Multi-scale sample entropy of electroencephalography during sevoflurane anesthesia

Abstract

The electroencephalogram (EEG) has been widely applied in the assessment of depth of anesthesia (DoA). Recent research has found that multi-scale EEG analysis describes brain dynamics better than traditional non-linear methods. In this study, we have adopted a modified sample entropy (MSpEn) method to analyze anesthetic EEG series as a measure of DoA. EEG data from a previous study consisting of 19 adult subjects undergoing sevoflurane anesthesia were used in the present investigation. In addition to the modified sample entropy method, the well-established EEG indices approximate entropy (ApEn), response entropy (RE), and state entropy (SE) were also computed for comparison. Pharmacokinetic/pharmacodynamic modeling and prediction probability (P k ) were used to assess and compare the performance of the four methods for tracking anesthetic concentration. The influence of the number of scales on MSpEn was also investigated using a linear regression model. MSpEn correlated closely with anesthetic effect. The P k (0.83 ± 0.05, mean ± SD) and the coefficient of determination R (2) (0.87 ± 0.21) for the relationship between MSpEn and sevoflurane effect site concentration were highest for MSpEn (P k : RE = 0.73 ± 0.08, SE = 0.72 ± 0.07, ApEn = 0.81 ± 0.04; R (2): RE = 0.75 ± 0.08, SE = 0.64 ± 0.09, ApEn = 0.81 ± 0.10). Scales 1, 3 and 5 tended to make the greatest contribution to MSpEn. For this data set, the MSpEn is superior to the ApEn, the RE and the SE for tracking drug concentration change during sevoflurane anesthesia. It is suggested that the MSpEn may be further studied for application in clinical monitoring of DoA.