PatchCNNforDMRI

Diffusion MRI (dMRI) can probe tissue microstructure properties by acquiring certain amount of dMRI measurements and solving the inverse problem of a diffusion model. Conventional model fitting-based methods usually require large amount of dMRI measurements with long acquisition time, which can be highly accelerated by deep learning-based methods through neural networks.

Taking the diffusion tensor imaging (DTI) model as an example, this demo builds patch-based and voxel-based neural networks to effectively solve the inverse problem of estimating fractional anisotropy (FA) and mean diffusivity (MD) measures from under-sampled diffusion dataset. The objectives of this project are to obtain 1) basic understanding of deep learning approaches applied in dMRI, 2) practical knowledge of essential components in building neural networks for diffusion model fitting.

References [1] Gong, T., et al. (2018).  Efficient Reconstruction of Diffusion Kurtosis Imaging Based on a Hierarchical Convolutional Neural Network. Presented at: the 26th ISMRM Annual Meeting. [2] Li, Z., et al. (2019).  Fast and Robust Diffusion Kurtosis Parametric Mapping Using a Three-Dimensional Convolutional Neural Network. IEEE Access, 7, 71398-71411: https://doi.org/10.1109/access.2019.2919241

I. Guidance of environment setting and packages installation

1. Install Anaconda/Miniconda where you can use Conda commands to easily configure different environments

Follow the installation here: https://conda.io/projects/conda/en/latest/user-guide/install/index.html

2. Setting up the environment for Python version of 3.7

# Create a new environment with python version of 3.7
conda create --name your_env_name python=3.7   

# Activate the environment when you need to use it
conda activate your_env_name 

# Deactivate the enviroment when you have finished using it   
conda deactivate

3. Install necessary python packages

conda activate your_env_name

# Install tensorflow (which now includes keras) these two librarys are used for deep learning in python
pip3 install tensorflow==2.3.1

# Install scipy and numpy these libraries are used for performing mathmatical calculations on datasets 
pip3 install scipy
pip3 install numpy==1.17.0

# Install nipy, this library is used for loading NIfTI images (.nii/.nii.gz). This is how MRI images are normally saved
pip3 install nipy==0.4.2

4. Download MRIcron for convenient visualisation of Nifty image files

https://www.nitrc.org/frs/?group_id=152

II. Download the sample datasets and codes

1. Download the repository

You can download the repository to your local system via the following command:
git clone https://github.com/Tinggong/DTIwithNeuralNetworkDemo

2. Description about the datasets

The dataset can be downloaded here: https://drive.google.com/file/d/1aDoU8CJ695Xsm1C9uGnKvq0XOtgmgfwn/view?usp=sharing  

The DTI dataset contains data from two subjects (S1; S2) extracted from a publicly available multi-centre dataset. 
Please refer to the paper for full dataset and imaging parameters: https://www.nature.com/articles/s41597-020-0493-8

Folder structure:
# subject folder S1 
bval                 # the b-values associated with each image volume; 6 b0 images + 30 b=1000 s/mm2 images
bvec                 # the directions of diffusion gradients
diffusion.nii        # the diffusion dataset
nodif_brain_mask.nii # the binary brain mask
S1_FA.nii            # the FA and MD measures estimated from DTI model ...
S1_MD.nii            # using dtifit in FSL <https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FDT/UserGuide#DTIFIT>

III. Running the demos

Step 1: Choose the under-sampled image volumes from full diffusion dataset (the nifty files), and format and save training and testing data to .mat file; /datasets folder will be generated containing the formatted data.

# a. Formatting training dataset with the first 10 image volumes from the full diffusion dataset
    python3 FormatData.py --path /Your/Data/Dir/Data-DTI --subjects S1 --nDWI 10 --fc1d_train 

    # You can also format training dataset with a scheme file contained in the ${NetDir}/schemes folder, 
    # which are 1 for the target image volumes to choose and 0 for all other volumes. (see example file scheme1)
    # Use when you want to design new under-sampling schemes
    python FormatData.py --path /Your/Data/Dir/Data-DTI --subjects S1 --scheme scheme1 --fc1d_train 


# b. Formatting test dataset (add --Nolabel option if the dataset contains no available labels
    python3 FormatData.py --path $DataDir --subjects S2 --nDWI 10 --test

Step 2: network training; Check all available options and default values in /utils/model.py; /weights folder will be generated containing the trained model.

# Using the first 10 volumes; you can also use a scheme file to determined the input DWI volumes. 
    python3 Training.py --train_subjects S1 --DWI 10 --model fc1d --train 

Step 3: Test the model with dataset from S2; weights are saved from previous training; /nii folder will be generated containing the estimated parameters from testing data in nifty format, and you can view the results with MRIcron.

    python3 Testing.py --test_subject S2 --DWI 10 --model fc1d