Published October 15, 2025 | Version v1
Dataset Open

Circle of Willis Centerline Graphs and Morphometric Features

  • 1. ROR icon ZHAW Zurich University of Applied Sciences
  • 2. EDMO icon University of Zurich

Description

This dataset contains anatomically labeled centerline graphs, surface meshes, and derived morphometric features of the Circle of Willis (CoW). The data is based on the TopCoW 2024 training set, which consists of 250 (=125 paired) CTA and MRA images displaying brain vasculature, including the CoW.

The dataset complements the available TopCoW multi-class segmentations and supports research in vascular morphology, graph-based modeling, and automated centerline extraction. The dataset itself, along with the accompanying baseline algorithm for CoW centerline and feature extraction, is described in our manuscript:

Musio et al., “Circle of Willis Centerline Graphs: A Dataset and Baseline Algorithm” (2025)

A detailed description of the dataset structure and contents is provided in the README file included in the downloadable ZIP archive.

For access to the raw imaging data and multi-class masks, please refer to the TopCoW dataset repository:
https://zenodo.org/records/15692630

Note: The baseline algorithm can be applied on top of the TopCoW segmentation models to extract centerlines and features. The code is available on GitHub: https://github.com/fmusio/CoW_Centerline_Extraction

Contents

For each patient and modality (n=250), the dataset includes:

  • CoW Graph (cow_graphs/): Labeled CoW centerline graph (.vtp)
  • CoW Mesh (cow_meshes/): Labeled CoW surface mesh (.vtp)
  • CoW Variant (cow_variants/): Variant description encoding presence/absence of segments and fetal-type PCA (.json)
  • CoW Nodes (cow_nodes/): Node descriptions for start, end, bifurcation and segment boundary points (.json)
  • CoW Features (cow_features/): Morphometric feature descriptions for segments and bifurcations (.json)

The files follow the same naming pattern as introduced by TopCoW: topcow_{modality}_{pat_id}.{suffix}
Where modality is ct for CTA or mr for MRA, and pat_id is the patient ID (e.g., 001, 002, ...).

Citation

If you use this dataset in your work, please cite:

Musio et al., “Circle of Willis Centerline Graphs: A Dataset and Baseline Algorithm” (2025)

License

This dataset is released under the CC BY-NC (Attribution-NonCommercial) license.
By downloading the data, you agree to the license terms.

Contact

For questions or feedback, please contact:

  • Fabio Musio – fabio.musio@zhaw.ch
  • Prof. Dr. Sven Hirsch – sven.hirsch@zhaw.ch
    Zurich University of Applied Sciences (ZHAW)

Acknowledgements

The work on this dataset was supported by the EU project GEMINI (funded by the EU Horizon Europe R&I programme, Grant No. 101136438).

 

Files

CoW_Centerline_Data.zip

Files (373.1 MB)

Name Size Download all
md5:8e414a6cd50e85ef4a8f68e37fb25fc3
373.1 MB Preview Download

Additional details

Related works

Is derived from
Dataset: 10.5281/zenodo.15692630 (DOI)
Is supplement to
Software: 10.48550/arXiv.2312.17670 (DOI)
Is supplemented by
Software: https://github.com/fmusio/CoW_Centerline_Extraction/ (URL)

Dates

Created
2025-08-29

References

  • Musio, F., Juchler, N., Yang, K., Shit, S., Prabhakar, C., Menze, B., & Hirsch, S. (2025). Circle of Willis Centerline Graphs: A Dataset and Baseline Algorithm. arXiv. https://doi.org/10.48550/arXiv.2510.13720