SpatialOCT-Glaucoma

This repository contains the implementation of the paper entitled "Spatial-aware Transformer-GRU Framework for Enhanced Glaucoma Diagnosis from 3D OCT Imaging."

Please contact ashtari.mona@gmail.com for further information.

Project Structure

project/
│
├── OCT_data/
│   ├── class0
│   └── class1
│
├── RNN_main.py # Main script for RNN sequential processing
│
├── feature_extraction_main.py # Main script for feature extraction
│
├── RETFound_oct_weights.pth
│
└── split_index.pickle # Indices for cross-validation

Setup

1. Download the RETFound pre-trained weights for OCT data from RETFound_MAE GitHub and place it in the project folder.

2. Download the 3D OCT dataset from here and extract it. Separate the data into two classes and place them into the OCT_data/class0 and OCT_data/class1 folders.

Usage

3. Navigate to the project directory and execute the feature extraction script:

python feature_extraction_main.py

4. Execute the RNN sequential processing script:

python RNN_main.py

Citation

If you use this repository for your research, please cite our paper:

@article{ashtari2024spatial,
  title={Spatial-aware Transformer-GRU Framework for Enhanced Glaucoma Diagnosis from 3D OCT Imaging},
  author={Ashtari-Majlan, Mona and Dehshibi, Mohammad Mahdi and Masip, David},
  journal={arXiv preprint arXiv:2403.05702},
  year={2024}
}