Granger causality analysis of steady-state EEG during propofol-induced anaesthesia
Normal 0 false false false EN-GB X-NONE X-NONE
An important challenge in the neuroscience of consciousness is to discern neurophysiological correlates of changes in conscious level. Various theories suggest a key role for directed information flow between brain regions in modulating conscious level. However, measuring information flow is technically challenging, even given relevant data. A promising measure in this context is Granger causality (G-causality), which quantifies the extent to which past observations from one region help to predict future observations from another.
We developed a new method for applying G-causality to high-density, steady-state EEG data, and applied it to data from subjects undergoing propofol-induced anaesthesia. These data were source localized, furnishing low-artefact time-series from anterior and posterior cingulate cortices, these being areas showing large gamma-based power changes between waking and loss-of-consciousness (LOC). For each subject, we computed G-causalities for multiple short data segments taken from both the waking state and LOC. We used permutation analysis to eliminate statistical bias in individual observations, while allowing that G-causalities could vary during both waking and LOC. Inferences on changes in mean G-causality showed high inter-subject consistency, marked by a significant increase in bidirectional Granger causality during LOC in most subjects, especially in the gamma and beta bands. In contrast, changes in synchrony showed substantially lower consistency among subjects, though across subjects we confirmed a general increase in gamma synchrony during LOC.
Our results illustrate a methodological pipeline for rigorous G-causality analysis of steady-state EEG data, and underline the utility of G-causality in exposing functional neural interactions underlying different conscious levels.