ACVD Processed Lung Nodule CT Data

Processed data used for Anatomy-Constrained Voxel Diffusion for Controllable Synthesis of Complex Lung Nodules, accepted at IEEE BIBM 2026 (publication forthcoming).

Code: lakelk/ACVD.

Contents

Subset	CT series with each anatomical mask	Paired 64³ crops
LUNA16	888	1186
LUNA25	4069	6155

The release provides one ZIP per subset. Extract both ZIPs into the same data directory. Each archive preserves this structure:

Dataset_LUNA16/  # Dataset_LUNA25 uses the same structure
  annotations.csv
  airway_masks/*.nii.gz
  vessel_masks/*.nii.gz
  lung_masks/*.nii.gz
  bone_masks/*.nii.gz
  ControlNet_Data/*.npy

Full-volume anatomical masks

In addition to the paired 64³ crops, we release four anatomical channels covering the full CT grid: airway, pulmonary artery, lung parenchyma and bone. These masks support research on CT morphology, anatomical structure and spatial relationships over larger regions or entire scans.

Masks are binary NIfTI volumes and retain the corresponding CT grid and physical geometry, including anisotropic spacing. Original full-volume CT images and nodule masks must be obtained separately from the sources below. This release includes processed CT crops, but not the original full-volume CT images.

Paired crop format

Each .npy contains a dictionary:

Key	Meaning
uid	CT series UID
nodule_idx	Zero-based row index in the included annotation table
gt_image	float32 (1,64,64,64), normalized to [-1,1]
conditions	Binary float32 (5,64,64,64)

Condition order: nodule, vessel, airway, lung, bone. Spatial array order: Z, Y, X.

import numpy as np
sample = np.load("path/to/sample.npy", allow_pickle=True).item()
ct = sample["gt_image"]
conditions = sample["conditions"]

CT values are clipped to [-1000,400] HU before normalization. The dictionaries do not include a physical affine. Preserve annotation-table row order when interpreting nodule_idx. Preprocessing and experimental settings are described in the code repository.

Use and download

After downloading the repository, extract:

python -m zipfile -e archives/Dataset_LUNA16.zip ./ACVD-data
python -m zipfile -e archives/Dataset_LUNA25.zip ./ACVD-data

For training, evaluation and raw-volume data requirements, see the code repository.

Citation

Until the proceedings are published, cite the paper as:

Anatomy-Constrained Voxel Diffusion for Controllable Synthesis of Complex Lung Nodules. Accepted at IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2026. Publication forthcoming.

The author list and final proceedings/DOI citation will be added when available. Please also cite the upstream datasets and annotation sources relevant to your use, as listed in Sources and references below. Copyable BibTeX entries are grouped by data component at the end of this card.

License

The authors' generated four-channel anatomical masks are released under CC BY-NC 4.0: attribution is required and commercial use is not permitted. Third-party CT, coordinates and nodule masks retain their source-specific conditions; see LICENSE.md. ACVD original code is licensed separately under MIT.

Sources and references

Please cite ACVD and the sources corresponding to the data and channels used.

CT and nodule annotations

Component	Source and citation
LUNA16 CT and coordinates	LUNA16; Setio et al., Validation, comparison, and combination of algorithms for automatic detection of pulmonary nodules in computed tomography images: The LUNA16 challenge, Medical Image Analysis, 2017.
LUNA16 nodule masks	LIDC-IDRI; Armato et al., Data From LIDC-IDRI, TCIA, 2015 (dataset DOI).
LUNA25 CT	Peeters et al., The LUNA25 Challenge: Public Training and Development set - Imaging Data, 2025. DOI:10.5281/zenodo.14223624.
LUNA25 coordinates	Peeters et al., The LUNA25 Challenge: Public Training and Development set - Annotation Data, 2025. DOI:10.5281/zenodo.14673658.
LUNA25 nodule masks	LUNA25-MedSAM2; Ma et al., MedSAM2: Segment Anything in 3D Medical Images and Videos, 2025. Paper. Please cite LUNA25 and MedSAM2, following the upstream dataset card; the imaging and annotation records are linked above.

Automatically generated anatomical masks

Released channel	Generation source and reference
Lung and bone	Author-generated TotalSegmentator predictions; Wasserthal et al., TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images, Radiology: Artificial Intelligence, 2023.
Airway	Author-trained nnUNet on ATM22, applied to LUNA CT; Zhang et al., Multi-site, Multi-domain Airway Tree Modeling, Medical Image Analysis, 2023.
Vessel / pulmonary artery	Author-trained nnUNet on PARSE22, applied to LUNA CT; Luo et al., Efficient Automatic Segmentation for Multi-level Pulmonary Arteries: The PARSE Challenge, 2023; Chu et al., Deep learning-driven pulmonary artery and vein segmentation reveals demography-associated vasculature anatomical differences, Nature Communications, 2025 (DOI).
Segmentation framework	nnUNet; Isensee et al., nnU-Net: A self-configuring method for deep learning-based biomedical image segmentation, Nature Methods, 2021.

PARSE22 requests both Luo et al. (2023) and Chu et al. (2025). The ATM22 citation page lists CFDA (Zhang et al., 2022), class-wise gradient imbalance (Zheng et al., 2021), BREAK (Yu et al., 2022) and AirwayNet (Qin et al., 2019), and also recommends the 2023 challenge benchmark paper. Follow the upstream citation instructions linked above.

The ATM22/PARSE22 original training data and segmentation model weights are not included.

For processing and reproduction instructions, see the ACVD code repository. Source-specific licenses are recorded in LICENSE.md.

BibTeX by data component

Expand the group corresponding to the data or channel used. Each group explains the role of its references; the entries document upstream data and tools used to prepare this release. The ACVD paper citation will be added when its final author list and publication details are available.

LUNA16 CT and nodule coordinates — Setio et al. (2017)

For the LUNA16 CT images and annotation coordinates used in the LUNA16 crops.

@article{setio2017validation,
  title = {Validation, comparison, and combination of algorithms for automatic detection of pulmonary nodules in computed tomography images: The LUNA16 challenge},
  author = {Setio, Arnaud Arindra Adiyoso and Traverso, Alberto and De Bel, Thomas and Berens, Max S. N. C. and van den Bogaard, Cas and Cerello, Piergiorgio and Chen, Hao and Dou, Qi and Fantacci, Maria Evelina and Geurts, Bram and others},
  journal = {Medical Image Analysis},
  volume = {42},
  pages = {1--13},
  year = {2017},
  publisher = {Elsevier},
  doi = {10.1016/j.media.2017.06.015}
}

LUNA16 nodule annotations — LIDC-IDRI data record (2015)

For the LIDC-IDRI annotations underlying the LUNA16 nodule masks; this entry includes the TCIA dataset DOI.

@misc{armato2015lidcdata,
  author = {Armato III, Samuel G. and McLennan, Geoffrey and Bidaut, Luc and others},
  title = {Data From {LIDC-IDRI}},
  year = {2015},
  publisher = {The Cancer Imaging Archive},
  doi = {10.7937/K9/TCIA.2015.LO9QL9SX},
  url = {https://doi.org/10.7937/K9/TCIA.2015.LO9QL9SX}
}

LUNA25 CT and nodule coordinates — imaging and annotation records (2025)

The imaging record corresponds to CT intensities; the annotation record corresponds to nodule coordinates and annotations.csv.

@misc{luna25_imaging,
  author = {Peeters, Dré and Obreja, Bogdan and Antonissen, Noa and Jacobs, Colin},
  title = {The LUNA25 Challenge: Public Training and Development set - Imaging Data},
  year = {2025},
  publisher = {Zenodo},
  doi = {10.5281/zenodo.14223624}
}

@misc{luna25_annotations,
  author = {Peeters, Dré and Obreja, Bogdan and Antonissen, Noa and Jacobs, Colin},
  title = {The LUNA25 Challenge: Public Training and Development set - Annotation Data},
  year = {2025},
  publisher = {Zenodo},
  doi = {10.5281/zenodo.14673658}
}

LUNA25 nodule masks — MedSAM2 (2025)

For the MedSAM2 nodule masks. The upstream card requests attribution to both LUNA25 and MedSAM2; the two LUNA25 data records are provided in the group above.

@article{medsam2,
  author = {Ma, Jun and Yang, Zongxin and Kim, Sumin and Chen, Bihui and Baharoon, Mohammed and Fallahpour, Adibvafa and Asakereh, Reza and Lyu, Hongwei and Wang, Bo},
  title = {MedSAM2: Segment Anything in 3D Medical Images and Videos},
  journal = {arXiv preprint arXiv:2504.03600},
  year = {2025},
  url = {https://arxiv.org/abs/2504.03600}
}

Lung and bone channels — TotalSegmentator (2023)

For the TotalSegmentator predictions consolidated into the lung and bone channels. TotalSegmentator also requests the shared nnU-Net reference in the final group.

@article{wasserthal2023totalsegmentator,
  title = {TotalSegmentator: Robust Segmentation of 104 Anatomic Structures in CT Images},
  author = {Wasserthal, Jakob and Breit, Hanns-Christian and Meyer, Manfred T. and Pradella, Maurice and Hinck, Daniel and Sauter, Alexander W. and Heye, Tobias and Boll, Daniel and Cyriac, Joshy and Yang, Shan and others},
  journal = {Radiology: Artificial Intelligence},
  volume = {5},
  number = {5},
  pages = {e230024},
  year = {2023},
  publisher = {Radiological Society of North America},
  doi = {10.1148/ryai.230024}
}

Airway channel — ATM22 training-data references

Airway masks were generated by author-trained nnUNet models on LUNA CT.

@inproceedings{zhang2022cfda,
  title = {{CFDA: Collaborative Feature Disentanglement and Augmentation for Pulmonary Airway Tree Modeling of COVID-19 CTs}},
  author = {Zhang, Minghui and Zhang, Hanxiao and Yang, Guang-Zhong and Gu, Yun},
  booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2022},
  year = {2022},
  pages = {506--516},
  doi = {10.1007/978-3-031-16431-6_48}
}

@article{zheng2021gradient,
  title = {{Alleviating Class-Wise Gradient Imbalance for Pulmonary Airway Segmentation}},
  author = {Zheng, Hao and Qin, Yulei and Gu, Yun and Xie, Fangfang and Yang, Jie and Sun, Jiayuan and Yang, Guang-Zhong},
  journal = {IEEE Transactions on Medical Imaging},
  year = {2021},
  volume = {40},
  number = {9},
  pages = {2452--2462},
  doi = {10.1109/tmi.2021.3078828}
}

@inproceedings{yu2022break,
  title = {{Break: Bronchi Reconstruction by Geodesic Transformation and Skeleton Embedding}},
  author = {Yu, Weihao and Zheng, Hao and Zhang, Minghui and Zhang, Hanxiao and Sun, Jiayuan and Yang, Jie},
  booktitle = {2022 IEEE 19th International Symposium on Biomedical Imaging (ISBI)},
  year = {2022},
  pages = {1--5},
  doi = {10.1109/isbi52829.2022.9761697}
}

@inproceedings{qin2019airwaynet,
  title = {{AirwayNet: A Voxel-Connectivity Aware Approach for Accurate Airway Segmentation Using Convolutional Neural Networks}},
  author = {Qin, Yulei and Chen, Mingjian and Zheng, Hao and Gu, Yun and Shen, Mali and Yang, Jie and Huang, Xiaolin and Zhu, Yue-Min and Yang, Guang-Zhong},
  booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2019},
  year = {2019},
  pages = {212--220},
  doi = {10.1007/978-3-030-32226-7_24}
}

@article{zhang2023multisite,
  title = {Multi-Site, Multi-Domain Airway Tree Modeling},
  author = {Zhang, Minghui and Wu, Yangqian and Zhang, Hanxiao and Qin, Yulei and Zheng, Hao and Tang, Wen and Arnold, Corey and Pei, Chenhao and Yu, Pengxin and Nan, Yang and others},
  journal = {Medical Image Analysis},
  volume = {90},
  pages = {102957},
  year = {2023},
  publisher = {Elsevier},
  doi = {10.1016/j.media.2023.102957}
}

Vessel / pulmonary artery channel — PARSE22 references (2023 and 2025)

Vessel masks were generated by author-trained nnUNet models on LUNA CT.

@article{luo2023efficient,
  title = {Efficient Automatic Segmentation for Multi-Level Pulmonary Arteries: The PARSE Challenge},
  author = {Luo, Gongning and Wang, Kuanquan and Liu, Jun and Dong, Shumeng and Liang, Xinjie and Li, Xiangyu and Gan, Shifang and Wang, Wei and Wang, Suyu and Yu, Pengxin and others},
  journal = {arXiv preprint arXiv:2304.03708},
  year = {2023},
  doi = {10.48550/arXiv.2304.03708}
}

@article{chu2025vascular,
  title = {{Deep learning-driven pulmonary artery and vein segmentation reveals demography-associated vasculature anatomical differences}},
  author = {Chu, Yuetan and Luo, Gongning and Zhou, Longxi and Cao, Shaodong and Ma, Guolin and Meng, Xianglin and Zhou, Juexiao and Yang, Changchun and Xie, Dexuan and Mu, Dan and Henao, Ricardo and Setti, Gianluca and Xiao, Xigang and Wu, Lianming and Qiu, Zhaowen and Gao, Xin},
  journal = {Nature Communications},
  year = {2025},
  volume = {16},
  number = {1},
  pages = {2262},
  doi = {10.1038/s41467-025-56505-6}
}

Shared segmentation framework — nnU-Net (2021)

For nnUNet used to train the airway and vessel segmentation models. This is also the framework reference requested by TotalSegmentator.

@article{isensee2021nnu,
  title = {nnU-Net: A Self-Configuring Method for Deep Learning-Based Biomedical Image Segmentation},
  author = {Isensee, Fabian and Jaeger, Paul F. and Kohl, Sebastian A. A. and Petersen, Jens and Maier-Hein, Klaus H.},
  journal = {Nature Methods},
  volume = {18},
  number = {2},
  pages = {203--211},
  year = {2021},
  publisher = {Nature Publishing Group},
  doi = {10.1038/s41592-020-01008-z}
}