Application Of Anesthesia Target Controlled Infusion System Based On LSTM Neural Network PID Control Algorithm

Anesthetic target-controlled infusion(TCI)is a process in which the drug concentration in plasma or effect chamber is regulated by computer control program to reach the target concentration based on PB-PK physiological model.In the actual operation,the single-channel open-loop TCI system is usually used,and its defect is that the anesthesiologist needs to set the new drug concentration in time according to the changes of the patient's vital signs during the whole operation.This strategy will make the anesthetic effect very dependent on the anesthesiologist's professional level and can not monitor the patient's vital signs in real time.In order to reduce the risk of operation,this thesis proposes an anesthetic target-controlled infusion system based on LSTM neural network,which communicates and transmits data between single-chip microcomputer and TCI infusion pump and Narcotrend monitor.The host computer software collects,processes,displays and stores the data in real time and gives the predicted value of the monitor index,which can effectively reduce the dependence on the specialty of anesthesiologists and improve the accuracy of drug administration.In the selection of monitoring indicators of anesthetic depth,this thesis discusses the principles of electroencephalogram(EEG),EEG bispectral index(BIS)and Narcotrend index(NI),and selects NI index as the monitoring index of this experiment;in the selection of hardware,according to the actual design requirements and limited products,we finally choose BCP-100 infusion pump,Narcotrend monitor,router,STM32F103 single-chip microcomputer and notebook computer.In the aspect of algorithm selection,several neural networks and several different forms of PID control algorithms are compared,and finally the combination of LSTM neural network and PID control algorithm is chosen according to the actual situation.This experiment also collected 90 real patient data from Hospital of Chinese people's Liberation Army,and each case was composed of thousands of text data.therefore,this thesis also designed a MySQL database to store them,which also provided a sample for the training of LSTM neural network in this thesis.With STM32F103 single chip microcomputer and notebook computer as the core,an anesthetic target-controlled infusion system based on LSTM neural network PID control is built.It mainly includes five aspects: the construction of hardware platform,communication design,algorithm research and software program development,function realization and human-computer interface design.Based on the analysis of the functional requirements of the whole system,the principle of the main functional modules is studied,and the simulation experiments of the main algorithms are carried out.Finally,the stability and feasibility of the system are proved,and the related functions of the whole system are realized.