(96e) Adaptive Model Predictive Control for Automatic Regulation of Anesthesia by Simultaneous Administration of Two Intravenous Drugs

General anesthesia consists of producing loss of consciousness (hypnosis), inhibition of noxious stimuli reaching the brain (analgesia) and providing sufficient skeletal muscle relaxation. Anesthetists maintain these patient states within acceptable ranges by infusing suitable anesthetic drugs, analgesics, neuromuscular blocking agents and/or intravenous fluids. To analyze the state of the patient, the anesthesiologist must monitor and regulate a wide range of physiological states, such as Bispectral Index (BIS), Mean Arterial Pressure (MAP), Blood Pressure (BP), Cardiac Output (CO), endtidal carbon dioxide and oxygen levels, blood acidity, fluid levels, heart contractility, renal function, and more. Because there are no direct measurements available for hypnosis and analgesia, BIS and MAP are respectively taken as indirect measurements.

It is desirable to have an automated system that takes into account the physiological status of the patient and routinely deliver the right amount of drugs into the body to ensure patient's safety [1]. Such a strategy together with appropriate constraints can avoid over-dosing and under-dosing of the drug, suppress intra- and inter-patient variability, compensate for differences in surgical procedures and anesthetic regimes and importantly reduce post-operative side effects. Further, it relieves anesthetists for more important tasks during the surgery. When administering several drugs into the patient's body, interactions between the drugs play a critical role in maintaining the anesthetic state. The majority of the research in this area has mainly dealt with the automatic delivery of one drug and manual administration of the other drugs. The multivariable fuzzy control strategy (employed in [2]) has considered automatic regulation of anesthesia using two drugs [2]. Another work [3] considered adaptive predictive control strategy for the automatic regulation of one drug and the manual administration of the other drug [3]. The controller design in both the above works took into account the interaction between the drugs. However, the issue of simultaneously infusing two interacting drugs has not been fully resolved the literature. The dilemma of whether to vary hypnotic drug dosage or analgesic drug dosage still exists.

The objective of our work is to determine the best infusion rates of the hypnotic and analgesic drugs such that the patient's anesthetic state is well regulated and the side effects (due to over-dosage) are minimized. Manual administration of neuromuscular blocking drug is assumed for the skeletal muscle relaxation. A Model Predictive Control (MPC) strategy that incorporates a Pharmacokinetic (PK) model, a Pharmacodynamic (PD) model and a drug interaction model is devised for the simultaneous administration of the hypnotic and analgesic drugs. The infusion rates of both drugs are determined according to the hypnosis level and the surgical stimulus leading to a satisfactory regulation of the patient's anesthetic state. To account for inter- and intra- patient variability, an adaptation mechanism that updates the internal model of the controller, based on the variability between the predicted drug concentrations and the measured drug concentrations in the patient is developed. This is an extension of the work done in [4] for a single drug. This is incorporated into the MPC with State Estimation (MPCSE) algorithm that updates selected model parameters through a Kalman filter at each time step [5]. Several studies in the literature show that, inter-patient variability in the PD model is significantly higher than that in the PK model [6]. Hence, only the parameters of the PD model are updated in the designed MPCSE algorithm. Several simulation studies are conducted to test the performance of the proposed MPC. The results show that there is a considerable decrease in the amount of drugs infused when compared to automatic control of a single drug only.

References:

1.Struys M.M.R.F., Mortier E.P., Smet T.D., ?Closed loops in anesthesia?, Best Practice & Research Clinical Aneasthesiology?, 2005, 20(1):211-220.

2.Zhang X.S., Roy R.J., Huang J.W., ?Closed-loop system for total intravenous anesthesia by simultaneously administering two anesthetic drugs?, Proc. of the 20th annual international conference of the IEEE Engineering in Medicine and Biology Society, 1998, 20(6):3052-3055.

3.Mendonca T., Nunes C.S., Magalhaes H., Lemos J.M., Amorim P., ?Predictive adaptive control of unconsciousness - Exploiting remifentanil as an accessible disturbance?, Proc. of the IEEE international conference on control applications, 2006, 205-210.

4.Sartori V., Schumacher P.M., Bouillon T., Luginbuehl M., Morari M., ?On-line estimation of Propofol pharmacodynamic parameters?, Proc. of the IEEE Engineering in Medicine and Biology Society, 2005.

5.Parker R.S., Doyle III F.J., Peppas N.A., ?A model-based algorithm for blood glucose control in type I diabetis patients?, IEEE Transactions on Biomedical Engineering, 1999, 46(2): 148-157.

6.Furutani E., Sawaguchi Y., Shirakami G., Araki M., Fukuda K., ?A hypnosis control system using a model predictive controller with online identification of individual parameters?, Proc. of the IEEE international conference on control applications, 2005, 154-159.