Tuning of a predictive control scheme for intravenous anesthesia: influence on performance and computational requirements

This paper focuses on a Model Predictive Control (MPC) application for the multivariable control of depth of hypnosis in total intravenous anesthesia. The control system co-administers propofol and remifentanil to regulate the Bispectral Index (BIS). A key feature of the clinical practice is the setting of a ratio between the two drugs, which defines the opioid-hypnotic balance. This ratio must be explicitly controlled by the anesthesiologist to implement a balanced anesthesia technique. However, since propofol and remifentanil have a synergistic effect when they are co-administered, the response of the patient to drug administration changes according to the selected value of the ratio. Consequently, the MPC parameters must be carefully tuned for each specific ratio to ensure performance and robustness. This tuning is typically performed by means of optimization techniques, which are computationally demanding and require several hours to complete. This approach is feasible for a fixed, predefined ratio but becomes impractical when the ratio needs to be adjusted in real-time based on the clinical situation. To solve this problem, this paper proposes a methodology based on precomputed lookup tables and fitting techniques. This solution provides near-optimal tuning parameters for any ratio within a defined clinical range, eliminating the need for real-time optimization. The performance and robustness of the proposed method are evaluated through a simulation study, showing results that are consistent with clinical requirements. The proposed approach represents a step towards the deployment of flexible MPC-based control solutions in the clinical practice.