Deep folding

The project aims to study cortical folding patterns thanks to deep learning tools. MRIs are processed through BrainVISA/Morphologist tools.

Prerequisites

Brainvisa parts (deep_folding.brainvisa) must run with brainvisa installer (usually using brainvisa singularity container).

Deep learning part (preprocessing and utils) must run with python3 and works with pytorch.

Package documentation can be found at https://neurospin.github.io/deep_folding/index.html.

Generates datasets of crops

Deep learning pipelines to investigate folding patterns are not working on the whole brain (or the whole hemisphere) but on brain crops. Several processings are required, as drawn here:

We give a step-by-step description of the pipeline in deep_folding/brainvisa/README.rst.

Development

git clone https://github.com/neurospin/deep_folding.git

# Install for development
bv bash
cd deep_folding
virtualenv --python=python3 --system-site-packages venv
. venv/bin/activate
pip3 install -e .

# Tests
python3 -m pytest  # run tests

If you want to install the package:

python3 setup.py install

Notebooks are in the folder notebooks, access using:

bv bash # to enter brainvisa environnment
. venv/bin/activate
jupyter notebook # then click on file to open a notebook

If you want to build the documentation and pushes it to the web:

bv bash # to enter brainvisa environnment
. venv/bin/activate
pip3 install -e .[doc]
cd docs
./make_docs.sh

If you want to clean the documentation:

cd docs/source
make clean