Skip to content

gliff-ai/robust-3d-unet-macular-holes

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

27 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Robust 3D U-Net Segmentation of Macular Holes

This repository contains the official code for the paper Robust 3D U-Net Segmentation of Macular Holes.

Code was written by Jonathan Frawley ORCID iD iconhttps://orcid.org/0000-0002-9437-7399.

If you use this software, please cite it as below:

@misc{frawley2021robust,
      title={Robust 3D U-Net Segmentation of Macular Holes}, 
      author={Jonathan Frawley and Chris G. Willcocks and Maged Habib and Caspar Geenen and David H. Steel and Boguslaw Obara},
      year={2021},
      eprint={2103.01299},
      archivePrefix={arXiv},
      primaryClass={eess.IV}
}

Setup

Requires Python 3.8 or newer. To install dependencies:

pip install -r requirements.txt

Dataset

We cannot make public the dataset used for the paper due to privacy concerns. The dataloader expects three folders:

  • train
  • validation
  • test

each with a folder im and gt within them, corresponding to the OCT image and ground truth image respectively. All images and ground truths are of the following dimensions: 321x376x49

For convenience, we provide a script to generate synthetic data, to demonstrate this layout and file format:

python3 generate_macular_holes.py

Running

To train the models for the paper:

cd bin
./run_train.sh

Perf metrics on train, validation and test sets as it is trained will be in CSV files in out/cli-seg-results.

Inference

To run inference on the trained models for the paper:

cd bin
./run_inference.sh

The output 3D TIFF images will be in out/cli-seg-infer.