This dissertation develops physiology-guided machine learning digital twins and a person-centered decision support system to improve personalized diabetes management. By combining mechanistic models of glucose–insulin regulation with hybrid modeling techniques, the proposed digital twins accurately simulate glucose dynamics for individuals with both type 1 and type 2 diabetes, enabling reliable in silico evaluation of treatment and lifestyle interventions. The work also introduces a personalized recommendation framework for meals and physical activity that adapts to individual goals and preferences, demonstrating the potential of hybrid artificial intelligence and digital twin technologies to support person-centered diabetes care.
