Dual-energy CT of pancreatic adenocarcinoma: reproducibility of primary tumor measurements and assessment of tumor conspicuity and margin sharpness

Purpose: To determine the inter- and intra-reader agreement of size, conspicuity, and margin sharpness of pancreatic adenocarcinoma on monochromatic, polychromatic, and iodine map dual-energy CT (DECT) images. Methods: Retrospective review of DECT images from 61 patients with untreated pancreatic adenocarcinoma was performed by three radiologists independently. Pancreatic parenchymal phase images were generated as 50 and 70 keV, 140 kVp quality control (QC), and iodine map images. These were analyzed in a blinded randomized order during four reading sessions separated by 5–7 days. For each image set, readers recorded the longest axial and perpendicular primary tumor dimensions, and qualitatively scored tumor conspicuity and edge sharpness on 5-point scales. Linear mixed model was used to estimate and compare tumor measurements, tumor conspicuity, and tumor edge sharpness scores between readers and image sets. Kappa statistics were used to determine inter-observer agreement for tumor conspicuity and edge sharpness. Results: The range of tumor measures (mean of longest dimension ± standard deviation) was 3.18 ± 1.41 to 3.83 ± 1.57 cm. Reproducibility of tumor measurements was very high with mild variability (s² = 0.01–0.10) between readers for the different image sets. Inter-observer agreement values for tumor conspicuity (κ = 0.01–0.17) and edge sharpness (κ = 0.12–0.25) were low for all image sets, although two of three readers scored tumor conspicuity and edge sharpness higher on monochromatic and iodine map DECT images than on 140 kVp QC images (p < 0.05). Conclusions: Pancreatic adenocarcinoma measurements were highly reproducible on DECT images, and subjective reader preference trended toward monochromatic and iodine images rather than polychromatic images.