A foot motor imagery brain-computer interface with realistic visual feedback: preliminary evaluation in healthy and stroke subjects

This paper presents the development and preliminary evaluation of a non-invasive brain-computer interface (BCI) based on motor imagery (MI) aimed at lower limb stroke rehabilitation. The BCI considers the subject’s ability, concentration, and motivation through a classification threshold adapted by the operator. It presents realistic visual feedback, which consisted of an avatar that mimics foot dorsiflexion. Ten healthy and three chronic stroke volunteers participated in one session of use; all were naive regarding BCI. Two evaluations were performed. The first one considered the area under the curve (AUC) in the receiver operating characteristics (ROC) space to assess user’s ability to modulate his/her sensorimotor rhythms. The second evaluation addressed the MI-BCI operation and computed the accuracy (ACC), the true positive rate (TPR), and false positive rate (FPR). These last two metrics were represented in the ROC space. The outcomes for healthy subjects (AUC = 0.66 ± 0.12, ACC = 64.4 ± 6.9%) resulted to be above the chance level. In stroke subjects, the results could have been affected by the brain injury (AUC = 0.53 ± 0.03, ACC = 54 ± 6%). The results of TPR and FPR for healthy (TPR = 58 ± 10%, FPR = 29 ± 18%)and stroke (TPR = 51 ± 21%, FPR = 45 ± 28%) subjects suggested a difficulty inherent to the idle state identification. It was shown that the developed MI-BCI aimed at motor recovery could adapt to the SMR modulation of the user-patient and enable to control avatar dorsiflexion despite subjects’ inexperience or stroke sequelae in a single session of use.