GEPAR3D 3D CBCT Tooth Segmentation Dataset

This dataset accompanies the MICCAI 2025 paper titled:
“GEPAR3D: Geometry Prior-Assisted Learning for 3D Tooth Segmentation.”

Part 1: Re-annotated Segmentation Maps (150 Volumes) - 32class_labels.zip

This subset contains segmentation maps only (no imaging data), re-annotated into 32 semantic classes with Universal Numbering System based on segmentation maps introduced by Cui et al. (Nature Communications, 2022).

• The segmentation volumes are derived from CBCT data used in Cui et al.'s work.

• Imaging data (CBCT scans) are not included due to privacy and institutional data policies.

• Researchers interested in the corresponding raw CBCT scans must contact the original data providers:

  • Z. Cui et al. (A fully automatic AI system for tooth and alveolar bone segmentation from cone-beam CT images) at cuizm.neu.edu@gmail.com.

  • Data requests are subject to institutional approval and privacy review, typically within 15 working days.

  • Cite the original Cui et al. publication when using any associated imaging data.

Part 2: External Test Set (Centers A and B, Poland) - GEPAR3D_dataset.zip

This part includes CBCT scans with segmentation labels from two independent Polish clinical centers, used as external test data in our MICCAI study.

• Data consists of anonymized CBCT volumes and voxel-wise segmentation masks for 3D tooth structures. Use of data was approved by the Institutional Review Board (IRB Approval ID: OKW-623/2022).

• These data were collected, processed, and labeled under ethical approval and conform to institutional privacy requirements.

• This dataset is released under the Non-Commercial Academic Research Use License – Redistribution Prohibited (NCRU-NR). Use of this dataset is permitted solely for non-commercial academic research purposes. Redistribution, reproduction, or adaptation of the dataset, whether in whole or in part, is strictly prohibited without prior written permission from the dataset owners. Commercial use is not permitted under this license. Users must provide appropriate attribution in any academic publications or outputs derived from the dataset.

• Contact: For questions, data access requests, collaborations regarding this dataset, inquiries, or support, please contact:
  Tomasz Szczepański: t.szczepanski@sanoscience.org.

Preprocessing and Annotation Details

Detailed information on:

• Annotation protocols,

• Preprocessing steps (resampling, normalization),

• Label structures (32-class scheme),

• Evaluation splits (training/test IDs)

… is provided in the GitHub repository accompanying the MICCAI paper. We strongly recommend users consult the repository to ensure proper usage and reproducibility.

Licence Requirements
When applying for access to this dataset, license as Non-Commercial Academic Research Use License:

Citation Requirement

If you use this dataset in any publication, please cite our MICCAI 2025 paper (full citation to be provided upon proceedings release). A BibTeX citation will be included in the repository and dataset page once available.

@misc{szczepański2025gepar3dgeometrypriorassistedlearning,
      title={GEPAR3D: Geometry Prior-Assisted Learning for 3D Tooth Segmentation}, 
      author={Tomasz Szczepański and Szymon Płotka and Michal K. Grzeszczyk and Arleta Adamowicz and Piotr Fudalej and Przemysław Korzeniowski and Tomasz Trzciński and Arkadiusz Sitek},
      year={2025},
      eprint={2508.00155},
      archivePrefix={arXiv},
      primaryClass={eess.IV},
      url={https://arxiv.org/abs/2508.00155}, 
}