Objective: This study assesses artifactual luminal distortion, or eccentricity, that affects measurement of stenosis on CT angiography performed with a variety of helical protocols.
Materials and methods: A 32-vessel carotid artery phantom was built with five known grades of stenoses (25%, 50%, 75%, 88%, and 94%) and three lengths of stenosis (1, 3, and 5 mm). This phantom was scanned with conventional and 1.0-, 1.5-, and 2.0-pitch helical CT with slice thicknesses of 2, 4, and 8 mm, and three vessel orientations: parallel, 45 degrees oblique, and perpendicular to the z-axis. Oblique multiplanar reconstruction was performed with the latter two vessel orientations to produce images similar to the parallel to z-axis orientation. The cross-sectional images were then used to measure the maximum and minimum (longest and shortest) luminal diameters in and out of each stenosis at a computer workstation by a single investigator who was unaware of the scanning technique. Percentage of stenosis was assessed by three methods: cross-sectional area in and out of the stenosis, maximum diameter out of stenosis and minimum in stenosis (North American Symptomatic Carotid Endarterectomy Trial method), and minimum diameter in and out of the stenosis. Comparisons were made with the gold standard using the equation (measured percentage of stenosis-actual percentage of stenosis) based on known luminal diameters of the phantom. Luminal eccentricity was assessed for each of the vessels and scanning parameters as a ratio of minimum to maximum diameters.
Results: All three methods of measuring stenoses were strongly affected by luminal eccentricity. The North American Symptomatic Carotid Endarterectomy Trial method overestimated percentage of stenosis an average of 1.64%. The most accurate results were obtained when using the minimum diameter in and out of the stenoses (-0.45% from the gold standard). Eccentricity was significantly greater in stenoses than in normal lumen (p < .0001) and when the vessels were oriented perpendicular to the z-axis (p = .0003). A progressive increase in eccentricity was seen in the 4- and 8-mm slice thicknesses and the 3- and 5-mm-long stenoses (p < .001; p < .001).
Conclusion: Artifactual luminal eccentricity has significant implications for measuring percentage of stenosis revealed by CT angiography. Eccentricity increases in longer stenoses, thicker slices, and vessels oriented perpendicular to the z-axis. With CT angiography, measurement of minimum diameters in and out of a stenosis provides the most accurate assessment of percentage of stenosis.