Non-linear forecasting measurements of multichannel EEG dynamics.

This work presents a new method for studying the underlying dynamics of multichannel EEG on the basis of the mathematical theory of dynamical systems. It computes the local loss of predictability and Kolmogorov entropy of the dynamics reconstructed from brain electrical activity. This reconstruction uses multichannel recordings in order to quantify an equivalent of spatio-temporal mapping. Five experimental conditions have been studied: closed eyes at rest, closed eyes and counting even numbers, staring at a spotlight, passive and active auditive odd-ball tasks. The entropy is positive for all the experimental conditions which proves that the underlying EEG dynamics are chaotic. Moreover, on the basis of the dynamical signature it is possible to differentiate 3 types of EEG activity: the rest closed eyes activity, the task closed eyes activity (counting and odd-ball tasks) and the open eyes activity (staring at a spotlight). It is inferred that this index could characterize task-related changes in brain activity.