“Dual Estimation of States and Disturbances in Anesthesia and Hemodynamic Systems" DYNAMICAL SYSTEMS AND CONTROL RESEARCH GROUP Bouchra Khoumeri Promotor: Dr. Dana Copot, Prof. Clara M. Ionescu Contact
[email protected] Universiteit Gent @ugent Ghent University [1] Bernard, P., Andrieu, V., & Astolfi, D. (2022). Observer design for continuous-time dynamical systems. Annual Reviews in Control, 53, 224-248 [2] Ionescu, C. M., De Keyser, R., Copot, D., Yumuk, E., Ynineb, A., Othman, G. B., & Neckebroek, M. (2024). Model extraction from clinical data subject to large uncertainties and poor identifiability. IEEE Control Systems Letters, 8, 2151-2156. [3] Yin, Le, et al. "Simultaneous input and state estimation: From a unified least-squares perspective." Automatica 171 (2025): 111906. [4] Bighamian, R., Kinsky, M., Kramer, G., & Hahn, J. O. (2017). In-human subject-specific evaluation of a control-theoretic plasma volume regulation model. Computers in Biology and Medicine, 91, 96-102. [5] Yin, W., Tivay, A., & Hahn, J. O. (2022). Hemodynamic monitoring via model-based extended Kalman filtering: Hemorrhage resuscitation and sedation case study. IEEE Control Systems Letters, 6, 2455-2460. Introduction •Reliable estimation is essential in biomedical systems where many relevant variables cannot be measured directly[1]. In addition, surgical stimulation, hemorrhage act as external disturbances that alter the observed responses. Without suitable estimation, these unobserved dynamics and disturbances may compromise both patient safety and closedloop drug dosing. •The estimation for anesthesia and hemodynamic systems present particular challenges. First, input excitation is often weak, limiting parameter identifiability [2]. Second, complicated separation of true drug sensitivity changes from external effects [3]. Finally, real-time blood loss estimation remains a critical challenge, as hemorrhage detection relies on indirect measurements and delayed vital signs changes that may not reveal symptoms during early stages[4,5]. Hemorrhage Detection Patient Model Joint recovery of hidden states and disturbances in both anesthesia and hemodynamic subsystems. 1 Informativity-driven adaptation of patient sensitivity parameters, where updates are emphasized during periods of informative input. 2 Integration of two coupled estimators with information sharing mechanism between them , one for anesthesia and for hemodynamics. 3 •Investigate theoretical properties of the proposed estimators, including observability, stability, and identifiability analysis, to ensure clinical reliability. •Validate the framework on large-scale virtual patient cohorts, quantifying accuracy in drug effect tracking and hemorrhage detection across diverse physiological scenarios. •Integrate probabilistic observer outputs into predictive control schemes (e.g., stochastic MPC) to move towards closed-loop decision support for anesthesia and hemodynamic management. Future steps •Kalman filters and estimation methods have seen a steady rise in publications over the past two decades, with growing interest in biomedical applications for patient monitoring and decision support. The patient is represented as the interaction of two subsystems: •Anesthesia subsystem: a pharmacokinetic model with a nonlinear pharmacodynamic model of BIS response. •Hemodynamic subsystem: a lumped cardiovascular model including fluid exchange. Formally, each subsystem is described by: where 𝑥𝑘are system states, 𝑢𝑘the measured inputs (drug infusion, fluids), 𝑑𝑘the disturbance, and 𝜃𝑘the slowly varying patient parameters. Two estimators are employed in parallel: •Anesthesia estimator: recovers PK states, effect-site concentration, output disturbance, and PD parameters (𝐶50,𝛾.). •Hemodynamic estimator: reconstructs cardiovascular states and disturbance. And a bidirectional information-sharing mechanism is introduced •Hemorrhage is the leading cause of preventable death following trauma, responsible for up to 25% of trauma fatalities. Early detection and accurate estimation of blood loss are critical, as missed hemorrhage significantly increases mortality. Figure 2Yearly distribution of the proportion of in-hospital trauma deaths caused by haemorrhagic shock from 2016 to 2023 in Europe. Figure 1Annual number of Kalman filter publications in medical and non-medical fields from 2000 to 2024, based on Web of Science records.
