A Robust and Explainable Deep CNN Framework for Multi-Disease Detection in Retinal Fundus Images

The current paper has proposed a useful and explainable deep convoluted neural network (CNN) architecture of automated detection and classification of different eye diseases using retinal fundus images. The primary issues related to computer-aided ophthalmic diagnosis that the given approach was expected to address were the multi-disease classification, lack of data, generalization of the model, and non-interpretability. The framework is a composite of modern data augmentation strategies, and a deep CNN model, and thus, the framework manages to identify discriminatory properties of different retinal appearances, including normal, diabetic retinopathy, glaucoma, cataract, and age-related macular degeneration. The results of the experiment reveal that the proposed framework is efficient in regard to high classification accuracy, precision, recall, and F1-score across all disease categories compared to the existing state-of-the-art methods. The use of the strong augmentation strategies is associated with significant improvements in the generalization and reduction of overfitting in cases where the datasets are insufficient and unbalanced. In addition, the implementation of the visual explanations according to Grad-CAM will provide meaningful information regarding the process of decision-making of the model by demonstrating the existence of the clinically relevant retinal areas, hence increasing the transparency and adding to the clinical trust. The proposed explainable CNN system has high chances of being an effective decision-supporting system to carry out automated screening of the retinal diseases as well as early diagnosis. The framework will be extended to larger and more heterogeneous data in the future directions, multi-modal ophthalmic data, and validated in clinical settings to enhance its usefulness and credibility.

1.    S. Ejaz, R. Baig, Z. Ashraf, M. M. Alnfiai, M. M. Alnahari, and R. M. Alotaibi, “A deep learning framework for the early detection of multi-retinal diseases,” PLOS ONE, vol. 19, no. 7, p. e0307317, 2024, doi: 10.1371/journal.pone.0307317. 2.    G. Boscia et al., “Long-term impact of diabetic retinopathy on response to anti-VEGF treatment in neovascular AMD,” Investigative Ophthalmology & Visual Science, vol. 65, p. 6, 2024. 3.    H. Bilal, A. Keles, and M. Bendechache, “Advances in disease detection through retinal imaging: A systematic review,” Computers in Biology and Medicine, vol. 194, p. 110412, 2025, doi: 10.1016/j.compbiomed.2025.110412. 4.    L. A. Everett and Y. M. Paulus, “Laser therapy in the treatment of diabetic retinopathy and diabetic macular edema,” Current Diabetes Reports, vol. 21, p. 35, 2021. 5.    N. Devi, P. L. Rani, and A. R. Guru Gokul, “Multi-disease classification of retinal images using convolutional neural network,” International Journal of Computer Applications, vol. 186, no. 31, pp. 35–42, Jul. 2024, doi: 10.5120/ijca2024923850. 6.    D. Wang, J. Lian, and W. Jiao, “Multi-label classification of retinal disease via a novel vision transformer model,” Frontiers in Neuroscience, vol. 17, p. 1290803, 2024. 7.    P. Li, J. Tao, Q. Yuan, R. Zhang, and P. Gao, “Research progress in deep learning-based fundus image analysis for the diagnosis and prediction of hypertension-related diseases,” Frontiers in Cell and Developmental Biology, vol. 13, p. 1608994, 2025, doi: 10.3389/fcell.2025.1608994. 8.    M. M. Aghabeigi Alooghareh, M. M. Sheikhey, A. Sahafi, H. Pirnejad, and A. Naemi, “Deep learning for comprehensive analysis of retinal fundus images: Detection of systemic and ocular conditions,” Bioengineering, vol. 12, no. 8, p. 840, 2025, doi: 10.3390/bioengineering12080840. 9.    Y. Zhou et al., “A foundation model for generalizable disease detection from retinal images,” Nature, vol. 622, pp. 156–163, 2023. 10. S. Takahashi et al., “Comparison of vision transformers and convolutional neural networks in medical image analysis: A systematic review,” Journal of Medical Systems, vol. 48, p. 84, 2024.