Automated Brain Artery Segmentation Analysis from Time-of-Flight MRA

Figure 1: ArteryX Evaluation Pipeline for arterial centerline extraction and analysis

This project focuses on developing advanced tools for the automated segmentation, classification, and feature extraction of brain arteries from time-of-flight (TOF) MR angiography. Accurate characterization of intracranial arteries is essential for understanding vascular health, assessing stroke risk, and studying cerebrovascular contributions to neurological diseases. However, manual analysis is labor-intensive and prone to variability, limiting its use in large-scale and multi-site research.

We are building a fully integrated pipeline that leverages deep learning and super-resolution reconstruction to produce high-quality arterial maps and quantitative vascular features. Our models automatically identify and segment major arterial territories, classify vessel types and branching patterns, and extract clinically meaningful metrics such as diameter, tortuosity, curvature, and topology. Super-resolution techniques further enhance TOF-MRA beyond its native resolution, enabling finer delineation of small distal vessels and improving the accuracy of downstream analyses.

A key innovation from our group is ArteryX, a semi-automated, high-precision framework for intracranial artery analysis. ArteryX combines a vessel-fused neural network with graph-based anatomical priors to generate continuous and anatomically consistent artery traces. The platform supports rapid landmarking, automated feature extraction, and standardized reporting of vascular metrics, including diameter, tortuosity, branching complexity, and territorial patterns. Additionally, ArteryX provides tools for synthesizing realistic arterial profiles and validating segmentation pipelines—addressing a major gap in reproducible vascular imaging. The ArteryX toolbox is freely available and can be accessed here.

By integrating state-of-the-art neural networks with vascular graph analysis and anatomical priors, this project aims to deliver a fast, reproducible, and scalable solution for brain-artery quantification. These tools enable robust investigation of cerebral small vessel disease, aging, stroke, and neurodegeneration, ultimately contributing to more precise assessment of cerebrovascular health across diverse populations.