Department of Computer, Control and Management Engineering

Brain-Computer Interfaces based on Motor Imagery (MI-BCI) have been validated as promising systems to support rehabilitative protocols for the post-stroke motor recovery of the upper limb. To date, the long-term effects of MI-BCI training in post-stroke patients need to be clarified. In order to fill this gap, we analysed electroencephalographic (EEG) data collected during the neurophysiological assessment sessions (longitudinal study) in a context of a two-arm Randomized Controlled Trial. Seventeen subacute stroke patients were enrolled and randomly assigned to the interventions: MI training supported by BCI (BCI group) and MI training without feedback (CTRL group). The aim of this work is to investigate for each intervention group the long-term effects of the MI training on the EEG sensorimotor responsiveness. Therefore, for each group we analysed the alpha and beta frequency band activity for the motor imagery task of grasping and finger extension, performed both with the affected and the healthy side separately collected, before and after the rehabilitation treatment and 1, 3 and 6 months later. Results obtained in this preliminary study show that in the BCI group the activity in alpha and beta band are located over the sensorimotor areas and that such pattern is maintained over time while the amplitude decreases. In the CTRL group the alpha activity is evident on the sensorimotor channels, the amplitude decreases with time till T4 when the pattern turns to be again widespread over the scalp and in the beta band no temporal modifications are reported. Our findings suggest that the MI-BCI induces a sensorimotor EEG reactivity rebalance in hemispheres.