ARVO: reference images boost glaucoma AI

- An ARVO poster reported that providing reference retinal images to a general-purpose large language model improved its glaucoma detection performance. - Poster authors showed higher diagnostic accuracy when the model received comparator images alongside target images versus receiving no visual context. - The finding highlights that AI performance can be highly prompt‑ and workflow‑dependent; exact test conditions must be reported for clinical claims. (healio.com)