metadata

license: cc-by-nc-sa-4.0

ACDC Dataset

This is a pre-processed version of the original Automated Cardiac Diagnosis Challenge (ACDC) dataset. Short-axis view images have been resampled to 1mm x 1mm x 10mm. Images have also been center cropped at the left ventricle mask center. The size of each slice is 192 x 192 pixels.

Download the dataset using the following command:

from huggingface_hub import snapshot_download
data_dir = snapshot_download(repo_id="mathpluscode/ACDC", allow_patterns=["*.nii.gz", "*.csv"], repo_type="dataset")

where data_dir is the directory where the dataset will be downloaded. The data structure is as follows:

train.csv
train/
├── patient001/
│   ├── patient001_frame01.nii.gz
│   ├── patient001_frame01_gt.nii.gz
│   ├── patient001_frame12.nii.gz
│   ├── patient001_frame12_gt.nii.gz
├── patient002/
├── ...
test/
├── patient101/
├── patient102/
├── ...