1. Permutation entropy of the electroencephalogram: a measure of anaesthetic drug effect 
E. Olofsen, J.W. Sleigh, A. Dahan 
British Journal of Anaesthesia 
Volume 101, Issue 6, Pages (December 2008)
DOI: /bja/aen290
Copyright © 2008 British Journal of Anaesthesia Terms and Conditions

2. Fig 1
Extraction of ordinal patterns from the EEG signal. As the algorithm moves sequentially through the EEG signal, the sections (‘motifs’ consisting of three data points’ length) are classified as one of the six possible patterns, depicted in the second row. A histogram of the relative numbers of each motif in the signal is shown on the top row of the diagram. The paler dashed-line motif is a demonstration of the operation of the τ=2 lag.
British Journal of Anaesthesia  , DOI: ( /bja/aen290)
Copyright © 2008 British Journal of Anaesthesia Terms and Conditions

3. Fig 2
The frequency dependence of PEs with different lags (PEτ=1 and PEτ=2) (a), and response to added white noise (b), or a second sine-wave oscillation (c). Because the signals were artificially generated, no noise threshold was used in the calculation of the PE. PE, permutation entropy; τ, lag (see text for explanation); f2, second frequency (12 Hz). The figures show the results of 1000 simulation runs.
British Journal of Anaesthesia  , DOI: ( /bja/aen290)
Copyright © 2008 British Journal of Anaesthesia Terms and Conditions

4. Fig 3
An example of how the changes in EEG frequency content relate to different raw PEs and EEG permutation index (CPEI) during an inhalation induction of anaesthesia with sevoflurane. The top diagram (a) shows the spectrogram. (The maximum spectrogram frequency=30 Hz for display purposes only.) Dark red indicates high spectral power and blue low power. The spectra are computed on 1 s segments of data. The dashed line shows the changes in the end-tidal sevoflurane concentration (scaled by 10 for illustrative purposes) superimposed on the spectrogram. The lower diagram (b) shows how the different PE indices reflect the changes in EEG spectral power. The text annotations are descriptions of the dominant EEG pattern at that time. The double arrow demonstrates the ‘gap’ between the PEτ=1 and the PEτ=2, indicative of an EEG rich in spindle-like activity [see corresponding region in the spectrogram (a)]. ApEn, approximate entropy/2, m=2, lag=1. The ApEn value is halved to simplify the display. Because of the disturbance of blink artifacts, it was not possible to reasonably fit a PK–PD model using the ApEn parameter. The CPEI parameter was calculated as described in the text. Both PEs and CPEI were calculated including the noise threshold (tie <0.5 μV). SE is the state entropy index obtained from the output of the commercial M-Entropy system. For display purposes, it is scaled (0.5+SE/200).
British Journal of Anaesthesia  , DOI: ( /bja/aen290)
Copyright © 2008 British Journal of Anaesthesia Terms and Conditions

5. Fig 4
Detailed segments of EEG data from Figure 3. Raw EEG signals are shown on the left, their power spectra (√amplitude2) in the middle, and their permutation motif histograms on the right (see text for explanation). Note the variations in y-axis scaling for illustrative purposes. In the spindle example, the power spectrum show peaks at about 13 Hz superimposed on about 2 Hz low-frequency activity. The white sections of the bar graphs depict the difference between the PEτ=1 and the PEτ=2 histograms. They are most marked in the spindle EEG example, where long ‘slope’ sequences (motifs #1 and #5) present in the PEτ=1 statistic, are less frequent in the PEτ=2 histograms (i.e. the white tops of the motifs #1 and #5 bars are removed to fill the white sections of the other motif bars, giving a flatter frequency distribution and hence a higher PE value). This is because the pattern matched by the PEτ=2 statistic is almost double the length that of the PEτ=1 statistic, and hence the PEτ=2 statistic quickly traverses the short slopes of the spindles, without recording prolonged series of ‘slope’ motifs. In contrast, the very large amplitude of the delta EEG pattern means that long runs of ‘slope’ motifs are achieved by both the PEτ=1 and the PEτ=2 statistics (the motif distribution is dominated by #1 and #5). In the delta EEG pattern, the values of PEτ=1 and PEτ=2 are very similar (e.g. about 7 min).
British Journal of Anaesthesia  , DOI: ( /bja/aen290)
Copyright © 2008 British Journal of Anaesthesia Terms and Conditions

6. Fig 5
Examples of PK–PD modelling of sevoflurane effects. The dots are the experimental CPEI (left) or M-entropy (right) data points. The lines through the data are the model fits.
British Journal of Anaesthesia  , DOI: ( /bja/aen290)
Copyright © 2008 British Journal of Anaesthesia Terms and Conditions

7. Fig 6
Examples of PK–PD modelling of propofol effects. The dots are the experimental CPEI (left) or BIS (right) data points. The lines through the data are the model fits.
British Journal of Anaesthesia  , DOI: ( /bja/aen290)
Copyright © 2008 British Journal of Anaesthesia Terms and Conditions

8. British Journal of Anaesthesia 2008 101, 810-821DOI: (10
British Journal of Anaesthesia  , DOI: ( /bja/aen290)
Copyright © 2008 British Journal of Anaesthesia Terms and Conditions