[Dual energy CT of the peripheral arteries: a phantom study to assess the effect of automatic plaque removal on...

To evaluate the accuracy of dual energy (DE)-based plaque removal in a vessel phantom.

Acrylic vessel phantoms of different diameters (3, 5, 8 mm), degrees of stenoses (25 - 100%) and plaque densities (300 - 750 HU) were filled with contrast-enhanced blood (150 - 450 HU). Dual source CT was used for simultaneous image acquisition at 80 and 140 kV. Beside a DE-based plaque-subtracted dataset (DE-PS), a virtual 120 kV non-plaque subtracted dataset (N-PS) was generated. Agreement between the known and measured luminal diameter in both datasets was determined using Lin's concordance correlation coefficient (kappaLin).

A total of 8260 measurements were taken. The correlation of measured diameter in DE-PS images was excellent (kappaLin = 0.83 - 0.96) for 5 - 8 mm vessel phantoms with high luminal enhancement (300 - 450 HU) and plaque density (500 - 750 HU), moderate (kappaLin = 0.6 - 0.67) for 5 mm vessels with lower luminal enhancement and plaque density and poor (kappaLin = 0.10 - 0.64) in the 3 mm vessels. The correlation of N-PS-based stenosis quantification was excellent (kappaLin = 0.86 - 0.99) for 5 - 8 mm vessel phantoms if the contrast between lumen and plaque was above 100 HU. The correlation decreased in 3 mm vessels (kappaLin = 0.45 - 0.93), while the lowest correlation was observed for the lowest contrast between plaque and vessel lumen.

Automatic DE-based plaque removal is highly effective for heavily calcified plaques and high luminal enhancement in larger diameter vessels > or = 5 mm). However, accuracy is limited for low density calcified plaque, lower luminal enhancement and smaller caliber vessels mainly due to poor specificity.