Benchmarking and fine-tuning vision-language models on a visual question answering dataset for myopic maculopathy
Asia-Pacific Journal of Ophthalmology · 2026
DOI: 10.1016/j.apjo.2026.100345 ↗Kurz gefasst
Fachartikel in der Zeitschrift Asia-Pacific Journal of Ophthalmology, 2026. Verfasst von Tsun Hei Yip und 12 weiteren Autor:innen. Beteiligte Einrichtung: Department of Ophthalmology, LKS Faculty of Medicine, The University of Hong Kong, Hong Kong, China. Hauptthemen laut MEDLINE: Benchmarking, Myopia, Degenerative, Macular Degeneration. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: University of Hong Kong, Health and Medical Research Fund.
Abstract (Originalsprache)
PURPOSE: To build a visual question answering (VQA) dataset for fine-tuning and evaluating vision-language models (VLMs) in myopic maculopathy (MM). DESIGN: Cross-sectional study. METHODS: Colour fundus photographs (CFPs) from two publicly available datasets were graded using META-PM classification system. GPT-5 was used to generate clinical captions, true/false [TFQ] and open-ended [OEQ] question-answer pairs, all of which were manually verified. InternVL3-8B was fine-tuned on this dataset and evaluated against Gemini 3 Pro, Claude Sonnet 4.5, Qwen3-VL-30B-A3B-Instruct, and pre-trained InternVL3-8B. OEQ responses was evaluated by GPT-5 using a three-level scoring system (0, completely incorrect; 0.5, partially correct; 1, fully correct) and summarized as weighted accuracy. Overall accuracy was defined as the arithmetic mean of the TFQ and OEQ accuracies. RESULTS: MM-VQA comprises 2591 CFPs and 19,648 question-answer pairs. Fine-tuned InternVL3-8B model achieved an overall accuracy of 0.746, surpassing Claude Sonnet 4.5 (0.596), Qwen3-VL-30B-A3B-Instruct (0.566), and pre-trained InternVL3-8B (0.428) (all P < 0.001), while showing no significant difference compared with Gemini 3 Pro (0.724, P = 0.642). For TFQ, the fine-tuned model reached an accuracy of 0.919, outperforming Gemini 3 Pro (0.881), Qwen3-VL-30B-A3B-Instruct (0.834), Claude Sonnet 4.5 (0.796), and the pre-trained model (0.696) (all P < 0.001). On OEQ, it also ranked highest (0.572), outperforming Gemini 3 Pro (0.567, P = 0.044), Claude Sonnet 4.5 (0.395, P < 0.001), Qwen3-VL-30B-A3B-Instruct (0.297, P < 0.001) and the pre-trained model (0.160, P < 0.001). CONCLUSION: This study provides a valuable VQA dataset for MM, supporting the development of disease-specialised VLMs in ophthalmology.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Ophthalmology, LKS Faculty of Medicine, The University of Hong Kong, Hong Kong, China.
- Department of Ophthalmology, LKS Faculty of Medicine, The University of Hong Kong, Hong Kong, China; Department of Ophthalmology, Queen Mary Hospital, Pok Fu Lam, Hong Kong, China; Department of Ophthalmology, Grantham…
- Tianjin Eye Hospital, Tianjin, China.
- The Primasia International Eye Research Institute (PIERI), School of Medicine, The Chinese University of Hong Kong (Shenzhen), Shenzhen, China; The C-MER Dennis Lam & Partners Eye Center, C-MER International Eye Care Gr…
Förderung
- University of Hong Kong
- Health and Medical Research Fund
Zitieren
APA
Yip, T. H., Xu, P., Liu, Z., Zaabaar, E., Fong, A. H. C., Guo, Y., Xu, Z., Zhang, W., Chan, Y. K., Lam, D. S. C., Shih, K. C., Leung, C. K. S., & Zhang, X. J. (2026). Benchmarking and fine-tuning vision-language models on a visual question answering dataset for myopic maculopathy. Asia-Pacific Journal of Ophthalmology. https://doi.org/10.1016/j.apjo.2026.100345
Vancouver
Yip TH, Xu P, Liu Z, Zaabaar E, Fong AHC, Guo Y, et al. Benchmarking and fine-tuning vision-language models on a visual question answering dataset for myopic maculopathy. Asia-Pacific Journal of Ophthalmology. 2026. doi:10.1016/j.apjo.2026.100345
BibTeX
@article{Yip2026,
author = {Yip, Tsun Hei and Xu, Pusheng and Liu, Zirong and Zaabaar, Ebenezer and Fong, Angie H C and Guo, Yatu and Xu, Ziming and Zhang, Wei and Chan, Yau Kei and Lam, Dennis S C and Shih, Kendrick C and Leung, Christopher K S and Zhang, Xiu Juan},
title = {Benchmarking and fine-tuning vision-language models on a visual question answering dataset for myopic maculopathy},
journal = {Asia-Pacific Journal of Ophthalmology},
year = {2026},
doi = {10.1016/j.apjo.2026.100345}
}
EUVIMED