retinalysis-vascx: An explainable software toolbox for the extraction of retinal vascular biomarkers

summary

The gist

Automatic extraction of retinal vascular biomarkers from color fundus images (CFI) is crucial for large-scale studies of the retinal vasculature, and this paper presents VascX, an open-source

In short

VascX is an open-source software toolbox designed to automatically extract explainable and flexible biomarkers from retinal artery-vein segmentations using color fundus images. It uses a four-stage pipeline—skeleton, undirected graph, directed digraph, and resolved vessels—to compute metrics like vascular density and tortuosity across different anatomical regions defined by the optic disc and fovea.

Key concepts

Skeleton: This initial step cleans the predicted vessel map into single-pixel-wide centerlines. If an optic disc mask is available, it excludes that region to avoid false centerlines, ensuring the resulting structure accurately reflects the network's topology.
Undirected Graph: The skeleton is converted into a NetworkX Graph where nodes are vessel endpoints or junctions and edges connect adjacent points along the centerline. This allows for segment-level analysis and aggregation before any flow direction is assigned to the vessels.
Directed Digraph: The undirected graph is oriented by rooting each component at the optic disc and directing edges away from it. This turns vessel segments into directed edges, allowing for calculations of geometry-based properties like length and curvature along specific trajectories.
Region-Aware Computation: VascX uses predefined grids, such as DiscCenteredGrid or CircleGrid, anchored to landmarks like the optic disc and fovea. This enables researchers to compute biomarkers specifically within anatomically relevant areas, ensuring measurements are consistent regardless of the image's orientation.

Terminology used across episodes

Reinforcement Learning (RL) — explained in 24 episodes
Retrieval-Augmented Generation (RAG) — explained in 16 episodes
Interpretability — explained in 15 episodes
Sample Complexity — explained in 15 episodes
Federated Learning — explained in 14 episodes
Generalization — explained in 13 episodes

This episode discusses

The paper

retinalysis-vascx: An explainable software toolbox for the extraction of retinal vascular biomarkers · Read on arXiv

Jose D. Vargas-Quiros, Michael J. Beyeler, Sofia Ortin-Vela, Sven Bergmann, Caroline C.W. Klaver, Bart Liefers

Department of Ophthalmology, Erasmus University Medical Center, Rotterdam, the Netherlands · Department of Epidemiology, Erasmus University Medical Center, Rotterdam, the Netherlands · Department of Ophthalmology, Radboud University Medical Center, Nijmegen, the Netherlands · Institute of Molecular and Clinical Ophthalmology, University of Basel · Dept. of Computational Biology, University of Lausanne · Swiss Institute of Bioinformatics · Dept. of Integrative Biomedical Sciences, University of Cape Town

Automatic extraction of retinal vascular biomarkers from color fundus images (CFI) is crucial for large-scale studies of the retinal vasculature. We present VascX, an open-source Python toolbox that extracts biomarkers from CFI artery-vein segmentations. VascX starts from vessel segmentation masks, extracts their skeletons, builds undirected and directed vessel graphs, and resolves vessel segments into longer vessels. A comprehensive set of biomarkers is derived, including vascular density, central retinal equivalents (CREs), and tortuosity. Spatially localized biomarkers may be calculated over grids placed relative to the fovea and optic disc. VascX is released via GitHub and PyPI with comprehensive documentation and examples. Our test-retest reproducibility analysis on repeat imaging of the same eye by different devices shows that most VascX biomarkers have moderate to excellent agreement (ICC > 0.5), with important differences in the level of robustness of different biomarkers. Our analyses of biomarker sensitivity to image perturbations and heuristic parameter values support these differences and further characterize VascX biomarkers. Ultimately, VascX provides an explainable and easily modifiable feature-extraction toolbox that complements segmentation to produce reliable retinal vascular biomarkers. Our graph-based biomarker computation stages support reproducible, region-aware measurements suited for large-scale clinical and epidemiological research. By enabling easy extraction of existing biomarkers and rapid experimentation with new ones, VascX supports oculomics research. Its robustness and computational efficiency facilitate scalable deployment in large databases, while open-source distribution lowers barriers to adoption for ophthalmic researchers and clinicians.

DOI: 10.1167/tvst.15.9.11

Transcript

Introduction to the show: ident: Genomics Radio. Generated commentary on the latest computational biology and genomics papers.

Ines: I'm Ines, and with me are Marcus and Yuki, guest researcher.

Marcus: Today's paper: "retinalysis-vascx: An explainable software toolbox for the extraction of retinal vascular biomarkers".

Ines: Automatic extraction of retinal vascular biomarkers from color fundus images (CFI) is crucial for large-scale studies of the retinal vasculature, and this paper presents VascX,

Marcus: First, who's behind it and why it matters.

Title and authors: Ines: So we're diving into the paper titled "retinalysis-vascx: An explainable software toolbox for the extraction of retinal vascular biomarkers." What does that title actually mean when you look at it from a computational biology standpoint?

Marcus: I think the name itself suggests a focus on creating a tool that is transparent, which is important because we're dealing with complex biological signals.

Yuki: From my perspective, I’m interested in how this tool helps us connect these retinal features to the broader history of human cardiovascular health and population genetics.

Ines: Exactly, Yuki, and from a computational biology viewpoint, I want to know what kind of actual biological information this toolbox is designed to recover from color fundus images.

Marcus: And I'm thinking about the data side—how robust are these extracted features when we look at them across different patient cohorts or imaging devices?

The paper's summary: Ines: Okay, so the paper summarizes VascX as an open-source software toolbox that extracts biomarkers from artery-vein segmentations. It goes through a pretty structured pipeline involving skeletonization, graph building, and resolving vessel segments.

Marcus: That structured pipeline sounds like a solid engineering approach for dealing with noisy image data; it moves away from just taking one big output from a deep learning model and gives us traceable steps.

Yuki: And the summary mentions they derive things like vascular density, central retinal equivalents, and tortuosity, which are all classic markers that link retinal structure to systemic health outcomes.

Ines: Precisely, Yuki; the core value here is in providing an explainable way to calculate these complex metrics rather than just giving us a final number.

Marcus: I agree with Ines on the explainability; and from a statistics standpoint, knowing exactly how the graph is built helps us understand where noise might be introduced during that process, which affects our batch effects in genomics data.

The paper's improvements: Ines: The authors suggest several ways to improve this toolbox, focusing on region-aware computation using grids anchored at the optic disc and fovea locations. This is a big step because it moves beyond just calculating metrics across the whole image in a generic way.

Marcus: I see that region-awareness as a way to make the output more biologically meaningful, since we know those specific anatomical landmarks relate directly to visual processing and central retinal structures.

Yuki: That spatial localization is really interesting for population genetics because it allows us to compare measurements consistently across different individuals or even different clinics who might have slightly varied image centers.

Ines: And the paper also discusses sensitivity analysis and cross-device reproducibility testing, which is crucial for validating any tool we build before trusting its output in a large study.

Marcus: I'm keen on the reproducibility part; if we can quantify how stable these biomarkers are across different devices, that directly impacts how much weight we should give them when trying to find associations with traits like hypertension.

Conclusion: Ines: To wrap things up, the paper on "retinalysis-vascx: An explainable software toolbox for the extraction of retinal vascular biomarkers" presents a systematic method for extracting detailed morphology and topology metrics from fundus images through a defined computational pipeline.

Marcus: It’s important to remember that while most biomarkers show moderate to excellent agreement with ICC greater than zero point five, there are clear differences in how robust they are against image noise or segmentation errors <ref:2602.08580#pg1>.

Yuki: And from a population perspective, the ability to use spatially localized grids helps us harmonize data across different imaging centers, which is essential for longitudinal studies connecting retinal structure to long-term health trends.

Ines: Exactly; this tool gives us the framework to see not just *what* the measurements are, but *how* they were calculated and *where* they are located anatomically.

Marcus: So, moving forward, this kind of explainable extraction method could significantly strengthen our ability to use CFI data in large-scale epidemiological studies by providing more trustworthy inputs for our statistical models.

More episodes

← Home