A Novel Methodology to Study Synchrony, Causality and Delay in EEG Data

Abstract

Synchrony and causality analyses performed in EEG data have improved the understanding of complex interactions within the human brain. However, few attempts have been conducted for using the delay magnitude as a feature of synchrony events. A new methodology for studying synchrony in EEG data is presented here. It includes synchrony detection and a novel mechanism to estimate delay magnitude and sign - hence causality - between narrow bandwidth signals. Synchrony detection and delay estimation are separated in two steps: first, significant synchrony is detected using a measure based on phase differences, then, the delay is estimated by analyzing the dispersion of measure maxima in the space spanned by time, frequency and delays. Synthetic EEG data is used to validate the methodology using a synchrony spectral model with controlled bandwidth and multivariate autoregressive models (MVAR). The proposed methodology achieves a superior performance in causality estimation than state-of-the-art techniques in accuracy and robustness to noise. We also present an analysis of data from a psychophysiological experiment of figure categorization. This methodology provides a reliable method to estimate the delay magnitude of synchrony events and it is a better alternative for studying causality than the state-of-the-art techniques employed here.