Volume 5 Supplement 1
Assessment of sedation level and EEG recovery after major operation by spectral entropy
© The Author(s) 2001
Received: 15 January 2001
Published: 2 March 2001
Entropy quantifies the amount of disorder in a system and characterizes chaotic behaviour. The complexity of a signal can be characterized by spectral entropy, which gives the amount of disorder in frequency space . If an EEG signal includes a wide spectrum of frequencies, its spectral entropy has a high value (near one), and in case of few relevant frequencies spectral entropy is low (near zero). Spectral entropy has been shown to be an effective tool in measuring depth of anaesthesia .
In this study, we investigated whether spectral entropy can distinguish between the different sedation levels corresponding to the Ramsay Scores 2, 4, and 6. In order to study spectral entropy during different sedation levels, EEG was recorded from 26 patients scheduled for an elective cardiopulmonary bypass operation with propofol/alfentanil/isoflurane/pancuronium anaesthesia . Postoperative sedation was maintained with propofol to keep the sedation level at Ramsay Score 6 (not responding to any commands) until the patients were hemodynamically stable. EEG was recorded 5 times for each patient: 1 day before the operation (Ramsay Score 2), after premedication 1 hour before the operation (Ramsay Score 2 or 3), immediately after the operation (Ramsay Score 6), after the patient had opened his eyes for the first time (Ramsay Score 4), and the following morning (Ramsay Score 2 or 3). The EEG signal was recorded bipolarly between electrodes Fz-M1, Cz-M2, C3-P3 and C4-P4. It was amplified and digitized continuously at 100 Hz using the Datex-Ohmeda EEG module and stored to a PC for off-line analysis. Spectral entropy values were evaluated for 5 s epochs in two frequency bands: 0.5-32 Hz and 7-32 Hz. Epochs including artefacts were removed from the data before the calculation.
Spectral entropy for the range 0.5-32 Hz differentiated statistically significantly whether the patient was awake (Score 2) or asleep (Scores 4-6) (P < 0.05). Spectral entropy for 7-32 Hz was able to differentiate the sedation levels 4 and 6 (P < 0.001). Sedation levels 2, 4, and 6 could thus be distinguished by using spectral entropy. For comparison, we analyzed whether spectral edge frequency or auditory evoked potentials can distinguish between these levels. These methods failed in separating levels 6 and 4. There was considerable variation in spectral entropy values between the patients having the same Ramsay Score. This may be due to the physiological variation of different EEG-patterns between individuals.
We divided the patients into two groups according to how the EEG, measured the following morning after the operation, was recovered compared to the EEG 1 day before the operation. In both recordings the patients were awake. The spectral entropy values 1 day after the operation were significantly lower in the group in which the EEG was not at all recovered compared to the group in which the EEG was almost recovered.
Our results indicate that spectral entropy can be a useful tool for assessment of the sedation level of a patient. The performance of spectral entropy in distinguishing Ramsay Score levels 2, 4, and 6 was superior in comparison to spectral edge frequency and auditory evoked potentials. However, in patients with postoperative EEG significantly slower compared to preoperative EEG, spectral entropy remained at a low level and was not able to indicate whether the patient had waken up. A relation between postoperative EEG slowing and mild subclinical cerebral injuries has been discussed by Vanninen et al . Our results suggest that spectral entropy might provide diagnostic information of such a state.
- Rezek IA, et al: IEEE Trans Biomed Eng. 1998, 45: 1186-10.1109/10.709563.PubMedView ArticleGoogle Scholar
- Viertiö-Oja H, et al: Anesthesiology. 2000, 93: A1369-View ArticleGoogle Scholar
- Jakob S, et al: Comp Methods Programs Biomed. 2000, 63: 161-10.1016/S0169-2607(00)00106-1.View ArticleGoogle Scholar
- Vanninen R, et al: Arch Neurol. 1998, 55: 618-10.1001/archneur.55.5.618.PubMedView ArticleGoogle Scholar