Assessing impact of channel selection on decoding of motor and cognitive imagery from MEG data

Objective. Magnetoencephalography (MEG) based brain–computer interface (BCI) involves a largenumber of sensors allowing better spatiotemporal resolution for assessing brain activity patterns.There have been many efforts to develop BCI using MEG with high accuracy, though an increase inthe number of channels (NoC) means an increase in computational complexity. However, not allsensors necessarily contribute significantly to an increase in classification accuracy (CA) andspecifically in the case of MEG-based BCI no channel selection methodology has been performed.Therefore, this study investigates the effect of channel selection on the performance of MEG-basedBCI. Approach. MEG data were recorded for two sessions from 15 healthy participants performingmotor imagery, cognitive imagery and a mixed imagery task pair using a unique paradigm.Performance of four state-of-the-art channel selection methods (i.e. Class-Correlation, ReliefF,Random Forest, and Infinite Latent Feature Selection were applied across six binary tasks in threedifferent frequency bands) were evaluated in this study on two state-of-the-art features, i.e.bandpower and common spatial pattern (CSP). Main results. All four methods provided astatistically significant increase in CA compared to a baseline method using all gradiometersensors, i.e. 204 channels with band-power features from alpha (8–12 Hz), beta (13–30 Hz), orbroadband (α + β) (8–30 Hz). It is also observed that the alpha frequency band performed betterthan the beta and broadband frequency bands. The performance of the beta band gave the lowestCA compared with the other two bands. Channel selection improved accuracy irrespective offeature types. Moreover, all the methods reduced the NoC significantly, from 204 to a range of1–25, using bandpower as a feature and from 15 to 105 for CSP. The optimal channel number alsovaried not only in each session but also for each participant. Reducing the NoC will help todecrease the computational cost and maintain numerical stability in cases of low trial numbers.Significance. The study showed significant improvement in performance of MEG-BCI with channelselection irrespective of feature type and hence can be successfully applied for BCI applications.