Unofficial mirror of the SynthRAD2025 Grand Challenge training data, Zenodo record 15373853 (v2: CT and CBCT clipped to [-1024, 3071] HU), extracted and checked so that cases can be downloaded selectively. The files are the original .mha images, unchanged; each archive was verified against the MD5 published by Zenodo and every uploaded file against its local size.

Licence: CC BY-NC 4.0, as the original record: non-commercial use only, with attribution. Cite the SynthRAD2025 dataset (Thummerer et al., 2025, doi:10.5281/zenodo.15373853); challenge site: synthrad2025.grand-challenge.org.

Layout

Task1/<region>/<case_id>/{ct,mask,mr}.mha     MR → CT
Task2/<region>/<case_id>/{ct,mask,cbct}.mha   CBCT → CT
Task<k>/<region>/overview/...                 overview images and acquisition parameters of the release
Task<k>/zenodo.json                           source archive, MD5 and counts
manifest.csv                                  integrity check of every case (geometry, spacing, mask, HU range)

Regions: AB (abdomen), HN (head and neck), TH (thorax); case_id = task, region, centre letter, number.

Task	AB	HN	TH	Total
Task1 (MR → CT)	175	156	182	513
Task2 (CBCT → CT)	256	260	258	774

1287/1287 cases pass the integrity check of manifest.csv.

Download

from huggingface_hub import snapshot_download

snapshot_download("developper669/synthrad2025", repo_type="dataset", revision="v2", local_dir="synthrad2025",
                  allow_patterns=["manifest.csv", "Task1/HN/**"])   # e.g. task 1, head and neck only

With sct-range: sctrange-download --root synthrad2025 --task 1.

Downloads last month: -

Total file size:

27 GB