2346. Intra-Individual Reproducibility of Virtual Noncontrast Measurements: Analysis with Three Different Dual-Energy CT Scanner Types
Authors * Denotes Presenting Author
  1. Simon Lennartz *; Massachusetts General Hospital
  2. Anushri Parakh; Massachusetts General Hospital
  3. Jinjin Cao; Massachusetts General Hospital
  4. Avinash Kambadakone; Massachusetts General Hospital
To evaluate the intra-individual, longitudinal reproducibility of attenuation measurements performed in virtual non contrast (VNC) images derived from three different dual-energy CT (DECT) scanner types.

Materials and Methods:
140 patients who underwent two repeated portal venous phase DECT examinations for clinical purposes were retrospectively included in this IRB-approved, HIPAA-compliant study: 51 patients with dual-source DECT (dsDECT), 46 patients with rapid switching keV DECT (rsDECT) and 43 patients with dual-layer detector DECT (dlDECT). VNC as well as virtual monoenergetic images at 65 keV (VMI65keV) were reconstructed, the latter serving as contrast-enhanced reference images. Hounsfield unit (HU) attenuation in VNC and VMI65keV was measured by three readers who placed regions of interest in the liver, spleen, renal cortex, aorta, portal vein and subcutaneous fat. Only parenchyma with physiologic imaging appearance and without interval changes between examinations was included. Absolute differences (|HUScan1-HUScan2|) and relative variation (|(HUScan1-HUScan2)/((HUScan1+HUScan2)x0.5)x100|) between repeated examinations were calculated and compared between VNC and VMI65keV using the Wilcoxon signed-rank test. Comparisons between scanner types were performed using the Steel Dwass test.

Averaged over all scanner types, absolute HU differences between repeated examinations were significantly higher in VMI65keV compared to VNC for all organs, vessels and fat (all p<0.05). Relative inter-scan variation was comparable between VNC and VMI65keV in the liver (p=0.19) and in fat (p=0.32), while it was lower for VNC in the spleen (p<0.05). In contrast, VNC showed a higher relative inter-scan variation than VMI65keV in the aorta, the renal cortex of both kidneys and the portal vein (all p<0.05). Averaged over all assessed regions, absolute attenuation differences in VNC were 7.6 ± 6.1 HU for dsDECT, 4.3 ± 3.8 HU for rsDECT and 4.3 ± 4.5 HU for dlDECT, showing significant differences between dsDECT and the other two scanner types (p<0.05). The corresponding relative inter-scan differences were highest in dsDECT (20.6 ± 21.7) compared to 9.2 ± 8.2 in rsDECT and 8.0 ±9.8 in dlDECT (all p<0.05).

Variation between VNC measurements in repeated dual-energy scans seems to be dependent on the iodine load of the corresponding region: it was higher than in contrast-enhanced VMI65keV in vessels and the renal cortex and comparable/lower in the liver/spleen, respectively. Absolute and relative longitudinal variations showed certain differences between dual-energy CT scanner types. Both findings should be acknowledged when performing VNC measurements in a longitudinal setting, e.g. for renal lesion characterization.