MARVAL

Morphological Analysis of ceRebroVAscuLature

MARVAL is an integrated platform for quantitative cerebrovascular morphometry from time-of-flight magnetic resonance angiography (TOF-MRA). It was developed to make brain-vessel morphology measurable at population scale, from automated vessel segmentation to imaging-derived phenotype (IDP) extraction, standard-template construction, and group-level association analysis.

Publication

A comprehensive normative atlas of cerebrovascular morphology and its associations with stroke prognosis
Yarong Ding, Longhao Cao, Zhe Wang, Dawei Li, Junyu Chen, Yong Jiang, Jing Jing, Yuesong Pan, Weibin Gu, Cang Guo, Siding Chen, Chaoyue Wang, Xia Meng, Hao Li, Zixiao Li, Liping Liu, Yilong Wang, Xingquan Zhao, Yongjun Wang

Published in Science Bulletin.

• Article page: https://www.sciencedirect.com/science/article/pii/S2095927326005463
• DOI: https://doi.org/10.1016/j.scib.2026.05.044

What MARVAL Provides

MARVAL combines four core components:

1. A high-quality annotated TOF-MRA dataset built from healthy participants and stroke patients, with expert consensus labels refined for distal and small vessels.
2. An automated cerebrovascular segmentation pipeline designed to improve sensitivity to small vessels while preserving vascular topology.
3. A deep-phenotyping module that extracts scalar and spatial cerebrovascular IDPs, including diameter, segment length, curvature, tortuosity, torsion, branching patterns, vessel volume, fractal dimension, and lacunarity.
4. Group-analysis tools and a standard template framework supporting population-level morphometry, including the Chinese Human Cerebrovascular Template (CHC-Template).

Study Highlights

Using MARVAL, the study:

• constructed a high-resolution cerebrovascular morphology reference in an East Asian population;
• quantified normative aging trajectories of cerebrovascular morphology in 3,035 community participants from the PRECISE cohort;
• showed that vascular aging is scale-dependent, with distinct patterns across large, medium, and small vessels;
• linked cerebrovascular morphology to non-imaging phenotypes such as renal-function markers, smoking and alcohol exposure, plaque burden, hypertension, diabetes, and cognition;
• analyzed 8,994 acute ischemic stroke patients with high-quality TOF-MRA from CNSR-III and identified cerebrovascular morphometric markers associated with functional disability, recurrence, and mortality.

Figures

The assets/figures/ directory contains public-facing figures associated with the MARVAL study:

• graphical_abstract.jpg: graphical abstract of the study.
• figure1_platform_overview.jpg: MARVAL dataset construction, segmentation, phenotyping, and group-analysis workflow.
• figure2_segmentation_idp_maps.jpg: representative segmentation and spatial IDP maps.
• figure3_normative_template.jpg: normative trajectories and CHC-Template visualization.
• figure4_nonimaging_associations.jpg: associations between cerebrovascular morphology and non-imaging phenotypes.
• figure5_stroke_prognosis.jpg: associations between cerebrovascular morphology and stroke prognosis.

Please cite the paper when reusing the figures or describing the MARVAL framework.

Code Availability

The MARVAL codebase is being prepared for public release.

Because patent-related applications and institutional review processes are currently in progress, implementation code, model checkpoints, and detailed deployment scripts will be released after those processes are complete. This repository is therefore a public project page at this stage, with the analysis code marked as coming soon.

See code/COMING_SOON.md for the planned release scope.

Planned Release Scope

The planned public release will include:

• preprocessing utilities for TOF-MRA images;
• vessel segmentation inference workflow;
• post-processing and centerline-refinement utilities;
• scalar and spatial cerebrovascular IDP extraction;
• group-analysis examples and template-space visualization utilities;
• documentation and example commands.

Clinical data, protected cohort information, and non-public model assets will only be shared in forms permitted by the relevant ethics, data-use, patent, and institutional policies.

Citation

A formal BibTeX entry will be updated as additional publication metadata become available.

@article{ding_marval_2026,
  title   = {A comprehensive normative atlas of cerebrovascular morphology and its associations with stroke prognosis},
  author  = {Ding, Yarong and Cao, Longhao and Wang, Zhe and Li, Dawei and Chen, Junyu and Jiang, Yong and Jing, Jing and Pan, Yuesong and Gu, Weibin and Guo, Cang and Chen, Siding and Wang, Chaoyue and Meng, Xia and Li, Hao and Li, Zixiao and Liu, Liping and Wang, Yilong and Zhao, Xingquan and Wang, Yongjun},
  journal = {Science Bulletin},
  year    = {2026},
  doi     = {10.1016/j.scib.2026.05.044}
}

Contact

For scientific questions about the study, please contact the corresponding author listed in the publication.

For repository updates, please watch this GitHub repository.