ACCURACY AND PRECISION OF CT ANGIOGRAPHY IN A MODEL OF CAROTID-ARTERYBIFURCATION STENOSIS

PURPOSE: To determine optimal acquisition parameters and measurement t echniques for CT angiography of the carotid bifurcation. METHODS: Anat omic phantoms were created in which the diameter of the carotid artery stenoses ranged from 15% to 95%. Initially, we compared the accuracy of stenosis determination obtained by using various values of section collimation and table pitch, Subsequently, applying the combination of collimation and pitch that yielded the greatest longitudinal coverage without degradation in accuracy, we compared the accuracy of measurem ents performed with various display algorithms, including axial, magni fied axial, maximum intensity projection (MIP), and shaded surface dis play (SSD) images. Last, we determined the effect on accuracy of varyi ng both window and level settings. The standard of reference for all m easurements was considered to be caliper measurements made of the mode ls at the time of their construction. RESULTS: CT angiography was high ly accurate for determining the percentage of stenosis; the average di fference between CT angiographic measurements and the standard of refe rence was less than 1% for all parameter combinations and measurement techniques. Precision varied among the measurement techniques. Magnifi ed axial images provided more precise measurements than either the MIP or SSD images, Although there was a trend toward improved precision w ith the use of magnified versus unmagnified axial images and MIP versu s SSD images, neither of these comparisons reached statistical signifi cance. Systematic error was produced by changing the level setting fro m that halfway between the luminal density and vessel wall density. Ra ndom error was introduced by using window settings greater than zero. CONCLUSION: CT angiography was highly accurate and precise for determi ning percentage of stenosis. The highest precision was attained by usi ng magnified axial images with the level halfway between luminal densi ty and vessel wall density and with the window set to zero.