Real-Time Insulin Bolusing for Unannounced Meals Using CGM Measurements

Post-prandial hyperglycemia can occur more frequently in artificial pancreas control systems that do not have meal announcements that trigger insulin boluses. Meal announcements are manual feedforward inputs to the artificial pancreas. We have developed a meal-detection and meal bolusing algorithm based on continuous glucose measurements that does not require any manual information from patients. Bergman's minimal model is modified and used in an Unscented Kalman filter for state estimations. The estimated rate of appearance of glucose is used for meal detection and calculation of meal boluses. The proposed algorithm is tested on the UVa/Padova metabolic simulator. The results indicate that the proposed algorithm decreases the frequency, duration and magnitude of hyperglycemia significantly without causing any additional hypoglycemia. The algorithm is developed as a new module of an integrated multivariable adaptive artificial pancreas control system.