Shivani Sista

Background: Diabetic retinopathy (DR) is a microangiopathy of the retina resulting from chronic hyperglycemia.¹ It is the leading cause of blindness in the working-age population, and the fifth leading cause of preventable blindness worldwide.^1,2 DR is characterized by progressive capillary damage and retinal ischemia, and advanced stages are marked by neovascularization on the optic disc or retina which contributes to severe vision loss.^2,8 In many cases, DR remains asymptomatic until these advanced stages, highlighting the need for early screening to enable timely intervention.³ Traditional screening methods like dilated fundoscopic exams can be time-consuming and prone to human error, prompting investigation into computational image analysis methods.^2,3

Methods: Literature review was conducted through a PubMed database search using the keywords “diabetic retinopathy,” “diabetic retinopathy computational methods,” and “diabetic retinopathy detection.” Results were filtered from 2020-2025.

Results: Deep learning detection systems using ultra-wide fundus (UWF) images have achieved an accuracy of 83.38% and an area under the curve (AUC) of 91.5%, with use of the Early Treatment Diabetic Retinopathy Study 7-standard field (ETDRS 7SF) region producing the most optimal results.⁴ While UWF images most effectively capture the retinal surface, they are costly and less widely available in low-resource settings. Other deep learning studies used more easily-accessible standard fundus images to train a convolutional neural network with lesion-level detection, and obtained an accuracy of 99.1% and an AUC of 99.9%.⁵ Apart from deep learning, 3D optical coherence tomography computer-aided diagnosis (3D-OCT CAD) systems based on a higher-order spatial appearance model had an accuracy of 96.88%, outperforming other machine learning classifiers on the same dataset across multiple cross-validation settings.⁶ Finally, stacked auto-encoders have been used to detect five different stages of DR with an accuracy of 88% on a 75:25 train-test ratio.⁷ This use of unsupervised learning enabled feature extraction from even noisy and limited data, and multi-stage DR classification provided more detailed, clinically-relevant predictions over binary classifiers.⁷

Conclusion: Recent advances in early DR detection have shown significant improvements from traditional screening methods. The accuracy of 3D-OCT CAD methods and deep learning frameworks demonstrate high diagnostic potential, while unsupervised machine learning models have shown promise in extracting meaningful features from limited annotated datasets. Overall, these findings showcase the value of integrating computational methods into current DR screening procedures to ensure timely intervention and prevent vision loss.

Works Cited:

1. Kropp M, Golubnitschaja O, Mazurakova A, et al. Diabetic retinopathy as the leading cause of blindness and early predictor of cascading complications-risks and mitigation. EPMA J. 2023;14(1):21-42. Published 2023 Feb 13. doi:10.1007/s13167-023-00314-8
2. Chong DD, Das N, Singh RP. Diabetic retinopathy: Screening, prevention, and treatment. Cleve Clin J Med. 2024;91(8):503-510. Published 2024 Aug 1. doi:10.3949/ccjm.91a.24028
3. Fung TH, Patel B, Wilmot EG, Amoaku WM. Diabetic retinopathy for the non-ophthalmologist. Clin Med (Lond). 2022;22(2):112-116. doi:10.7861/clinmed.2021-0792
4. Oh K, Kang HM, Leem D, Lee H, Seo KY, Yoon S. Early detection of diabetic retinopathy based on deep learning and ultra-wide-field fundus images. Scientific Reports. 2021;11(1). doi:https://doi.org/10.1038/s41598-021-81539-3
5. Erciyas A, Barışçı N. An Effective Method for Detecting and Classifying Diabetic Retinopathy Lesions Based on Deep Learning. Rajinikanth V, ed. Computational and Mathematical Methods in Medicine. 2021;2021:1-13. doi:https://doi.org/10.1155/2021/9928899
6. Elsharkawy M, Sharafeldeen A, Soliman A, et al. A Novel Computer-Aided Diagnostic System for Early Detection of Diabetic Retinopathy Using 3D-OCT Higher-Order Spatial Appearance Model. Diagnostics (Basel). 2022;12(2):461. Published 2022 Feb 11. doi:10.3390/diagnostics12020461
7. Almas S, Wahid F, Ali S, et al. Visual impairment prevention by early detection of diabetic retinopathy based on stacked auto-encoder. Scientific Reports. 2025;15(1). doi:https://doi.org/10.1038/s41598-025-85752-2
8. Wong T, Sun J, Kawasaki R, et al. Guidelines on diabetic eye care the International Council of Ophthalmology recommendations for screening, follow-up, referral, and treatment based on resource settings. Ophthalmology. 2018;125(10):1608-1622. doi:https://doi.org/10.1016/j.ophtha.2018.04.007