SCLEROSIS, ISBI 2015, ljubljana, MICCAI 2008, MICCAI 2016, HEALTHY, KIRBY, OASIS-3

Dataset Overview

Dataset Content

• SCLEROSIS: 109 images
• ISBI 2015: 21 images, training set with 5 patients, test set with 14 patients
• Ljubljana: 30 images, training only, 30 patients
  • Demographic Features:
    • Age
    • Sex (7 male, 23 female)
    • MS type (24 relapsing‑remitting, 2 clinically isolated syndrome, 1 primary progressive, 2 secondary progressive, 1 undiagnosed)
• MICCAI 2008: 51 images, training set with 20 patients, test set with 25 patients, no additional attributes
• MICCAI 2016: 15 images, training only, 15 patients
  • Demographic Features:
    • Age
    • Sex (7 male, 8 female)
• HEALTHY: 21 images
• KIRBY: 42 images, training set with 21 patients
  • Demographic Features:
    • Age
    • Sex (10 male, 11 female)
• OASIS‑3: forthcoming

Dataset Download Instructions

• ISBI 2015: Register and download here
• Ljubljana: Download here
• MICCAI 2008: Register and download here
• MICCAI 2016: Register and download here
• KIRBY: Download here
• OASIS‑3: Register and download here

Dataset Pre‑processing Protocol

• Pre‑processing Steps:
  • Rigid registration to T1 space
  • Alignment to MNI template
  • Anisotropic filtering
  • Skull stripping
  • Bias‑field (intensity non‑uniformity) correction

Dataset Preparation Workflow

1. Run build_directory.py to create input/output folder structures.
2. Download images and unzip into the appropriate folders.
3. Execute change_extensions.py to convert images to .nii.gz format.
4. Run create_json.py to generate a JSON dictionary of image paths.
5. Execute main.py for image pre‑processing.
6. Run Manipulating_data.py to split JSON data into training/testing sets (75 %/25 %), ensuring balanced representation of healthy and MS subjects.