Detection of condition-based changes in cross-frequency coupling with MEG

We present a method for the study of condition-based variations of functional connectivity in the brain with MEG data. The method is an implementation, for multivariate inputs, of the concept of psychophysiological interactions (PPI). We obtain spectral representations of estimates of brain activity, and use the signal power at specific frequency bands as inputs to our PPI model. PPI can be formulated as a multivariate linear model, which allows us to compute statistics measuring the amount of coupling change. The resulting PPI maps can be thresholded for significance controlling the familywise error rate. These thresholds are directly applicable to tests to identify the frequency bands that cause a significant effect, which are modifications of the original PPI model. Applying our method to simulations and to data from a MEG visuomotor study, we demonstrate that it is able to accurately detect modulations in interaction across space as well as the frequency bands that contribute to these modulations.