Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Eye Disease Diagnosis with Deep Learning and Synthetic Data Augmentation

Saideep Kilaru, Kothamasu Jayachandra, Tanishka Yagneshwar, Suchi Kumari

TL;DR

This work tackles automated diabetic retinopathy (DR) diagnosis from fundus images by proposing an ensemble framework that combines DenseNet121 and InceptionV3 with a multi-label encoding strategy and data balancing via oversampling. A stacking meta-model fuses the base predictions, trained on the public APTOS dataset, achieving a validation accuracy of $0.99$, Cohen's Kappa of $0.98$, and F1-score of $0.99$, outperforming several established CNN architectures. The approach captures both local and global retinal features through dual architectures and transfer learning, offering a robust tool for early DR screening and potential generalization to other eye-disease classification tasks. These results suggest practical benefits for clinical decision support by enabling accurate, scalable DR detection and staging.

Abstract

In recent years, the focus is on improving the diagnosis of diabetic retinopathy (DR) using machine learning and deep learning technologies. Researchers have explored various approaches, including the use of high-definition medical imaging, AI-driven algorithms such as convolutional neural networks (CNNs) and generative adversarial networks (GANs). Among all the available tools, CNNs have emerged as a preferred tool due to their superior classification accuracy and efficiency. Although the accuracy of CNNs is comparatively better but it can be improved by introducing some hybrid models by combining various machine learning and deep learning models. Therefore, in this paper, an ensemble learning technique is proposed for early detection and management of DR with higher accuracy. The proposed model is tested on the APTOS dataset and it is showing supremacy on the validation accuracy ($99\%)$ in comparison to the previous models. Hence, the model can be helpful for early detection and treatment of the DR, thereby enhancing the overall quality of care for affected individuals.

Enhancing Eye Disease Diagnosis with Deep Learning and Synthetic Data Augmentation

TL;DR

This work tackles automated diabetic retinopathy (DR) diagnosis from fundus images by proposing an ensemble framework that combines DenseNet121 and InceptionV3 with a multi-label encoding strategy and data balancing via oversampling. A stacking meta-model fuses the base predictions, trained on the public APTOS dataset, achieving a validation accuracy of , Cohen's Kappa of , and F1-score of , outperforming several established CNN architectures. The approach captures both local and global retinal features through dual architectures and transfer learning, offering a robust tool for early DR screening and potential generalization to other eye-disease classification tasks. These results suggest practical benefits for clinical decision support by enabling accurate, scalable DR detection and staging.

Abstract

In recent years, the focus is on improving the diagnosis of diabetic retinopathy (DR) using machine learning and deep learning technologies. Researchers have explored various approaches, including the use of high-definition medical imaging, AI-driven algorithms such as convolutional neural networks (CNNs) and generative adversarial networks (GANs). Among all the available tools, CNNs have emerged as a preferred tool due to their superior classification accuracy and efficiency. Although the accuracy of CNNs is comparatively better but it can be improved by introducing some hybrid models by combining various machine learning and deep learning models. Therefore, in this paper, an ensemble learning technique is proposed for early detection and management of DR with higher accuracy. The proposed model is tested on the APTOS dataset and it is showing supremacy on the validation accuracy ( in comparison to the previous models. Hence, the model can be helpful for early detection and treatment of the DR, thereby enhancing the overall quality of care for affected individuals.
Paper Structure (16 sections, 4 equations, 7 figures, 2 tables)

This paper contains 16 sections, 4 equations, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Proposed Approach
  4. Data Acquisition and Preprocessing
  5. Convolutional Neural Network (CNN)
  6. Transfer Learning with CNNs
  7. Resampling
  8. Multi Label
  9. Architecture of the proposed ensemble model
  10. Training details and hyperparameters
  11. Evaluation and Refinement
  12. Results and Analysis
  13. Dataset description
  14. Evaluation of the proposed ensemble model on the dataset
  15. Discussion of results
  16. ...and 1 more sections

Figures (7)

  • Figure 1: Dataset Before Preprocessing
  • Figure 2: Dataset After Preprocessing
  • Figure 3: Flow diagram of the steps involved from data pre-processing to Ensemble model evaluation of the diabetic retinopathy disease.
  • Figure 4: Model Architceture
  • Figure 5: Validation loss and validation Accuracy Curve
  • ...and 2 more figures