Variability of CT Attenuation Measurements in Virtual Unenhanced Images Generated Using Multimaterial Decomposition from Fast Kilovoltage-switching Dual-energy CT.

RATIONALE AND OBJECTIVES: To compare Hounsfield unit (HU) data obtained from true-unenhanced (TUE) and virtual-unenhanced (VUE) imaging obtained with a fast kv-switching dual-energy computed tomography (CT) scanner using multimaterial decomposition algorithm.

MATERIALS AND METHODS: In this Institutional Review Board-approved, Health Insurance Portability and Accountability Act-compliant, retrospective cohort study, CT scans of 19 patients undergoing multiphasic renal protocol abdominal CT on a fast kv-switching dual-energy CT scanner were reviewed. CT numbers were measured on the matched TUE and VUE generated using a multimaterial decomposition algorithm with selective iodine suppression, and postcontrast images at predefined locations in seven organs. Six hundred sixty regions of interest were placed at 132 locations. Agreement was assessed with paired t test, Pearson's correlation, and Bland-Altman analysis.

RESULTS: Mean TUE and VUE measurements were not significantly different in the corticomedullary (P = 0.25) or nephrographic (P = 0.10) phases. There was a strong correlation between TUE and VUE CT numbers (corticomedullary: r = 0.90, nephrographic: r = 0.90, each P < 0.001). Discrepancies ≥5 HU occurred 46 times (35%, 46 of 132) in the corticomedullary phase and 44 times (33%, 44 of 132) in the nephrographic phase. Discrepancies ≥10 HU occurred in 7% (9 of 132 in both corticomedullary and nephrographic phases). Interphase, intrasubject VUE CT numbers were strongly correlated (r = 0.93, P < 0.001), but discrepancies ≥5 HU (22% [29 of 132]) and ≥10 HU (2% [3 of 132]) occurred. There was no significant correlation between the true postcontrast CT number and the magnitude of VUE-TUE discrepancy (r = -0.04, P = 0.6).

CONCLUSION: CT numbers on VUE images generated from fast kv-switching dual-energy CT scans strongly correlate with TUE CT numbers on a population basis, but commonly vary 5-9 HU on a per-patient basis.