Advancements in ophthalmology healthcare using multimodal AI: a systematic review of methods, applications, and future directions

Recent advancements in artificial intelligence are revolutionizing ophthalmology. Specifically, the integration of structural, functional, and clinical datasets is achieving new heights in diagnostic accuracy and decision-making support. This review addresses advancements in multimodal AI from 2020 to 2025, focusing on the integration of fundus photography, OCT, OCTA, and clinical metadata. This integration not only enhances the detection and grading of diabetic retinopathy, glaucoma, and macular degeneration, but also improves prognostication. This work delineates important multimodal architectures–early, late, hybrid and transformer-based fusion models–and discusses their increasing importance in the field of precision ophthalmology. Essential hurdles still remain, with no agreements on standardized datasets, protocols on variations for imaging, interpretability gaps, and hurdles to real-world implementation caused by privacy and regulatory barriers. New techniques including federated learning, explainable AI, and generative AI, support overcoming data shortages and building confidence in AI-powered diagnostics. The manuscript also emphasizes the need for cross-institutional validation and the development of ethically sound frameworks that will allow for implementation in practice. This review seeks to consolidate knowledge on the field of multimodal AI in ophthalmology to point out both the technological and clinical gaps and steer future work on the development of clinically reliable, interpretable, and equitable AI for vision care.