Virtual non-contrast for evaluation of liver parenchyma and vessels: results from 25 patients using multi-phase spectral-detector CT

Abstract

Background

In abdominal imaging, contrast-enhanced computed tomography (CT) examinations are most commonly applied; however, unenhanced examinations are still needed for several clinical questions but require additional scanning and radiation exposure.

Purpose

To evaluate accuracy of virtual non-contrast (VNC) from arterial and venous phase spectral-detector CT (SDCT) scans compared to true-unenhanced (TNC) images for the evaluation of liver parenchyma and vessels.

Material and Methods

A total of 25 patients undergoing triphasic SDCT examinations were included. VNC was reconstructed from arterial and venous phases and compared to TNC images. Quantitative image analysis was performed by region of interest (ROI)-based assessment of mean and SD of attenuation (HU) in each liver segment, spleen, portal vein, common hepatic artery, and abdominal aorta. Subjectively, iodine subtraction and diagnostic assessment were rated on 5-point Likert scales.

Results

Attenuation and image noise measured in the liver from VNC were not significantly different from TNC (TNC: 54.6 ± 10.8 HU, VNC arterial phase: 55.7 ± 10.8 HU; VNC venous phase: 58.3 ± 10.0 HU; P > 0.05). VNC also showed accurate results regarding attenuation and image noise for spleen, portal vein, and abdominal aorta. Only iodine subtraction in the common hepatic artery in the arterial phase was insufficient which was confirmed by the subjective reading. Apart from that, subjective reading showed accurate iodine subtraction and comparable diagnostic assessment.

Conclusion

VNC from the arterial and venous phases were very similar to TNC yielding mostly negligible differences in attenuation, image noise, and diagnostic utility. Inadequate iodine subtraction occurred in hepatic arteries in the arterial phase.

Keywords

X-ray computed tomography dual-energy computed tomography spectral-detector computed tomography virtual non-contrast iodine

Introduction

In abdominal imaging, contrast-enhanced computed tomography (CT) is broadly applied, e.g. in oncologic imaging or to detect vessel pathologies (1,2). Unenhanced examinations of the abdomen are less commonly applied in clinical routine; however, they are still required for several indications, e.g. for characterization of renal, adrenal, and certain liver lesions such as hepatocellular carcinoma (3,4) as well as evaluation of vascular disease or abdominal bleeding (1,5). Non-invasive determination of liver steatosis using attenuation is another field of application for unenhanced images that continuously gains importance in clinical patient care (6,7). More than 30% of liver fat fraction is considered to have relevant clinical implications, e.g. post-surgery prognosis after liver surgery (8), poor liver graft outcome/failure (9), and as a biomarker for cardiovascular events (10). Non-alcoholic liver steatosis is the most common type of fatty liver disease with an incidence of around 15% in the general population and highly elevated in obese patients, apart from inflammation or excessive alcohol consumption that can also cause steatosis or steatohepatitis (7,11). CT imaging offers two different methods to determine elevated fatty liver content. The first one uses a simple attenuation cut-off, meaning <40 Hounsfield Units (HU) being predictive for a liver fat content of >30% (12). The second method proposes a liver attenuation index based on the difference between liver and spleen attenuation which is also predictive for a liver fat fraction >30% when the difference is <–10 HU (13). Both methods are highly specific but only yield a moderate sensitivity for the diagnosis of liver steatosis (14). Each, however, is reliant on unenhanced examinations that are frequently unavailable in clinical routine.

Material decomposition from dual-energy CT (DECT) offers an opportunity to derive virtual unenhanced or non-contrast images (VNC) from contrast-enhanced CT examinations by identifying and then subtracting iodine. Image properties of VNC correspond to images acquired with true unenhanced examinations (3,15,16). Promising results have been shown in abdominal imaging presenting similar and comparable attenuation and image noise properties (1,3,7). There are different technical solutions to DECT (1,17) that can be divided into detector- and emission-based solutions. For clinical application, several emission-based DECT techniques are available: dual-source; split or twin beam; and rapid kilovolt-switching. The only detector-based approach for clinical application is a dual-layer spectral-detector CT (SDCT). This scanner uses a dual-layer detector and a single X-ray source. Low-energetic photons are detected at the upper layer while high-energetic photons are detected at the lower layer (17,18). Data are registered in both temporal and spatial coherence facilitating postprocessing of dual-energy information and creation of VNC images. Post-processing in projection compared to image domain could potentially offer higher accuracies (1,17).

The introduction to the clinical application of SDCT has been more recent compared to other dual-energy approaches. Therefore, studies investigating VNC imaging facilitated by SDCT are still rather rare and there is only one study evaluating VNC performance of the entire abdomen in patients on a clinical SDCT scanner (1). Hence, the purpose of this study was to evaluate the SDCT-based application of VNC for the assessment of liver parenchyma and vessels as well as to test whether iodine subtraction operates comparably to other DECT approaches as elucidated in prior studies. For that matter, we tested the accuracy of VNC iodine subtraction in arterial and venous phase acquisitions by comparison to TNC images.

Material and Methods

This retrospective study was approved by the institutional review board. Standards of the Health Insurance Portability and Accountability Act were followed. Informed written consent was waived under the Code of Federal Regulations (title 45, §46.116d). Criteria for inclusion were: (i) contrast-enhanced abdominopelvic triphasic examinations including unenhanced, arterial and venous phase images, acquired in a supine head-first position on a clinical spectral-detector CT (IQon, Philips Healthcare, Cleveland, OH, USA) between June 2017 and June 2018; (ii) patient age at date of the examination ≥18 years; and (3) scans including the imaging protocol described below. Twenty-five patients fulfilled all criteria. Clinical indication was required for each examination.

Imaging protocol

For contrast-enhanced examinations, a bodyweight-adapted amount of iodinated contrast media (1.5 mL/kg; Optiray 350, Guerbet, Bloomington, IN, USA; maximum amount of contrast media injected was set to 150 mL) with a flow rate of 3.5 mL/s was intravenously administered. The arterial phase acquisitions were started 8–12 s (delays were adjusted to the clinical question; see the first paragraph of the “Results” section) after reaching a 100-HU threshold in the descending aorta; the venous phase acquisitions were started 70–80 s after contrast media injection. Scan parameters were set as follows: matrix = 512 × 512; collimation = 64 × 0.625 mm; rotation time = 0.40 s; and pitch = 1.02. Tube voltage was set to 120 kVp and anatomic-based tube current modulation was applied (DoseRight 3D-DOM, Philips, Cleveland, OH, USA). Conventional images from the unenhanced examinations (TNC), as well as arterial and venous phase, were reconstructed with the vendor’s hybrid-iterative reconstruction algorithm (iDose 4, level 3, Philips Healthcare, Cleveland, OH, USA). VNC from the arterial and venous phase were reconstructed using the vendor’s spectral reconstruction algorithm (Spectral B, level 3). Slice thickness was set to 3 mm.

Objective analysis

Image analysis was carried out using the vendor’s proprietary image viewer (IntelliSpace Portal v9, Philips, Cleveland, OH, USA) on axial images by placing circular regions of interest (ROI) with a standardized size of 200 mm². Size of the ROI was reduced only for the common hepatic artery to account for its small size. Placement of the ROI was performed in venous phase acquisitions first as anatomical orientation, and delineation of organs as well as vessels were easiest here. A comparable ROI was then placed on each of the unenhanced and arterial phase images. Transfer to corresponding VNC of arterial and venous phases was performed using the copy-and-paste functionality. Due to respiratory motion, ROI positions could differ slightly between the three acquisitions. To reduce bias from differences in ROI positioning, the readers used screenshots of each ROI in the venous phase to maintain the exact anatomic location in the other acquisitions. Two ROIs were placed in each liver segment (I, II, III, IVa, IVb, V, VI, VII, VIII), portal vein, common hepatic artery, and abdominal aorta. As well as in the liver, two ROIs were placed in the spleen, as liver parenchyma and its attenuation are often evaluated in context with the spleen (13,19). This resulted in a total of 3250 ROIs (26 ROIs in five reconstructions for each of the 25 patients). Mean and SD of attenuation (HU) were recorded. SD was considered indicative for image noise (20). Values from each pair of ROIs were averaged. To obtain attenuation and noise for the entire liver, values from all ROIs placed in each liver segment were also averaged. Differences in the attenuation of each structure (pair of ROIs) in VNC and TNC were determined as in earlier studies (1,21) to evaluate reliability of VNC.

Visual analysis

Visual assessment was performed by two board-certified fellowship-trained radiologists with seven and 14 years of experience in abdominal imaging. Iodine subtraction in VNC, image noise, and diagnostic utility were rated on 5-point Likert scales. Further, readers were asked to evaluate image distortion, i.e. contour softening and artificial image impression. A detailed description of the criteria for visual assessment is provided in Table 1. Readers were free to adjust window settings. The readers were unaware of the results from the objective analysis while performing the visual assessment.

Table 1.

Visual analysis.

Iodine subtraction of VNC in arterial and venous phases compared to TNC in the (i) right and (ii) left liver lobe, (iii) spleen, (iv) smaller liver vessels (smaller intrahepatic vessels, common hepatic artery), as well as (v) larger liver vessels (portal vein, left/middle/right hepatic vein)	5 = complete removal of contrast media,4 = almost complete removal of contrast media,3 = moderate removal of contrast with incomplete parenchymal areas or vessels,2 = partly sufficient removal of contrast media with larger incomplete areas,1 = insufficient removal of contrast media
Image noise in TNC and VNC of arterial and venous phases	5 = visually not perceptible, 4 = mild, 3 = moderate,2 = severe, 1 = excessive
Image distortion in VNC of arterial and venous phases (contour softening, artificial image impression)	3 = no image distortion, 2 = little image distortion without hampering diagnostic assessment, 1 = image distortion with hampering diagnostic assessment
Visual impression and diagnostic assessment of VNC of arterial and venous phases compared the TNC regarding iodine subtraction, image noise and image distortion in the (i) right and (ii) left liver lobe, (iii) spleen, as well as (iv) liver vessels (smaller intrahepatic vessels, common hepatic artery, portal vein, left/middle/right hepatic vein)	5 = improved visual impression and diagnostic assessment,4 = identical/almost identical visual impression and diagnostic assessment,3 = smaller differences in visual impression and diagnostic assessment,2 = differences in visual impression and hampered diagnostic assessment,1 = clear differences in visual impression and insufficient diagnostic assessment

TNC, true non-contrast; VNC, virtual non-contrast.

Statistical analysis

JMP Software was used for the statistical analysis (V12, SAS Institute, Cary, NC, USA). Quantitative results are shown as mean ± SD. Qualitative results from visual assessment are displayed as median and 10/90 percentile. Shapiro–Wilk test was used to evaluate the presence of normal distribution. Wilcoxon signed-rank test was used to test for any difference. The statistical significance was set to P <0.05. Intraclass correlation coefficient (ICC) was calculated and evaluated as proposed earlier, agreement is poor at <0.40, fair at 0.40–0.59, good at 0.60–0.75, and excellent at 0.75–1.0.

Results

A total of 25 patients were included (14 women, 11 men; mean age = 55.8 ± 16.1 years; age range = 26–93 years). Examinations were performed for the following clinical reasons; internal bleeding, n = 4; acute aortic aneurysm/dissection, n = 2; mesenteric ischemia, n = 2; evaluation infrarenal graft repair, n = 7; and kidney donor examination, n = 10.

Objective assessment

In the liver, VNC attenuation of arterial and venous phase images of the liver were very similar to TNC without significant differences averaging over the entire liver and in each liver segment alone (Table 2, Figs. 1 and 2a). Only liver segment 6 showed significant attenuation differences in VNC images from venous phase acquisition compared to TNC. Differences of mean attenuation values between VNC and TNC of the whole liver were 1.1 HU for arterial phase and 3.7 HU for venous phase acquisitions (Table 2, Fig. 1). Only one patient showed liver attenuation <40 HU, indicating a relevant liver steatosis in TNC (Fig. 3). Comparable liver attenuation was displayed in VNC arterial and venous phases, as shown in Fig. 1. Image noise in VNC from the whole liver and each individual liver segment were also not significantly different from TNC (Table 2). Only image noise of the VNC venous phase of liver segment 1 showed significant differences from TNC. In the spleen, mean attenuation differences between VNC from the arterial (3.5 HU) and venous phase (4.7 HU) and TNC were small but attenuation differed significantly (Table 2, Fig. 1; P <0.05).

Table 2.

Objective analysis of liver, spleen, and vessels.

	Liver parenchyma		Spleen parenchyma		Portal vein		Common hepatic artery		Abdominal aorta
	Attenuation	Image noise	Attenuation	Image noise	Attenuation	Image noise	Attenuation	Image noise	Attenuation	Image noise
TNC (mean ± SD)	54.6 ± 10.8	15.5 ± 2.1	46.6 ± 5.1	15.3 ± 2.5	39.2 ± 5.9	17.0 ± 2.6	37.3 ± 8.3	16.3 ± 3.6	39.7 ± 7.3	18.5 ± 2.6
VNC arterial phase (mean ± SD)	55.7 ± 10.8	14.8 ± 2.0	50.1 ± 5.6	15.7 ± 2.1	39.4 ± 6.3	16.5 ± 2.6	58.0 ± 14.5	19.1 ± 5.9	44.8 ± 8.4	19.3 ± 2.8
VNC venous phase (mean ± SD)	58.3 ± 10.0	16.4 ± 2.1	51.3 ± 5.9	15.2 ± 2.0	42.2 ± 7.0	18.5 ± 2.4	41.4 ± 10.9	18.0 ± 5.0	39.6 ± 8.3	19.6 ± 2.7
Arterial phase* (mean ± SD)	63.1 ± 12.4	18.6 ± 2.2	92.6 ± 24.1	24.4 ± 5.0	73.4 ± 24.0	21.1 ± 2.9	297.6 ± 62.5	41.6 ± 15.2	344.4 ± 74.7	24.0 ± 3.5
Venous phase^† (mean ± SD)	100.5 ± 20.0	21.3 ± 2.7	99.4 ± 21.6	19.7 ± 2.5	136.9 ± 28.5	23.0 ± 3.1	115.8 ± 24.6	22.9 ± 6.2	131.6 ± 24.8	24.0 ± 3.0
Difference of mean attenuation of VNC arterial phase and TNC	1.1	–0.7	3.5	0.4	0.2	–0.5	20.7	2.8	5.1	0.8
VNC venous phase and TNC	3.7	0.9	4.7	–0.1	3.0	1.5	4.1	1.7	–0.1	1.1
Arterial phase: VNC and TNC differences (%)
<5 HU	96		68		76		8		52
<10 HU	100		92		96		16		80
VNC and TNC differences (%)
<5 HU	64		56		56		56		76
<10 HU	92		84		88		80		92
P values
VNC arterial phase vs. TNC	>0.05	>0.05	0.006	>0.05	>0.05	>0.05	<0.001	>0.05	>0.05	>0.05
VNC venous phase vs. TNC	>0.05	>0.05	0.038	>0.05	>0.05	>0.05	>0.05	>0.05	>0.05	>0.05

Units for attenuation and image noise are HU. Values in bold represent significant differences between values.

*Conventional arterial phase reconstruction.

^†Conventional venous phase reconstruction.

TNC, true non-contrast; VNC, virtual non-contrast.

Fig. 1.

Box plot diagram displaying liver and spleen attenuation in true non-contrast (TNC), arterial phase, virtual non-contrast (VNC) arterial phase, venous phase, and VNC venous phase. Contrast media information is effectively removed by VNC of both displayed phases approximating the attenuation in TNC.

Attenuation and image noise of the portal vein and abdominal aorta in VNC images from the arterial and venous phases as well as attenuation of the common hepatic artery from the venous phase were not significantly different from TNC. VNC from the arterial phase of the common hepatic artery did, however, differ significantly from TNC, with a considerable 20.7 HU difference of mean attenuation (Table 2, Fig. 4).

Individual differences in attenuation of VNC from arterial and venous phases compared to TNC were mostly <10 HU in the liver, spleen, portal vein, and abdominal aorta (in the range of 80%–100%, Table 2). Only VNC in the common hepatic artery created from the arterial phase showed high individual attenuation differences mostly >10 HU (Table 2).

Visual assessment

Corresponding to the results of the objective analysis, VNC showed excellent capabilities for removing contrast media in the subjective assessment of the right and left liver lobe, the spleen, as well as larger liver vessels (Table 3, Fig. 2a), which resulted in a mostly comparable overall visual impression and diagnostic assessment to TNC (Table 3, Fig. 2a). Only iodine subtraction by VNC images in smaller liver vessels in the arterial phase was considered only as moderate, because of the incomplete subtraction of iodine in the hepatic arteries (Table 3, Fig. 2b). Inter-rater agreement was good, indicated by an ICC of 0.63.

Table 3.

Subjective assessment of iodine subtraction and diagnostic assessment in the liver, spleen, and vessels.

	Iodine subtraction					Image noise	Image distortion	Visual/Diagnostic assessment
	Right liver lobe	Left liver lobe	Spleen	Smaller liver vessels*	Larger liver vessels^†	Image noise	Image distortion	Right liver lobe	Left liver lobe	Spleen	Liver vessels^‡
TNC	N/A	N/A	N/A	N/A	N/A	4 (3/5)	N/A	N/A	N/A	N/A	N/A
VNC arterial phase	5 (4/5)	5 (5/5)	4 (1/5)	3 (2/4)	5 (3/5)	5 (4/5)	3 (2/3)	5 (4/5)	5 (4/5)	3 (2/5)	4 (2/5)
VNC venous phase	5 (4/5)	5 (4/5)	5 (4/5)	4 (3/5)	4 (2/5)	5 (4/5)	2 (1/3)	4 (3/5)	4 (3/5)	4 (4/5)	3 (2/4)
P values
TNC vs. VNC arterial phase	N/A	N/A	N/A	N/A	N/A	<0.001	N/A	N/A	N/A	N/A	N/A
TNC vs. VNC venous phase	N/A	N/A	N/A	N/A	N/A	<0.001	N/A	N/A	N/A	N/A	N/A

Values are given as median and 10/90 percentiles.

*Intrahepatic vessels, common hepatic artery.

^†Portal vein, left/middle/right hepatic vein.

^‡Smaller and larger liver vessels.

N/A, not applicable; TNC, true non-contrast; VNC, virtual non-contrast.

Fig. 2.

(a) Effective removal of iodine by VNC offering comparable visual impression and diagnostic assessment compared to TNC. Axial CT images in soft-tissue window settings (window level = 60, window width = 350) of the upper abdomen in a 26-year-old man receiving a kidney donor examination. Depicted are true non-contrast (TNC), arterial phase, virtual non-contrast (VNC) arterial phase, venous phase, and VNC venous phase. (b) Iodine subtraction by VNC in arterial phase is insufficient in hepatic arteries. Zoomed-in axial CT images of the same patient on a different level. VNC is not able to completely remove contrast media in hepatic arteries in the arterial phase (red arrows). Contrary VNC of the venous phase completely removes contrast media in hepatic arteries. Apparently the combination of small vessel diameter and high–contrast media concentration is problematic for iodine subtraction algorithm.

Subjective image noise was perceived as significantly lower in VNC compared to TNC (Table 3). Image distortion in VNC from the arterial phase was noticed in 6% (3/50) of the patients but was not considered to affect diagnostic assessment in any of the cases. In VNC from venous phase acquisition, image distortion was noticed more often. Image distortion was registered in 68% (34/50) of patients but was considered to negatively affect diagnostic assessment in only 10% (5/50) of these patients (Table 3).

Discussion

The present study evaluated the accuracy of iodine subtraction provided by VNC derived from SDCT compared to true unenhanced examinations with a focus on liver parenchyma and vessel depiction. Attenuation of VNC generated from arterial phase and venous phase data was very similar to values from TNC images with regards to organ parenchyma. Accuracies were highest in the liver and slightly less accurate in the spleen, possibly due to more inhomogeneous perfusion of the spleen (22). VNC accuracies regarding the portal vein and abdominal aorta were also high compared to TNC. The subjective analysis supported these findings of effective iodine subtraction and established comparable image quality and diagnostic utility of VNC images created from both contrast phases.

On the contrary, VNC iodine subtraction performed insufficiently in arteries of the liver in the arterial phase. This was shown by larger attenuation differences of the common hepatic artery in the arterial phase and supplemented by the subjective reading that showed inadequate iodine subtraction in the hepatic arteries in the arterial phase (Fig. 2b). Apparently, the combination of small vessel size (e.g. common hepatic artery, intrahepatic arteries) and the high contrast media concentration in the arterial phase in these vessels led to insufficient iodine subtraction by the algorithm. Contrastingly, hepatic vessels in the venous phase showed accurate iodine subtraction. Further, in the arterial phase, the abdominal aorta as a larger vessel was also not affected and iodine subtraction worked well. Another factor that might have impacted objective analysis of the common hepatic artery is its relatively small size which made accurate ROI placement more difficult and made it necessary to reduce the ROI size considerably to avoid inclusion of unrepresentative tissue. Larger studies are needed to confirm if iodine subtraction by SDCT does not work in hepatic arteries in the arterial phase. For dual-source and rapid kilovolt-switching CT, this problem of incomplete iodine removal in hepatic arteries is not described in the literature (23,24). SDCT separates photon energy on the detector level such that data are accurately matched temporally and spatially offering the advantage of material decomposition in the projection domain, which is not the case for emission-based DECT solutions. This is thought to be an advantage but might also create new challenges compared to material decomposition in the image domain (1,17). In addition, the overlap between photon energy spectra is relatively high in the detector-based approach; thereby, separation of low and high photon energies could potentially be less accurate (17). These technical differences should at least partly explain differing performance in iodine subtraction accuracies.

In the literature, differences in attenuation <10 HU were considered clinically negligible (1,21). In the present study, differences between VNC and TNC attenuation were <10 HU in >90% of measurements (excluding the common hepatic artery) and mean attenuation differences in liver were in the range of 1.1–3.7 HU. Studies investigating emission-based DECT mostly showed lower VNC accuracies (<10 HU in 75% of patients (25), mean attenuation differences of 5.4 HU for the liver (7)); however, a more recent study analyzing a third-generation dual-source CT reported comparable to even further improved VNC accuracy (mean differences in the range of 0–2 HU for the liver and spleen) (26). Compared to detector-based SDCT studies, our results have been more accurate (differences <10 HU in 75%–80% (1,21)). These improvements might be explained by using a clinical and not a prototype SDCT scanner (21) and by a more advanced and detailed ROI placement trying to minimize any measuring bias (1). Each ROI was carefully copied from the venous phase to the other reconstructions using screenshots for documentation to maintain anatomical positions despite respiratory motion. Further, two ROIs were placed and then averaged for each structure, and measurements were applied in each liver segment.

One patient in the present study showed reduced liver attenuation (Fig. 3). Liver attenuation was less than the cut-off value of 40 HU (12) and liver attenuation index was <–10 HU (19), both indicating liver steatosis and a fat fraction >30% with high specificity and moderate sensitivity (14). In our study cohort, VNC was slightly less accurate in the spleen. This could be relevant whenever the liver attenuation index is used as it comprises spleen attenuation as a reference measurement; therefore, inaccurate VNC imaging of the spleen might negatively influence accuracy in this setting. Liver steatosis is relatively common in the general population (7,11) and >30% of liver fat fraction is considered to have relevant clinical implications (9). Non-invasive, attenuation-based diagnosis with CT gains more relevance; however, it historically requires unenhanced examinations (6,7). One solution could be VNC when used as a replacement for TNC. Non-invasive diagnosis of liver steatosis would then be possible in all contrast-enhanced examinations allowing screening for fatty liver disease in clinical practice. Additional proof for SDCT is needed, preferably by pathology-based studies. Haji-Momeninan et al. (7) evaluated VNC using a dual-source CT scanner for the diagnosis of liver steatosis and reported promising results; however, overestimation of attenuation in VNC of the liver and spleen worsened diagnostic accuracy. Higher reliability of VNC in liver parenchyma as presented in our study and by Durieux et al. (26) in a third-generation dual-source scanner might increase diagnostic accuracies and thereby allow for broader clinical application for diagnosis of liver steatosis. This could also be applied to other pathologies that require unenhanced images or attenuation for their characterization, e.g. characterization of adrenal lesions, vessel pathologies, or acute bleeding (1,3,5). Iodine subtraction also worked well throughout the entire liver and each liver segment alone, supplemented by the subjective reading that showed complete iodine subtraction in both liver lobes. This is relevant as liver steatosis can occur asymmetrically and heterogeneously, resulting in the requirement for reliable attenuation-based diagnosis in each segment (6,7).

Fig. 3.

Patient with reduced liver attenuation indicating liver steatosis. Axial CT images in soft-tissue window settings (window level = 60, window width = 350) of a 56-year-old patient receiving a kidney donor examination. Depicted are true non-contrast (TNC), arterial phase, virtual non-contrast (VNC) arterial phase, venous phase, and VNC venous phase. In TNC as well as VNC arterial and venous phases, liver attenuation is <40 HU and liver attenuation index is <–10 HU. Both values indicate liver steatosis and are predictive for a fat fraction >30% as well as relevant clinical implication.

Fig. 4.

Box plot diagram displaying the portal vein, common hepatic artery, and abdominal aorta attenuation in true non-contrast (TNC), arterial phase, virtual non contrast (VNC) arterial phase, venous phase, and VNC venous phase. Contrast media information is effectively removed by VNC of both displayed phases approximating their true attenuations in TNC. Only subtraction of iodine in the VNC arterial phase was limited and attenuation values increased compared to TNC.

The gold standard for the diagnosis of liver steatosis and quantification of liver fat content remains histopathologic analysis (6). Potential disadvantages consist in the invasive procedure and potential sampling errors, as liver fat content can be distributed heterogeneously. Imaging-based assessment of liver fat allows for an evaluation of the entire liver, thereby sampling errors can be avoided. CT offers accurate and reliable quantification of liver fat, but results might be confounded by liver cirrhosis or depositional diseases. MRI examinations using chemical shift imaging also offer high diagnostic accuracies for the assessment of liver fat (6). Recently, MR elastography and proton density fat fraction have proven their dedicated value for further improved quantification of liver cirrhosis and steatosis as a potential replacement for liver sampling (27). Ultrasonography is less expensive and allows for effective detection of moderate or severe fatty infiltration, but reproducibility and reliability of results were shown to be lower (6).

In abdominal imaging, contrast-enhanced CT examinations are most commonly applied. VNC derived from contrast-enhanced scans could serve as a potential replacement for the characterization of renal, adrenal, and liver lesions (3,4) as well as the evaluation of vascular disease and/or abdominal bleeding (1,5), allowing for a relevant reduction of radiation exposure, scan time, and costs. Further, VNC offers great potential to replace TNC images of the liver for the characterization of liver lesions, e.g. in primary diagnosis of hepatocellular carcinoma or after radiofrequency ablation of liver lesions, VNC can be used to identify real hyperenhancing lesions and separate them from hyperdense structures without contrast agent enhancement (4,28). Similarly, abdominal bleeding could be identified with higher diagnostic certainty if VNC images are used as an addition to arterial and venous phase acquisitions. Further, VNC images have shown potential to improve the rate of detection of smaller hypodense liver lesions compared to TNC images (24).

In SDCT, dual-energy acquisitions are always enabled due to the detector-based separation of photons, thereby VNC images are always created and can easily be used whenever needed, e.g. for ROI-based assessment of liver fat content (17). This should benefit the workflow in clinical practice as an additional assessment of liver fat can be achieved relatively quickly without almost no time loss.

Image noise as measured from TNC and both sets of VNC was not statistically different, yet visual, subjective assessment of the image noise indicated it was perceived to be lower in the VNC images created from both the arterial and venous phases compared to TNC. These findings are consistent with earlier studies that concluded noise in VNC images was the same or less than noise in TNC (1,7). In the subjective reading, image distortion or artificial image impression was noticed in VNC, mainly in VNC from the venous phase. This, however, negatively affected image quality in only a very low percentage of cases.

This study has several limitations. We applied a ROI-based evaluation of attenuation and image noise to objectively compare VNC contrast performance as applied in prior studies (1,3,7,29); however, a volumetric assessment of attenuation in the analyzed structures might be less susceptible to any measuring bias. Still, we applied an advanced and detailed ROI placement to account for any mismeasurement. In addition, our objective analysis did not cover any intrahepatic arteries as these were too small for precise ROI placement and to maintain adequate ROI sizes. To account for limitations of the objective analysis, we included a detailed subjective reading by two experienced and independent radiologists evaluating iodine removal and diagnostic utility. Good inter-rater agreement and clear conclusions validate their results. Finally, our assessment did not focus on a certain pathology due to limited data available.

In conclusion, SDCT derived virtual non-contrast images created from both the arterial and venous phases accurately removed iodine in liver parenchyma and larger vessels. Objective and subjective assessment showed comparable image properties to TNC. Inadequate iodine subtraction occurred in hepatic arteries in the arterial phase.

Footnotes

Acknowledgement

The authors thank Sandra Halliburton for reviewing the manuscript and her technical advice.

Declaration of conflicting interests

The authors declared the following conflicts of interst with respect to the research, authorship, and/or publication of this article: NGH and JB: On the speaker’s bureau of Philips Healthcare; SL: Received travel expense reimbursement from Philips Healthcare and is exempt from clinical duties as part of a research agreement between University Hospital Cologne and Philips Healthcare. The other authors state that they have nothing to disclose.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by Philips Healthcare under a research agreement with University Hospitals Cleveland Medical Center and Case Western Reserve University.

ORCID iDs

Kai Roman Laukamp

Verena Carola Obmann

References

Sauter

Muenzel

Dangelmaier

, et al. Dual-layer spectral computed tomography: Virtual non-contrast in comparison to true non-contrast images. Eur J Radiol 2018; 104:108–114.

Haneder S, Siedek F, Doerner J, et al. Thoracic-abdominal imaging with a novel dual-layer spectral detector CT: intra-individual comparison of image quality and radiation dose with 128-row single-energy acquisition. Acta Radiol 2018;59:1458–1465.

Connolly

McInnes

MDF

El-Khodary

, et al. Diagnostic accuracy of virtual non-contrast enhanced dual-energy CT for diagnosis of adrenal adenoma: A systematic review and meta-analysis. Eur Radiol 2017; 27:4324–4335.

Willatt

Ruma

Azar

, et al. Imaging of hepatocellular carcinoma and image guided therapies - how we do it. Cancer Imaging 2017; 17:9.

Stolzmann

Frauenfelder

Pfammatter

, et al. Endoleaks after endovascular abdominal aortic aneurysm repair: detection with dual-energy dual-source CT. Radiology 2008; 249:682–691.

Holalkere

N-S

, et al. Imaging-based quantification of hepatic fat: methods and clinical applications. RadioGraphics 2009; 29:1253–1277.

Haji-Momenian

Parkinson

Khati

, et al. Single-energy non-contrast hepatic steatosis criteria applied to virtual non-contrast images: is it still highly specific and positively predictive? Clin Radiol 2018; 73:594.e7–594.e15.

Veteläinen

van Vliet

Gouma

, et al. Steatosis as a risk factor in liver surgery. Ann Surg 2007; 245:20–30.

Chu

MJJ

Dare

Phillips

ARJ

, et al. Donor hepatic steatosis and outcome after liver transplantation: a systematic review. J Gastrointest Surg 2015; 19:1713–1724.

10.

Pickhardt

Hahn

del Rio

, et al. Natural history of hepatic steatosis: observed outcomes for subsequent liver and cardiovascular complications. Am J Roentgenol 2014; 202:752–758.

11.

Browning

Szczepaniak

Dobbins

, et al. Prevalence of hepatic steatosis in an urban population in the United States: Impact of ethnicity. Hepatology 2004; 40:1387–1395.

12.

Kodama

, et al. Comparison of CT methods for determining the fat content of the liver. AJR Am J Roentgenol 2007; 188:1307–1312.

13.

Limanond

Raman

Lassman

, et al. Macrovesicular hepatic steatosis in living related liver donors: correlation between CT and histologic findings. Radiology 2004; 230:276–280.

14.

Bohte

van Werven

Bipat

, et al. The diagnostic accuracy of US, CT, MRI and 1H-MRS for the evaluation of hepatic steatosis compared with liver biopsy: a meta-analysis. Eur Radiol 2011; 21:87–97.

15.

Ananthakrishnan

Duan

Rajiah

, et al. Phantom validation of spectral detector computed tomography–derived virtual monoenergetic, virtual noncontrast, and iodine quantification images. J Comput Assist Tomogr 2018; 42:959–964.

16.

Lundin

Lidén

Magnuson

, et al. Virtual non-contrast dual-energy CT compared to single-energy CT of the urinary tract: a prospective study. Acta Radiol 2012; 53:689–694.

17.

McCollough

Leng

, et al. Dual- and multi-energy CT: principles, technical approaches, and clinical applications. Radiology 2015; 276:637–653.

18.

Laukamp KR, Zopfs D, Lennartz S, et al. Metal artifacts in patients with large dental implants and bridges: combination of metal artifact reduction algorithms and virtual monoenergetic images provides an approach to handle even strongest artifacts. Eur Radiol 2019;29:4228–4238.

19.

Park

Kim

, et al. Macrovesicular hepatic steatosis in living liver donors: use of CT for quantitative and qualitative assessment. Radiology 2006; 239:105–112.

20.

Bamberg

Dierks

Nikolaou

, et al. Metal artifact reduction by dual energy computed tomography using monoenergetic extrapolation. Eur Radiol 2011; 21:1424–1429.

21.

Ananthakrishnan

Rajiah

Ahn

, et al. Spectral detector CT-derived virtual non-contrast images: comparison of attenuation values with unenhanced CT. Abdom Radiol 2017; 42:702–709.

22.

Talakić

Schaffellner

Kniepeiss

, et al. CT perfusion imaging of the liver and the spleen in patients with cirrhosis: Is there a correlation between perfusion and portal venous hypertension? Eur Radiol 2017; 27:4173–4180.

23.

Lehti

Söderberg

Höglund

, et al. Reliability of virtual non-contrast computed tomography angiography: comparing it with the real deal. Acta Radiol Open 2018; 7:2058460118790115.

24.

De Cecco

Muscogiuri

Schoepf

, et al. Virtual unenhanced imaging of the liver with third-generation dual-source dual-energy CT and advanced modeled iterative reconstruction. Eur J Radiol 2016; 85:1257–1264.

25.

Toepker

Moritz

Krauss

, et al. Virtual non-contrast in second-generation, dual-energy computed tomography: reliability of attenuation values. Eur J Radiol 2012; 81:e398–405.

26.

Durieux

Gevenois

Muylem

A Van

, et al. Abdominal attenuation values on virtual and true unenhanced images obtained with third-generation dual-source dual-energy CT. AJR Am J Roentgenol 2018; 210:1042–1058.

27.

Imajo

Kessoku

Honda

, et al. Magnetic resonance imaging more accurately classifies steatosis and fibrosis in patients with nonalcoholic fatty liver disease than transient elastography. Gastroenterology 2016; 150:626–637.e7.

28.

Vandenbroucke

Van Hedent

Van Gompel

, et al. Dual-energy CT after radiofrequency ablation of liver, kidney, and lung lesions: a review of features. Insights Imaging 2015; 6:363–379.

29.

Laukamp KR, Lennartz S, Neuhaus V-F, et al. CT metal artifacts in patients with total hip replacements: for artifact reduction monoenergetic reconstructions and post-processing algorithms are both efficient but not similar. Eur Radiol 2018;28:4524–4533.