A new study led by researchers at Beijing Tiantan Hospital, Capital Medical University, Changping Laboratory and Peking University has produced a detailed atlas of human brain-vessel morphology and shown that subtle changes in the small-vessel network may help explain why some stroke patients face worse recovery and long-term outcomes.
The work introduces MARVAL, short for Morphological Analysis of ceRebroVAscuLature, an integrated platform that automatically segments intracranial vessels from time-of-flight magnetic resonance angiography (TOF-MRA), extracts quantitative imaging-derived phenotypes, and supports group-level analysis. The study, titled "A comprehensive normative atlas of cerebrovascular morphology and its associations with stroke prognosis," is published in Science Bulletin.
Blood vessels are fundamental to brain health: they deliver oxygen and nutrients, maintain perfusion and help support neural function. Their shape can change with aging and disease, but until now researchers have had limited tools for measuring the full three-dimensional vascular network in large populations. Manual review is labor-intensive, and conventional automated workflows often miss distal or small vessels where early pathological changes may appear.
MARVAL was designed to close this gap. The team first created a high-quality annotated dataset from 100 healthy participants in the PRECISE community cohort and 50 stroke patients in the nationwide CNSR-III cohort. Each vascular label was refined through a two-tier expert consensus process, with particular attention to small branches. The resulting pipeline combines standardized TOF-MRA preprocessing, deep-learning-based segmentation, post-processing, vascular feature extraction and statistical analysis. An overview of MARVAL and the study design is presented in Figure 1.
From each scan, MARVAL can quantify vessel diameter, length, curvature, tortuosity, torsion, bifurcation patterns, volume fraction and fractal features, producing both scalar measurements and spatial maps. This turns non-invasive TOF-MRA images into a richer description of the cerebrovascular network, including smaller distal branches that are often difficult to assess consistently in large cohorts.
Using MARVAL in 3,035 healthy participants from the PRECISE cohort, the researchers calculated 218 cerebrovascular morphology features and modeled how these measures change with age. Overall vessel volume declined with aging, while average diameter and curvature increased. The aging pattern was scale-dependent: large vessels remained relatively stable in diameter, whereas medium and small vessels showed distinct and more sensitive trajectories.
The researchers also constructed the Chinese Human Cerebrovascular Template, or CHC-Template, from 300 healthy participants. Similar in purpose to widely used brain templates, the CHC-Template provides a standard coordinate space for East Asian cerebrovascular anatomy. It preserves fine tubular structures and captures a gradient of spatial consistency, from highly conserved major arteries to more variable distal branches.
Beyond aging, the study found broad links between brain-vessel morphology and non-imaging health measures. Renal-function indices such as serum creatinine and uric acid, smoking and alcohol exposure, plaque burden in peripheral or coronary arteries, cognitive scores, hypertension and diabetes were all associated with specific vascular features. These findings suggest that the morphology of intracranial vessels may serve as an imaging window onto systemic vascular health.
To test clinical relevance, the team analyzed the CNSR-III cohort of 15,166 patients with acute ischemic cerebrovascular events, including 8,994 patients with high-quality TOF-MRA scans. Multiple vascular features were significantly associated with functional disability and all-cause mortality, and the strongest patterns came from small- and medium-vessel measurements and global features reflecting vascular network complexity.
Sparse principal component analysis further identified six interpretable vascular morphology components. One component, PC 2, was dominated by small-vessel tortuosity, diameter, segment length and fractal complexity. Lower PC 2 scores were consistently associated with higher risks of early functional disability, recurrence and mortality after first-ever ischemic stroke, with the strongest disability association observed at the 3-month follow-up.
The findings point to a new way of reading information already present in non-invasive angiographic images. Rather than focusing only on visible stenosis or a few large-vessel measurements, MARVAL enables a richer assessment of the vascular network, including distal branches that may reflect cerebrovascular reserve and network robustness.
"By making brain-vessel morphology measurable at population scale, MARVAL provides a reproducible framework for studying vascular aging and stroke prognosis," said corresponding author Yongjun Wang of Beijing Tiantan Hospital. "Small-vessel integrity and network complexity may offer clinically useful information beyond conventional imaging markers."
The authors note that the current template was derived from Chinese participants and should be extended to more diverse populations. They also emphasize that 3.0 T TOF-MRA is affected by flow-related signal loss and cannot resolve the smallest vessels. Future validation using CT angiography, contrast-enhanced MRA or 7.0 T TOF-MRA could help distinguish imaging artifacts from true structural changes.
Together, MARVAL and the CHC-Template provide a reusable platform for quantitative cerebrovascular assessment. The work lays the foundation for future studies of vascular aging, systemic vascular health, stroke risk stratification and precision medicine in cerebrovascular disease.
Science Bulletin
Imaging analysis
The authors declare that they have no conflicts of interest.