Synthetic Patient Database of Drug Effect in General Anesthesia for Evaluation of Estimation and Control Algorithms

This paper describes a database of synthetic patients for the use in estimation and control design in closed-loop anesthesia. The synthetic patients are represented by pharmacokinetic-pharmacodynamic (PKPD) Wiener models for the Depth of Anesthesia estimated from clinical data. The input of the Wiener model is given by the flow rates of propofol and remifentanil while the output is the bispectral index. A positive stable realization of the Wiener model describing the system dynamics is adopted to ensure a biologically feasible behavior of the PKPD system. Both time-varying and time-invariant versions of the models are available. An Extended Kalman filter (EKF) is applied to the clinical data to estimate the patient-dependent parameters of the Wiener model. The time-invariant version of a model is obtained by averaging of the time-varying estimates produced by the EKF. The performance of the Wiener model with estimated parameters is assessed and discussed. To illustrate the utility of the database, a PID controller is evaluated over the synthetic patient cohort.