Event-based GPC for depth of hypnosis in anesthesia for efficient use of propofol

Abstract

This work presents a simulation study of an event-based predictive control system for depth of hypnosis in anesthesia using bispectral index as a controlled variable. The developed control structure uses a Wiener model structure to exploit the linear model predictive approach. Due to this architecture it is possible to use a well-established model predictive controller for linear system taking advantage of constraints handling mechanism and keeping the computational effort in reasonable limits. In such a scheme, the predictive controller is implemented within adjustable virtual deadband on actuator to limit changes in control signal and preserve the control system resources. The presence of the virtual deadband permits to establish the tradeoff between control performance and the use of the control resources (propofol administration). This feature could reduce the risk of drug overdosis during the anesthesia reducing negative effects on patients health with postoperative delirium. The analyzed control system is evaluated for different values of the actuator deadband in order to test its influence on the controlled variable. Additionally, a comparison with a standard time-based predictive controller is performed.