EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Performance of General-Purpose Vision Language Models and Ophthalmology Foundation Models in Glaucoma Detection and Function Prediction

Jalil Jalili, Justin Huynh, Evan Walker, Benton Gabriel Chuter, Christopher Bowd, Anna Heinke, Akram Belghith, Michael Henry Goldbaum, Massimo Antonio Fazio, Christopher A Girkin, Carlos Gustavo De Moraes, Jeffrey M Liebmann und 4 weitere

Translational Vision Science & Technology · 2025

DOI: 10.1167/tvst.14.11.31 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Translational Vision Science & Technology, 2025. Verfasst von Jalil Jalili und 15 weiteren Autor:innen. Beteiligte Einrichtung: Division of Ophthalmology Informatics and Data Science, Viterbi Family Department of Ophthalmology, Shiley Eye Institute, University of California, S…. Hauptthemen laut MEDLINE: Glaucoma, Visual Fields, Ophthalmology. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 52 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: NEI NIH HHS, NIH HHS, NLM NIH HHS.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Purpose

To evaluate the performance of vision-language models (VLMs), in glaucoma detection and visual field (VF) mean deviation (MD) prediction tasks using optical coherence tomography (OCT) images.

Methods

A total of 27,610 SPECTRALIS OCT images from 1025 participants (1690 eyes), collected between 2008 and 2021 as part of the Diagnostic Innovations in Glaucoma Study (DIGS) and the African Descent and Glaucoma Evaluation Study (ADAGES), were included. Vision components of LLaVA and PaliGemma, as well as RETFound and ResNet-50 models, were fine-tuned for glaucoma classification and VF MD prediction. Models were trained using OCT circle scans centered on the optic nerve head. Three training configurations were compared. Performance was evaluated using area under the receiver operating characteristic curve (AUC), mean absolute error (MAE), and related metrics.

Results

The LLaVA model, when both vision encoder and multi-layer projector were fine-tuned, achieved the best performance with an AUC of 0.92 (95% confidence interval [CI], 0.86-0.95) for glaucoma classification and an MAE of 1.79 dB (95% CI, 1.55-2.00) for VF MD prediction. RETFound and PaliGemma also performed well, with AUCs of 0.91 and 0.90 and MAEs of 1.87 dB and 1.84 dB, respectively. Models with frozen vision encoders showed reduced accuracy. Stratified analysis showed better glaucoma classification in older individuals and moderate-to-advanced cases. VF MD prediction was more accurate in younger individuals, with higher errors in advanced glaucoma.

Conclusions

Fine-tuned VLMs demonstrated high performance in glaucoma detection and VF MD prediction, matching or exceeding specialized foundation models and traditional convolutional neural network (CNN)-based methods. Translational Relevance: This study highlights the potential of general-purpose AI models to be adapted for glaucoma care, enabling scalable decision support from OCT imaging.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

GlaucomaVisual FieldsOphthalmologyHumansTomography, Optical CoherenceMaleMiddle AgedFemaleAgedROC CurveOptic Disk

Autor:innen und Institutionen

  • Division of Ophthalmology Informatics and Data Science, Viterbi Family Department of Ophthalmology, Shiley Eye Institute, University of California, San Diego, La Jolla, CA, USA.
  • Hamilton Glaucoma Center, Viterbi Family Department of Ophthalmology, Shiley Eye Institute, University of California, San Diego, La Jolla, CA, USA.
  • School of Medicine, University of Illinois Urbana-Champaign, Urbana, IL, USA.
  • Department of Ophthalmology and Vision Sciences, University of Alabama at Birmingham, Birmingham, AL, USA.
  • Department of Ophthalmology, Harkness Eye Institute, Bernard and Shirlee Brown Glaucoma Research Laboratory, New York, NY, USA.

Förderung

  • NEI NIH HHS
  • NIH HHS
  • NLM NIH HHS

Zitieren

APA

Jalili, J., Huynh, J., Walker, E., Chuter, B. G., Bowd, C., Heinke, A., Belghith, A., Goldbaum, M. H., Fazio, M. A., Girkin, C. A., De Moraes, C. G., Liebmann, J. M., Baxter, S. L., Weinreb, R. N., Zangwill, L. M., & Christopher, M. (2025). Performance of General-Purpose Vision Language Models and Ophthalmology Foundation Models in Glaucoma Detection and Function Prediction. Translational Vision Science & Technology. https://doi.org/10.1167/tvst.14.11.31

Vancouver

Jalili J, Huynh J, Walker E, Chuter BG, Bowd C, Heinke A, et al. Performance of General-Purpose Vision Language Models and Ophthalmology Foundation Models in Glaucoma Detection and Function Prediction. Translational Vision Science & Technology. 2025. doi:10.1167/tvst.14.11.31

BibTeX
@article{Jalili2025,
  author = {Jalili, Jalil and Huynh, Justin and Walker, Evan and Chuter, Benton Gabriel and Bowd, Christopher and Heinke, Anna and Belghith, Akram and Goldbaum, Michael Henry and Fazio, Massimo Antonio and Girkin, Christopher A and De Moraes, Carlos Gustavo and Liebmann, Jeffrey M and Baxter, Sally L and Weinreb, Robert N and Zangwill, Linda M and Christopher, Mark},
  title = {Performance of General-Purpose Vision Language Models and Ophthalmology Foundation Models in Glaucoma Detection and Function Prediction},
  journal = {Translational Vision Science & Technology},
  year = {2025},
  doi = {10.1167/tvst.14.11.31}
}
Kostenlos starten →