Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Jul 13, 2026

A Gaze-Contingent Display Framework for Perceptual Learning Research with Simulated Central Vision Loss

A Gaze-Contingent Display Framework for Perceptual Learning Research with Simulated Central Vision Loss

Published on: April 11, 2025

Clinically grounded retinal representation learning from minimal supervision.

Boa Jang^1,2, Youngbin Ahn^2,3, Eun Kyung Choe⁴

¹Interdisciplinary Program in Bioengineering, College of Engineering, Seoul National University, Seoul, Republic of Korea.

Scientific Reports

|July 11, 2026

Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Intraocular Lens Positional Instability and Refractive Prediction Error after Cataract Surgery in Eyes with Retinitis Pigmentosa: A Comparative Study.

Korean journal of ophthalmology : KJO·2026

Same author

Impact of gestational diabetes on future diabetic retinopathy and microvascular complications: a nationwide cohort study.

BMC medicine·2026

Same author

Gas-Lesion Contact and Postural Compliance After Vitrectomy With Tamponade: A Continuous Monitoring and 3D Quantitative Analysis.

Translational vision science & technology·2026

Same author

Correction: Clinical variant interpretation comparing two saturation genome editing-based functional studies for <i>BRCA2</i>.

Frontiers in genetics·2026

Same author

Clinical and genetic characteristics of RPE65-associated inherited retinal degeneration in Koreans.

Scientific reports·2026

Same author

Stepwise analysis of the scleral encircling learning curve for clinical fellows using surgical video review.

International journal of retina and vitreous·2026

This summary is machine-generated.

M-FunFound, a new framework, uses binary labels from routine eye screenings to learn from retinal images. This approach improves early detection of vision loss by enabling robust representation learning with reduced annotation needs.

Area of Science:

Ophthalmology
Medical Imaging
Computer Vision

Background:

Early detection of retinal abnormalities is crucial for preventing vision loss.
Large-scale screening is hindered by limited specialists, varied imaging, and scarce annotations.
Binary classification (normal vs. abnormal) is the primary decision in routine practice.

Purpose of the Study:

To develop a clinically grounded retinal representation learning framework (M-FunFound).
To leverage binary abnormality labels for pretraining.
To evaluate M-FunFound's performance on downstream tasks compared to other methods.

Main Methods:

Developed M-FunFound, a framework pretrained on 113,645 fundus images with binary abnormality labels.
Evaluated on abnormality classification (internal, JSIEC, RFMiD datasets).

Keywords:

Foundation model Fundus photography Retinal disease Transfer learning Weakly supervised learning

Related Experiment Videos

Last Updated: Jul 13, 2026

A Gaze-Contingent Display Framework for Perceptual Learning Research with Simulated Central Vision Loss

A Gaze-Contingent Display Framework for Perceptual Learning Research with Simulated Central Vision Loss

Published on: April 11, 2025

Assessed multi-label disease classification (internal dataset) and vessel segmentation (FIVES dataset).

Main Results:

M-FunFound achieved high AUCs for binary abnormality classification (0.945 internal, 0.958 JSIEC, 0.829 RFMiD).
Secured the highest weighted F1-score (0.725) for multi-label disease classification.
Obtained the top Dice score (0.853) for vessel segmentation.

Conclusions:

Binary abnormality supervision enables robust and transferable retinal representation learning.
This approach significantly reduces annotation burden in real-world screening.
M-FunFound offers a scalable solution for improving ophthalmic screening.