Multimodal artificial intelligence in retinopathy of prematurity: A comprehensive narrative review.

Retinopathy of prematurity (ROP) remains a leading cause of preventable childhood blindness globally, particularly in regions with limited screening resources. Traditional diagnosis relying on subjective interpretation of fundus images faces challenges of inter-observer variability and manual analytical limitations. Recent advances in artificial intelligence (AI) have demonstrated substantial potential to achieve intelligent ROP management. We present a comprehensive review that synthesizes contemporary advances in AI-driven ROP management and reviewed the AI technologies based on retinal imaging, clinical text data, and smartphone-captured images transform ROP diagnosis, risk prediction, and treatment suggestion. We also illustrate the applications of AI in intelligent diagnosis and treatment of ROP through multimodal imaging. Despite remarkable progress, challenges persist in data heterogeneity, model generalizability, and real-world integration. By mapping technical breakthroughs to unmet clinical needs, we provide actionable insights to accelerate the development of equitable, clinically deployable AI solutions for preventing ROP-related vision loss.