Comparing MEG and EEG measurement set-ups for a brain--computer interface based on selective auditory attention

Auditory attention modulates auditory evoked responses to target vs. non-target sounds in electro- and magnetoencephalographic (EEG/MEG) recordings. Employing whole-scalp MEG recordings and offline classification algorithms has been shown to enable high accuracy in tracking the target of auditory attention. Here, we investigated the decrease in accuracy when moving from the whole-scalp MEG to lower channel count EEG recordings and when training the classifier only from the initial part of the recording instead of extracting training samples throughout the recording. To this end, we recorded simultaneous MEG (306 channels) and EEG (64 channels) in 18 healthy volunteers while they were presented with concurrent streams of spoken "Yes"/"No" words and instructed to attend to one of them. We then trained support vector machine classifiers for predicting the target of attention from unaveraged trials of MEG/EEG. Classifiers were trained either on 204 MEG gradiometers or on EEG with 64, 30, 9 or 3 channels and with samples extracted randomly across or only from the beginning of the recording. The highest classification accuracy, 73% on average across the subjects for 1.0-s trials, was obtained with MEG when the training samples were randomly extracted throughout the recording. With EEG, the accuracies were 69%, 69%, 67%, and 63% when using 64, 30, 9, and 3 channels, respectively. When training the classifiers with the same amount of data but extracted only from the beginning of the recording, the accuracy dropped by 12 %-units on average, causing the result from the 3-channel EEG to fall below the chance level. Combining five consecutive trials partially compensated for this drop such that it was 1-8 %-units. While moving from whole-scalp MEG to EEG reduces classification accuracy, a usable auditory-attention-based brain-computer interfaces can be implemented with a small set of optimally-placed EEG channels.