A simple measure of correlation across time, frequency and space between continuous brain signals

This paper introduces a simple but systematic method to estimate correlations between the spectral energy of two continuous electrophysiological signals in such a way that it can detect relationships between different frequencies and different latencies. From two series of signals (e.g. electroencephalogram, magnetoencephalogram or local field potentials) recorded from two sites in response to repeated sensory stimulations, the method computes the time-frequency energy of each signal. Then, it computes the Spearman rank order correlation coefficient across all the trials between the energy of the first signal series in one time-frequency region and the energy of the second signal series in a second time-frequency region. The method was designed to analyze interactions between frequency bands, in an effort to describe how the main brain rhythms interact with each other across time and space. It was applied to two simulations and to intracranial electro-encephalogram (EEG) recordings obtained from an epileptic patient performing two verbal discrimination tests (a phonological and a semantic task). It led to the identification of different correlations patterns in the gamma band depending on the level of semantic analysis performed by the patient.