Towards the control of depth of anaesthesia: Identification of patient variability

Depth of anaesthesia (DOA) is usually assessed through the Bispectral Index (BIS) and State Entropy (SE), which derived EEG signals. Studying the effect of drug interaction on these signals is of great importance for the development of a suitable drug infusion system designed to control DOA. In this paper, two renowned pharmacokinetic (PK) models for the anaesthetic drug propofol are considered, and their influence on the fitting and prediction abilities of a drug interaction model for BIS and SE is assessed. This interaction model is fitted to the individual patient data during anaesthesia induction and tested for prediction during surgery. Two identification methods are considered for the fitting purpose: a hybrid method and a nonlinear least squares curve-fitting algorithm. The results obtained for 7 patients show that the choice of the PK model has influence on the overall performance of the interaction model; in particular, only one PK model leads to good results in the prediction phase. The choice of the identification method is equally important, being the hybrid method the better suited. The successful identification of patient variability here obtained is a key step towards the control of DOA.