Work

Quantitative chest computed tomography: regional differences in dual-energy-derived virtual vs. true non-contrast scans

Year: 2025

Type: article

Abstract: Aim Quantitative analysis of chest computed tomography (CT) usually requires non-contrast scans. Multi-energy CT-derived virtual non-contrast (VNC) images may obviate additional true non-contrast (TNC) scans. We aimed to quantitatively compare virtual and true non-contrast images, with a focus on regional differences in pulmonary tissue. Materials and Methods This bi-center retrospective, IRB-approved study included consecutive patients who received contrast-enhanced dual-energy and non-contrast chest CT scans within one year between April 2018 and December 2022. Scans were co-registered and semi-manually segmented into various volumes of interest: whole lung, center/periphery, ventral/dorsal half and upper/middle/lower third. The mean lung density, high and low attenuation volumes and the lower 15th percentile were computed from the VNC and TNC scans for each volume. Correlation and agreement were assessed using Pearson's and intraclass correlation coefficients and Bland–Altman analysis. Subgroup analyses involved body mass index and markers of pulmonary emphysema. Results Forty-eight patients were included (26 females, median age: 63 years, interquartile range: 58–69). Pearson's and intraclass correlation coefficients > 0.9 demonstrated high correlation and agreement for the mean lung density, low attenuation volume and the lower 15th percentile across the whole lung. The agreement for high attenuation volume was slightly less pronounced (intraclass correlation coefficient, 0.84 [95% confidence interval: 0.73, 0.91]). Regional differences were observed, for example, concerning the mean lung density with mean differences (HU VNC minus HU TNC ) of -5.71 HU [-11.07, -0.35], 1.62 HU [-2.98, 6.23] and 5.77 HU [-0.47, 12.01] for the upper, middle and lower third, respectively. Conclusion Although slight regional differences were observed, quantitative analysis can be reliably performed using VNC images.

Source: Clinical Radiology

DOI: 10.1016/j.crad.2025.106912