+91-9561854529 editorinchief.ijama@gmail.com submitpaper@ijama.in Contact Form

Deep Learning-Based Diabetic Retinopathy Detection Using Convolutional Neural Networks

Author(s):Vikram Singh Chauhan

Affiliation: Department of Computer Science and Engineering, Rajasthan Institute of Technology, Jaipur, Rajasthan, India

Page No: 42-48

Volume issue & Publishing Year: Volume 3, Issue 6, June 2026

published on: 2026/06/12

Journal: International Journal of Advanced Multidisciplinary Application.(IJAMA)

ISSN NO: 3048-9350

DOI:

Download PDF

Abstract:
Diabetic Retinopathy (DR) is a leading cause of preventable blindness worldwide, affecting approximately 463 million people with diabetes globally, of whom an estimated 77.2 million reside in India. Early automated detection of DR from digital fundus photographs using deep learning methods offers a scalable, cost-effective solution for mass screening programmes in resource-constrained settings. This study presents a comparative evaluation of five convolutional neural network (CNN) architectures — VGG-16, ResNet-50, InceptionV3, DenseNet-121, and a custom-designed lightweight CNN — for four-class DR severity grading (No DR, Mild, Moderate, Severe/Proliferative) on the publicly available APTOS-2019 retinal fundus image dataset (3,662 images). All models were fine-tuned using transfer learning with ImageNet pretrained weights. Preprocessing steps including contrast-limited adaptive histogram equalization (CLAHE), green channel extraction, circular cropping, and data augmentation were applied uniformly. The proposed lightweight CNN achieves 95.3% accuracy, 94.1% precision, 94.7% recall, and 94.4% F1-score on the held-out test set, outperforming all benchmark architectures. Gradient-weighted Class Activation Mapping (Grad-CAM) visualizations confirm the model's attention to clinically relevant regions including microaneurysms, haemorrhages, and hard exudates. The model attains a mean AUC of 0.977 across all four classes. These results demonstrate that the proposed lightweight architecture, with 6.2 million parameters versus ResNet-50's 25.6 million, achieves superior diagnostic accuracy with substantially reduced computational overhead, making it suitable for deployment on mobile screening devices.

Keywords: diabetic retinopathy, deep learning, convolutional neural network, transfer learning, retinal fundus image, CLAHE, Grad-CAM, APTOS-2019, medical image classification

Reference:

  • [1] Gargeya, R., & Leng, T. (2017). Automated identification of diabetic retinopathy using deep learning. Ophthalmology, 124(7), 962-969.
  • [2] Graham, B. (2015). Kaggle diabetic retinopathy detection competition report. University of Warwick.
  • [3] Gulshan, V., Peng, L., Coram, M., Stumpe, M. C., Wu, D., Narayanaswamy, A., ... & Webster, D. R. (2016). Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA, 316(22), 2402-2410.
  • [4] He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Proceedings of the IEEE CVPR, 770-778.
  • [5] Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017). Densely connected convolutional networks. Proceedings of the IEEE CVPR, 4700-4708.
  • [6] IDF. (2021). IDF Diabetes Atlas, 10th edition. International Diabetes Federation, Brussels.
  • [7] Kauppi, T., Kalesnykiene, V., Kamarainen, J. K., Lensu, L., Sorri, I., Raninen, A., ... & Uusitalo, H. (2007). DIARETDB1 diabetic retinopathy database and evaluation protocol. BMVC.
  • [8] Selvaraju, R. R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., & Batra, D. (2017). Grad-CAM: Visual explanations from deep networks via gradient-based localization. Proceedings of the IEEE ICCV, 618-626.
  • [9] Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
  • [10] Sinthanayothin, C., Boyce, J. F., Williamson, T. H., Cook, H. L., Mensah, E., Lal, S., & Usher, D. (2002). Automated detection of diabetic retinopathy on digital fundus images. Diabetic Medicine, 19(2), 105-112.
  • [11] Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016). Rethinking the inception architecture for computer vision. Proceedings of the IEEE CVPR, 2818-2826.
  • [12] Ting, D. S. W., Cheung, C. Y. L., Lim, G., Tan, G. S. W., Quang, N. D., Gan, A., ... & Wong, T. Y. (2017). Development and validation of a deep learning system for diabetic retinopathy and related eye diseases using retinal images from multiethnic populations with diabetes. JAMA, 318(22), 2211-2223.
  • [13] Zhao, Z., Zhang, K., Hao, X., Tian, J., Chua, M. C. H., Chen, L., & Xu, X. (2019). Bira-net: Bilinear attention net for diabetic retinopathy grading. IEEE International Conference on Image Processing, 1385-1389.