Model predictive control for automated anesthesia

The main challenge of anesthesia is the maintenance of the desired sedation level before, during, and after the induction. We developed an in-silico model-based control loop for the administration of an analgesic opioid, remifentanil, and of an anesthetic, propofol. The patients' response (i.e. the real process) is predicted in-silico by a physiologically-based pharmacokinetic model, conjugated with suitable pharmacodynamic models that simulate the dynamic response of heart rate, arterial pressure, and bispectral index. We simulated the induction phase of anesthesia and obtained a fast and safe patient's response by setting the arterial pressure and bispectral index as controlled variables, and implementing proper bounds on both the plasma concentration and the controlled variables.