Generative artificial intelligence in ophthalmology: a scoping review of current applications, opportunities, and challenges.

Generative artificial intelligence (AI) has the potential for enhancing diagnostic accuracy, augmenting image datasets, and simulating disease progression in ophthalmology. There is a paucity in the literature regarding a comprehensive mapping of the broad range of generative AI models. A scoping review was conducted to (1) identify and categorise generative AI models applied in ophthalmology, (2) examine their primary clinical and research applications, (3) evaluate the feasibility and challenges of deploying these models in clinical practice, and (4) outline future research directions. Primary research articles published in English over the past 10 years, explicitly referencing a generative AI method for ophthalmic imaging modalities were included. Searches were conducted on, MEDLINE, Embase, and Web of Science up to December 24, 2024. One reviewer independently screened titles, abstracts, and full texts and subsequently charted data using a calibrated extraction form. Data regarding study design, generative AI type, and clinical outcomes or applications were included. 40 studies were included, with most studies focused on data augmentation (n = 11) and predictive modelling (n = 11), followed by image enhancement (n = 8), segmentation (n = 7), and education/interpretability (n = 3). GAN-based approaches predominated, but diffusion models, VAEs, and flow-based models have recently gained traction. Implementation barriers included data availability, regulatory considerations, and model stability. This scoping review mapped the current applications of generative AI in ophthalmology and identified emerging trends, such as diffusion models. Although these methods hold promise, robust real-world validation and clearer regulatory pathways are needed to fully integrate generative AI into clinical practice.