Large language models (LLMs) show significant promise in glaucoma for patient education and clinical decision support, though their performance varies widely.
A scoping review analyzed how LLMs and vision language approaches are applied in glaucoma research and early clinical settings. The study screened 316 records, identifying 27 studies focused on glaucoma-related tasks.
The strongest evidence for LLM application lies in patient education, where models can explain complex concepts and translate clinical jargon into understandable terms. Text-based clinical decision support also shows potential, assisting with information summarization and differential diagnosis.
However, the review stresses that current LLMs are best suited as assistive tools, not autonomous decision-makers. Key limitations include variable accuracy, a lack of multimodal integration for image interpretation, and insufficient domain-specific fine-tuning for ophthalmology.
Future advancements should focus on domain-trained and retrieval-augmented models, robust evaluation standards, and clear ethical and regulatory frameworks for safe implementation. Clinical deployment must be guided by validation, transparency, and careful oversight.