Accuracy of Dual-Energy CT Virtual Unenhanced and Material-Specific Images: A Phantom Study

The purpose of this study was to determine the quantification accuracy of virtual unenhanced images and establish the lower limit of iodine quantification as a function of dose. A large elliptical and cylindric phantom mimicking the patient abdomen was scanned on two commercial dual-energy CT scanners, an IQon Spectral CT (Philips Healthcare) and a Revolution CT with Gemstone Spectral Imaging Xtream suite (GE Healthcare). The phantom contained simulated soft tissue, blood, and bone with known elemental composition. It also contained simulated iodine concentrations (0.2-15.0 mg/mL) and iodine-enhanced blood (0.5-5.0 mg/mL). The mean absolute error in CT value for virtual unenhanced images and mean absolute percent error in iodine, calcium, and fat-specific images were measured. For virtual unenhanced images, when excluding the simulated bone, the mean absolute error in CT value was 8.0 ± 5.0 (SD) HU and 9.0 ± 6.2 HU for the IQon and the Revolution CT, respectively ( = 0.61). The mean error in CT value of the simulated bone was -90.5 ± 111.6 HU and -98.5 ± 117.8 HU on the IQon and the Revolution CT, respectively ( = 0.08). For iodine-specific images, the mean absolute percent error was 13.7% and 8.3% for the IQon and the Revolution CT, respectively, above 0.5 mg/mL iodine concentration, and 150% and 100% at less than 0.5 mg/mL iodine concentration. The mean absolute percent error increased from 16.2% at 100% radiation dose to 18.9% and 24% at 75% and 50% dose, respectively, on the IQon; and from 8.8% at 100% dose to 11.1% and 17.8% at 75% and 50%, respectively, on the Revolution CT. Virtual unenhanced images are reasonably accurate for simulated soft tissues and contrast materials, except for simulated bone. The lower limit of iodine quantification is radiation-dose dependent. For typical dose levels, 0.5 mg/mL iodine concentration is the lower threshold for iodine detection accuracy.