Awesome

[Paper] [Project Page]

Abstract

Optical Coherence Tomography Angiography (OCTA) is a crucial imaging technique for visualizing retinal vasculature and diagnosing eye diseases such as diabetic retinopathy and glaucoma. However, precise segmentation of OCTA vasculature remains challenging due to the multi-scale vessel structures and noise from poor image quality and eye lesions. In this study, we proposed OCTAMamba, a novel U-shaped network based on the Mamba architecture, designed to segment vasculature in OCTA accurately. OCTAMamba integrates a Quad Stream Efficient Mining Embedding Module for local feature extraction, a Multi-Scale Dilated Asymmetric Convolution Module to capture multi-scale vasculature, and a Focused Feature Recalibration Module to filter noise and highlight target areas. Our method achieves efficient global modeling and local feature extraction while maintaining linear complexity, making it suitable for low-computation medical applications. Extensive experiments on the OCTA 3M, OCTA 6M, and ROSSA datasets demonstrated that OCTAMamba outperforms state-of-the-art methods, providing a new reference for efficient OCTA segmentation. Code is available at https://github.com/zs1314/OCTAMamba

Overview

Quantitative results

Qualitative visualization

💎Let's Get Started!

A. Installation

Note that the code in this repo runs under Linux system. We have not tested whether it works under other OS.

The repo is based on the VMama repo, thus you need to install it first. The following installation sequence is taken from the VMamba repo.

Step 1: Clone the repository:

Clone this repository and navigate to the project directory:

git clone https://github.com/zs1314/OCTAMamba.git
cd OCTAMamba

Step 2: Environment Setup:

It is recommended to set up a conda environment and installing dependencies via pip. Use the following commands to set up your environment:

Create and activate a new conda environment

conda create -n octamamba
conda activate octamamba

Install dependencies

pip install -r requirements.txt
cd kernels/selective_scan && pip install .

B. Data Preparation

If you want to run your own dataset, you can configure it in datasets.py, in function prepareDatasets:

def prepareDatasets():
    all_datasets = {}
    
    // Add your datasets here
    // ......

    return all_datasets

Note that your dataset should follow the given structure:

--root
    --dataset
        |
        |--Your Dataset
            |
            |--train
            |--val
            |--test

where each folder in train, val, test should follow the given format:( take train as an example)

--train
    |
    |--image
    |    |
    |    |--......(images)
    |    |--......
    |    |.......
    |--label
        |
        |--......(labels)
        |--......
        |......

The OCTA500_3M, OCTA500_6M and ROSSA datasets are formatted as follows:

--root
    --dataset
        |
        |--OCTA500_3M
            |
            |--train
            |    |
            |    |--image
            |    |    |--......(images)
            |    |    |--......
            |    |    |.......
            |    |--label
            |        |--......(labels)
            |        |--......
            |        |......
            |--val
            |--test
        |--OCTA500_6M
        |--ROSSA

You can download the dataset from here: Baidu Netdisk or Google Drive

C. Model Training and Testing

If you want to run your own model, please modify the models variable in settings_benchmark.py:

models = {
    # More models can be added here......
}

Each item in models must be of type ObjectCreator, in which your model can be created.

The current version of the code contains 11 models, and 3 datasets: ROSSA, OCTA_500 (3M) and OCTA_500 (6M).

By running run_benchmark.py, the 11 models on 3 datasets will be trained and evaluated at once (that is, a total of 11x3=33 results).

The results will be saved in json format to the result folder.

python run_benchmark.py

🐥: Before training and testing, configure the relevant parameters in the script

D. Get model weights

You can download the model weights (OCTAMamba) from here: Baidu Netdisk or Google Drive

E. Model Predict

python predict.py

🤝Acknowledgments

This project is based on VMamba (paper, code). Thanks for their excellent works!!

🙋Q & A

For any questions, please feel free to contact us.

📜Reference

If this code or paper contributes to your research, please kindly consider citing our paper and give this repo ⭐️ 🌝

@article{zou2024octamamba,
  title={OCTAMamba: A State-Space Model Approach for Precision OCTA Vasculature Segmentation},
  author={Zou, Shun and Zhang, Zhuo and Gao, Guangwei},
  journal={arXiv preprint arXiv:2409.08000},
  year={2024}
}