Analyzing synchronization of neural populations from the EEG – Model considerations

Objective: In order to gain a better understanding of the underlying neuronal processes, the interdependence between brain regions is studied for a variety of applications using surface (scalp EEG) and/or intracranial recordings (ECoG or depth EEG). For example, in the field of seizure prediction, the synchronization between EEG channels is analyzed in order to evaluate whether consistent precursors for epileptic seizures can be found. Yet, the question arises to what extent changes in the interaction of the underlying neuronal populations can be detected by analyses of macroscopic EEG recordings.

Methods: We used a computational, generative model of EEG signals, which combines a physiologically relevant simulation of the activity of cortical neuronal populations with a biophysically relevant description of the transmission of cortical activity to the recording electrodes (Cosandier-Rimélé et al. 2008, Neuroimage 42:135). Simulations were based on a topographic reconstruction of the epileptic source of a patient with frontal epilepsy. To simulate different levels of synchronous activity, neuronal populations within the epileptic cortical source were connected with a varying coupling degree. In order to evaluate to which extent changes in the coupling between neuronal populations could be reconstructed from macroscopic EEG, we analyzed the mean phase coherence of signals simulated for the actual positions of invasive (ECoG) and surface contacts (scalp EEG).

Results: In general, significant increases of the synchrony estimate for increasing coupling degree between the underlying neuronal populations could be observed for electrodes close to the epileptic cortical source. As expected, this could be detected more easily from intracranial EEG than from scalp EEG. Yet, it was found that the type of activity generated is a decisive factor in this regard: in the non-spiking regime, changes in the coupling structure of the epileptic cortical source could only hardly be detected.

Discussion: While the analysis of macroscopic EEG recordings is a key for studying higher brain functions, the relationship to the underlying neuronal sources is still not thoroughly understood. In our study, we could show that the degree of coupling between neuronal populations could be estimated by analyzing interactions in the signals recorded by macroscopic contacts, for spiking regimes. Additionally, we discuss potential limitations of these analyses on a macroscopic scale.