Periodic Reporting for period 1 - AMICAS (Adaptive Multi-Drug Infusion Control System for General Anesthesia in Major Surgery)
Reporting period: 2022-10-01 to 2025-03-31
The ERC AMICAS project proposes an Adaptive Multi-drug Infusion Control system for general Anesthesia in major Surgery. A major challenge in anesthesia is to adapt the drug infusion rates from observed patient response to complex surgical actions. The patient models are based on nominal population characteristic response and lack specific surgical effects. In major surgery (for instance, cardiac, transplant, bariatric surgery) modelling uncertainty arises from significant blood losses, anomalous drug diffusion, drug effect synergy/antagonism, anesthetic-hemodynamic interactions, etc. This complex interplay requires superhuman abilities of the anesthesiologist, acquired along many years of training and practice. How can we mimic this large amount of expertise? How to provide support for their critical decisions? These are but only few of the answers we provide in AMICAS.
The overall objectives are classified as follows:
Objective 1: Design and validate a time-and-data efficient online estimation of a patient-based model to include cardiovascular system dynamical changes with predictable disturbance profiles.
Objective 2: Design and validate a stable control methodology and algorithm for optimal multi-drug infusion rates to mitigate dynamic interactions and uncertainty.
Objective 3: Provide a demonstrator as an integrated technological solution for computer based surgery under general anesthesia with hemodynamic stabilization.
The overall objectives are classified as follows:
Objective 1: Design and validate a time-and-data efficient online estimation of a patient-based model to include cardiovascular system dynamical changes with predictable disturbance profiles.
Objective 2: Design and validate a stable control methodology and algorithm for optimal multi-drug infusion rates to mitigate dynamic interactions and uncertainty.
Objective 3: Provide a demonstrator as an integrated technological solution for computer based surgery under general anesthesia with hemodynamic stabilization.
Significant advancements have been made in the development of novel methodologies for anesthesia management, nociception monitoring, and system control. These innovations bridge various disciplines, including biomedical engineering, clinical medicine, and control theory, fostering knowledge transfer between research and clinical practice.
Several innovative methodologies were developed to enhance anesthesia monitoring, control, and patient care:
• Nociception/Antinociception Index: Novel approach for the pain/analgesia indices using the AnspecPro device was developed and validated against commercial devices (Medasense, Medstorm). This allows for accurate measurement of nociceptive responses to surgical stimuli during anesthesia, ensuring better control and patient safety.
• Multivariable Patient Models: Advanced multivariable models were created to capture the interaction of hypnotic and opioid drugs (Propofol and Remifentanil) with hypnosis (Bispectral Index) and analgesia. Moreover, a novel system modeling technique to adapt patient model parameters under conditions of uncertainty were proposed, addressing the challenges of poor identifiability in clinical data due to ethical and safety constraints. These models improve drug dosing strategies and enhance the accuracy of patient-specific anesthesia management.
• Augmented Patient Models: An augmented pharmacokinetic (PK) model that accounts for non-uniform drug uptake and clearance as a nonlinear function of body mass index (BMI), addressing drug trapping risks in long-term anesthesia. The model is validated through both open-loop and closed-loop control, optimizing drug dosages to minimize overdosing and protect vital organs such as the liver and kidneys.
• Model predictive control with integrated disturbance (surgical) profiles: A novel method in multi-drug control optimization involves the development and integration of surgical disturbance profiles—tailored to major, moderate, and minor surgeries—into a Model Predictive Control (MPC) strategy. This approach mimics closely the surgical stimuli to provide information to the controller and minimize intraoperative uncertainty, thereby enhancing patient safety and reducing variability in response.
Several innovative methodologies were developed to enhance anesthesia monitoring, control, and patient care:
• Nociception/Antinociception Index: Novel approach for the pain/analgesia indices using the AnspecPro device was developed and validated against commercial devices (Medasense, Medstorm). This allows for accurate measurement of nociceptive responses to surgical stimuli during anesthesia, ensuring better control and patient safety.
• Multivariable Patient Models: Advanced multivariable models were created to capture the interaction of hypnotic and opioid drugs (Propofol and Remifentanil) with hypnosis (Bispectral Index) and analgesia. Moreover, a novel system modeling technique to adapt patient model parameters under conditions of uncertainty were proposed, addressing the challenges of poor identifiability in clinical data due to ethical and safety constraints. These models improve drug dosing strategies and enhance the accuracy of patient-specific anesthesia management.
• Augmented Patient Models: An augmented pharmacokinetic (PK) model that accounts for non-uniform drug uptake and clearance as a nonlinear function of body mass index (BMI), addressing drug trapping risks in long-term anesthesia. The model is validated through both open-loop and closed-loop control, optimizing drug dosages to minimize overdosing and protect vital organs such as the liver and kidneys.
• Model predictive control with integrated disturbance (surgical) profiles: A novel method in multi-drug control optimization involves the development and integration of surgical disturbance profiles—tailored to major, moderate, and minor surgeries—into a Model Predictive Control (MPC) strategy. This approach mimics closely the surgical stimuli to provide information to the controller and minimize intraoperative uncertainty, thereby enhancing patient safety and reducing variability in response.
The Adaptive Multi-drug Infusion Control system for general Anesthesia in major Surgery project integrates human expertise with computer optimization to create a successful solution for breakthrough into clinical practice.
The following elements as key advancements that extend beyond the current state-of-the-art:
• Adaptive Multivariable Patient Modeling: A complex multivariable model developed using clinical data captures unique pharmacokinetic and pharmacodynamic interactions, especially between hypnotics and analgesics (e.g. Propofol and Remifentanil). The multivariable model introduces a more personalized approach to anesthesia by integrating hypnosis levels with nociception level marking a significant advance in patient-specific anesthesia control. The model is further augmented with heterogeneous drug diffusion patterns and hemodynamic variable interaction through cardiac output and mean arterial pressure levels.
• Model Predictive Control Integrated with Surgical Disturbance Profiles: By embedding tailored disturbance profiles for major, moderate, and minor surgeries into a control framework, the project has achieved a novel predictive and adaptive control method that minimizes intraoperative variability. This integration significantly improves patient safety and response stability during surgery, a leap forward in closed-loop anesthesia systems.
• Digital Twin Simulator for Real-World Scenario Testing: The creation of a digital twin simulator for general anesthesia management, which combines patient databases, pharmacological models, and surgical disturbance profiles, supports experimentation with anesthesia strategies in a realistic, controlled environment. This hybrid simulation framework extends beyond traditional open-loop and closed-loop setups, providing a groundbreaking tool for research and clinical training.
The following elements as key advancements that extend beyond the current state-of-the-art:
• Adaptive Multivariable Patient Modeling: A complex multivariable model developed using clinical data captures unique pharmacokinetic and pharmacodynamic interactions, especially between hypnotics and analgesics (e.g. Propofol and Remifentanil). The multivariable model introduces a more personalized approach to anesthesia by integrating hypnosis levels with nociception level marking a significant advance in patient-specific anesthesia control. The model is further augmented with heterogeneous drug diffusion patterns and hemodynamic variable interaction through cardiac output and mean arterial pressure levels.
• Model Predictive Control Integrated with Surgical Disturbance Profiles: By embedding tailored disturbance profiles for major, moderate, and minor surgeries into a control framework, the project has achieved a novel predictive and adaptive control method that minimizes intraoperative variability. This integration significantly improves patient safety and response stability during surgery, a leap forward in closed-loop anesthesia systems.
• Digital Twin Simulator for Real-World Scenario Testing: The creation of a digital twin simulator for general anesthesia management, which combines patient databases, pharmacological models, and surgical disturbance profiles, supports experimentation with anesthesia strategies in a realistic, controlled environment. This hybrid simulation framework extends beyond traditional open-loop and closed-loop setups, providing a groundbreaking tool for research and clinical training.