Meal Size Estimation Using a Personalized Glucose-Insulin Model for Diabetic Patients

Abstract

Adaptive model predictive control methods are used to estimate the model parameters of an FDA-approved model of glucose-insulin interaction and the meal size (in grams of glucose) given patient blood glucose concentration data. The parameter estimation achieves personalization of this model for a given patient and the meal size estimation facilitates the use of model in a separate on-going medication dosing project. The current project investigates parameter and meal size estimation when measurement noise (from the blood glucose meter) is present. The success of the algorithm is measured through a comparison of the measured glucose data and simulated response and the accuracy of the meal size estimation.