ORIGINAL RESEARCH article
Front. Ophthalmol.
Sec. Retina
Volume 5 - 2025 | doi: 10.3389/fopht.2025.1572586
This article is part of the Research TopicReal-World Effectiveness and Challenges of Novel Anti-VEGF Therapies in Macular DiseasesView all articles
Predictive factors for treating diabetic macular edema with anti-vascular endothelial growth factor based on machine learning and deep learning
Provisionally accepted
- 1Southeast University, Nanjing, Jiangsu Province, China
- 2Zhongda Hospital, Southeast University, Nanjing, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Background To investigate the importance of predictors for the efficacy of anti-vascular endothelial growth factor (VEGF) in the treatment of diabetic macular edema (DME) based on deep learning and machine learning. Method A total of 182 eyes with VEGF treatment were enrolled in this study. Patients’ clinical information and optical coherence tomography (OCT) scanning images were collected and divided into an 80% training set and a 20% testing set. Predictive models were established based on one deep learning algorithm and six machine learning algorithms (gradient boosting, random forest, support vector machine, decision tree, K-nearest neighbor, and logistic regression) respectively, and the importance of features were compared. Result The decision tree model had the best performance with an area under the curve of 0.90 in predicting decreased central macular thickness (CMT). The gradient boosting model demonstrated the best performance, achieving an area under the curve of 0.82 in predicting the best corrected visual acuity (BCVA). Decision tree, gradient boosting, and random forest models predicted the quantity of retinal hyperreflective foci(HRF) and baseline CMT as the top 2 important predictive factors for CMT. Gradient boosting and random forest models predicted the extent of edema and the impairment of IS/OS layer as the top 2 important predictive factors for BCVA. Conclusion Predictive models based on machine learning and deep learning can better predict the decrease of CMT after anti-VEGF treatment compared to improvement of BCVA, further supported data may be required to strengthen the models’ predictive capability for BCVA improvement. Baseline CMT and the quantity of retinal HRF may be reliable predictive features of anatomical prognosis, and the extent of edema and the impairment of IS/OS layer may indicate BCVA outcomes after anti-VEGF treatment.
Keywords: Diabetic macula edema, Anti-vascular endothelial factor (anti-VEGF), machine learning (ML), deep learning, prognosis
Received: 07 Feb 2025; Accepted: 11 Jun 2025.
Copyright: © 2025 Xu, Li, Ding, Wang and Ma. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Fenglan Li, Zhongda Hospital, Southeast University, Nanjing, China
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.