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.
-
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).
- Notebook:
-
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.
- Notebook:
-
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.
- Notebook:
These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.
Before you begin, ensure you have the following installed:
-
Clone the Repository
git clone https://github.com/GuiCamargoX/synthetic-retina-amd.git cd synthetic-retina-amd
-
Downloading Model Weights:
Before running the application, download the required model weights for DenseNet, ResNet18, StyleGAN2:
Place these weights in the specified directory (e.g., weights/).
-
Create Conda environment:
conda env create -f environment.yml
-
Activate the Conda environment and Launch Jupyter Notebook:
conda activate retina jupyter notebook
We welcome contributions from the community.
This project is licensed under the MIT License.
If you use this code, please cite the following papers:
-
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.
-
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
We would like to thank Fu et al. for sharing their code at https://github.com/HzFu/EyeQ.