Can off-the-shelf visual large language models detect and diagnose ocular diseases from retinal photographs?
Although GPT-4V default mode showed generally high sensitivity in abnormality detection, all evaluated VLLMs were inadequate in providing accurate diagnoses for ocular diseases. These findings emphasise the need for domain-customised VLLMs and suggest the continued need for human oversight in clinical ophthalmology.
Authors
- Jocelyn Hui Lin Goh (ORCID: https://orcid.org/0000-0002-5052-6081)
- Xiaofei Wang (ORCID: https://orcid.org/0000-0002-3175-0344)
- Ya Xing Wang (ORCID: https://orcid.org/0000-0003-2749-7793)
- Zhi Da Soh (ORCID: https://orcid.org/0000-0002-1182-3489)
- David Ziyou Chen (ORCID: https://orcid.org/0000-0002-2153-3100)
- Ching‐Yu Cheng (ORCID: https://orcid.org/0000-0003-0655-885X)
- Hongwei Ji
- Wendy Wong (ORCID: https://orcid.org/0000-0002-4514-250X)
- Tien Yin Wong (ORCID: https://orcid.org/0000-0002-8448-1264)
- Krithi Pushpanathan (ORCID: https://orcid.org/0000-0001-6263-8489)
- Sahana Srinivasan (ORCID: https://orcid.org/0009-0004-9355-6717)
- Yih Chung Tham (ORCID: https://orcid.org/0000-0002-6752-797X)
- Weizhi Ma (ORCID: https://orcid.org/0000-0001-5604-7527)
Institutions
- National University of Singapore (SG)
- Singapore National Eye Center (SG)
- Singapore Eye Research Institute (SG)
- Duke-NUS Medical School (SG)
- National University Hospital (SG)
- National University Health System (SG)
- Beijing Tsinghua Chang Gung Hospital (CN)
- Beihang University (CN)
- Tsinghua University (CN)
Publication Details
- Journal
- BMJ Open Ophthalmology
- Published
- 2025-04-01
- DOI
- https://doi.org/10.1136/bmjophth-2024-002076
- Citations
- 3
- Primary Topic
- Retinal Imaging and Analysis
- Type
- article
- Field-Weighted Citation Impact
- 4.08
Funders
- National University of Singapore
- Tsinghua University
- National Tsing Hua University
- Medical Research Council
- National Medical Research Council