RDE-DR: robust deep ensemble CNNs for automated diabetic retinopathy detection from fundus images
Clinical Snapshot
PICO Framework
| P — Population | Fundus images from APTOS 2019 dataset for diabetic retinopathy screening |
| I — Intervention | Robust Deep Ensemble (RDE-DR) system using four pre-trained CNNs with CLAHE enhancement |
| C — Comparator | Individual CNN models (ResNet50, VGG16, VGG19, DenseNet121) and different fusion strategies |
| O — Outcomes | Diagnostic accuracy metrics: sensitivity, specificity, accuracy, precision, recall, F1-score, and AUC |
Bottom Line
This study presents a deep learning ensemble system achieving impressive diagnostic accuracy (98.64%) for diabetic retinopathy detection using fundus images. While the technical performance appears excellent, several methodological concerns limit clinical confidence. The study lacks confidence intervals, external validation, and detailed reference standard methodology. The single-dataset validation on APTOS 2019 raises questions about generalisability to real-world screening populations. Despite high reported metrics, the absence of clinical utility assessment and cost-effectiveness analysis limits immediate clinical application. The ensemble approach shows promise for automated screening, but requires rigorous external validation, regulatory approval, and integration studies before clinical deployment. Australian clinicians should await further validation studies and TGA assessment before considering implementation in routine diabetic retinopathy screening programs.
Key Findings
P Value: Not reported
Effect Size: 98.64% accuracy with best ensemble configuration
Primary Outcome: Diagnostic accuracy for diabetic retinopathy detection
Nnt Or Sensitivity: Sensitivity (recall): 98.92%, Specificity: Not explicitly reported
Confidence Interval: Not reported
Clinical Application
Requires high-quality fundus images and computational infrastructure for CNN ensemble processing Could support RACGP diabetic retinopathy screening guidelines and reduce ophthalmologist referral burden, but requires TGA approval and integration with existing screening programs Patients requiring diabetic retinopathy screening with fundus photography
Abstract
Diabetic retinopathy (DR) is a leading cause of preventable blindness, motivating the development of reliable automated screening systems. This work proposes a Robust Deep Ensemble for Diabetic Retinopathy detection (RDE-DR) by analyzing ensemble fusion strategies. Four pre-trained convolutional neural networks (ResNet50, VGG16, VGG19, and DenseNet121) are trained using CLAHE-enhanced APTOS 2019 fundus images and integrated through seven heterogeneous fusion mechanisms, including voting-based, rank-based, and fuzzy-integral-inspired strategies. A consistent evaluation protocol is adopted, incorporating threshold optimization and probabilistic calibration analysis to validate robustness, decision margins, and accuracy-precision trade-offs. Experimental results show that multiple fusion techniques achieve comparable high performance and stable behavior on the APTOS 2019 benchmark, with the best configuration reaching 98.64% accuracy, 98.40% precision, 98.92% recall, 98.66% F1-score, and 99.78% Area-Under-Curve (AUC). Beyond peak accuracy, the study provides insights into ensemble reliability, calibration characteristics, and practical design choices for medical image classification systems. These results show that integrating transfer learning with CLAHE preprocessing and ensemble fusion yields stable experimental performance on the APTOS 2019 benchmark, suggesting potential for future medical decision support.
References
- 1.Aiche, I., Brik, Y., Attallah, B., Bouguerra, O., Rabehi, A., Habib, M., Khafaga, D. S., & El-Kenawy, E. M. (2026). RDE-DR: robust deep ensemble CNNs for automated diabetic retinopathy detection from fundus images. Scientific Reports. https://doi.org/10.18196/jrc.v6i1.25331
Related Research
Indian journal of ophthalmology
Internet of things in assistive technology for people with visual impairment: A scoping review
1 Aug 2026
BMJ open ophthalmology
Screening for diabetic retinopathy with artificial intelligence in a primary care setting: a comparative cost analysis
1 Aug 2026
Science advances
Multispectral infrared-to-full-color upconversion expanding human vision
1 Aug 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service