MPC for Propofol Anesthesia: the Noise Issue

The design of automatic control systems for general anesthesia is a challenging task due to the severe safety requirements and process constraints. This is even more complex when model-based control techniques are used due to the significant variability of the process model. Additionally, issues like noisy measurements and interference also influence the control system overall performance. In this context, adequate filtering and control system sampling period selection should be analyzed to test their influence on the controller. In this paper, an MPC system for the depth of hypnosis, where the BIS signal is used as a controlled variable, is analyzed. The main purpose is to test and evaluate how the process noise affects the performance of the control system. The analysis is performed in a simulation study using a dataset of virtual patients representative of a wide population. Results show that a satisfactory performance is obtained when the noise is explicitly taken into account in the controller tuning procedure for a specific sampling period.