Quantification of carotid stenosis on CT angiography.

PURPOSE Carotid artery stenosis quantification uses percent diameter ratios from conventional angiography. Multidetector high-speed CT angiography (CTA) allows direct millimeter measurement of carotid stenosis. We hypothesize a linear relationship between millimeter stenosis measurements and derived percent, alleviating cumbersome ratio calculations. METHODS Two neuroradiologists separately reviewed CTAs of 268 carotid arteries, blinded to other information. The narrowest portion of each carotid stenosis was measured in millimeters from axial source images. Distal internal carotid arteries (ICAs) were measured beyond the bulb, where walls are parallel. North American Symptomatic Carotid Endarterectomy Trial (NASCET)-style ratios were calculated for each ICA, except for suspected near-occlusions. Interobserver agreement was calculated for all measurements. Correlation coefficients were calculated comparing millimeter and derived percent stenosis, followed by regression analysis. Sensitivity and specificity values tested validity. RESULTS Interobserver agreement correlations were excellent, from 0.78 to 0.89 (2-tailed P </= .01). Excellent reader correlation permitted averaging of millimeter stenosis and distal ICA measurements, which were then used to calculate mean percent stenosis. Stenosis assessment was confidently performed from source images even with calcification. Regression analysis of mean percent as a function of mean millimeter stenosis shows a linear relationship, correlating 1.3 mm to 70% and 2.2 mm to 50% NASCET-style stenosis (Pearson's correlation of -0.95; n = 136). Direct millimeter measurement defining severe stenosis (70% or greater) has sensitivity of 88.2%, specificity of 92.4%, and negative predicted value of 98.2%. CONCLUSION There is a linear relationship between millimeter carotid stenosis diameter and derived percent stenosis. This allows prediction of NASCET-type percent from millimeter stenosis.