GitLocker: The Coding Marketplace

Description:

deepparcellation 1.0.35

DeepParcellation Package
DeepParcellation: fast and accurate brain MRI parcellation by deep learning
Contributions

The project was initiated by Dr. Lim (abysslover) and Dr. Choi (yooha1003).
The code is written by Dr. Lim at Gwangju Alzheimer's & Related Dementias (GARD) Cohort Research Center (GARD CRC), Chosun University.
This research was conducted in collaborations with the following people:
Eun-Cheon Lim1, Uk-Su Choi1, Yul-Wan Sung2, Kun-Ho Lee1 and Jungsoo Gim1.

Gwangju Alzheimer's & Related Dementias (GARD) Cohort Research Center, Chosun University, Gwangju, Republic of Korea
Kansei Fukushi Research Institute, Tohoku Fukushi University, Sendai, Miyagi, Japan

The manuscript will be available in the future.

Getting Started
A step-by-step tutorial is provided in the following sections.
Prerequisites
You should install CUDA-enabled GPU cards with at least 8GB GPU memory manufactured by nVidia, e.g., Titan XP.
Prepare T1-weighted MR images

Convert MR images to Neuroimaging Informatics Technology Initiative (NIfTI) format.
The parent directory name of a NIfTI file path will be used as Subject Id during prediction.
You can specify either an input path of the NIfTI file or input direcotry of many NIfTI files.

Install DeepParcellation

Install Anaconda

Download an Anaconda distribution: Link

Create a Conda environment

conda create -n deepparc python=3.8 -y

Install DeepParcellation (CPU mode)

conda activate deepparc
pip install deepparcellation

Install DeepParcellation (GPU mode)

conda activate deepparc
pip install deepparcellation
conda install cudnn=7.6.5 -c anaconda -y
conda install cudatoolkit=10.1.243 -c conda-forge -y
pip uninstall tensorflow -y
pip install tensorflow-gpu==2.2.0
pip uninstall keras -y
conda install keras-gpu=2.4.3 -c anaconda -y

Install DeepParcellation (Force)

conda activate deepparc
pip install --force --no-dependencies deepparcellation

Run DeepParcellation

conda activate deepparc
deepparcellation -o=/tmp/test --i=./subject-0-0000/test.nii.gz
or
deepparcellation -o=/tmp/test --i=./

NOTE:

You must always activate the conda enviroment before running DeepParcellation if you opened a new console.
You should install DeepParcellation with force parameter when all the other dependencies were manually met, i.e., you have installed tensorflow-macos, tensorflow-metal dependencies for Apple M1 chips.

Contact
Please contact abysslover@gmail.com if you have any questions about DeepParcellation.