Interpretable Diabetic Retinopathy Classification Using Hybrid ConvNeXt–MLP-Mixer and ConvNeXt–PoolFormer Architectures
DOI:
https://doi.org/10.63158/journalisi.v8i5.1863Keywords:
Diabetic Retinopathy, Fundus Image Classification, ConvNeXt, MLP-Mixer, PoolFormer, Explainable AI, Medical Image Analysis, Multi-class ClassificationAbstract
Diabetic retinopathy (DR) severity grading from retinal fundus images is challenging because retinal abnormalities vary in size, distribution, and severity. This study evaluates two hybrid architectures, ConvNeXt–MLP-Mixer and ConvNeXt–PoolFormer, for five-class DR classification by combining convolutional feature extraction with global feature mixing. Experiments used the APTOS 2019 Blindness Detection dataset from Kaggle. After exact-duplicate filtering, 3,596 images remained: 2,502 for training, 544 for validation, and 550 for held-out testing. The imbalanced test set comprised five DR severity classes. The proposed models were compared with EfficientNet, Vision Transformer, ConvNeXt, MLP-Mixer, and PoolFormer using accuracy, macro- and weighted F1-scores, and macro- and weighted-average one-versus-rest (OvR) ROC-AUC. ConvNeXt–PoolFormer achieved 82.7% test accuracy and 82.7% weighted F1-score, while ConvNeXt–MLP-Mixer achieved 82.4% accuracy and 82.5% weighted F1-score, with the latter obtaining the highest macro F1-score of 67.5%. Moderate DR showed the clearest class-specific improvement, whereas Severe and Proliferative DR remained difficult to classify. Layer-wise Relevance Propagation (LRP) and Grad-CAM generated post-hoc visual explanation maps for representative predictions. These visualizations were treated as exploratory because they were not independently validated using lesion-level annotations or by clinical experts, including ophthalmologists.
Downloads
References
[1] H. Li et al., “Current research and future strategies for the management of vision-threatening diabetic retinopathy,” Asia-Pacific J. Ophthalmol., vol. 13, no. 5, p. 100109, Sep. 2024, doi: 10.1016/j.apjo.2024.100109.
[2] Y. Meng et al., “Global, Regional, and National Burden of Blindness due to Diabetic Retinopathy, 1990–2021,” Ophthalmol. Ther., vol. 14, no. 10, pp. 2599–2615, Oct. 2025, doi: 10.1007/s40123-025-01230-y.
[3] T. Zikhali, C. Kalinda, and Z. N. Xulu-Kasaba, “Screening of Diabetic Retinopathy Using Teleophthalmology to Complement Human Resources for Eye Health: A Systematic Review and Meta-Analysis,” Clin. Pract., vol. 12, no. 4, pp. 457–467, Aug. 2022, doi: 10.3390/clinpract12040050.
[4] M. A. Raza, M. Dua, and K. Berwal, “Early Detection of Diabetic Retinopathy Using Pretrained Models: A Focus on Initial Stages,” in 2024 1st International Conference on Advanced Computing and Emerging Technologies, ACET 2024, Institute of Electrical and Electronics Engineers Inc., 2024. doi: 10.1109/ACET61898.2024.10730746.
[5] Y. Miao, S. Tang, P. Du, and Z. Li, “Research on Deep Learning in the Detection and Classification of Diabetic Retinopathy,” in 2021 IEEE International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology, CEI 2021, Institute of Electrical and Electronics Engineers Inc., Sep. 2021, pp. 107–113. doi: 10.1109/CEI52496.2021.9574509.
[6] S. Zhu, C. Xiong, Q. Zhong, and Y. Yao, “Diabetic Retinopathy Classification With Deep Learning via Fundus Images: A Short Survey,” IEEE Access, vol. 12, pp. 20540–20558, 2024, doi: 10.1109/ACCESS.2024.3361944.
[7] N. S. Kumar and B. Ramaswamy Karthikeyan, “Diabetic Retinopathy Detection using CNN, Transformer and MLP based Architectures,” in ISPACS 2021 - International Symposium on Intelligent Signal Processing and Communication Systems: 5G Dream to Reality, Proceeding, Institute of Electrical and Electronics Engineers Inc., 2021, pp. 1–2. doi: 10.1109/ISPACS51563.2021.9651024.
[8] M. Tan and Q. V. Le, “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” in Proceedings of the 36th International Conference on Machine Learning, ICML 2019, 2019, pp. 6105–6114.
[9] A. Dosovitskiy et al., “An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale,” in International Conference on Learning Representations, 2021.
[10] N. S. Kumar, R. K. Balasubramanian, and M. R. Phirke, “Image Transformers for Diabetic Retinopathy Detection from Fundus Datasets,” Revue d’Intelligence Artificielle, vol. 37, no. 6, pp. 1617–1627, Dec. 2023, doi: 10.18280/ria.370626.
[11] Z. Liu, H. Mao, C. Y. Wu, C. Feichtenhofer, T. Darrell, and S. Xie, “A ConvNet for the 2020s,” in Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, IEEE Computer Society, 2022, pp. 11966–11976. doi: 10.1109/CVPR52688.2022.01167.
[12] Devanshi, S. K. Baliarsingh, and P. P. Dev, “An Early diagnosis of diabetic retinopathy using ConvNeXt,” in Proceedings of the 10th International Conference on Signal Processing and Integrated Networks, SPIN 2023, Institute of Electrical and Electronics Engineers Inc., 2023, pp. 739–743. doi: 10.1109/SPIN57001.2023.10116214.
[13] I. Tolstikhin et al., “MLP-Mixer: An all-MLP Architecture for Vision,” in Advances in Neural Information Processing Systems, 2021, pp. 24261–24272.
[14] W. Yu et al., “MetaFormer is Actually What You Need for Vision,” in Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, IEEE Computer Society, 2022, pp. 10809–10819. doi: 10.1109/CVPR52688.2022.01055.
[15] H. Jiang et al., “A Multi-Label Deep Learning Model with Interpretable Grad-CAM for Diabetic Retinopathy Classification,” in Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, 2020, pp. 1560–1563. doi: 10.1109/EMBC44109.2020.9175884.
[16] R. Singh, “Understanding Image Classification Tasks Through Layerwise Relevance Propogation,” in Proceedings - 2022 IEEE 18th International Conference on Intelligent Computer Communication and Processing Conference, ICCP 2022, Institute of Electrical and Electronics Engineers Inc., 2022, pp. 199–203. doi: 10.1109/ICCP56966.2022.10053963.
[17] I. Ullah, A. Rios, V. Gala, and S. McKeever, “Explaining Deep Learning Models for Tabular Data Using Layer-Wise Relevance Propagation,” Applied Sciences, vol. 12, no. 1, p. 136, Jan. 2022, doi: 10.3390/app12010136.
[18] X. Luo et al., “A deep convolutional neural network for diabetic retinopathy detection via mining local and long-range dependence,” CAAI Trans. Intell. Technol., vol. 9, no. 1, pp. 153–166, 2024, doi: 10.1049/cit2.12155.
[19] R. Katarya, P. Sharma, N. Soni, and P. Rath, “A Review of Interpretable Deep Learning for Neurological Disease Classification,” in 8th International Conference on Advanced Computing and Communication Systems, ICACCS 2022, Institute of Electrical and Electronics Engineers Inc., 2022, pp. 900–906. doi: 10.1109/ICACCS54159.2022.9785321.
[20] A. M. N. Hirushan and S. P. K. Arachchi, “Integrate Interpretable AI into Chest X ray Diagnosis through a Comparative Analysis of CNN Architectures with Grad CAM Visualizations,” in 9th International Conference on Advances in Technology and Computing, ICATC 2024, Institute of Electrical and Electronics Engineers Inc., 2024. doi: 10.1109/ICATC64549.2024.11025232.
[21] Asia Pacific Tele-Ophthalmology Society, “APTOS 2019 blindness detection,” Kaggle Competition. Accessed: Aug. 08, 2026. [Online]. Available: https://www.kaggle.com/competitions/aptos2019-blindness-detection
[22] O. Cîmpeanu, I. Mihaela Liliac, M. Mămuleanu, Ștefan-V. Voinea, M. Olteanu, and C.-T. Streba, “Using Convoluted Neural Networks in Diagnosing Lung Cancer on Computed Tomography Scans,” Curr. Health Sci. J., vol. 51, no. 1, pp. 87–95, 2025, doi: 10.12865/chsj.51.01.09.
[23] C. Hu, N. Cao, H. Zhou, and B. Guo, “Medical Image Classification with a Hybrid SSM Model Based on CNN and Transformer,” Electronics (Basel)., vol. 13, no. 15, p. 3094, 2024, doi: 10.3390/electronics13153094.
[24] X. Huang et al., “A CNN-Transformer fusion network for Diabetic retinopathy image classification,” Comput. Med. Imaging Graph., vol. 126, p. 102655, 2025, doi: 10.1016/j.compmedimag.2025.102655.
[25] P. Sekar, R. Bhoopalan, N. Nagaprasad, T. R. Mamo, S. P. Dhanabal, and R. Krishnaraj, “D-TNet: a hybrid Dense Net-transformer model for robust diabetic retinopathy detection,” Sci. Rep., vol. 15, no. 1, p. 39594, 2025, doi: 10.1038/s41598-025-23234-1.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Journal of Information Systems and Informatics
This work is licensed under a Creative Commons Attribution 4.0 International License.
Author's Declaration
- The Authors certify that they have read, understood, and agreed to the Journal of Information Systems and Informatics (JournalISI) submission guidelines, policies, and submission declaration. The submission has been prepared using the provided template.
- The Authors certify that all authors have approved the publication of this manuscript and that there is no conflict of interest.
- The Authors confirm that the manuscript is their original work, has not received prior publication, is not under consideration for publication elsewhere, and has not been previously published.
- The Authors confirm that all authors listed on the title page have contributed significantly to the work, have read the manuscript, attest to the validity and legitimacy of the data and its interpretation, and agree to its submission.
- The Authors confirm that the manuscript is not copied from or plagiarized from any other published work.
- The Authors declare that the manuscript will not be submitted for publication in any other journal or magazine until a decision is made by the journal editors.
- If the manuscript is finally accepted for publication, the Authors confirm that they will either proceed with publication immediately or withdraw the manuscript in accordance with the journal’s withdrawal policies.
- The Authors agree that, upon publication of the manuscript in this journal, they transfer copyright or assign exclusive rights to the publisher, including commercial rights
