Multimodal Artificial Intelligence for Precision Health in Diabetes Care

Diabetes Mellitus (DM) is a chronic metabolic disorder that affects over 529 million people worldwide and remains one of the leading causes of morbidity, mortality, and healthcare expenditure. Its management requires continuous monitoring, lifestyle adaptation, and multidisciplinary coordination across primary, secondary, and tertiary care settings. Despite advances in diagnostics and therapeutics, maintaining stable glucose control and preventing long-term complications such as diabetic retinopathy, diabetic kidney disease, and diabetic foot ulcers remain major challenges, as existing care pathways struggle to integrate the growing volume and diversity of data needed to support timely and personalized decision-making. Artificial intelligence (AI) and machine learning (ML) offer new and exciting possibilities for addressing this complexity by analyzing large, heterogeneous datasets. However, current AI applications in DM care are fragmented, focusing mainly on single data types or isolated clinical settings. The absence of multimodal integration, limited generalizability, and re-use of approaches and data hinder development. This PhD project aims to develop and evaluate multimodal and cross-setting AI systems that reflect the real-world complexity of DM care. The research will develop a conceptual framework that structures and connects AI approaches in relation to data modalities, clinical functions, and healthcare settings. Empirical studies will use multiple sources of data, including structured, unstructured, and temporal information from primary, secondary, and tertiary care, to create predictive and decision-support models for early detection and improved management of DM and its complications. The project will also address equity of AI models by incorporating underrepresented populations and diverse data sources. Collectively, the findings will advance the development of AI across the DM care continuum.