Abstract
BACKGROUND: Investigating functional connectivity between brain networks has become an area of interest in neuroscience. Several methods for investigating connectivity have recently been developed, however, these techniques need to be applied with care. We demonstrate that global field synchronization (GFS), a global measure of phase alignment in the EEG as a function of frequency, must be applied considering signal processing principles in order to yield valid results.
NEW METHOD: Multichannel EEG (27 derivations) was analyzed for GFS based on the complex spectrum derived by the fast Fourier transform (FFT). We examined the effect of window functions on GFS, in particular of non-rectangular windows.
RESULTS: Applying a rectangular window when calculating the FFT revealed high GFS values for high frequencies (>15Hz) that were highly correlated (r=0.9) with spectral power in the lower frequency range (0.75-4.5Hz) and tracked the depth of sleep. This turned out to be spurious synchronization. With a non-rectangular window (Tukey or Hanning window) these high frequency synchronization vanished. Both, GFS and power density spectra significantly differed for rectangular and non-rectangular windows.
COMPARISON WITH EXISTING METHOD(S): Previous papers using GFS typically did not specify the applied window and may have used a rectangular window function. However, the demonstrated impact of the window function raises the question of the validity of some previous findings at higher frequencies.
CONCLUSIONS: We demonstrated that it is crucial to apply an appropriate window function for determining synchronization measures based on a spectral approach to avoid spurious synchronization in the beta/gamma range.
Rusterholz, Thomas