MOVE

From EEG signals to brain connectivity: model-based evaluation of interdependence measures and applications in epilepsy

In the field of brain research, the past decades have witnessed a considerable increase of interest for methods aimed at estimating the functional connectivity between spatially-distributed regions, under normal (cognitive research) or patho-logical (clinical research) conditions. The reason for this increased interest is related to the commonly-admitted assump-tion according to which most of the brain functions are based on interactions between neuronal assemblies distributed within and across different cerebral regions. Indeed, it has been shown that specific networks activate in response to a particular cognitive task, although underlying processes such as the way coordination between distant areas is achieved are not resolved yet. In the context of epilepsy, it has also been shown that paroxysmal activity may occur in epileptogenic networks extending over rather large regions. The identification of such networks from electrophysiological (scalp EEG, MEG or intracerebral EEG) and imaging (fMRI) data available in human subjects is still considered as a difficult and un-solved problem. In this context, the main objectives of this presentation are: (i)to review some of the methods that can be used to estimate interdependences between field potentials recorded from distant brain structures, (ii) to highlight some key points emerging from the quantitative comparison of these methods, not only on real intracerebral EEG data but also on signals simulated from models in which the underlying connectivity patterns are known a priori (ground truth), (iii) to provide concrete examples of application in the context of drug-refractory epilepsies.