Multiscale permutation entropy analysis of EEG recordings during sevoflurane anesthesia

Abstract

Electroencephalogram (EEG) monitoring of the effect of anesthetic drugs on the central nervous system has long been used in anesthesia research. Several methods based on nonlinear dynamics, such as permutation entropy (PE), have been proposed to analyze EEG series during anesthesia. However, these measures are still single-scale based and may not completely describe the dynamical characteristics of complex EEG series. In this paper, a novel measure combining multiscale PE information, called CMSPE (composite multi-scale permutation entropy), was proposed for quantifying the anesthetic drug effect on EEG recordings during sevoflurane anesthesia. Three sets of simulated EEG series during awake, light and deep anesthesia were used to select the parameters for the multiscale PE analysis: embedding dimension m, lag tau and scales to be integrated into the CMSPE index. Then, the CMSPE index and raw single-scale PE index were applied to EEG recordings from 18 patients who received sevoflurane anesthesia. Pharmacokinetic/pharmacodynamic (PKPD) modeling was used to relate the measured EEG indices and the anesthetic drug concentration. Prediction probability (P(k)) statistics and correlation analysis with the response entropy (RE) index, derived from the spectral entropy (M-entropy module; GE Healthcare, Helsinki, Finland), were investigated to evaluate the effectiveness of the new proposed measure. It was found that raw single-scale PE was blind to subtle transitions between light and deep anesthesia, while the CMSPE index tracked these changes accurately. Around the time of loss of consciousness, CMSPE responded significantly more rapidly than the raw PE, with the absolute slopes of linearly fitted response versus time plots of 0.12 (0.09-0.15) and 0.10 (0.06-0.13), respectively. The prediction probability P(k) of 0.86 (0.85-0.88) and 0.85 (0.80-0.86) for CMSPE and raw PE indicated that the CMSPE index correlated well with the underlying anesthetic effect. The correlation coefficient for the comparison between the CMSPE index and RE index of 0.84 (0.80-0.88) was significantly higher than the raw PE index of 0.75 (0.66-0.84). The results show that the CMSPE outperforms the raw single-scale PE in reflecting the sevoflurane drug effect on the central nervous system.

Complexity of resting-state EEG activity in the patients with early-stage Parkinson’s disease

“…. ; m!Þ to denote the probability distribution of permutation p j , the PE for this scalar series was defined as (Cao et al 2004;Li et al 2010;Yi et al 2014)…”

“…Based on this definition, the relevant normalized PE of the series can be written as (Cao et al 2004;Li et al 2010;Yi et al 2014) PE ¼ H p ðmÞ lnðm!Þ ð 2Þ…”

“…Further, the calculation of each complexity measure was dependent on the embedding dimension m and lag s (Ouyang et al 2010;Cao et al 2004;Keller et al 2014;Li et al 2008Li et al , 2010. In our previous study (Yi et al 2014), we have systematically investigated how the values of dimension m, lag s, and data length L influence PE and OI.…”

“…Earlier studies have shown that this kind of complexity measures is particularly suitable for non-stationary time series (Cao et al 2004;Keller et al 2014). They have been successfully applied to analyze epileptic and anesthetic EEG data (Ouyang et al 2010;Cao et al 2004;Keller et al 2014;Li et al 2007Li et al , 2008Li et al , 2010. In our previous study (Yi et al 2014), we have also used them to detect the complexity changes of EEG activity associated with manual acupuncture at acupoint ST 36 in healthy subjects.…”

Permutation Entropy: New Ideas and Challenges

“…Entropy measures of that type are the fine-grained Permutation entropy proposed by Liu and Wang [7], the weighted Permutation entropy and the robust Permutation entropy introduced by Fadlallah et al [8] and Keller et al [9], respectively, Bian et al [10] have adapted Permutation entropy to the situation of delay vectors with equal components which are relevant when the number of possible values in a time series is small. Other variants are to consider Tsallis or Renyi entropies instead of the Shannon entropy which goes back to the work of Zunino et al [11] and Liang et al [12], or to integrate information of different scales as done by Li et al [13], Ouyang et al [14] and Azami and Ascodero [15]. Unakafov and Keller [16] have proposed the Conditional entropy of (successive) ordinal patterns as a measure which has been shown to perform better than the Permutation entropy itself in many cases.…”