Virtual non-contrast detector-based spectral CT predictably overestimates tissue density for the characterisation of adrenal lesions compared to true non-contrast CT.

Abstract

To establish if virtual non-contrast (VNC) images generated from contrast-enhanced detector-based spectral CT could replace true non-contrast (TNC) imaging for the characterisation of adrenal masses. TNC and VNC images were retrospectively reviewed for 39 patients with one or more adrenal lesions who underwent contrast-enhanced spectral CT of the upper abdomen. Lesions were categorised as either 'adenoma' or 'indeterminate/other lesion' based on current reference standards. The CT density of each lesion was measured on both image sets by two readers and compared using Wilcoxon signed-rank test. ROC analysis with Youden's J index method was performed to determine the optimal attenuation cut-off for diagnosing benign adenoma on VNC images. Forty-four lesions were included, 37 of which were diagnosed as adenomas. There were significant differences between TNC and VNC measurements for both readers (mean difference 9.1 HU for reader 1; 9.8 HU for reader 2; p < 0.01). Optimal attenuation thresholds for diagnosing adenomas on VNC were 25.3 HU (reader 1) and 23.9 HU (reader 2) for the entire population, and 18.3 HU (reader 1) and 19.7 HU (reader 2) for lipid-rich adenomas < 10 HU on TNC imaging. There is insufficient evidence to support the use of VNC as a substitute for TNC images in the characterisation of adrenal lesions. VNC using a detector-based spectral CT scanner shows a predictable increase in attenuation values compared to TNC. Thus, future studies might be better directed towards finding a new threshold value for diagnosing benign adrenal adenomas on VNC imaging.