Dr Daniele Marinazzo
Machine Learning applications to EEG data: phase synchronization, predictability and causality in multi-EEG recordings
Monday May 4, 2009. 11.00 am
The fundamental mechanisms underlying biological systems can be understood by studying complexity of recorded data and measuring in what proportion the individual components exchange information among each other. The study of phase synchronization may, in particular, reveal interactions which are highly nonlinear and do not depend on amplitude changes. I will show how to detect phase synchronization in evoked potentials as a reaction to a visual stimulus in migraine patients. Furthermore, I will show that it is possible to analyze the issue of regression and predictability of time series with the main goal of finding the sources of predictability itself, that is the correlations and inner structures, strongly nonlinear effects. I will address the issue of how a specific stimulus (in this case a painful stimulus) resets the temporal characteristics of a time series (in this case EEG’s from migraine patients and healthy subjects). Finally, a key feature of collective behavior is the concept of causality, an operative definition of which can be easily derived when dealing with models and predictions. I will describe a method of analysis of dynamical networks based on a recent measure of Granger causality between time series and I will be happy to discuss interactively possible applications to EEG signals.