The Research On Modeling Of The Closed-Loop Anesthesia Control System

General anesthesia is essential in clinical surgery. To offer a real-time monitoring and control of depth of anesthesia(Do A), a closed-loop anesthetic delivery(CLAD) system based on Renyi permutation entropy was established. The CLAD system proposed in this paper consists of: a three-compartment pharmacokinetic model; a pharmacodynamic model identified by partical searm optimization(PSO); a PID controller optimized by ant colony optimization(ACO) procedure and MPC controller. The performance of two control algorithms was evaluated with the rise time, percent overshoot after the onset of RPE setpoint change, the performance error, median performance error, median absolute performance error and wobble statistically. The results showed that RPE could be used to measure the Do A in the CLAD system and the response of the two controllers can react rapidly to the sudden changes and maintain the desired Do A by manipulating the anesthetic infusion rate to meet the clinical requirement.To produce EEG signal in the simulation and improve the closed-loop anesthetic delivery system closer to the clinical application, a thalamo-cortical neural mass model(TCNMM) was established to simulate anesthesia EEG. This model consists of a thalamocortical relay cell population(TCR) and its related thalamic reticular nucleus(TRN), pyramidal neurons, excitatory interneurons, inhibitory interneurons with slow and fast kinetics. Three modulatory inputs, namely inputs reaching the TCR, the TRN and the pyramidal populations, were modulated to switch between vigilance states. The modulations were provided to make sure the power spectrum of produced signal meets the power spectrum of real EEG in different states.Meanwhile, since depth of anesthesia indexes in closed-loop anesthesia control system are usually single-scale index and the neural signals are usually multiscale signals, to explore a multiscale depth of anesthesia indexes, six multiscale permutation entropy(MSPE) measures were proposed to quantify the anesthetic drug effect on EEG recordings. Six MSPE algorithms were derived from Shannon permutation entropy(SPE), Renyi permutation entropy(RPE) and Tsallis permutation entropy(TPE) combined with the decomposition procedures of coarse-graining(CG) and moving-average(MA) analysi. To validate its relative effectiveness, the performance of these indexes was evaluated from the ability of each measure to track the dynamical changes in both simulated and real EEG data, as well as the performance in anti-noise ability and state discrimination.