Computational NeuroscienceDetecting millisecond-range coupling delays between brainwaves in terms of power correlations by magnetoencephalography

- A new method to detect power correlations (PCs) between brainwaves is presented.
- The millisecond resolution of the PC method is demonstrated by simulations.
- The PC method is applied successfully to three healthy subjects measured by MEG.
- The PC method revealed diverse and meaningful delays in real MEG data.

BackgroundThe spatiotemporal coupling of brainwaves is commonly quantified using the amplitude or phase of signals measured by electro- or magnetoencephalography (EEG/MEG). To enhance the temporal resolution for coupling delays down to millisecond level, a new power correlation (PC) method is proposed and tested.New methodThe cross-correlations of any two brainwave powers at two locations are calculated sequentially through a measurement using the convolution theorem. For noise suppression, the cross-correlation series is moving-average filtered, preserving the millisecond resolution in the cross-correlations, but with reduced noise. The coupling delays are determined from the delays of the cross-correlation peaks.ResultsSimulations showed that the new method detects reliably power cross-correlations with millisecond accuracy. Moreover, in MEG measurements on three healthy volunteers, the method showed average alpha-alpha coupling delays of around 0-20 ms between the occipital areas of two hemispheres. Lower-frequency brainwaves vs. alpha waves tended to have a larger lag; higher-frequency waves vs. alpha waves showed delays with large deviations.Comparison with existing methodsThe use of signal power instead of its square root (amplitude) in the cross-correlations improves noise cancellation. Compared to signal phase, the signal power analysis time delays do not have periodic ambiguity. In addition, the novel method allows fast calculation of cross-correlations.ConclusionsThe PC method conveys novel information about brainwave dynamics. The method may be extended from sensor-space to source-space analysis, and can be applied also for electroencephalography (EEG) and local field potentials (LFP).