What are electroencephalogram entropies really measuring

Abstract Entropy is a theoretical framework rather than a single concept or value. Two families of EEG entropies are typified by the ‘approximate entropy’ and the ‘spectral entropy’. Both are measures of uncertainty, or regularity, of a signal. They have maximal values when the signal is Gaussian random noise; and minimal values when the signal is completely predictable (e.g. single-frequency sine wave). EEG entropies correlate closely with cortical activity—a drug (or disease state) that depresses cortical activity will correlate with a change in EEG entropy. For example, GABAergic anaesthetic agents cause the EEG to become more regular, more sinusoidal, and the EEG entropy to decrease in proportion to anaesthetic effect and level of consciousness. Conversely, many adjunctive anaesthetic drugs (nitrous oxide and opioids) have little effect on cortical activity and consequently no significant effect on the entropy of the frontal EEG. In such cases, the entropy of the EEG will tend to give falsely high values in relation to the level of consciousness. Correct interpretation of entropy in the setting of clinical anaesthesia requires an understanding of these limitations.