Study On The Simulation Of Closed Loop Control With Propofol Anesthesia

The safety of anesthetic administration is a necessary condition for clinical operation.With the gradual improvement of medical level,people pay more and more attention to the problem of anesthesia control,and how to achieve accurate anesthetic administration has become the focus of more and more scholars.However,in the current actual operation,anesthetic administration is still mainly controlled by the anesthesiologist,and there is no necessary link between the anesthetic observation index and the control of anesthetic administration.Therefore,it has become an urgent problem to realize the precise control of anesthetic administration.In this study,the closed-loop control of propofol anesthesia is simulated from the controlled object model,state observer and control algorithm.Firstly,for the state observer,the observation indicators' performance of EEG is studied.The curve trend and correlation analysis of approximate entropy,sample entropy,synchfastslow,and other indexes with BIS indexes,which have been used in closed-loop control.The results show that the approximate entropy,? ratio,Shannon permutation entropy,sample entropy and Renyi permutation entropy all show similar trends to BIS index.In the correlation analysis,the correlation between Renyi permutation entropy and BIS index is the highest(correlation coefficient: 0.72±0.14).Secondly,the controlled object model is established by two means.On the one hand,the classical Pharmacokinetic–pharmacodynamic model is selected as the controlled object.Due to the unknown parameters of pharmacodynamic model,particle swarm optimization algorithm is used to identify the parameters of the model.On the other hand,the drug metabolism model is constructed by deep learning method,and the convolutional neural networks' training model is selected to replace the traditional Pharmacokinetic–pharmacodynamic model for control simulation.The combination of index parameters extracted from EEG signal and the data of BIS index are used as the input for training.The results show that the prediction of BIS index is statistically good.Finally,Ant Colony Optimization PID and model predictive control are selected as controllers to simulate the two drug metabolism models.The results show that thesimulation results of closed-loop anesthesia control with Ant Colony Optimization PID based on convolutional neural networks are the best(rise time: 1.68 ± 0.05 min,adjust time:2.68 ± 0.22 min,overshoot: 0.15 ± 0.04%).It provides a theoretical basis for accurate administration of anesthesia during surgery.