Modulation of MEG signals during overt and imagined wrist movement for brain-computer interfaces

This work uses Magnetoencephalography (MEG) to investigate movement-related neural activity in the cerebral cortex. MEG is an efficient non-invasive tool to study cortical activity because it has higher temporal and spatial resolutions than other non-invasive methods, such as fMRI and EEG. One objective of the proposed study is to characterize MEG signal modulation during overt and imagined movements. Such characterization can then be implemented to study motor control and cortical plasticity. In the future, this information can be used to aid the mapping of motor regions of the brain prior to surgical implantation of electrodes for a brain-computer interface (BCI). For the current experiments, four right-handed subjects were asked to perform wrist movements with their dominant hand in four directions (radial deviation, ulnar deviation, flexion, and extension) following a visual cue (up, down, left, and right, respectively). In separate sessions, subjects were then asked to imagine performing the same movements following the visual cue. Frequency-domain analysis of the MEG signals reveals consistent modulation during both overt and imagined movements on sensors overlaying sensorimotor areas of the brain. Modulation preceded movement onset and was characterized as an inhibition in low frequency bands (10-30Hz) and excitation of lower bands (0-10Hz), starting 200 ms after the visual cue and lasting 500 ms, which was accompanied by an increase of power in the 65-90Hz band during the same period. This sequence is followed by an increase in power in the 10-30Hz band. Several of these modulations in cortical activity were also significantly tuned (p < 0.05) to the direction of movement in both overt and imaginary tasks. Two methods were used for decoding: Optimal Linear Estimator (OLE) and Bayesian. The decoding accuracy of a given target for the imagined wrist movement data varied among subjects from 29.4% to 49.75% (mean: 41.4%) correct trials for OLE, and 30.1% to 50.9% (mean: 41.5%) for Bayesian. For overt wrist movement data, decoding accuracy for a given target ranged from 34.1% to 67.4% (mean: 48.3%) correct trials for OLE, and 33.1% to 66.9% (mean: 48.0%) for Bayesian. MEG can detect cortical areas that show directionally tuned modulation during overt and imagined wrist movement. We conclude that MEG may be an important tool for the development of BCIs, and for the identification of regions for future insertion of electrodes for neuroprosthetic control