Evaluation of a motor imagery electroencephalography brain-computer interface as a communication modality for children with cerebral palsy
House, Sarah 1,2; Chau, Tom 1,2
1 Institute of Biomaterials and Biomedical Engineering, University of Toronto ; 2 Bloorview Research Institute
Cerebral palsy (CP) is the most common cause of childhood disability, limiting the ability of affected children to coordinate their movements. The majority of affected children also have speech and language impairments rendering typical communication methods difficult to use or impossible. Brain-computer interfaces (BCIs) controlled by imagination of movement of different body parts, known as motor imagery (MI), may serve as a communication solution. This study aims to create a MI-based pediatric BCI using electroencephalography (EEG) and to evaluate BCI performance when used by 20 typically developing children between the ages of 8 and 16. Classification accuracies above 70% indicate successful BCI control. To accomplish this, a MI-BCI has been created by bandpass filtering EEG signals to encompass mu and beta frequency ranges (7-30 Hz). Features in the signal are extracted using the common spatial pattern algorithm and are then classified with participant specific classification algorithms. BCI performance is tested during one offline session, where the acquired data is used to train the classifier, and four online sessions during which visual neurofeedback is provided while participants play a MI computer game. Participants, thus far, have achieved an online classification accuracy between 42-67% indicating that traditional MI features are not adequate to provide children with BCI control. Further post hoc analysis may uncover features that better distinguish MI patterns of activity in developing brains. The results of this study will speak, for the first time, to the potential of a MI EEG-BCI being translated as a communication method for children with CP.