Effect of anesthesia in rats on measures of complexity, differentiation, and integrated information

To investigate proposed mechanisms behind loss of consciousness, it is important to extend the current state of the field in human research to rodents. Perturbational complexity index (PCI) is a promising metric of “capacity for consciousness” in humans. PCI is based on a perturbational approach that allows inferring a system’s capacity for causal integration and differentiation of information. However, PCI is technically demanding and time-consuming. Thus, measures based on spontaneous EEG recordings may be more practical for certain clinical purposes. Furthermore, while a perturbational approach might better assess a system’s capacity for information processing, analysis of spontaneous EEG might be more indicative of current ongoing dynamics. Here, we compare PCI to a range of spontaneous EEG-based measures of signal diversity and integrated information in rats undergoing propofol, sevoflurane, and ketamine anesthesia.We find that Lempel-Ziv complexity, amplitude coalition entropy, and geometric integrated information are able to distinguish between awake and anesthesia, and that all implemented measures (except synchrony coalition entropy) were rank-order correlated with PCI. In addition, while the measures of signal diversity decreased in anesthesia (except for ketamine), several measures of integrated information increased, contrary to expectations. Based on a divergence in network properties estimated from directed functional connectivity (current results) and effective connectivity (earlier results), we speculate that while anesthesia may disrupt global cortico-cortical information transfer, as suggested by perturbation-based approaches, spontaneous activity may appear to suggest the opposite because of driving subcortical projections from e.g. the thalamus, or suppressed encoding specificity of information.