An Automated Framework for Large-Scale Graph-Based Cerebrovascular Analysis (Arxiv)
Accepted at ISBI2026!
CaravelMetrics is an open-source Python framework for comprehensive, automated analysis of cerebrovascular networks from medical imaging data. It transforms binary vessel segmentation masks into quantitative morphometric, topological, and geometric features across anatomically defined brain territories.
Understanding cerebrovascular morphology is critical for detecting disease and distinguishing pathological changes from normal aging. CaravelMetrics enables large-scale population studies by automating vessel feature extraction through graph-based analysis.
- Modular Architecture: Three independent processing modules (Atlas Registration, Graph Extraction, Metric Computation)
- Batch Processing: Automated multi-subject pipeline with parallel execution
- Automated Graph Construction: Converts vessel segmentation masks to mathematical graph representations via skeletonization
- Multi-Scale Feature Extraction: Computes 15 complementary features across four categories
- Dual Analysis Modes: Supports both whole-brain and atlas-based regional analysis
- Regional Analysis: Integrates arterial atlas for territory-specific measurements across 30 brain regions
- Population-Level Studies: Validated on 570 subjects (ages 20-86) from the IXI dataset
- Reproducible Pipeline: End-to-end automation from segmentation masks to statistical analysis
- Advanced Graph Processing: Geodesic distance computation, orphan node handling, artifact removal
# 1. Clone and install
git clone https://github.com/i-vesseg/CaravelMetrics.git
cd CaravelMetrics
pip install -r requirements.txt
# 2. Configure your data paths in CaravelMetrics.py
# Edit: segmentation_folder, image_folder, output_base_folder, use_atlas
# 3. Run the pipeline
python CaravelMetrics.py
# The pipeline will automatically process all subjects in your data foldersFor detailed configuration options and advanced usage, see the Usage Guide.
Full documentation and usage examples are available in docs/
- Pipeline Overview - Processing steps and modular architecture
- Installation Guide - Detailed setup instructions
- Usage Guide - Configuration and batch processing examples
- Features Documentation - Detailed feature definitions
- Analysis Guide - Statistical analysis and visualization
Details will be provided after acceptance.
This project is licensed under the MIT License - see the LICENSE file for details.
This work is co-funded by:
- European Research Council (ERC CARAVEL 101171357)
- ANR JCJC I-VESSEG (22-CE45-0015-01)
- TwinsUK Imaging: A Resource for Ageing Research (Chronic Disease Research Foundation)
- King's Health Partners Digital Health Hub and EPSRC
- IXI Dataset: https://brain-development.org/ixi-dataset/
- Arterial Atlas: Liu et al., Scientific Data (2023) - https://doi.org/10.1038/s41597-023-01967-w
- VesselVerse: Falcetta et al., MICCAI (2025) - Vessel segmentation resource. (Official Website)
- Daniele Falcetta - Data Science Department, EURECOM
- Liane S. Canas - School of Biomedical Engineering & Imaging Sciences, King's College London
- Lorenzo Suppa - Data Science Department, EURECOM & Politecnico di Torino
- Matteo Pentassuglia - Data Science Department, EURECOM
- Jon Cleary - School of Biomedical Engineering & Imaging Sciences, King's College London
- Marc Modat - School of Biomedical Engineering & Imaging Sciences, King's College London
- Sébastien Ourselin - School of Biomedical Engineering & Imaging Sciences, King's College London
- Maria A. Zuluaga - Data Science Department, EURECOM & School of Biomedical Engineering, King's College London
- Bogdan Ion - Data Science Department, EURECOM
- Duccio Lalli - Double Degree MSc Student in Data Science and Engineering at Politecnico di Torino & EURECOM
For questions about the code or research collaborations:
- Open an issue on GitHub
- Email: daniele.falcetta@eurecom.fr, maria.zuluaga@eurecom.fr