Federated multimodal AI for precision-equitable diabetes care

Area of Science:

• Medical Informatics
• Public Health
• Artificial Intelligence

Background:

• Type 2 diabetes mellitus (T2DM) is a growing global epidemic exacerbated by health inequities.
• Disadvantaged populations face disproportionately higher risks of T2DM onset and complications.

Purpose of the Study:

• To explore the potential of artificial intelligence (AI) to mitigate health disparities in T2DM care.
• To review AI applications in T2DM early detection, complication surveillance, and management.

Main Methods:

• Review of current AI applications in T2DM, including machine learning (ML) models integrating multimodal data.
• Analysis of AI's role in screening for diabetic retinopathy, forecasting renal decline, and detecting foot lesions.
• Examination of AI integration with continuous glucose monitoring (CGM), wearables, and telemedicine.

Main Results:

• AI models enhance early risk prediction by integrating diverse data sources.
• AI-powered systems achieve high accuracy in screening for diabetic complications.
• AI facilitates real-time glycaemic management and extends care via telemedicine.

Conclusions:

• AI offers transformative potential to reduce T2DM health inequities across the care continuum.
• Addressing data heterogeneity, algorithmic bias, and the digital divide is crucial for equitable AI implementation.
• Future AI development requires multimodal models, digital twins, and ethical governance for sustainable and equitable solutions.