Application of dynamic optimization-based parameter estimation to a diabetes mellitus patient model

This work presents a method for modeling the effect of insulin infusion and meal disturbances in a patient with diabetes mellitus. The model used was presented by Bergman and later modified by Lynch and Bequette. This modified Bergman model is a simple set of four differential equations which account for blood glucose, blood insulin, free insulin and subcutaneous glucose concentration profiles in the body of a person with diabetes mellitus. Using subcutaneous blood sugar measurements and data for the response to insulin of a patient, it is possible to develop an accurate dynamic model, leading to the possibility of applying model-predictive control to help regulate the blood glucose levels of a patient. The initial parameters for the modified Bergman model were recovered from simulated data by performing dynamic optimizations using MAPLE, Matlab^reg and IPOPT. In the future, this procedure could be tested using real data for a patient.