CTA-BET

Robust Brain Extraction Tool for non-enhanced CT and CT angiography: CTA-BET

This repository provides a tool based on nnUNet for automated brain mask creationg of CT angiography images.

If you are using CTA-BET, please cite the following publication:

Mustafa Ahmed Mahmutoglu, Aditya Rastogi, Yeong Chul Yun, Sanya Middha, Julius Kernbach, Martha Foltyn-Dumitru, Gianluca Brugnara, Philipp Vollmuth, Alexander Radbruch, Martin Bendszus, and Marianne Schell 
Robust Brain Extraction Tool for Non-enhanced CT and CT Angiography: CTA-BET
Radiology: Artificial Intelligence, e240847.
https://doi.org/10.1148/ryai.240847

Installation Instructions

Since our model is heavily dependend on nnUNet, please visit their repository for installation instructions and also cite their paper:

Isensee, F., Jaeger, P. F., Kohl, S. A., Petersen, J., & Maier-Hein, K. H. (2021). nnU-Net: a self-configuring 
method for deep learning-based biomedical image segmentation. Nature methods, 18(2), 203-211.

How to use it

Please install nnunetv2 following the instructions here:

https://github.com/MIC-DKFZ/nnUNet

Clone this repository and add your images to the input folder. Create input, output and model folders in the same folder as the run_CTA-DEFACE.py.

mkdir input
mkdir output
mkdir model

Download the trained model from the following link and put the Dataset001_CTABET folder inside the model folder.

https://drive.google.com/drive/folders/1_hnaxVUjEC9dqXifAhegT4lctZBf05Qa?usp=sharing

Make sure the CTA input image names end with _0000.nii.gz, which is important to be recognized by the model.

python run_CTA-DEFACE.py -i input -o output

The above command will look for all nifti files (*.nii.gz) in the input folder and save the defaced NIfTI files and the face mask in the output folder.

CAVE: Our model and python code was designed to run on Ubuntu in a preinstalled nnunet environment, please adjust accordingly in case you intend to use the code in other operating systems.