Research On Dynamic Closed Loop Insulin Control System

Diabetes is a kind of chronic disease caused by irregular operation of pancreas,and it has troubled more and more people in the world.Mainstream methods for treating diabetes in hospitals are mainly confined to manually injecting insulin to patients,which is inconvenient and highly expensive.Meanwhile,they are not fine-grained for doctors to accurately control insulin levels,so we try to improve the whole system for blood glucose-insulin control.Recent years have witnessed the rapid development of healthcare devices.The connections between patients and medical cyber-physical systems are becoming tighter and many patients have got enormous benefits from these convenient systems.Current medical systems combined with the features of information interactions and data processing of Internet can access public daily life more naturally,and we can find many popular products like bracelet,heart rate measuring and so on.Therefore,the concept of connected health has emerged.Our research,mainly refers to blood glucose insulin control system,is a hot topic in the field of connected health as the development of diabetes has attracted more and more attention during the past several decades.Medical cyber-physical system on blood glucose-insulin control consists of three parts;CGMS(Continuous Glucose Monitoring System),insulin pump and closed-loop control algorithm.CGSM and insulin pump have made great advances in recent years,but we are still trying to find a better method to decrease errors introduced by mechanical measurement.Closed loop control algorithm is vitally important and complex to study in this system,some algorithms like PID(Proportional–Integral–Derivative),MPC(Model Prediction Control)and machine learning algorithms are hot in this field.Thus,we introduce two algorithms which are based on MPC and BP(Back Propagation)neural networks.In addition,we add some cont constraints according to some problems in these algorithms,and they can better control blood glucose and insulin levels.To optimize this algorithm and solve storage problem,we also add back end analysis to this system.This system can communicate with the front end system which is related to patients like CGMS,insulin pump and insulin control algorithm.It can store data which is redundant in front end and optimized the control algorithm based on these data.Our experiments are based on real data from patients in Shanghai No.9 hospital,and we make some simulations based on our algorithms to predict patients’ insulin values.We have finally concluded that our system performs well.