DAMon

The team

The project facilitated collaboration among researchers from three French laboratories and one in Spain, as well as with anesthesiologists at the CHU Grenoble Alpes.  

Bob Aubouin–Pairault, PhD student at Gipsa-lab, University of Grenoble Alpes, Grenoble 
Daniel Denardi Huff, postdoctoral researcher at Gipsa-lab, University of Grenoble Alpes, Grenoble

Benjamin Meyer, intern at the Anesthesia and Critical Care Medicine Unit at CHU Grenoble Alpes
Mirko Fiacchini, Senior Researcher (CR) at the CNRS, Gipsa-lab, University of Grenoble Alpes, Grenoble 
Thao Dang, CNRS Researcher, Verimag, University of Grenoble Alpes, Grenoble
Kouther Moussa, associate professor at INSA Hauts-de-France, LAMIH, Valenciennes
Remi Wolf, University of Grenoble Alpes, Department of Anesthesia and Critical Care, Grenoble
Mazen Alamir, DR CNRS, Gipsa-lab, University of Grenoble Alpes, Grenoble 
Teodoro Alamo, Full Professor, University of Seville, Spain

In particular, the PERSYVAL-lab project funded the positions of B. Aubouin–Pairault and D.D. Huff.  

 

The objective

General anesthesia plays a fundamental role in providing surgeons with adequate conditions for surgery and in preventing discomfort or pain for the patient, while minimizing the negative postoperative effects of anesthesia. In medical practice, anesthesia involves monitoring and controlling the progression of areflexia (lack of movement), analgesia (lack of pain), and hypnosis (lack of consciousness) in the patient. Based on various physiological signals, such as the Bispectral Index (BIS), the electroencephalogram (EEG), and indicators of pain and neuromuscular blockade, the anesthesiologist adjusts the infusion rates of different drugs to achieve and maintain appropriate levels of anesthesia. In addition to controlling the patient’s level of sedation, the anesthesiologist is responsible for monitoring the hemodynamic status, measured by mean arterial pressure (MAP) and cardiac output (CO), since the cardiovascular system interacts significantly with the multi-drug anesthesia process. 

The primary objective of anesthesia is to maintain the desired level of hypnosis, areflexia, and analgesia to facilitate the surgeon’s tasks by avoiding both drug overdosing and underdosing and their potentially extremely severe consequences for the patient. To achieve this goal, automatic feedback control theory, formal verification, and machine learning can be of great help not only in increasing control efficiency and monitoring reliability but also in ensuring that anesthesiologists remain vigilant regarding potential critical events. Several sources of complexity, however, contribute to making the problem of monitoring, predicting, and controlling the anesthesia process extremely challenging. Although some studies have emerged proposing the application of automatic control to the anesthesia process, several key issues warrant further investigation. 

This research project aims to exploit the potential of advanced control theory, formal verification, and machine learning techniques to design and implement optimization and computation-oriented methods—
—to assist the anesthesiologist during surgery. The main challenges to be addressed are the high uncertainty affecting the system dynamics, its variability over time and from patient to patient, the high risk sensitivity of the application, the necessity of accurate validation and certification of the proposed solution, and the often partial information available regarding the evolution of such a complex process. Access to real surgical operation data plays a central role in developing, applying, and validating the proposed techniques, the real-time embedded implementation of which is the ultimate objective. The computational and implementation constraints inherent in the cyber-physical nature of the proposed anesthesiologist assistance are carefully considered in the theoretical developments.

 

Work methods or means

The following is a brief overview of the theoretical and computational tools used to address the challenges of monitoring, controlling, and predicting a patient’s physiological state during the surgical anesthesia process.

- Machine learning: In recent years, scientific research has witnessed a dramatic surge in the popularity of artificial intelligence methods, which can be summarized as a set of computational techniques capable of simulating the learning processes characteristic of living beings on digital devices. The widespread availability of large datasets and the increasing computational power of modern devices, combined with their proven efficiency, have substantially contributed to making these techniques accessible and promoting their application in many industrial and scientific contexts. Machine learning methods are a class of techniques whose objective is to infer properties of data as a result of solutions to optimization problems, in particular, the estimation of unknown functional relationships between certain data inputs and the corresponding outputs—i.e., the regression problem—or a criterion for distinguishing between data points—the classification problem.

- Moving horizon observers: The theory of observability for nonlinear dynamical systems has attracted the attention of researchers for several decades, and the proposed solutions have evolved over time. A popular approach, an alternative to classical observer methods, has recently emerged aimed at the practical application of observers to real systems and data-driven processes, defining the moving horizon observer paradigm. This family of approaches aims primarily to achieve the fundamental estimation objective of reducing the observation error as the result of solving an online optimization problem. Among the main benefits of moving horizon observers are a much smaller reliance on structural assumptions about the system and a more direct ability to handle parameter uncertainties, partial lack of knowledge regarding the dynamics, the presence of noise, and constraints. At the cost of a more significant computational burden during online processing, moving horizon observers—which can be viewed as the observation dual of model predictive controllers—provide a rather effective and more data-driven approach to estimating the state in real-world systems.

- Model Predictive Control: Model Predictive Control is a control technique whose popularity stems primarily from its ability to handle constraints and ensure performance optimization, as well as its suitability for practical application, while simultaneously guaranteeing desirable stability properties. The inherent goal of MPC in real-world control implementation has led to a focus on the effects of model uncertainties, disturbances, and noise on control performance and stability, resulting in robust and stochastic formulations of MPC in addition to the deterministic one. Recently, in order to reduce conservatism by exploiting the statistical structure of the uncertainty, stochastic MPC has attracted the attention of researchers. In this framework, the stochastic features of the predicted state evolution can be taken into account, and hard constraints can be relaxed in terms of satisfying chance constraints.

 

Results

The DAMon project has addressed a number of scientific challenges:

- Modeling: using the available datasets together with control theory and machine learning tools, we have been working to propose new and more accurate models, in particular: 
    - data-driven machine learning models
    - online joint state estimation and parameter identification

- Automatic drug control: we have been proposing new closed-loop methods for administering drugs during general anesthesia and improving performance compared to the current state of the art

- Critical event detection and prediction: machine learning and data-driven methods have been applied to predict hypotension 

- CHUGA data annotation: the challenge of processing the database to identify relevant alarms for use in alert generators for surgery has been addressed.

 

Leverage effect

The project provided an opportunity to establish scientific connections and collaboration with two international groups that are leaders in research on the control of anesthesia dynamics, specifically with the team of Prof. C. Ionescu at Ghent University in Belgium and Prof. Visioli at the University of Brescia in Italy. In this context, B. Aubouin-Pairault spent one month in Ghent. 

The collaboration led to a joint workshop on the topic in Grenoble in December 2024, the joint publication of a state-of-the-art simulator of anesthesia dynamics, and the associated journal article, currently under review.

 

The 5 main publications

B. Aubouin-Pairault, M. Fiacchini, T. Dang, (2024) Online identification of pharmacodynamic parameters for closed-loop anesthesia with model predictive control, Computers & Chemical Engineering, 191, 108837.

B. Aubouin-Pairault, M. Fiacchini, T. Dang, (2024) Data-based modeling of the pharmacodynamics for the effect of propofol and remifentanil during general anesthesia, Biomedical Signal Processing and Control, 98, 106728.

B. Aubouin-Pairault, M. Fiacchini, T. Dang, (2024) Comparison of multiple Kalman filters and moving horizon estimators for the anesthesia process, Journal of Process Control, 136, 103179.

D. D. Huff, M. Fiacchini, T. Dang, T. Alamo, (2024) Optimized coadministration of propofol and remifentanil during the induction phase of total intravenous anesthesia with statistical validation, IEEE Control Systems Letters, 8, 193-198.

K. Moussa, B. Aubouin-Pairault, M. Alamir, and T. Dang, "Data-Based Extended Moving Horizon Estimation for MISO Anesthesia Dynamics," in IEEE Control Systems Letters, vol. 7, pp. 3054–3059, 2023