An Automated Algorithm to Improve the Precision of Basilar Artery Diameter Measurements Before and After Subarachnoid Hemorrhage–Induced Vasospasm in an Animal Model

• Published in: Neurosurgery Vol. 66; no. 1; pp. 137 - 143
• Main Authors: Salvetti, David J., Sinha, Tuhin K., Wilson, Kevin J., Shay, Sheila D., Spratt, Daniel E., Mericle, Robert A.
• Format: Journal Article
• Language: English
• Published: United States Oxford University Press 01.01.2010; Wolters Kluwer Health, Inc
• Subjects: Algorithms; Animals; Automatic Data Processing - methods; Automation; Basilar Artery - pathology; Contrast Media; Diagnostic Imaging - methods; Disease Models, Animal; Neurosurgery; Rabbits; Subarachnoid Hemorrhage - complications; Surgical techniques; Vasospasm, Intracranial - etiology; Vasospasm, Intracranial - pathology; Veins & arteries; Animal model; Image segmentation; Artery diameter measurement; Subarachnoid hemorrhage; Vasospasm; Level-set formulation; Cerebral angiography
• ISSN: 0148-396X; 1524-4040
• DOI: 10.1227/01.NEU.0000362035.70735.63

Abstract OBJECTIVE Quantifying vasospasm has traditionally been performed manually, a method prone to imprecision and user bias. An alternative approach is to use computerized image analysis techniques to define and quantify the diameter of a vessel. The goal of this article is to demonstrate a novel automated vessel measurement algorithm specific to the needs of vasospasm studies and to compare it with traditional manual measurements in an animal model of vasospasm. METHODS A total of 576 arterial diameter measurements were collected by 4 independent, blinded examiners from 24 angiograms in a rabbit subarachnoid hemorrhage (SAH) model. Measurements were taken from 3 segments of the basilar artery in anteroposterior and lateral projections, both before SAH and after SAH-induced vasospasm. Means and standard deviations of 288 manual measurements were compared with 288 automated measurements. RESULTS The precision of automated measurements was significantly improved compared with standardized manual measurements (85.7% decrease in variation; P < .001). When using automated measurements, the precision was not affected by vessel size, but when using manual measurements, smaller arteries were less precise (P = .04). There was no significant difference in precision between 2 different contrast concentrations (P = .32). CONCLUSION Automated measurements of basilar artery diameters are more precise than manual measurements, both before and after SAH-induced vasospasm. The variability in the manual group worsens when the artery is smaller secondary to vasospasm, indicating a need for the use of this segmentation method.