跳到正文
原文
Medical Xpress·· 2026-10-21精选AI 评分64

单张眼底照片深度学习预测近视进展与高度近视

Deep learning using fundus photograph predicts myopia progression

AI 导读

北京同仁眼中心团队在 AAO 会议报告了基于非散瞳眼底照相预测一年近视发生及进展至中度或高度近视的深度学习模型。模型在内部及两个外部队列验证中,预测高度近视进展的 AUC 分别为 0.95、0.96 和 0.91,预测中度近视进展的 AUC 分别为 0.89、0.89 和 0.88。作者认为该技术可为早期临床干预提供强力工具。

推荐理由

研究展示了单张眼底照片预测近视进展的性能,为临床早期分层提供工具参考。

正文 · 原文
child wearing glasses
Credit: Unsplash/CC0 Public Domain

A deep learning algorithm analyzing one fundus photograph can predict progression to moderate or high myopia, according to a study presented at the annual meeting of the American Academy of Ophthalmology, held Oct. 9 to 12 in New Orleans.

Ang Ji, M.B.B.S., and Yin Guo, M.D., from Beijing Tongren Eye Center in China, developed and validated algorithms to predict one-year myopia onset and progression to moderate myopia and high myopia using nonmydriatic fundus imaging.

A two-stage transfer learning approach was pretrained on 150,213 cross-sectional images; fine-tuning/internal validation of models was performed on a longitudinal cohort of 16,384 children, and external validation was performed on two other cohorts (external validation cohort 1 [EC1], 2,309 participants; external validation cohort 2 [EC2], 2,783 participants).

The researchers found that for predicting progression to moderate myopia and high myopia, the algorithms demonstrated robust performance (area under the receiver operating characteristic curves: 0.89, 0.89, and 0.88 for internal, EC1, and EC2, respectively, for moderate myopia; 0.95, 0.96, and 0.91 for internal, EC1, and EC2, respectively, for high myopia). For myopia onset, prediction was moderate (area under the receiver operating characteristic curve: 0.73, 0.68, and 0.72 for internal, EC1, and EC2, respectively).

"Deep learning using a single fundus photograph can effectively stratify the one-year risk of progression to moderate and high myopia, offering a powerful tool for early clinical intervention," the authors write.

More information

Press Release

Key medical concepts

High myopia

Who's behind this story?

Sadie Harley

Sadie Harley

BSc Life Sciences & Ecology. Microbiology lab background with pharmaceutical news experience in oil, gas, and renewable industries. Full profile →

Andrew Zinin

Andrew Zinin

Master's in physics with research experience. Long-time science news enthusiast. Plays key role in Science X's editorial success. Full profile →

Copyright © 2026 HealthDay. All rights reserved.

Citation: Deep learning using fundus photograph predicts myopia progression (2026, October 21) retrieved 21 October 2026 from https://medicalxpress.com/news/2026-10-deep-fundus-myopia.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.

来源:Medical Xpress · medicalxpress.com