Automatic Brain Masking

Deep learning based proyect for the automatic masking of a fetal brain. Takes as input either the contents of the images directory or a directory marked with "--target_dir=path_to_dir"

Currently only a unet based model can be used. A mask rcnn model is available but only for use with a GPU, so all relevant maskrcnn parts have been commented out

It takes as input a path to a directory, and recursivly looks for all .nii files for each file it will save a new mask with the name name_mask.nii on the path of the original image it will skip files that end with mask.nii.

Requirementes

• Python 3
• pip

Following can be installed with the requirements.txt file

• tqdm==4.29.1
• opencv_python_headless==4.0.0.21
• numpy==1.16.2
• MedPy==0.3.0
• Keras==2.2.4

instalation

recommended install virtualenvwrapper Make a virtual environment as to not mix dependencies

$ pip install virtualenv virtualenvwrapper

Then, create a directory for your virtual environments e.g.:

$ mkdir ~/python-envs

You might want to add to your .bashrc file these two lines:

export WORKON_HOME=~/python-envs
source /usr/local/bin/virtualenvwrapper.sh

(Note depending on distro, the virtualenvwrapper.sh path might be

/usr/share/virtualenvwrapper/virtualenvwrapper.sh

Then you can source your .bashrc and create a new Python3 virtual environment:

$ source .bashrc
$ mkvirtualenv --python=python3 python_env

To activate or "enter" the virtual env:

$ workon python_env

To deactivate virtual env:

$ deactivate

Create a new environment and activate it the environments name should appear at the beginnig of the shell surrounded by parentheses

Download the source code, cd into your decired location

(env_name)$ git clone https://github.com/FNNDSC/brain-masking-tool
(env_name)$ cd brain-masking-tool

install requirements from requirements.txt

(env_name)$ pip install -r requirements.txt

to run the masking tool

(env_name)$ python brain_mask.py --target_dir = path_to_dir

if no --target_dir is specified it will target the images directory, where you can also include images. you will have to activate the environment everytime you would like to use the tool

limitation

• Unet can currently only work with 256x256 images

by Parikshit.G