CancerVerse: A Fully Open Longitudinal and Multimodal Dataset for Multicancer Screening

We present CancerVerse, a fully open longitudinal and multimodal CT dataset for multicancer screening. It contains 24,422 CT scans from 14,599 patients, followed over time for up to 14.4 years. CT scans were collected at hospitals in Switzerland and Turkey from 2010 to 2025. Human experts drew voxel-wise tumor annotations for 13 cancer types in the pelvis, abdomen, and chest. CT scans are paired with radiology reports, pathology results and clinical variables. CancerVerse also includes CT scans of verified healthy patients for measuring false positives.

As the first large-scale open dataset of its kind, CancerVerse offers a new benchmark for detecting cancer early at a low rate of false positives per scan. It is built and maintained by the BodyMaps team.

Paper

CancerVerse: A Fully Open Longitudinal and Multimodal Dataset for Multicancer Screening
Wenxuan Li†, Pedro R. A. S. Bassi†, Xinze Zhou†, Jakob Wasserthal♡, Ibrahim E. Hamamci♡, Sezgin Er♡, Bjoern Menze♡, Gulhan E. Akan♡, Szymon Płotka, Jakub Prządo, Qi Chen, Yucheng Tang, Daguang Xu, Arkadiusz Sitek, Kang Wang, Yang Yang, Alan L. Yuille, Zongwei Zhou★
Johns Hopkins University
MICCAI 2026
† Equal contribution   ♡ Open data contributors   ★ Corresponding author

CancerVerse Dataset

conda create -n huggingface python=3.11 -y
conda activate huggingface
python -m pip install --upgrade pip
python -m pip install --upgrade "huggingface_hub[cli]"
hf --version
hf auth login

Four lines of code to download the dataset

mkdir CancerVerse && cd CancerVerse
hf download BodyMaps/CancerVerse download_CancerVerse_data.sh download_CancerVerse_label.sh --repo-type dataset --local-dir .
bash download_CancerVerse_data.sh  # CT volumes and metadata, needs ~3 TB
bash download_CancerVerse_label.sh # tumor annotations, needs ~40 GB

After both scripts finish:

CancerVerse/
├── CancerVerse_dataset_metadata.csv
└── CancerVerse/
    ├── CV_00000001/
    │   ├── ct.nii.gz
    │   └── segmentations/   # 13 types of tumor annotations
    └── ...

Dataset at a glance

Voxel-wise tumor annotations

The annotation took 28 human experts more than 13 months: radiology residents drew the annotations, and eight board-certified radiologists reviewed them independently. Disagreements were resolved by adjudication.

License

CancerVerse is released under CC BY-NC-ND 4.0. Please review the license before use.

For commercial licensing, industry collaboration, or partnership inquiries related to BodyMaps datasets, please contact hello@thebodymaps.com or visit www.thebodymaps.com/contact.

Citation

@InProceedings{LiWen_CancerVerse_MICCAI2026,
        author = { Li, Wenxuan AND Bassi, Pedro R. A. S. AND Zhou, Xinze AND Wasserthal, Jakob AND Hamamci, Ibrahim E. AND Er, Sezgin AND Menze, Bjoern H. AND Akan, Gulhan E. AND Płotka, Szymon AND Prządo, Jakub AND Chen, Qi AND Tang, Yucheng AND Xu, Daguang AND Sitek, Arkadiusz AND Wang, Kang AND Yang, Yang AND Yuille, Alan L. AND Zhou, Zongwei},
        title = { { CancerVerse: A Fully Open Longitudinal and Multimodal Dataset for Multicancer Screening } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 16895},
        month = {September},
        page = {pending}
}