Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Diabetic Retinopathy Detection Using CNN with Residual Block with DCGAN

Debjany Ghosh Aronno, Sumaiya Saeha

TL;DR

Automated five-class diabetic retinopathy (DR) staging from retinal fundus images is addressed using a residual CNN architecture with DCGAN-based augmentation to mitigate class imbalance. The approach integrates advanced data synthesis for underrepresented DR stages and standard augmentation, reporting strong overall performance (about 0.987 accuracy) and robust No DR detection, while Proliferative DR remains more challenging. The study demonstrates the feasibility of scalable, automated DR screening in resource-constrained settings and highlights practical trade-offs between GAN quality, training complexity, and dataset size. Future work targets higher-fidelity GANs (including RGB variants) and larger datasets to further boost generalization and clinical utility.

Abstract

Diabetic Retinopathy (DR) is a major cause of blindness worldwide, caused by damage to the blood vessels in the retina due to diabetes. Early detection and classification of DR are crucial for timely intervention and preventing vision loss. This work proposes an automated system for DR detection using Convolutional Neural Networks (CNNs) with a residual block architecture, which enhances feature extraction and model performance. To further improve the model's robustness, we incorporate advanced data augmentation techniques, specifically leveraging a Deep Convolutional Generative Adversarial Network (DCGAN) for generating diverse retinal images. This approach increases the variability of training data, making the model more generalizable and capable of handling real-world variations in retinal images. The system is designed to classify retinal images into five distinct categories, from No DR to Proliferative DR, providing an efficient and scalable solution for early diagnosis and monitoring of DR progression. The proposed model aims to support healthcare professionals in large-scale DR screening, especially in resource-constrained settings.

Diabetic Retinopathy Detection Using CNN with Residual Block with DCGAN

TL;DR

Automated five-class diabetic retinopathy (DR) staging from retinal fundus images is addressed using a residual CNN architecture with DCGAN-based augmentation to mitigate class imbalance. The approach integrates advanced data synthesis for underrepresented DR stages and standard augmentation, reporting strong overall performance (about 0.987 accuracy) and robust No DR detection, while Proliferative DR remains more challenging. The study demonstrates the feasibility of scalable, automated DR screening in resource-constrained settings and highlights practical trade-offs between GAN quality, training complexity, and dataset size. Future work targets higher-fidelity GANs (including RGB variants) and larger datasets to further boost generalization and clinical utility.

Abstract

Diabetic Retinopathy (DR) is a major cause of blindness worldwide, caused by damage to the blood vessels in the retina due to diabetes. Early detection and classification of DR are crucial for timely intervention and preventing vision loss. This work proposes an automated system for DR detection using Convolutional Neural Networks (CNNs) with a residual block architecture, which enhances feature extraction and model performance. To further improve the model's robustness, we incorporate advanced data augmentation techniques, specifically leveraging a Deep Convolutional Generative Adversarial Network (DCGAN) for generating diverse retinal images. This approach increases the variability of training data, making the model more generalizable and capable of handling real-world variations in retinal images. The system is designed to classify retinal images into five distinct categories, from No DR to Proliferative DR, providing an efficient and scalable solution for early diagnosis and monitoring of DR progression. The proposed model aims to support healthcare professionals in large-scale DR screening, especially in resource-constrained settings.
Paper Structure (15 sections, 6 figures, 2 tables)

This paper contains 15 sections, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Problem Definition
  3. Literature Review and Gap Analysis
  4. Materials and Methods
  5. Dataset
  6. Preprocessing
  7. Data Augmentation
  8. Proposed Architecture
  9. Training Details
  10. Evaluation and Validation
  11. Results and Discussion
  12. Metrics Considered for Performance Evaluation
  13. Performance Evaluation
  14. Discussion
  15. Conclusion

Figures (6)

  • Figure 1: Image classification distribution by class
  • Figure 2: Real image
  • Figure 3: Generated fake image
  • Figure 4: Image classification distribution by class
  • Figure 5: Architecture for Image Classification
  • ...and 1 more figures