YongchengYAO/Ceph-Biometrics-400

About

This is a modified version of data for the paper: Fully Automatic System for Accurate Localisation and Analysis of Cephalometric Landmarks in Lateral Cephalograms

Dataset Summary:

• 400 head and neck X-ray scans and 19 landmarks for each image

What's new?

• We saved the original 2D .bmp images into pseudo 3D .nii.gz files in a manner that the original 2D sagittal slices can be extracted from the 3D NIfTI files correctly assuming the standard RAS+ orientation of .nii.gz files. A pseudo voxel size [0.1, 0.1, 0.1] is set in the NIfTI header, according to the data description in the paper: "The image resolution was 1935 × 2400 pixels with a pixel spacing of 0.1 mm."
• We used the landmark annotations from the folder 400_senior in the original dataset.
• We changed the landmarks coordinates to 3D, making it consistent with the new 3D NIfTI files.
• Figures of image and landmarks are added for visual inspection.
• Images (.nii.gz files), landmarks (.json files) and figures (.png files) are saved in Images, Landmarks, and Landmarks-fig folders.

Changelog 🔥

• [6 Mar, 2025] Update landmark coordinates to 0-based indices‼️
• [3 Mar, 2025] Update the data structure of JSON files within Landmarks.zip. Landmarks coordinates remain unchanged.
• [1 Mar, 2025]
  • Update Landmarks.zip: correct the coordinates in JSON files. It's recommended to redownload all compressed files.
  • Add a data preparation script get_dataset.py.
  • Add data usage agreement.

Data Usage Agreement

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

• You must cite the paper: Fully Automatic System for Accurate Localisation and Analysis of Cephalometric Landmarks in Lateral Cephalograms
• You must refer to the source of this dataset in any publication: https://huggingface.co/datasets/YongchengYAO/Ceph-Biometrics-400

Download from Huggingface

#!/bin/bash
pip install --upgrade huggingface-hub[cli]
huggingface-cli login --token $HF_TOKEN

# python
from huggingface_hub import snapshot_download
snapshot_download(repo_id="YongchengYAO/Ceph-Biometrics-400", repo_type='dataset', local_dir="/your/local/folder")

Landmarks Labels

landmarks_map = {
    "P1": "sella",
    "P2": "nasion",
    "P3": "orbitale",
    "P4": "porion",
    "P5": "subspinale",
    "P6": "supramentale",
    "P7": "pogonion",
    "P8": "menton",
    "P9": "gnathion",
    "P10": "gonion",
    "P11": "incision inferius",
    "P12": "incision superius",
    "P13": "upper lip",
    "P14": "lower lip",
    "P15": "subnasale",
    "P16": "soft tissue pogonion",
    "P17": "posterior nasal spine",
    "P18": "anterior nasal spine",
    "P19": "articulare",
}

From Raw Data

You can replicate the data processing step by running the script:

python get_dataset.py -d <datasets_folder> -n Ceph-Biometrics-400

License

This dataset is released under the CC BY-NC 4.0 license.