CBAM增强卷积网络在ODIR数据集上用于糖尿病视网膜病变二分类的评估
CBAM-enhanced convolutional networks for diabetic retinopathy classification in the ODIR dataset
一项基于公开ODIR数据集的研究评估了CBAM增强卷积网络在糖尿病视网膜病变二分类中的性能,其准确率、F1值均为92.5%,AUC达0.96。该模型文件仅35.2 MB,推理时间为28毫秒,证明了紧凑注意力分类器在受控图像级评估中的可行性。研究指出,由于缺乏外部验证和临床疗效证明,该结果不能确立完全的临床有效性。
- Article
- Open access
- Published: 13 October 2026
Scientific Reports (2026) Cite this article
We’re sharing this article early to provide faster access to peer-reviewed, accepted research. It is citable and carries a permanent DOI. This version is subject to further edits and will be replaced automatically by the final Version of Record. All legal disclaimers apply.
Abstract
Attention modules may improve fundus-image classification by reweighting informative channels and spatial regions. We evaluated a Convolutional Block Attention Module (CBAM)-enhanced convolutional network for binary diabetic retinopathy classification using preprocessed images from the Ocular Disease Recognition dataset and compared it with VGG-11, VGG-13, ResNet-18, ResNet-34, and ResNet-50 under the reported common training conditions. On the fixed 1,000-image test set, the CBAM model produced 465 true positives, 460 true negatives, 35 false negatives, and 40 false positives, corresponding to 92.5% accuracy, 92.1% precision, 93.0% sensitivity, and a 92.5% F1-score; the reported AUC was 0.96. The model also had a reported inference time of 28 ms, a 35.2 MB file size, and 1.8 billion floating-point operations. These results support the feasibility of a compact attention-enhanced classifier in this controlled image-level evaluation. They do not establish full clinical effectiveness because the study used one public dataset and one archived split, lacked external validation and component-wise ablation, and did not retain paired lesion-level explanation outputs.
Funding
This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF), funded by the Ministry of Education (NRF-RS-2023-00237287, 2026S1A5A01004224).
Author information
Authors and Affiliations
Worker’s Care & Digital Health Lab, Department of Future Technology, Korea University of Technology and Education, Cheonan, 31253, South Korea
Haewon Byeon
Authors
- Haewon Byeon
Corresponding author
Correspondence to Haewon Byeon.
Ethics declarations
Ethics approval and consent to participate
Source documentation states that direct patient identifiers were removed and that release followed applicable ethical and privacy rules in China31. This secondary analysis used only publicly available, de-identified records and involved no new participant recruitment, intervention, or access to direct identifiers.
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Byeon, H. CBAM-enhanced convolutional networks for diabetic retinopathy classification in the ODIR dataset. Sci Rep (2026). https://doi.org/10.1038/s41598-026-75629-3
Received: 09 September 2025
Accepted: 08 October 2026
Published: 13 October 2026
DOI: https://doi.org/10.1038/s41598-026-75629-3
Keywords
来源:Nature 健康护理主题 · nature.com