Room: Room 205
Purpose: Virtual non-contrast (VNC) imaging uses dual-energy CT (DECT) acquisition combined with image processing to remove iodine contrast in a contrast-enhanced CT examination. This study compares the accuracy of CT numbers in VNC imaging in different types of DECT scanners over a range of scan geometries.
Methods: Three different scanners were evaluated in this study corresponding to three different DECT technologies: dual-source, kV-switching, and split-filter. VNC image CT number accuracy was assessed using an adult-torso-sized phantom containing material inserts of differing iodine concentrations. Automated ROI analysis measured iodine insert ROI means and ROI variation along the scanner z-axis. VNC errors (deviation from 0 HU) were measured for all iodine inserts. Test conditions included varying material insert position within the phantom (8 and 16 cm from phantom center), and phantom vertical centering (-4.5, 0, and +4.5 cm offset from isocenter). A stable scan range was defined as the length of the scan which maintained CT numbers within a tolerance of 20 HU along the z-axis.
Results: The dual-source, split-filter, and kV-switching DECT VNC scans had root-mean-square (RMS) errors of 9.2, 12.1, and 10.9 HU, respectively, among iodine inserts. RMS differences due to insert position within phantom were 13.7, 6.7, and 5.3 HU DECT and RMS differences due to phantom centering were 4.2, 10.1, and 17.7 HU. The stable scan ranges for a 160 mm scan were 138 mm, 85 mm, and 135 mm.
Conclusion: VNC CT number accuracy is similar among the three DECT technologies. The position of the iodine material within the phantom and the phantom centering are more important factors impacting VNC image accuracy than the type of DECT technology. The VNC from the split-filter scanner had large deviations at the end slices leading to a shorter stable scan range.