Abstract
<title>Abstract</title> <p>Retinal diseases are major causes of global blindness, with emerging treatments and artificial intelligence (AI) rapidly reshaping care. Conventional bibliometrics relying on lifetime citations suffer from temporal bias, often failing to capture immediate research frontiers. We aimed to map the immediate trajectory of high-impact retinal disease research by comparing influential literature from 2023 and 2024. We performed a retrospective, large language model (LLM)–assisted bibliometric analysis of 100 highly cited OpenAlex articles (50 per year), ranked by 18-month citation impact. LLMs verified topical relevance, extracted concepts, and normalized them into 89 parent themes and 874 canonical topics. Diabetic retinopathy and diabetic eye disease remained the dominant parent theme (normalized frequency 0.48 to 0.50), while retinal degenerative and inherited diseases and diabetes mellitus and its complications rose modestly (+0.06 and +0.04, respectively). Diabetic retinopathy grading increased (+0.10), whereas convolutional neural network (−0.22) and fundus imaging (−0.20) declined; large language models and genome-wide association study emerged (+0.06 each). These findings indicate a shift from foundational imaging-based machine learning toward clinical diabetic retinopathy management, genomic studies, and deployed AI tools. A web app (https://pri.pepkio.com) supports data exploration.</p>