Propofol-induced unconsciousness is linked to decreased connectivity within frontoparietal networks (the default-mode and executive-control networks) and between the thalamus and these networks, with a negative correlation between thalamic and cortical activity emerging during unconsciousness. In contrast, connectivity in low-level sensory cortices (auditory and visual networks) is preserved, including their thalamocortical connections. Loss of consciousness is associated with a breakdown of cross-modal interactions between visual and auditory networks. These findings suggest that unconsciousness results from disrupted communication between sensory and higher-order frontoparietal cortices, preventing conscious perception.
Resting brain activity reveals two anticorrelated cortical systems linked to conscious awareness: an extrinsic system (lateral fronto-parietal areas) associated with external awareness and an intrinsic system (medial brain areas) associated with internal awareness. In 31 healthy volunteers, external and internal awareness were significantly anticorrelated, with a mean switching frequency of 0.05 Hz, similar to BOLD fMRI slow oscillations. In 22 volunteers, fMRI showed that precuneus/posterior cingulate, anterior cingulate/mesiofrontal cortices, and parahippocampal areas (intrinsic system) correlated with internal awareness, while lateral fronto-parietal cortices (extrinsic system) correlated with external awareness.
Loss of consciousness from propofol sedation reduces long-range temporal correlations in frontothalamic brain activity and weakens the link between functional connectivity and anatomical structure. A model based on phase transitions in complex systems reproduces these patterns and also explains the cortex's reduced sensitivity to external stimuli during unconsciousness. The findings suggest that these neural changes are universal across different causes of unconsciousness.
A mean-field model inspired by Integrated Information Theory and Tegmark's representation of consciousness analyzes order-disorder phase transitions on Curie-Weiss models generated from EEG signals recorded on healthy individuals undergoing deep sedation. A machine learning tool classifies mental states using critical temperatures computed from these models. The method discriminates between states of awareness and deep sedation. A state space representing the path between mental states is identified, with dimensions corresponding to critical temperatures over different EEG frequency bands. The method may have clinical applications.
A machine-learning classifier trained on fluorodeoxyglucose PET brain scans from 37 healthy controls and 13 patients in a vegetative state achieved 100% accuracy in distinguishing between conscious awareness and the vegetative state during cross-validation. When tested on 8 patients with locked-in syndrome, all scans were classified as "conscious" with a mean probability of .95. The authors conclude that relevance vector machine classification of cerebral metabolic images could become a useful tool for automated PET-based diagnosis of altered states of consciousness in coma survivors.