IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 9, SEPTEMBER 2008 2161 Robust Predictive Control Strategy Applied for Propofol Dosing Using BIS as a Controlled Variable During Anesthesia Clara M. Ionescu ∗ , Student Member, IEEE, Robin De Keyser, Bismark Claure Torrico, Tom De Smet, Michel MRF Struys, and Julio E. Normey-Rico Abstract—This paper presents the application of predictive con- trol to drug dosing during anesthesia in patients undergoing surgery. The performance of a generic predictive control strategy in drug dosing control, with a previously reported anesthesia-specific control algorithm, has been evaluated. The robustness properties of the predictive controller are evaluated with respect to inter- and intrapatient variability. A single-input (propofol) single-output (bispectral index, BIS) model of the patient has been assumed for prediction as well as for simulation. A set of 12 patient models were studied and interpatient variability and disturbances are used to assess robustness of the controller. Furthermore, the controller guarantees the stability in a desired range. The applicability of the predictive controller in a real-life environment via simulation studies has been assessed. Index Terms—Anesthesia, constraints, drug dosing control, model-based predictive control, nonlinear model, robustness. I. INTRODUCTION D URING the last decade, the control technology has suc- cessfully influenced modern medicine through robotic surgery, electrophysiological system life support, and image- guided therapy and surgery [1]. Another area of medicine suited for applications of control is clinical pharmacology in general, and a particular case is the anesthesia and critical care unit medicine. Within this particular group of applications, monitor- ing and controlling the depth of anesthesia for patients during surgery offers interesting challenges to the control engineer [2]. This topic captured the attention of engineers and clinicians already decades ago [3], starting with expert systems that offer advice to the anesthesiologist upon optimal drug infusion rate during clinical trials [4]. It soon became clear that control of Manuscript received October 15, 2007; revised February 4, 2008. Asterisk indicates corresponding author. ∗ C. Ionescu is with the Department of Electrical Energy, Systems and Automation, Ghent University, B9000 Ghent, Belgium (e-mail: clara@ autoctrl.UGent.be). R. De Keyser is with the Department of Electrical Energy, Systems and Automation, Ghent University, B9000 Ghent, Belgium (e-mail: rdk@ autoctrl.UGent.be). B. Claure Torrico and J. E. Normey-Rico are with the Department of Au- tomation and Systems, Federal University of Santa Catarina, Santa Catarina CEP 88040-970, Brazil (e-mail: bismark@das.ufsc.br; julio@das.ufsc.br). T. De Smet is with DEMED Engineering, B-9140 Temse, Belgium (e-mail: tom.desmet@demed.be). M. MRF Struys is with the Department of Anesthesia, University Medical Center Groningen, 9713 Groningen, The Netherlands and also with the De- partment of Anesthesiology, Ghent University, B9000 Ghent, Belgium (e-mail: michel.struys@ugent.be). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TBME.2008.923142 anesthesia poses a manifold of challenges, with multivariable characteristics [5], different dynamics depending on anesthetics substances [6], [7], and stability problems [8]. Further investiga- tions proved propofol to be an anesthetic tackled well in control problems for hypnosis [9], [10], while recent studies showed that the control performance may also depend on the controlled variable [11], [12]. This paper presents a simulation study for the control of ad- ministration of propofol using the BIS as controlled variable. Propofol is an intravenous hypnotic providing a rapid onset time and relatively short duration of action, while the BIS (As- pect Medical Systems, Norwood, USA) is a commercially avail- able measure of the effects of anesthetics and sedatives on the brain based on a bispectral analysis of the patient’s EEG. It is important to realize that the BIS, like other measures such as midlatency auditory evoked potentials or entropy analysis of the EEG, is a surrogate measure for the depth of anesthesia as there is no direct measure available. Therefore, BIS will only be one of multiple indicators used by the anesthetist to safeguard the patient’s wellbeing during an anesthetic procedure. In spite of these limitations, automated feedback control of propofol ad- ministration using the BIS as a controlled variable can serve as an automated pilot during long periods of surgery requir- ing a stable anesthetic, preserving the anesthetist’s vigilance for critical events. The nonlinear response profile and inter- and in- trapatient variation of the patient’s hypnotic state to infusion of propofol should be handled by a robust controller. From a clin- ical point of view, an ideal controller would guide the induction of anesthesia in order to reach the target as fast as possible with- out initial overshoot. Afterwards, the controller would simply maintain the desired target as well as possible. Therefore, from control engineering viewpoint, model predictive control (MPC) plays a crucial role in solving such complex problems. The original objective of the paper is threefold: 1) to compare the performance of a generic predictive control strategy, which is applied in drug dosing control, with a previously reported con- trol algorithm specifically developed for anesthesia; 2) to evalu- ate via extensive simulation studies the robustness properties of the predictive controller with respect to inter- and intrapatient variability; and 3) to assess the applicability of the predictive anesthesia controller in a real-life clinical environment. In this contribution, the predictive control strategy extended predic- tion self-adaptive control (EPSAC) [13] is compared to a model adaptive controller for the control of depth of anesthesia. An overview of the models used for prediction and for control is 0018-9294/$25.00 © 2008 IEEE