README.md

Data

Data availability

We plan to host the dataset on a publicly accessible server and make it available upon request. In the meantime, please contact Dr. Maria del C. Valdés-Hernández (M.Valdes-Hernan@ed.ac.uk) to inquire about the dataset used in this work.

Data folder structure

• Data should have the following structure

  clinical-super-mri/
      data/
          affine/
              SR_002_NHSRI_V0_affine.mat
              SR_002_NHSRI_V1_affine.mat
              SR_005_BRIC1_V0_affine.mat
              ...
          rigid/
              SR_002_NHSRI_V0_rigid.mat
              SR_002_NHSRI_V1_rigid.mat
              SR_005_BRIC1_V0_rigid.mat
              ...
          warp/
              SR_002_NHSRI_V0.mat
              SR_002_NHSRI_V1.mat
              SR_005_BRIC1_V0.mat
              ...
  

Convert .mat to .npy

• Here we provide a simple Python script mat2npy.py to convert .mat files to .npy as loading .npy files is much faster in Python.
• The follow command convert all .mat scans in affine/ to .npy files and store in affine/npy

  python mat2npy.py --input_dir affine --output_dir affine/npy
  

• Note that our data reader still support reading directly from .mat files though the training speed might be bottlenecked by reading .mat files.