Intra-patient and inter-patient adaptive control of hypnotic states during total intravenous anesthesia

Abstract

Total intravenous anesthesia refers to administering medications to induce a state of controlled unconsciousness or analgesia during surgical or medical procedures. These medications target the nervous system, producing various effects such as pain relief, amnesia, and muscle relaxation. General anesthesia renders the patient completely unconscious and unaware of their surroundings. The input to the total intravenous anesthesia system is the rate of Propofol, and the output is the bispectral index when considered as a control system. The patient's hypnotic state can be controlled by taking the BIS as feedback. During surgery, stimuli such as scalpel cuts affect the system as output disturbances, while blood loss affects the system as parameter uncertainty and time-variant parameters. The total intravenous anesthesia system can be considered one of the most complex control systems due to its time-varying characteristics, intra-patient and inter-patient variability, unpredictability, and model uncertainty. Controlling the patient's hypnotic state in anesthesia faces many challenges, including patient-dependent variables, uncertain time delays, drug-dependent dynamics, and stability issues. Despite the development of various control systems such as PID control, MPC, and various adaptive control methods, the inability to ensure intra-patient and inter-patient validity remains one of the biggest obstacles to applying anesthesia systems with automatic control features to broader patient groups. Therefore, new adaptive control structures with minimal patient parameter dependency are needed. In this thesis, two primary objectives can be mentioned for the modeling and control of the total intravenous anesthesia system. The first is to propose a model improvement method that enhances the correlation with real surgical results by adding output delays to the mathematical patient models of total intravenous anesthesia based on real surgical patient data. The second is to propose a model reference adaptive control structure for the inter-patient adaptive control of closed-loop total intravenous anesthesia systems and to prove that the proposed structure can achieve more robust results in anesthesia control compared to traditional PID controllers. The proposed methodology for determining the delay time of delayed patient models involves estimating patient model parameters obtained as parametric transfer functions using the least squares method from the VitalDB database containing real surgical data. The delay value that produces the highest correlation with real surgical data is accepted as the individual delay value of each patient. The validity of delayed models is proven by comparing the correlations between the outputs of delayed and non-delayed models and real surgical output data. The model reference adaptive control system structure has been integrated into closed-loop total intravenous anesthesia to ensure intra-patient and inter-patient control validity. The model reference adaptive control structure for total intravenous anesthesia is based on an observer-based state feedback controller. The adaptation laws created using Lyapunov's stability theory adjust the state feedback controller gains to ensure inter-patient validity. The model reference adaptive controller has been applied to total intravenous anesthesia and tested on a database of 24 patients and patients identified by VitalDB under conditions of stimulus disturbances and blood loss. Correlation analysis results show that patient models with output delays in pharmacokinetic-pharmacodynamic analysis, with the bispectral index as the output, exhibit stronger correlations with real patient data from VitalDB. The importance of this result increases, especially considering that the bispectral index is the only measurable output value. In addition, the model reference adaptive control system proposed in this thesis has improved the control of total intravenous anesthesia, producing more robust and desirable results compared to the common PID controllers. The model reference adaptive control system for total intravenous anesthesia suggested in this thesis can be used to develop and manage the patient's hypnotic state during total intravenous anesthesia.