Detection of movement-related cortical potentials associated with upper and low limb movements in patients with multiple sclerosis for brain-computer interfacing

Objectives: Brain-computer interface (BCI) training has been shown to be effective for inducing neuralplasticity and for improving motor function in stroke patients. BCI training could potentially have a positive effect on people with multiple sclerosis (MS) as well by pairing movement-related brain activity with congruent afferent feedback from e.g. functional electrical stimulation. In the current study, the aim was to detect movement-related cortical potentials (MRCPs) from single-trial EEG in people with MS across two separate days using different classifier calibration schemes to estimate the performance of a BCI that can be used for neurorehabilitation.

Approach: Fifteen individuals with MS performed 100 wrist movements and 100 ankle movements while continuous EEG was recorded. Also, idle brain activity was recorded. This was repeated on a separate day.The data were filtered and divided into epochs containing data prior to the movement onset. Temporal, spectral and template matching features were extracted and classified with a random forest classifier using different calibration schemes to estimate the performance when training the classifier on data from the same day and same participant, different day but same participant, and across different participants.

Main results: Clear MRCPs were elicited across both recording days, and it was possible to discriminate between idle activity and movement-related brain activity with accuracies between ~80-90% when training and testing the classifier on data from the same day and participant. The performance decreased when using data from a separate day but same participant (~70-80%) or data from separate participants (~70%) for training the classifier.

Significance: The results showed that it is feasible for people with MS to use a BCI for inducing neuralplasticity.