Li_MultiView_reproduction_tensorflow

This repository using tensorflow to reproduce an multi-view DL-based approach to segment the claustrum in T1-weighted MRI scans.

Reference of the original GitHub: https://github.com/hongweilibran/claustrum_multi_view

Reference of the original paper: https://arxiv.org/abs/1911.07515

Overview

@article{albishri2019automated,
  title={Automated human claustrum segmentation using deep learning technologies},
  author={Albishri, Ahmed Awad and Shah, Syed Jawad Hussain and Schmiedler, Anthony and Kang, Seung Suk and Lee, Yugyung},
  journal={arXiv preprint arXiv:1911.07515},
  year={2019}
}

Environments and Requirements

This test implementation is designed to run on CPU.

Python: Version 3.12

To set up the environment:

git clone https://github.com/ShutingXie/Li_MultiView_reproduction_tensorflow.git

Create a virtual environment (You can use other way to creat a virtual environment):

python -m venv myenv

Activate the virtual environment:

source myenv/bin/activate

Install all dependent packages

pip install -r requirements.txt

Dataset

Put your MRI data and labels in the data_org/ and labels_org folders respectively

Preprocessing

Preprocessing details can be checked in Li's Github: https://github.com/hongweilibran/claustrum_multi_view

1. Resampling the MR scans to 1 mm resolution.

python resampler.py

1. Skull-stripping

chmod +x skull_stripping.sh
./skull_stripping.sh

1. (I did not do this) "Image denoising using an adaptive nonlocal means filter for 3D MRI (ANLM, in Matlab). Unfortunately, we did not find the python version for this step. The default setting in Matlab was used in our work." -- From author's GitHub

Test

python test.py

Evaluation the prediction results

python compute_agreement.py