Saratov JOURNAL of Medical and Scientific Research
Home

Interpretable classifcation method for fundus image analysis in primary ophthalmological screening

Year: 2026, volume 22 Issue: №2 Pages: 145-149
Heading: Ophthalmology Article type: Original article
Authors: Dolinina O.N., Arutyunyan D.A., Kochkareva E.I., Kamenskikh T.G., Kolbenev I.O., Veselova E.V.
Organization: Ulyanovsk State Technical University, Ulyanovsk, Russia, Saratov State Medical University
Summary:

Objective: to develop and experimentally evaluate an interpretable fundus image classifcation method for primary ophthalmic screening based on the joint use of a deep learning model and visual explanation techniques. Material and methods. The study was conducted on a dataset of 2,500 fundus images balanced across 5 classes (500 images per class): degenerative retinal changes, macular pathology, neuro-ophthalmic pathology, retinal vascular pathology, and the absence of clinically signifcant pathology. A convolutional neural network with a ConvNeXt-like architecture was used as the classifer. Classifcation performance was assessed using Accuracy, while interpretability quality was evaluated using the Intersection over Union (IoU) metric. Results. The baseline model achieved Accuracy values ranging from 0.74 to 0.77, whereas the model incorporating the interpretability component achieved values between 0.79 and 0.82. The interpretability quality of the baseline model reached an IoU of approximately 0.79, while the proposed method achieved IoU values of 0.98–0.99. Conclusion. The proposed method provides not only fundus image classifcation but also a visual explanation of the obtained results. Incorporating expert annotations into the training process improves the correspondence between attention maps and clinically relevant image regions and may increase trust in clinical decision support systems for primary ophthalmic screening.

Keywords: saliency map, ophthalmologic screening, interpretable classifcation, fundus image, deep learning

Bibliography:
1. Burton MJ, Ramke J, Marques AP, et al. The Lancet Global Health Commission on Global Eye Health: Vision beyond 2020. Lancet Global Health. 2021;9(4):489-551.
2. Zheng S, Choo J, Chen J, et al. The barriers to global eye care equity and the role of digital innovations. Adv Ophthalmol Pract Res. 2021;1(2):100021. DOI:10.1016/j.aopr.2021.100021
3. Akhtar S, Aftab S, Ali O, et al. A deep learning-based model for diabetic retinopathy grading. Sci Rep. 2025;15(1):3763. DOI:10.1038/s41598-025-87171-9
4. Duan MM, Tu X. Deep learning-based classifcation of multiple fundus diseases using ultra-widefeld images. Front Cell Dev Biol. 2025;13:1630667. DOI:10.3389/fcell.2025.1630667
5. Mienye ID, Swart TG, Obaido G, et al. Deep convolutional neural networks in medical image analysis: A review. Information (Switzerland). 2025;16(3):195. DOI:10.3390/info16030195
6. Sebastian A, Elharrouss O, Al-Maadeed S, Al-Maade-ed NA. Survey on deep-learning-based diabetic retinopathy classifcation. Diagnostics (Basel). 2023;13(3):345. DOI:10.3390/diagnostics13030345
7. Tjoa E, Guan C. A survey on explainable artifcial intelligence (XAI): Toward medical XAI. IEEE Trans Neural Netw Learn Syst. 2021;32(11):4793-813. DOI:10.1109/TNNLS.2020.3027314
8. Abdallah HH, Nagibin SYa. Search and analysis of explainable artifcial intelligence methods in machine learning problems: Classifcation and interpretation of models. Information Technology and Applied Mathematics. 2024;1:113.
9. Boulaabi M, Ben Aïcha Gader T, Kacem Echi A, Mbarek S. Advanced Segmentation of Diabetic Retinopathy Lesions Using DeepLabv3+. 2024 IEEE/ACS 21st International Conference on Computer Systems and Applications (AICCSA 2024). 2024 arXiv:2504.17306 [cs.CV]. DOI:10.48550/arXiv.2504.17306
10. Holzinger A, Carrington A, Müller H. Measuring the quality of explanations: The System Causability Scale (SCS): Comparing human and machine explanations. KI – Künstliche Intelligenz. 2021;2(34):193-8. arXiv:1912.09024 [cs.AI]. DOI:10.48550/arXiv.1912.09024
11. Xie X, Niu J, Liu X, et al. A survey on incorporating domain knowledge into deep learning for medical image analysis. Med Image Anal. 2021;11(69):101985. DOI:10.1016/j. media.2021.101985
12. Cheng B, Girshick R, Dollár P, et al. Boundary IoU: Improving object-centric image segmentation evaluation. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2021. arXiv:2103.16562 [cs. CV]. DOI:10.48550/arXiv.2103.16562
13. Miller C, Portlock T, Nyaga DM., O’Sullivan JM. A review of model evaluation metrics for machine learning in genetics and genomics. Front Bioinform. 2024:4:1457619. DOI:10.3389/ fbinf.2024.1457619
14. Rainio O, Teuho J, Klén R. Evaluation metrics and statistical tests for machine learning. Sci Rep. 2024;14(1):6086. DOI:10.1038/s41598-024-56706-x
15. Esteva A, Robicquet A, Ramsundar B, et al. A guide to deep learning in healthcare. Nat Med. 2019;25(1):24-9. DOI:10.1038/ s41591-018-0316-z
16. 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):144-60.
17. Jiang H, Gao M, Liu Z, et al. GlanceSeg: Real time microaneurysm lesion segmentation with gaze map guided foundation model for early detection of diabetic retinopathy. IEEE Journal of Biomedical and Health Informatics. 2024;2(1):1-14.
18. Rieger L, Singh C, Murdoch WJ, Yu B. Interpretations are useful: penalizing explanations to align neural networks with prior knowledge. Proceedings of the 37th International Conference on Machine Learning (ICML). 2020;119:8116-26.
19. Ross AS, Hughes MC, Doshi-Velez F. Right for the right reasons: training diferentiable models by constraining their explanations. Proceedings of the 26th International Joint Conference on Artifcial Intelligence (IJCAI). 2022. arXiv:1703.03717 [cs.LG]. DOI:10.48550/arXiv.1703.03717
20. Sebastian A, Elharrouss O, Al-Maadeed S, Al-Maade-ed NA. A survey on diabetic retinopathy lesion detection and segmentation. Appl Sci. 2023;13(8):51. DOI:10.3390/app13085111

AttachmentSize
02_maket_145-149.pdf431.51 KB
All articles of the authors: Veselova E.V., Kolbenev I.O., Kochkareva E.I., Kamenskikh T.G., Dolinina O.N. , Arutyunyan D.A.

No votes yet