# Artificial intelligence-powered prediction of diabetic complications: from clinical data to molecular omics

## TL;DR

This paper reviews how AI can predict diabetic complications like retinopathy and kidney disease, using clinical and molecular data to improve early detection and treatment.

## Contribution

The paper introduces a six-step framework for clinical deployment of AI models and highlights cross-scale data fusion as a novel approach for better prediction.

## Key findings

- Multimodal data fusion improves predictive performance for diabetic complications compared to single-modality approaches.
- AI models have evolved from traditional machine learning to advanced systems like large language models and agent-based frameworks.
- A six-step clinical translation pathway is proposed to guide the development and deployment of AI in diabetes care.

## Abstract

Diabetic complications are a major cause of disability and mortality among patients, and early identification of high-risk individuals is essential for precision prevention and management. In recent years, the rapid advancement of artificial intelligence (AI) has provided transformative tools for risk prediction and clinical decision support in diabetes care. In this narrative review, we systematically surveyed studies published between January 2015 and June 2025 in PubMed, Web of Science, and Scopus that applied AI-based predictive modeling for three major diabetic complications: diabetic retinopathy (DR), diabetic nephropathy (DN), and diabetic cardiovascular disease (CVD). A total of 58 studies were included, encompassing models based on clinical features, molecular omics, medical imaging, and multimodal data integration. Cross-scale and multimodal data fusion has emerged as a promising new paradigm, demonstrating improved predictive performance over single-modality approaches in three major diabetic complications. We also summarize the evolution from traditional machine learning to deep learning and, more recently, to large language models and agent-based systems, comparing their methodological characteristics, strengths, and suitable application scenarios. Finally, we proposed an actionable six-step framework and clinical translation pathway for AI in diabetic complications, outlining key steps from data curation and model development to validation, regulatory compliance, and real-world implementation. Together, these insights provide a roadmap toward developing robust, transparent, and clinically deployable AI systems capable of transforming the prevention and management of diabetic complications.