EEG-based effective connectivity distinguishes between unresponsive states with and without report of conscious experience and correlates with brain complexity

Abstract

Objective methods for distinguishing conscious from unconscious states in humans are of key importance for clinical evaluation of general anesthesia and patients with disorders or consciousness. Here, we test the generalizability of a DTF-based algorithm - a measure of effective connectivity - as an objective measure of conscious experience during anesthesia and correlate it with a well-tested index of consciousness: the Perturbational Complexity Index (PCI). We reanalyzed EEG data from an experimental study in which 18 healthy volunteers were randomly assigned to one of three types of general anesthesia: propofol, xenon, and ketamine. EEG was recorded before and during anesthesia, and DTF was calculated from every 1-second segment of the EEG data to quantify the effective connectivity between channel pairs. This was used to classify the state of each participant as either conscious or unconscious, and the classifications were compared with the participant’s delayed report of experience, and the PCI. The algorithm was more likely to classify participants as conscious in the awake state than during propofol and xenon anesthesia (p<0.05), but not during ketamine anesthesia (p>0.05). Furthermore, the DTF-based confidence of being classified as conscious was highly correlated with PCI (r2=0.48, p<0.05). These results provide further support for the notion that effective connectivity measured between EEG electrodes can be used to distinguish between conscious and unconscious states in humans.