Skip to content

CT-RATE: non-contrast chest CTs paired with radiology reports (first 20 validation cases)

pl.load("hf_ct_rate_valid") loads this dataset. CT-RATE (Istanbul Medipol / University of Zurich): 50 188 chest CT volumes from 21 304 patients with radiology reports and multi-abnormality labels. The repo is gated: accept its terms on huggingface.co and log in (HF_TOKEN). This manifest lists the first reconstruction of the first 20 validation patients explicitly (one 200-500 MB NIfTI each); reports and labels are CSV files under dataset/ in the same repo.

import pardesline as pl

ds = pl.load("hf_ct_rate_valid")
print(ds.describe())  # roles, frames, units, labels, licence, citation
s = ds[0]
obj = s.ct  # fetched now, cached under ~/.cache/pardesline/datasets/hf_ct_rate_valid

Provenance and licence
Source Hugging Face Hub repo ibrahimhamamci/CT-RATE, paths dataset/valid/valid_1
Scheme hf
Size of the whole source 6.8 GB
Licence CC-BY-NC-SA-4.0 (gated: accept the terms at the source first)
Commercial use not allowed (non-commercial licence)
Redistribution allowed
Licence note CC BY-NC-SA 4.0, gated repository (checked 2026-10-14): non-commercial, share-alike, your own Hub token after accepting the terms.
Terms https://huggingface.co/datasets/ibrahimhamamci/CT-RATE
Homepage https://huggingface.co/datasets/ibrahimhamamci/CT-RATE
Cite Hamamci I.E., Er S., Almas F., et al. Developing Generalist Foundation Models from a Multimodal Dataset for 3D Computed Tomography. arXiv:2403.17834, 2024.
Samples grouped by patient
Samples listed in the manifest 20

Roles (what each sample contains)

Role Type Frame Units Labels / layout Pattern or file
ct volume (CT) from the file header from the file header dataset/valid/valid_1/valid_1_a/valid_1_a_1.nii.gz, dataset/valid/valid_10/valid_10_a/valid_10_a_1.nii.gz, dataset/valid/valid_11/valid_11_a/valid_11_a_1.nii.gz
Non-contrast chest CT (NIfTI), first reconstruction of the study

Standardized layout

Files are fetched on first access into the dataset cache and stay in their original format; ds.export(folder) writes every sample as standard files (PLY for points and meshes, MetaImage for volumes and label maps) with a .pardesline.json sidecar carrying frame, units, labels, licence and provenance, plus a manifest for the folder. Label maps become exact surfaces with to_mesh(label=k), DICOM series become one volume, raw arrays are decoded through the layout above.

ds = pl.load("hf_ct_rate_valid").take(2)
ds.export("./standardized")        # PLY / MetaImage + sidecars + manifest.json

Algorithms and models that run on it

Generated from src/pardesline/data/manifests/hf_ct_rate_valid.json on 2026-10-14.