Multi-Input Multi-Output Model Predictive Control for Drugs Delivery in Clinical Anesthesia

Automatic drugs delivery during anesthesia can lead to faster sedation, increase patient’s safety, and reduce the burden on anesthesiologists. This preliminary study proposes a Model Predictive Control (MPC) algorithm to control the induction and maintenance of the states of hypnosis, analgesia and neuromuscular blockade by means of automatic infusion of three drugs: Propofol, Remifentanil, and Atracurium. The study is conducted in a simulated environment based on a non- linear model describing the patient physiology that includes synergic effects among drugs and interpatient variability, as well as measurement noise and a variable surgical stimulation disturbance. Simulations on 24 virtual subjects show fast and safe induction of anesthesia. In terms of median [5th , 95th ] percentiles, the MPC algorithm maintained hypnosis in a suitable range for 98.9 [93.7, 99.3]% of the time, analgesia for 100 [65.4, 100]% of the time, and neuromuscular blockade for 100 [100, 100]% of the time.