Limited value of a common spatial patterns approach to online discrimination of left- and right-hand motor imagery in a pediatric sample

Applications of brain-computer interfaces (BCIs) in pediatric rehabilitation are expanding. However, it is unclear whether popular BCI paradigms developed for adults are feasible in children. This study evaluated, in a typically developing pediatric sample, a time-honored, adult, motor imagery BCI paradigm that discriminates between imagined left- and right-hand movements. Methods: We developed an electroencephalographic pediatric BCI with visual- auditory feedback through a game interface controlled by left- and right-hand motor imagery (MI). The BCI was evaluated in one offline (with sham feedback) and four online (with real-time classifier feedback) sessions with 11 typically developing children aged 9–14 years. The BCI was personalized to each child, via a well-established adult pipeline, namely, a regularized linear discriminant classifier with selected common spatial patterns in mu and beta bands as inputs. Results: Unlike in adults, the online child-specific BCI demonstrated limited discrimination between left and right-hand MI using spatial features (52 ±9%). Only left-hand MI versus rest in a retrospective analysis with personalized feature sets reached 70 ±3%. Conclusions: Our findings suggest that cortical activity corresponding to MI in our pediatric sample departed from well-documented, conspicuously lateralized adult patterns. Further investigation of developmental MI patterns is warranted to identify a pediatric approach to MI BCI.