Emergence EEG pattern classification in sevoflurane anesthesia

Significant spectral electroencephalogram (EEG) pattern characteristics exist in individual patients during the re-establishment of consciousness after general anesthesia. However, these EEG patterns cannot be quantitatively identified using commercially available depth of anesthesia (DoA) monitors. This study proposes an effective classification method and indices to classify these patterns among patients.

Four types of emergence EEG patterns were identified based on the EEG data set from 52 patients undergoing sevoflurane general anesthesia from two hospitals. Then, the relative power spectrum density (RPSD) of five frequency sub-bands of clinical interest (delta, theta, alpha, beta and gamma) were selected for emergence state analysis. Finally, a genetic algorithm support vector machine (GA-SVM) was used to identify the emergence EEG patterns. The performance was reported in terms of sensitivity (SE), specificity (SP) and accuracy (AC).

The combination of the mean and mode of RPSD in the delta and alpha band (P (delta)/P (alpha) performed the best in the GA-SVM classification. The AC indices obtained by GA-SVM across the four patterns were 90.64  ±  7.61, 81.79  ±  5.84, 82.14  ±  7.99 and 72.86  ±  11.11 respectively. Furthermore, the emergence time of the patients with EEG emergence patterns I and III increased as the patients' age increased. However, for patients with EEG emergence pattern IV, the emergence time positively correlates with the patients' age when they are under 50, and negatively correlates with it when they are over 50.

The mean and mode of P (delta)/P (alpha) is a useful index to classify the different emergence EEG patterns. In addition, these patterns may correlate with an underlying neural substrate which is related to the patients' age. Highlights ► Four emergence EEG patterns were found in γ-amino-butyric acid (GABA)-ergic anesthetic drugs. ► A genetic algorithm combined with a support vector machine (GA-SVM) was proposed to identify the emergence EEG patterns. ► The relative power spectrum density (RPSD) was used as a feature to classify the emergence EEG patterns and good accuracy was achieved. ► The statistics shows that the emergence EEG patterns are age-related and may have value in assessing postoperative brain states.