Abstract — Manual or open-loop administration does not take into account patient’s individual dose or dose-response relationship; hence they represent sub-optimal solutions for optimizing individual drug titration. This may lead to under- or over- sedation, increasing the time on mechanical ventilation, the length of intensive care unit (ICU) stay and mortality. Model based predictive control can mitigate with this problem, improving the efficiency of drug delivery and patient safety. A multiple-input single-output (MISO) patient model is identified and validated in this paper. The inputs are two drugs commonly used for general anesthesia, propofol and remifentanil, and the output is the Bispectral Index (BIS). Wavelet time-frequency analysis was used to filter the measured signals. The parameters of the interaction model which relates the effect-site concentrations of these drugs to BIS are identified based on least-squares algorithm, using data from real-life clinical tests. Keywords — Anesthesia, modeling, Wavelet, filtering. I. INTRODUCTION ENERAL anesthesia plays an important role in surgery and intensive care unit (ICU) and requires critical assessment of induced quantities of drugs into the patient. There are three major interactive parts in anesthesia: sedation, analgesia and neuromuscular blockade. Nowadays, drug dosing control during anesthesia is changing from manual control to automated control. Some of the advantages of automated drug dosing control are: patient safety, a significant decrease in overall costs and improved healthcare. This topic captured the attention of engineers and clinicians already some time ago, starting with expert systems that offer advice to the anesthetist upon optimal drug infusion rate during clinical trials. For many control techniques, compartmental models are used to represent the drug distribution in the body for patients undergoing anesthesia. Single-input single-output (SISO) patient models for control already exist in the literature for propofol [1], as well as for neuromuscular blockade agents [2], while the analgesic effects remain one of the most difficult problems to identify. Analgesia is a very challenging aspect of general anesthesia and requires special attention, since its effects are dependent on the The paper was supported by the project "Improvement of the doctoral studies quality in engineering science for development of the knowledge based society-QDOC” contract no. POSDRU/107/1.5/S/78534, project co-funded by the European Social Fund through the Sectorial Operational Program Human Resources 2007-2013 R. Hodrea, C. Darab and I. Nascu are with the Technical University of Cluj-Napoca, Automation Department, Observatorului 2, Cluj-Napoca, Romania (corresponding author; e-mail address: ramona.hodrea@aut.utcluj.ro). drug used in the patient. Some previous reports from literature consider remifentanil as a suitable input for inducing analgesia into the patient. This paper presents identification results of a multiple- input single-output (MISO) patient model for sedation and analgesia components used in ICU. The final purpose is to use this model for prediction in a model based predictive control strategy. The structure of the model is presented in section 2, followed by signal filtering depicted in section 3. The identification results are discussed in section 4 and the conclusions and further work are summarized in a final section. II. PATIENT MODEL Several drugs and ways of administration (intravenously, by inhalation, etc.) can be used to provide sedation and analgesia. Nowadays, the most clinically used drugs are propofol and remifentanil, due to their beneficial pharmacological profile. These two drugs are the inputs of the model proposed in the paper and the output is the Bispectral Index (BIS), a signal derived from the electroencephalogram (EEG). Using EEG, several derived, computerized parameters have been tested and validated as a promising measure of the hypnotic component of anesthesia [3]. BIS combines features extracted from EEG including higher order spectra of the signal which can reveal phase coupling of single waveforms. Multivariate statistics were used to combine the different features into a single indicator value. BIS values lie in the range of 0 (equivalent to EEG silence) – 100 (equivalent to fully awake patient). A generic BIS value between 40 and 60 indicates an appropriate level for general anesthesia. For patient safety and fast recovery time, BIS should not decrease below 30. The general block diagram of the MISO patient model is depicted Fig. 1. PK – PD Propofol model PK – PD Remifentanil model Propofol delivery rate Remifentanil delivery rate Ce Propofol CeRemifentanil Interaction model BIS Fig. 1. Block diagram of the patient model In the figure above the pharmacokinetic- pharmacodynamic (PK-PD) blocks denote compartmental models used to represent the distribution of drugs in the body, i.e. mass balance. These models rely on a conservation principles applied to the exchange of chemicals among coupled macroscopic systems called Modeling of Drug Delivery in General Anesthesia Ramona Hodrea, Cosmin Darab, Ioan Nascu G 20th Telecommunications forum TELFOR 2012 Serbia, Belgrade, November 20-22, 2012. 978-1-4673-2984-2/12/$31.00 ©2012 IEEE 891