EEG entropy monitoring of depth of anaesthesia: Pharmacokinetic and dynamic modelling

Because of the difficulty in analyzing raw electroencephalographic signal, several electroencephalographic monitors have been developed to aid anaesthetists on their task to maintain adequate anaesthesia. Spectral Entropy is used as a measure of electroencefalographic effects of drugs in human patients, and is a valuable tool to predict depth of anaesthesia. Monitors with implemented entropy algorithms, process the electroencephalogram (EEG) and are in current use at the operating room. In this study we used the EEG collected in rats and applied the Shannon entropy over the signal. The information obtained was used as an indicator of depth of anaesthesia. The main objective was to model the relation between the depth of anaesthesia in rats (entropy) and the propofol infusion rates, with the purpose of obtaining a closed-loop control for propofol infusions. Five adult rats were sedated with isoflurane, cannulated and equipped for the EEG collection. After the preparation, anaesthesia was induced with propofol infusions, using different infusion rates on each rat. The collected EEG (125Hz) was processed using an entropy algorithm developed in MATLAB® 7 that determined the entropy value at each second using the preceding 15s of signal. Pharmacokinetic models were fitted for each rat using bi and tri-compartmental models; the pharmacodynamic phase was also modelled for each rat. The relation between obtained propofol effect-site concentrations and the entropy values was modelled by a Hill Equation. The model obtained for the relation between infusion rates and entropy values was evaluated using the mean absolute deviation (MAD) and the relative mean square error (RMSE) for models comparative analysis.