheless, exoskeletons exploited in rehabilitation are still not capable of fully adapting and properly interacting with the final user to build an optimal human-machine interface. To contribute to obtain this goal, physiological signals recorded by electroencephalography (EEG) and electromyography (EMG) together with signals derived from inertial measurement units (IMU) can be used to investigate and more deeply understand the effects of an exoskeleton training on brain and motor learning. In this study, participants were asked to sit with eyes opened in a quiet environment to collect a resting state EEG as a baseline, then they were asked to walk self-paced on a 15 m pathway. They underwent gait training with an exoskeleton and, after the device removal, each participant walked for three times on the same pathway. Only EMG and IMUs data were collected while they were walking without the exoskeleton. Then a new resting state EEG recording was collected. From this dataset, we aim to evaluate short-term neuro-muscular plasticity induced by a single training session of exoskeleton. To meet this purpose, we processed and evaluated EEG patterns comparing the power spectral density (PSD), the power spectra in the canonical EEG frequency bands and the Higuchi fractal dimension (HFD) pre- and post-training session. Moreover, we processed data collected by IMU using them firstly to obtain time parameters of gait and walking speed, and then to have information about stability, complexity and smoothness of gait. In particular, from inertial sensors data we evaluated the multiscale entropy (MSE), the Higuchi fractal dimension (HFD) and the tortuosity (T) to evaluate complexity and regularity of motion, the improved harmonic ratio (iHR) to assess symmetry, and the spectral arc length (SPARC) and the normalized jerk (NJ) for movement smoothness assessment. From the processing and manipulation of EMG data we computed the root mean square (RMS), the center of activity (CoA) and the co-contraction index (CI) to evaluate muscle activation amplitude and timing. Indexes obtained from IMU and EMG data were compared between pre-training session and post-training session and, in particular, data related to the first track that the subjects traveled before undergoing exoskeleton training session (T0), data collected immediately after the exoskeleton training (T1) and data regarding the fifth track crossed after the walk (T2) were analyzed. Results show that, even after a single training session of an exoskeleton, there is an influence in the neural processing at the level of sensorimotor areas in control subjects (i.e., a statistically significant increase in the power spectra of alpha and beta EEG frequency bands mirrored by an increase in regularity in the same area). Moreover, a new walking pattern is observed, in particular this is statistically significant comparing the two post-training conditions, together with a lower walking speed immediately after exoskeleton training that increases again later during the walking; the lumbar acceleration along the medio-lateral and the vertical directions, directions that were more stressed during the exoskeleton-training, is less complex and less rich of oscillatory components post-exoskeleton training with respect to the pre-training condition. Timing of muscles’ activation throughout the gait cycle and a higher muscles’ co-contraction at leg after exoskeleton training is the other observed major change in participants walking, which seem to persist over time highlighting the presence of short-term motor memory. These preliminary results encourage to deepen the understanding of human-machine interaction mechanisms and to promote the development of more efficient and suitable exoskeletons for walking assistance.