Objective: To evaluate accuracy of virtual-non-contrast images (VNC) compared to true-unenhanced-images (TNC) for evaluation of liver attenuation acquired using spectral-detector CT (SDCT). Methods: 149 patients who underwent multiphase transcatheter-aortic-valve-replacement (TAVR) SDCT-examinations [unenhanced-chest (TNC), CT-angiography chest (CTA-chest, early arterial-phase) and abdomen (CTA-abdomen, additional early arterial-phase after a second injection of contrast media)] were retrospectively included. VNC of CTA-chest (VNC-chest) and CTA-abdomen (VNC-abdomen) were reconstructed and compared to TNC. Region of interest-based measurement of mean attenuation (Hounsfield unit, HU) was applied in the following regions: liver, spleen, abdominal aorta and paraspinal muscle. Results: VNC accuracy was high in the liver, spleen, abdominal aorta and muscle for abdomen-scanning. For the liver, average attenuation was 59.0 ± 9.1 HU for TNC and 72.6 ± 9.5 HU for CTA-abdomen. Liver attenuation in VNC-abdomen (59.1 ± 6.4 HU) was not significantly different from attenuation in TNC (p > 0.05). In contrast, VNC was less accurate for chest-scanning: Due to the protocol, in CTA-chest no contrast media was present in the liver parenchyma as indicated by the same attenuation in TNC (59.0 ± 9.1 HU) and CTA-chest (58.8 ± 8.9 HU, p > 0.05). Liver attenuation in VNC-chest (56.2 ± 6.4 HU, p < 0.05) was, however, significantly lower than in TNC and CTA-chest implying an artificial reduction of attenuation. Conclusion: VNC performed well in a large cohort of TAVR-examinations yielding equivalent mean attenuations to TNC; however, application of this technique might be limited when no or very little contrast media is present in parenchyma, more precisely in an early arterial-phase of the liver. Advances in knowledge: This study showed that VNC can be reliably applied in cardiac protocols when certain limitations are considered