Abstract:Generative artificial intelligence(Gen-AI), an important branch of artificial intelligence(AI), has attracted increasing attention in medical image analysis and clinical decision support.Different from traditional discriminative models, Gen-AI can generate high-quality and diverse new data by learning the underlying distribution features of data. It exhibits unique advantages in alleviating insufficient training samples, mitigating uneven data distribution, and improving the generalization ability of models. Ophthalmology is highly dependent on multimodal imaging, making it a particularly suitable field for the application of Gen-AI. This review summarizes recent advances in major Gen-AI techniques, including generative adversarial networks, diffusion models, and large language models, and discusses their applications in ophthalmic disease screening and diagnosis, image synthesis and data augmentation, prognostic prediction, and individualized treatment planning. It also analyzes challenges in the course of clinical translation, including data quality, model credibility, ethical supervision and other issues, so as to provide references for its standardized application.
