Current Trends in AI and Eye Disease Diagnostics

Abstract Purpose Artificial intelligence (AI) has rapidly advanced as an approach for ophthalmic disease detection, driven by the widespread use of high-dimensional imaging modalities (e.g., optical coherence tomography). This review summarises the machine learning and AI approaches for disease detection in ophthalmology and discusses emerging paradigms and highlights key challenges impacting clinical translation. Recent Findings AI-based systems have demonstrated suitably high diagnostic performance across major ophthalmic diseases, including diabetic retinopathy (DR), diabetic macular oedema, glaucoma, age-related macular degeneration and retinopathy of prematurity. Several tools have even received regulatory approval for commercial DR screening. More recently, foundation models trained using self-supervised learning have enabled more generalisable and data-efficient disease detection across datasets and imaging modalities. In parallel, multimodal large language model systems are emerging that integrate imaging and clinical data to support more comprehensive diagnostic workflows. Early agentic AI systems extend this paradigm further by coordinating multiple models to perform disease detection, provide clinical decision support and generate reports. Summary AI-based disease detection in ophthalmology has achieved substantial technical progress but only limited translation into routine clinical practice. Key barriers include technical, clinical, ethical, economic and regulatory concerns. Future efforts should prioritise prospective evaluation in real-world settings, addressing model transparency and bias and alignment with clinical and regulatory requirements. With continued advances, AI has the potential to expand access to care, improve diagnostic accuracy and reduce the global burden of vision loss.