dwi_angular

PyTorch implementation of the paper "Super-Resolved q-Space Deep Learning" (MICCAI 2019, MedIA 2021)

Dependencies

• Pytorch
• numpy
• nibabel
• sklearn_extra

Usage

1. Clone the repository

• git clone https://github.com/Snailpong/dwi_angular.git

2. Dataset download

• Download from Human Connectome Project (HCP) Young Adult Database

• Download data.nii.gz, nodif_brain_mask.nii.gz, bval, bvec files for each subject

3. Preprocess data

• python vector_scatter.py: select index for extracting LR dimension (diffusion)

• python preprocess.py: make h5 file for training and testing

4. Train

• python train.py

• arguments

  • load_model: True/False

  • cuda_visible: CUDA_VISIBLE_DEVICES (e.g. 1)

  • batch_size: set batch size

5. Test

• python test.py

• arguments

  • load_model: True/False

  • cuda_visible: CUDA_VISIBLE_DEVICES (e.g. 1)

  • image_path: set validation path