Prof Markus Müller, physicist by training and now leading a group at the Autonomous University of Mexico in Cuernavaca and the Centre for Science for Complex Systems, will be talking about his EEG analyses techniques using cross-correlation patterns.

13.12.2018 | Hella Kastbjerg

Detecting Meaningful Fluctuations Around Stable Characteristic Correlation Patterns in Electroencephalographic Recordings.

Electroencephalographic (EEG) recordings constitute one of the most popular techniques of brain research. However, EEG signals are highly non-stationary. Thus, one might expect the averaged cross-correlation coefficient between channels, if assuming positive and negative values with equal probability, to turn out to be close to zero over long enough time windows. 

Instead, we find that the average zero-lag cross-correlation matrix estimated in healthy participants shows a characteristic correlation pattern containing pronounced non-zero values that is strikingly stable over time and different states. A similar correlation structure has been found even in EEG recordings of focal onset epileptic seizures. We conclude that this structure is independent of brain states. Because of the clear similarity across individuals, we believe that this correlation pattern depicts a generic feature of brain dynamics. In other words, we interpret this pattern as a manifestation of a dynamical "ground state" of brain activity: a prerequisite to maintain an efficient operational mode. Expressed in terms of dynamical system theory, we interpret this state as a ‘‘shadow’’ of the evolution toward an attractor pattern in phase space.

Taking this interpretation further, we would expect non-stationary dynamical aspects of higher cerebral processes to manifest in deviations from this stable pattern. We test and confirm this hypothesis in correlation analyses of EEG recordings from 10 healthy participants during night sleep, 20 recordings of 9 epilepsy patients, and 42 recordings of 21 healthy subjects in resting state during eyes-open and eyes-closed conditions. Furthermore, we consider intracranial data from epilepsy patients as well as MEG data from healthy participants during rest and during a working memory task.

The main aim of the talk is to show that the estimation of deviations from the stable characteristic (stationary) correlation structure provides a meaningful differentiation of physiological states with fairly homogeneous results across subjects.