YongchengYAO/PDDCA-Lite

About

This is a preprocessed redistribution of PDDCA v1.4.1 (from the TCIA Head-Neck Cetuximab collection), which is released under the N/A (public domain) / CC BY 3.0 license.

Dataset summary: 48 head-and-neck CT scans with 9-class organ-at-risk masks, and craniofacial landmarks for 33 of them.

Contents of this repository:

• Images/ — 48 files
• Masks/ — 48 files

📝 Landmark annotations, visualization figures and the benchmark plan files live in 🔥MedVision🔥, where you can load the complete images and annotations from dataset configs.

Relation to the source dataset

In the source	48 head-and-neck CT cases with per-structure OAR masks; 33 also ship landmark .fcsv files
Excluded here	nothing — this is a complete mirror
In this repo	48 Images + 48 Masks

Why -Lite? The suffix marks this as a derived redistribution rather than a copy of the source. These are preprocessed volumes — every case has been format-converted, geometry-normalised and reoriented to RAS+ — and for some sources cases or modalities are excluded as well (see the table above). Use it to reproduce MedVision, not as a substitute for the original release. See Preprocessing below for exactly what was changed.

Preprocessing

• NRRD images and per-structure masks converted to nii.gz and standardized to RAS+ orientation.

• The per-structure masks are merged into a single multi-label mask (see labels_map).

• LPS fix: PDDCA NRRDs declare space: left-posterior-superior. Copying that direction matrix verbatim into a (RAS-by-definition) NIfTI affine mirrors the volume L-R/A-P and makes the RAS+ reorientation a no-op. The affine is corrected before reorientation.

• Landmarks were parsed from the 3D Slicer .fcsv files and converted to 0-based voxel indices.

• Structure availability is ragged upstream: 6 of the 9 structures are present in all 48 cases, the submandibular glands and mandible in fewer. Missing structures are skipped, not asserted.

📝 Only 33 of the 48 cases ship landmark annotations upstream, so the landmark-biometry task uses a 33-case subset (Images-landmark/) while segmentation and detection use all 48.

Segmentation Labels

labels_map = {
    "1": "mandible",
    "2": "brainstem",
    "3": "left parotid gland",
    "4": "right parotid gland",
    "5": "left submandibular gland",
    "6": "right submandibular gland",
    "7": "left optic nerve",
    "8": "right optic nerve",
    "9": "optic chiasm"
}

Landmarks

landmarks_map = {
    "P1": "chin (most anterior-inferior point of the mandibular symphysis)",
    "P2": "right mandibular condyle",
    "P3": "left mandibular condyle",
    "P4": "odontoid process (dens) of the axis (C2)",
    "P5": "occipital bone (basion/occiput)"
}

News

• [25 Jul, 2026] Initial release. This dataset is integrated into 🔥MedVision🔥, where you can use these config names to load data in python:

  • PDDCA_BiometricsFromLandmarks_Task01_Axial_Test
  • PDDCA_BiometricsFromLandmarks_Task01_Axial_Train
  • PDDCA_BiometricsFromLandmarks_Task01_Sagittal_Test
  • PDDCA_BiometricsFromLandmarks_Task01_Sagittal_Train
  • PDDCA_BoxSize_Task01_Axial_Test
  • PDDCA_BoxSize_Task01_Axial_Train
  • PDDCA_BoxSize_Task01_Coronal_Test
  • PDDCA_BoxSize_Task01_Coronal_Train
  • PDDCA_BoxSize_Task01_Sagittal_Test
  • PDDCA_BoxSize_Task01_Sagittal_Train
  • PDDCA_MaskSize_Task01_Axial_Test
  • PDDCA_MaskSize_Task01_Axial_Train
  • PDDCA_MaskSize_Task01_Coronal_Test
  • PDDCA_MaskSize_Task01_Coronal_Train
  • PDDCA_MaskSize_Task01_Sagittal_Test
  • PDDCA_MaskSize_Task01_Sagittal_Train

Data Usage Agreement

By using the dataset, you agree to the terms as follow.

• You must comply with the original N/A (public domain) / CC BY 3.0 license terms of the source dataset.
• You are recommended to refer to the source of this dataset in any publication: https://huggingface.co/datasets/YongchengYAO/PDDCA-Lite
• You must cite the original publication(s):
  • https://doi.org/10.1002/mp.12197

Official Release

For more information, please go to the official site: http://www.imagenglab.com/newsite/pddca/

Download from Huggingface

# python
from huggingface_hub import snapshot_download
snapshot_download(repo_id="YongchengYAO/PDDCA-Lite", repo_type='dataset', local_dir="/your/local/folder")