Rationale and objectives: To compare dual-energy CT virtual monoenergetic images (VMI) and standard reconstructions for reliability of quantitative size measurements and diagnostic accuracy for the detection of colorectal liver metastases (CRLM).
Materials and methods: We retrospectively included 98 patients (mean age, 61.1±11.5 years) with colorectal cancer, of whom 49 subjects had CRLM. All patients underwent a portal-venous phase dual-energy CT examination. Standard linearly-blended reformats and 40-keV VMI were reconstructed. For both reconstruction techniques, two blinded readers performed measurements of CRLM twice in a preset sequence. Three additional radiologists independently assessed all liver lesions in terms of dignity (benign vs. malignant). Sensitivity, specificity and diagnostic accuracy were calculated on a per-patient basis using MRI as reference standard. Readers scored the suitability for metric measurements and their diagnostic confidence using 5-point Likert scales. Inter-rater agreement was evaluated using intraclass correlation coefficient (ICC).
Results: Inter-rater agreement for lesion size measurements was higher for 40-keV VMI (ICC, 0.88) compared to standard linearly-blended series (ICC, 0.80). Sensitivity and diagnostic accuracy for the detection of CRLM were significantly higher for VMI at 40-keV compared to standard reconstructions (90.6% vs. 80.6%, and 89.1% vs. 81.3%; p < 0.001). Reader scores indicated that 40-keV VMI were more suitable for metric lesion measurements and provided greater diagnostic confidence compared to standard reformats (median, 5 vs. 3, and 5 vs. 4; both p < 0.001).
Conclusion: Low-keV VMI reconstructions improve reliability of quantitative size measurements and diagnostic accuracy for the assessment of CRLM compared to standard linearly-blended images.
Keywords: Diagnostic accuracy; Dual-energy computed tomography; Liver metastases; Measurement reliability; Virtual monoenergetic imaging.
Copyright © 2019. Published by Elsevier Inc.