Clinician-Friendly Foundation Models for Ophthalmic Image Diagnostics without Fine-Tuning or Technical Barriers

Meng Wang · 2025

Read Clinician-Friendly Foundation Models for Ophthalmic Image Diagnostics without Fine-Tuning or Technical Barriers by Meng Wang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-15928.txt.

Description

Meng Wang and co-authors build an eye-image AI that clinics can use without retraining, in this 2025 paper. Most medical AI models must be refit for each new hospital, which blocks their spread. Their platform, GlobeReady, runs on a foundation model called RetiGlobe trained to read eye images broadly from the start.

Training came in two stages: first self-teaching on thirty-eight million synthetic eye images, then learning from nearly half a million real image-text pairs spanning many peoples, devices, and regions. The team tested the system on close to half a million images from centers in four countries. Thirty-one eye doctors also tried it in prospective use tests. Side studies checked how it handles new settings, flags its own doubt, and finds like cases.

Catalog record

CREATORMeng Wang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE69 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-15928

Browse the full library · About the Tower of Babel