Advanced model-based control studies for the induction and maintenance of intravenous anaesthesia.

This paper describes strategies toward model-based automation of intravenous anaesthesia employing advanced control techniques. In particular, based on a detailed compartmental mathematical model featuring pharmacokinetic and pharmacodynamics information, two alternative model predictive control strategies are presented: a model predictive control strategy, based on online optimization, the extended predictive self-adaptive control and a multiparametric control strategy based on offline optimization, the multiparametric model predictive control. The multiparametric features to account for the effect of nonlinearity and the impact of estimation are also described. The control strategies are tested on a set of 12 virtually generated patient models for the regulation of the depth of anaesthesia by means of the bispectral index (BIS) using Propofol as the administrated anaesthetic. The simulations show fast response, suitability of dose, and robustness to induce and maintain the desired BIS setpoint.

Modeling and analysis of individualized pharmacokinetics and pharmacodynamics for volatile anesthesia.

The presented procedure aims to establish an in-depth understanding of a derived mathematical model for drug distribution, pharmacokinetics, and drug effect, pharmacodynamics, during volatile anesthesia. A physiologically based, patient-specific model is derived, where the pharmacokinetic (PK) part consists of multiple blood and tissue compartmental models, each adjusted to the weight, height, gender, and age of the patient. The pharmacodynamic (PD) part is described by an effect site compartment and the Hill equation both linking the hypnotic effect measured by the Bispectral Index (BIS) to the arterial anesthetic concentration. Via a global sensitivity analysis the patient-specific PK and PD variables and parameters are analyzed regarding their influence on the measurable outputs, which are the end-tidal concentration of the volatile anesthetic and the BIS. Via this analysis, the uncertainty introduced by PD variability is identified to be more significant than the uncertainty introduced by PK variability. A case study of isoflurane-based anesthesia shows that the simulation results of the individualized PK variables are in good accordance with the measured end-tidal concentration. However, the PD parameters need to be estimated online to predict the hypnotic depth, measured by the BIS, correctly. As a result of this study, the aim should be to focus on the individual identification of the PD parameters before and during anesthesia with future application in safe and robust model predictive control.