Real-time magnetoencephalography for neurofeedback and closed-loop experiments

Abstract

Magnetoencephalography (MEG) provides millisecond-scale temporal resolution and can thus track human cortical processes at the speed they occur. Compared to EEG, MEG offers considerably higher spatial resolution which enables better separation of simultaneously active neural sources. Both features make MEG an attractive technology for noninvasive brain–computer/machine interfaces (BCI/BMI). As in EEG, machine-learning algorithms play a central role in optimally applying MEG for BCI/BMI. Although MEG is expensive and nonportable, it could serve as a rapid development platform for eventual inexpensive EEG-based BCI systems that could be applied to patients. In addition, BCI-type approaches may also be used in basic neuroscientific research as they allow unique “closed-loop” experiments where subject’s brain activity influences the stimulus presented to the subject in real time. Such setups may open new windows to human brain function. This chapter introduces the reader to MEG; signal genesis, instrumentation, data preprocessing, and modeling approaches are briefly discussed. Thereafter, real-time analysis of MEG signals is motivated with examples, and specific algorithmic and technical requirements for implementing such setups are covered and practical solutions referred to.