Retinal Image Classification and Synthetic Data Generation

Welcome to our repository dedicated to advancing the field of ophthalmology through deep learning. We've compiled three comprehensive Jupyter notebooks, each focusing on a different aspect of retinal image analysis and synthetic data generation.

Overview of Features

1. AMD Classification using ResNet-18:

   • Notebook: resnet.ipynb
   • Details: Utilizes the ResNet-18 architecture, pretrained by Oliveira et al.. This model is unique in its training approach, combining both real and synthetic data created by StyleGAN-2. This method enhances the model's ability to generalize, providing more accurate classifications in the detection of Age-related Macular Degeneration (AMD).

2. Retinal Image Quality Assessment:

   • Notebook: eyeq.ipynb
   • Details: Employs a DenseNet architecture trained by Fu et al. This model classifies retinal images into three categories: "Good," "Usable," and "Reject." Images categorized as "Reject" typically exhibit issues such as significant blurring, low contrast, or inadequate illumination, indicating poor quality.

3. Synthetic Fundus Eye Generation with StyleGAN-2:

   • Notebook: synthetic.ipynb
   • Details: Utilizes StyleGAN-2, pretrained by Oliveira et al., to generate synthetic images of the retina fundus. These images depict various stages of age-related macular disease, contributing to the dataset's diversity and aiding in the training of more robust AMD classification models.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

Before you begin, ensure you have the following installed:

• Anaconda or Miniconda
• Git

Installation

1. Clone the Repository

   git clone https://github.com/GuiCamargoX/synthetic-retina-amd.git
   cd synthetic-retina-amd
   

2. Downloading Model Weights:

   Before running the application, download the required model weights for DenseNet, ResNet18, StyleGAN2:

   • DenseNet Weights

   • ResNet18 Weights

   • StyleGAN2 Weights

   Place these weights in the specified directory (e.g., weights/).

3. Create Conda environment:

    conda env create -f environment.yml
   

4. Activate the Conda environment and Launch Jupyter Notebook:

    conda activate retina
    jupyter notebook
   

Contributing

We welcome contributions from the community.

License

This project is licensed under the MIT License.

Reference

If you use this code, please cite the following papers:

1. Oliveira, G. C., Rosa, G. H., Pedronette, D. C., Papa, J. P., Kumar, H., Passos, L. A., & Kumar, D. (2024). Robust deep learning for eye fundus images: Bridging real and synthetic data for enhancing generalization. Biomedical Signal Processing and Control, 94, 106263.

2. Fu, H., Wang, B., Shen, J., Cui, S., Xu, Y., Liu, J., Shao, L., (2019). Evaluation of Retinal Image Quality Assessment Networks in Different Color-Spaces. In: Shen, D., et al. Medical Image Computing and Computer Assisted Intervention – MICCAI 2019. MICCAI 2019. Lecture Notes in Computer Science(), vol 11764. Springer

Acknowledgement

We would like to thank Fu et al. for sharing their code at https://github.com/HzFu/EyeQ.