Approaches to Modeling Blood Glucose Dynamics In Type 1 Diabetic Patients for Insulin Pump Automation

With subcutaneous pump insulin therapy, due to the high delay in insulin action, it is necessary to predict the glucose dynamics. Model predictive control requires the development of sufficiently accurate mathematical models of blood glucose dynamics. The work considers the existing empirical, neural network-based, physiological-inspired and combined approaches to glucose modeling. We consider in detail the model used in the T1DMS simulator, which is a stiff system of nonlinear differential equations that includes a number of empirical dependencies. It describes the absorption of glucose during food intake, the dynamics of insulin and glucagon, the processes of utilization and general endogenous production of glucose. The paper proposes an approach to the description of glucose regulation processes, which consists in mapping the enzymes involved in catalysis and inhibition of glycolysis, gluconeogenesis, formation and degradation of glycogen, as well as in the application of the Michaelis-Menten equations for their description.