Seeing Beyond Borders: Evaluating LLMs in Multilingual Ophthalmological Question Answering
David S. Restrepo, Luis Filipe Nakayama, Robyn Gayle Dychiao, Chenwei Wu, Liam G. McCoy, Jose Carlo M. Artiaga, Marisa Cobanaj, João Carlos Ramos Gonçalves de Matos, Jack Gallifant, Danielle S. Bitterman, Vincenz Ferrer, Yindalon Aphinyanaphongs, Leo Anthony Celi · 2024
Large Language Models (LLMs), such as GPT-3.5 [1] and GPT-4 [2], have significant potential for transforming several aspects of patient care from clinical note summarization to performing board-level clinical question-answering tasks [3], [4]. Ophthalmology, is a field with high patient volume and therefore holds high documentation burden for physicians but great opportunities for leveraging LLMs. Furthermore, given the critical and permanent nature of negative disease outcomes like blindness and their ensuing social and financial damage to patients, the need for reliable, accessible, and robust tools is urgent. Several studies have already showcased the practicality of GPT applications in ophthalmology [5], [6], and in specific ophthalmology subspecialties, such as glaucoma and retina [7], [8]