1539. Quality Comparison of Portal Venous Phase Derived Virtual Noncontrast Images Generated by Photon-Counting CT with Dual-Source Dual-Energy CT
Authors * Denotes Presenting Author
  1. Andrew Ruff *; New York University
  2. Thomas O’Donnell; New York University
  3. Luke Ginocchio; New York University
  4. Vinay Prabhu; New York University
  5. Alexander El-Ali; New York University
  6. Alec Megibow ; New York University
  7. Bari Dane; New York University
Conventional energy-integrating CT creates images from all x-ray photon hitting the detector while photon counting CT (PCCT) bins the specific energies of photons to create the image. PCCT is more sensitive to lower energy photons, and by distinguishing them, can boost their contribution to the overall image. The purpose of this study was to compare the image quality of portal venous phase derived virtual noncontrast (VNC) images created using a PCCT with a dual-source dual-energy CT (dsDECT) in the same patient using quantitative and qualitative analysis.

Materials and Methods:
A total of 74 consecutive patients (27 men, mean [SD] age: 63 [13] years) with available portal venous phase derived VNC images from a PCCT and dsDECT were retrospectively identified by PACS search (median {IQR}: 209{465} days between CTs). Three fellowship-trained radiologists blinded to the source of VNC derivation qualitatively assessed deidentifed VNC images on a 5-point Likert scale in terms of overall image quality, image noise, delineation of small structures, noise texture, artifacts, and degree of iodine removal. Quantitative assessment required region of interest (ROI) to be drawn in the aorta at 4 different standard locations, within both psoas muscles, both renal cortices, spleen, air, retroperitoneal fat, and the inferior vena cava. Attenuation values (HU), quantitative noise (HU standard deviation, HUsd), contrast-to-noise (CNR) (CNRvascular, CNRkidney, CNRspleen, CNRfat), signal-to-noise (SNR) (SNRvascular, SNRkidney, SNRspleen, SNRfat), and CTDIvol were compared for PCCT and dsDECT with the Mann-Whitney U test. A p < 0.05 indicated statistical significance.

CTDIvol for PCCT and dsDECT were 9.2 [3.5]mGy and 9.4 [9.0]mGy, respectively (p = 0.06). Portal venous phase derived VNC images from PCCT had lower attenuation (all p<0.05), quantitative noise (HUsd, p = 0.02), and higher CNR for all computed values (p <0.0001-0.04). Contrast-enhanced structures had lower SNR on PCCT than dsDECT (p = 0.001-0.003) reflecting improved contrast removal on VNC images. SNRfat, a nonenhancing structure, showed higher SNR with PCCT than dsDECT (p < 0.00001). Qualitatively, PCCT VNC images had better overall image quality (4.1 [0.6] vs. 2.8[0.5]), image noise (3.9 [0.5] vs. 2.7 [0.5]), delineation of small structures (4.1 [0.8] vs. 3.1 [0.6]), noise texture (4.0 [0.5] vs. 2.8 [0.5]), and fewer artifacts (4.3 [0.5] vs. 3.5 [0.9]), all p < 0.0001. Visually, dsDECT VNC images had a more ideal degree of contrast removal in the aorta than PCCT images (0.1 [0.2] vs. 0.6 [0.4], p < 0.00001).

In conclusion, portal venous phase derived VNC images created with PCCT had improved image quality, lower noise, improved CNR and SNR than those derived from dsDECT with similar CTDIvol. With excellent image quality, PCCT VNC images can replace true noncontrast images to thereby reduce patient radiation in certain protocols, improve radiologist confidence, and afford accurate diagnosis of incidental findings.