ArteryX: Enhancing Sensitivity in Brain Artery Feature Extraction Using Arterial Graph Network

Motivation: Automated artery feature extraction algorithms come with validation challenges. Present validation toolboxes doesn't minimize operator variability in the extracted features which leads to loss of artery feature accuracy and sensitivity. Goal(s): Propose an easy-to-use artery-labelling workflow via artery-graph-network. Reduce operator variability; thus enable artery-feature reliability. Compare and validate with state-of-the-art in cerebrovascular-disease population. Approach: Graph artery network was computed in an unsupervised approach. This network helps the operator to landmark the graph nodes reliably on a 3D viewer (VR optional). The GUI tracks artery traces and generates artery-features. Results: Our features demonstrated higher-sensitivity to state-of-the-art(iCafe) in cerebrovascular-disease, achieving 5-fold reduced landmarking-time, and user-variability. Impact: Demonstrated increase in sensitivity and usability of artery features, hold promise of our approach's reliable application in other cerebrovascular studies. Features generated from this approach will be used for training and validation of automated artery extraction technique.