Multiple Model Predictor for Type 1 Diabetes

Abstract: The availability of reliable glucose-insulin models is a key point for both glucose control and hypo/hyperglycemia prevention in type 1 diabetes. This paper introduces an enhanced personalized Multiple Model Predictor (MMP) for glucose prediction, addressing both inter- and intra-patient variability. The MMP is identified using a clinically based cost function to weight differently hypo, hyper, and normoglycemia and a smooth switching mechanism between models is implemented to improve the glucose prediction. Tested on the UVA/Padova simulator, the MMP showed a 113% improvement in prediction accuracy with respect to the state-of-art Daily Model, especially during breakfast, demonstrating its potential for improving glucose management in T1D patients.