Abstract
Background
Electroencephalogram (EEG) entropy rate estimation has been proposed as a measure of anesthetic depth.
The authors explore the principles of entropy rates measurement for
the purpose of extracting depth of anesthesia information from the EEG, and apply entropic measures to EEG
data from patients undergoing general anesthesia. These are compared with traditional spectral indices.
Methods
The following were assessed: Conditional Entropy (CEn); Corrected Conditional Entropy (CCEn); Approximate Entropy (ApEn); Coarse-grained Entropy rates (CGEn); Gaussian Process Entropy rates (GPEn); and Spectral Entropy (SpEn). For comparison, Spectral Edge 95 (SE95) and Bispectral Index Scale (BIS) were used.
EEG data were logged using an Aspect A-2000 BIS monitor from adult patients undergoing elective surgery. Eight
EEG series were investigated. Two representative parts of these EEG series were used to quantify discriminative power of each method: a series containing moderate and light anesthesia; and one containing emergence
from anesthesia. Discrimination of each measure was measured in units of baseline variance.
All measures (except BIS) were calculated off-line. Where data quantization was required two methods were compared.
A range of subparameters were evaluated where relevant.
Results
All measures gave some indication of depth of anesthesia. A high level of correlation
among the all entropy rates measures was observed (Spearman's ranked correlation coefficient SRC > 0.8).
The entropy rates measures were as good as, or better than the spectral methods at distinguishing light from moderate anesthetic depth.
CGEn showed the greatest discriminative power. Equiquantization improved discrimination in all cases.
Conclusions
This study confirms entropy rate estimations as potentially useful measures of
anesthetic depth. Correction terms for limited data should be employed and equiquantization is preferred.
CGEn was consistently the best measure in this study and is worthy of further investigation. .
Go back