Abstract

Motivation: In this work, we address the challenge of cerebral artery segmentation when time-of-flight (TOF) imaging is unavailable. Goal(s): Develop an automatic data-driven segmentation of large cerebral arteries. Approach: Arteries were identified within the fMRI signal by leveraging large pulsation-driven fluctuations. Results: In the local subjects with TOF images, the approach displayed high levels of agreement with TOF-derived segmentation. Additionally, the segmentation demonstrated high scan-to-scan reproducibility in 430 subjects with four repeated fMRI scans from the HCP aging cohort. Lastly, the segmentation performed robustly across two different scanning protocols supporting its potential to be used for datasets with various acquisition parameters. Impact: Our robust data-driven approach reliably automatically segments the large cerebral arteries of fMRI datasets. This work enables more accessible large cerebral artery segmentation in existing MRI datasets, independent of TOF images.