An integrated information theory index using multichannel EEG for evaluating various states of consciousness under...

The integrated information theory (IIT) of consciousness introduces a measure Φ to quantify consciousness in a physical system. Directly related to this, general anesthesia aims to induce reversible and safe loss of consciousness (LOC). We sought to propose an electroencephalogram (EEG)-based IIT index Φ_EEG to evaluate various states of consciousness under general anesthesia.

Based on the definition of mutual information, we estimated the Φ_EEG by maximizing the integrated information under various time lags. We used the binning method to cut the nonGaussian EEG data for estimating mutual information. We tested two EEG databases collected from propofol- (n=20) and sevoflurane-induced (n=15) anesthesia, and especially, we compared the Φ_EEG of drowsy (n=7) and responsive participants (n=13) under propofol anesthesia. We compared the effectiveness of Φ_EEG with the estimated bispectral index (eBIS).

In all EEG frequency bands, we observed a negative correlation between Φ_EEG and end-tidal sevoflurane concentration under sevoflurane-induced anesthesia (p<0.001,BF₁₀>6000). Under propofol-induced anesthesia, drowsy participants in moderate sedation (6.96±0.26(mean±SD)) showed decreased alpha-band Φ_EEG compared with baseline (7.40±0.53,p=0.016,BF₁₀=3.58), no significant difference was observed for responsive participants. Oppositely, the responsive participants in moderate sedation (-5.32±0.38) showed decreased eBIS compared with baseline (-4.94±0.40,p=0.03,BF₁₀=2.41).

These findings may enable monitors of the anesthetic state that can distinguish consciousness and unconsciousness rather than the changes of anesthetic concentrations. The alpha-band Φ_EEG is promising for deriving the gold standard for depth of anesthesia monitoring.