Purpose: We studied which set of CT parameters and modeling parameters yielded accurate measurements of three graded artificial renal artery stenoses.
Method: An acrylic phantom resembling the abdominal aorta and renal arteries was constructed. Stenotic segments had diameters of 1.8, 3.2, and 4.8 mm; nonstenotic segment diameter was 6.3 mm. Helical scans were done using 1 and 3 mm collimation at pitches of 1, 1.5, and 2. 3D renderings were produced and measured. Multifactorial and regression tree analysis were used to determine the accuracy of the 3D renderings. Mean squared error (MSE) was used to compare true diameter with measured diameter.
Results: Collimation of 1 mm produced an MSE of 0.55 versus an MSE of 1.35 for 3 mm collimation. Stenosis grade was the next most important parameter in the 1 mm subgroup and viewing direction in the 3 mm collimation subgroup. In the 1 mm subgroup, high and mid grade stenoses had an MSE of 0.52 versus low grade stenosis that had an MSE of 0.61. Pitch was a fourth-order effect.
Conclusion: Collimation of 1 mm combined with a pitch ratio as high as 2:1 is superior to 3 mm collimation. Shaded surface modeling was the single best choice for rendering 3D data. Stenosis grade interacted strongly with user-controllable parameters.