Abstrakt: |
Objective: Perform multireader analysis of objective and subjective lesion conspicuity for small pancreatic adenocarcinomas using rapid switching dual energy CT (rsDECT). Materials and methods: With IRB approval, 51 abdominal multiphasic rsDECT scans in 46 subjects with small (<3 cm) pancreatic adenocarcinomas were retrospectively reviewed by three independent readers for objective and subjective lesion conspicuity. Measured variables during individual, blinded interpretive sessions of separate low (52) keV, PACS-equivalent (70) keV, and iodine material density (MD) image sets included Hounsfield units (HU) and mg/cc iodine for tumor, nontumoral pancreas, and subcutaneous fat. Objective measures included absolute lesion contrast (LC) and contrast to noise ratios (CNR). Subjective measures included image quality, lesion conspicuity, and reader confidence. Reader agreement was measured with kappa statistic; correlation with truth by Pearson coefficient, CNR with repeated mANOVA; subjective quality measures utilized Tukey-Cramer corrections for multiple testing, p < 0.05 considered significant. Results: Demographics: 26 F, 20 M, mean age 68 years, weight 75 kg, tumor size of 2.3 cm. LC was highest on 52 keV images for all three readers (mean 90.1 HU). Mean CNR for iodine MD images (4.87) was significantly higher than 52 keV (4.13) or 70 keV (3.9). Very high to near-perfect kappa values were observed for all individual measured objective variables but were best for 52 keV images (52 keV 0.89-0.95, 70 keV 0.76-0.83, iodine 0.87-0.92). 70 keV images scored best for subjective image quality; iodine MD images scored best for lesion conspicuity and reader confidence. Conclusion: We observed very high reader agreement for independent objective rsDECT image variables and subjective rsDECT image scores in patients with small pancreatic adenocarcinomas. Maximal objective tumor to nontumoral LC was depicted on 52 keV and highest CNR on iodine MD images; readers scored the iodine MD images best for lesion conspicuity and confidence. [ABSTRACT FROM AUTHOR] |