Comparison of different multivariate methods for the estimation of cortical connectivity: simulations and applications to EEG data

The problem of the definition and evaluation of brain connectivity has become a central one in neuroscience during the latest years, as a way to understand the organization and interaction of cortical areas during the execution of cognitive or motor tasks. Among various methods established during the years, the directed transfer function (DTF), the partial directed coherence (PDC) and the direct DTF (dDTF) are frequency-domain approaches to this problem, all based on a multivariate autoregressive modeling of time series and on the concept of Granger causality. In this paper we propose the use of these methods on cortical signals estimated from high resolution EEG recordings, a non invasive method which exhibits a higher spatial resolution than conventional cerebral electromagnetic measures. The principle contribution of this work are the results of a simulation study, testing the capability of the three estimators to reconstruct a connectivity model imposed, with a particular eye on the capability to distinguish between direct and indirect causality. An application to high resolution EEG recordings during a foot movement is also presented