Automatic image processing pipeline for tracking longitudinal vessel changes in magnetic resonance angiography. Issue 4 (7th March 2019)

• Record Type: Journal Article
• Authors: Hsu, Chih‐Yang; Li, Yimei; Han, Yuanyuan; Elijovich, Lucas; Sabin, Noah D.; Abuelem, Tarek; Torabi, Radmehr; Faught, Austin; Hua, Chia‐Ho; Klimo, Paul; Merchant, Thomas E.; Lucas, John T.
• Is Part Of: Journal of magnetic resonance imaging. Volume 50:Issue 4(2019)
• Page Start: 1063
• Page End: 1074
• Publication Date: 2019-03-07
• Subjects: cerebral vasculopathy -- MRA -- vessel segmentation -- image processing -- vessel diameter; Magnetic resonance imaging -- Periodicals; 616
• DOI: 10.1002/jmri.26699
• Languages: English
• ISSNs: 1053-1807

Abstract

Abstract : Background: Cerebral vessel diameter changes objectively and automatically derived from longitudinal magnetic resonance angiography (MRA) facilitate quantification of vessel changes and further modeling. Purpose: To characterize longitudinal changes in intracranial vessel diameter using time‐of‐flight (TOF) MRA. Study Type: Retrospective longitudinal study. Subject Population: IN all, 112 pediatric patients, aged 9.96 ± 4.59 years, with craniopharyngioma from 2006–2011 scanned annually. Field Strength/Sequence: 1.5T and 3T TOF MRA. Statistical Tests: Chi‐square and Wilcoxon–Mann–Whitney tests. Assessment: Manual measurements using interventional angiography was established as a reference standard for diameter measurements. Constant and linear quantile regression with absolute difference, percentage difference, and relative difference was used for outlier detection. Results: Major vessels surrounding the circle of Willis were successfully segmented except for posterior communicating arteries, mostly due to disease‐related hypoplasia. Diameter measurements were calculated at 1‐mm segments with a median computed vessel diameter of 1.25 mm. Diameter distortion due to registration was within 0.04 mm for 99% of vessel segments. Outlier detection using quantile regression detected less than 4.34% as being outliers. Outliers were more frequent in smaller vessels and proximity to bifurcations ( P < 0.001). Data Conclusion: Using the proposed method, objective changes in vessel diameter can be acquired noninvasively from routine longitudinal imaging. High‐throughput analyses of imaging‐derived vascular trees combined with clinical and treatment parameters will allow rigorous modeling of vessel diameter changes. Level of Evidence : 2 Technical Efficacy : Stage 2 J. Magn. Reson. Imaging 2019;50:1063–1074.