Prototype of a remote automated screening system for diabetic retinopathy

The article presents the results of the development and testing of a prototype automated system for remote screening of diabetic retinopathy (DR), featuring preprocessing and classification functionalities for digital images acquired by fundus cameras from various manufacturers. Experimental testing of the prototype was performed on several thousand real-world images from Belarusian medical institutions and open-access sources. Objective assessment of retinal image quality is computed through analysis of the central fragment based on the Weibull distribution scale parameter applied to local quality estimates of the green channel of the image. The prototype was developed on the Streamlit platform and employs the YOLOv8 and YOLOv11 neural network architectures for classifying five stages of DR, achieving classification accuracy of 84.0 % and 82.7 %, respectively. The combination of low-quality image filtering and neural network-based classification enhances the reliability of screening diagnostics and can be deployed in telemedicine systems with limited computational resources.

1. World Health Organization. Regional Office for Europe. Diabetic retinopathy screening: a short guide: Increase effectiveness, maximize benefits and minimize harm. 2020. Available at: https://iris.who.int/handle/10665/336660 (accessed 10 April 2026).

2. GBD 2019 Blindness and Vision Impairment Collaborators; Vision Loss Expert Group of the Global Burden of Disease Study. Causes of blindness and vision impairment in 2020 and trends over 30 years, and prevalence of avoidable blindness in relation to VISION 2020: the Right to Sight: an analysis for the Global Burden of Disease Study. Lancet Global Health. 2021;9(2):e144–e160. https://doi.org/10.1016/S2214-109X(20)30489-7

3. Skol'ko patsientov s sakharnym diabetom nakhodyatsya pod mednablyudeniem v Belarusi, rasskazala ehndokrinolog [How many patients with diabetes mellitus are under medical supervision in Belarus, the endocrinologist said]. BELTA. Available at: https://belta.by/society/view/skolko-patsientov-s-saharnym-diabetom-nahodjatsja-pod-mednabljudeniem-v-belarusi-rasskazala-746976-2025/ (accessed 10 April 2026) (in Russian).

4. Lukashevich M.M., Golub Y.I. Classification of diabetic retinopathy stages based on neural networks. System Analysis and Applied Information Science. 2022;(3):12–21 (in Russian). https://doi.org/10.21122/2309-4923-2022-3-12-21.

5. Abràmoff M.D., Lavin P.T., Birch M., Shah N., Folk J.C. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. npj Digital Medicine. 2018;1:39. https://doi.org/10.1038/s41746-018-0040-6.

6. Scanlon P.H. The English National Screening Programme for diabetic retinopathy 2003–2016. Acta Diabetologica. 2017;54:515–525. https://doi.org/10.1007/s00592-017-0974-1.

7. Gangwani R.A., Lian J.X., McGhee S.M., Wong D., Li K.K.W. Diabetic retinopathy screening: global and local perspective. Hong Kong Medical Journal. 2016;22(5):486–495. http://dx.doi.org/10.12809/hkmj164844.

8. Li T., Gaoet Y., Wang K., Guo S., Liu H., Kang H. Diagnostic assessment of deep learning algorithms for diabetic retinopathy screening. Information Sciences. 2019;501:511–522. https://doi.org/10.1016/j.ins.2019.06.011.

9. Golub Y., Lukashevich M., Starovoitov V. Retinal image analysis approach for diabetic retinopathy grading. Pattern Recognition and Information Processing. PRIP 2021. Communications in Computer and Information Science, vol. 1562. Springer, Cham; 2022. p. 152–165. https://doi.org/10.1007/978-3-030-98883-8_11.

10. Starovoitov V.V., Golub Yu.I., Lukashevich M.M. A universal retinal image template for automated screening of diabetic retinopathy. Pattern Recognition and Image Analysis. 2022;32:322–331. https://doi.org/10.1134/S1054661822020195.

11. Lukashevich M.M., Golub Yu.I., Starovoitov V.V. Klassifikacii stadij diabeticheskoj retinopatii na osnove algoritmov mashinnogo obucheniya i nabora priznakov [Classifications of diabetic retinopathy stages based on machine learning algorithms and feature set]. Informacionnye sistemy i tehnologii = Information Systems and Technologies : materialy mezhdunar. nauch. kongressa po informatike [Proceedings of the International Scientific Congress on Informatics], 27–28 Oct. 2022, Republic of Belarus, Minsk. In 3 parts. Part 2. Minsk; 2022. p. 169-176 (in Russian).

12. Golub Yu.I., Starovoitov V. Objective quality assessment of digital retinal images in screening study. System Analysis and Applied Information Science. 2025;(3):47–58 (in Russian). https://doi.org/10.21122/2309-4923-2025-3-47-58.

13. APTOS 2019 Blindness Detection. Kaggle. 2019. Available at: https://www.kaggle.com/c/aptos2019-blindness-detection (accessed 10 April 2026).

14. Farahat Z., Zrira N., Souissi N., Belmekki M., Ngote M.N., Megdiche K., et al. Diabetic retinopathy screening through artificial intelligence algorithms: A systematic review. Survey of Ophthalmology. 2024;69(5):707–721. https://doi.org/10.1016/j.survophthal.2024.05.008.

15. Pavlov V.G., Sidamonidze A.L., Petrachkov D.V. Current trends in the screening for diabetic retinopathy. Russian Annals of Ophthalmology. 2020;136(4):300–309 (in Russian). https://doi.org/10.17116/oftalma2020136042300.