Optically Pumped Magnetometers for Practical MEG-Based Brain-Computer Interfacing

Brain-computer interfaces (BCIs) analyse neural signatures to decode the user’s intention and control an external device. In support of a wide applicability, a reliable non-invasive tool for capturing neural signals with high information content is needed. Currently, the most prominent non-invasive technique is scalp-recorded electroencephalography (scalp-EEG). However, despite being cost-effective and delivering promising results, its limited spatial resolution hampers access to more sophisticated BCI applications. Magnetoencephalography (MEG) might be a better alternative, but is currently vastly underrepresented in the literature as costly and confining acquisition hardware hampers its adoption. Recently, a new generation MEG sensor based on optically pumped magnetometers (OPMs) has been introduced and shown to overcome many of the practical limitations of traditional MEG hardware. However, it is currently unclear whether the OPM-recorded signals are sufficiently stable when used in a BCI context. In this work, we report on a real-time OPM-MEG-based ‘mind-spelling’ BCI, with which three participants were able to spell words with an average accuracy of 97.7%. This demonstration confirms that single-trial neural responses can be reliably decoded from OPM-MEG and demonstrates its potential for the development of practical MEG-based BCI applications.