Reducing Visceral-Motion-Related Artifacts on the Liver with Dual-Energy CT: A Comparison of Four Different CT Scanner Techniques

Purpose: To assess the influence of different dual-energy CT (DECT) scanner techniques on the severity of visceral-motion-related artifacts on the liver. Methods: Two independent readers retrospectively evaluated visceral-motion-related artifacts on the liver on 120-kVp(-like), monoenergetic low- and high-keV, virtual non-contrast (VNC), and iodine images acquired on a dual-source, twin-beam, fast kV-switching, and dual-layer spectral detector scanner. Quantitative assessment: Depth of artifact extension into the liver, measurements of Hounsfield Units (HU) and iodine concentrations. Qualitative assessment: Five-point Likert scale (1 = none to 5 = severe). Artifact severity between image reconstructions were compared by Wilcoxon signed-rank and paired t-tests. Results: 615 contrast-enhanced routine clinical DECT scans of the abdomen were evaluated in 458 consecutive patients (mean age: 61 ± 14 years, 331 men). For dual-source and twin-beam scanners, depth of extension of artifacts into the liver was significantly shorter and artifact severity scores significantly lower for 120-kVp-like images compared with the other image reconstructions (p < 0.001, each). For fast kV-switching and spectral detector scanner images, depth of extension of artifacts was significantly shorter and artifact severity scores significantly lower for iodine images (p < 0.001, each). Dual-source 120-kVp-like and spectral detector iodine images reduced artifacts to an extent that no significant difference in HU or iodine concentrations between artifacts (dual-source: 97 HU, spectral detector: 1.9 mg/mL) and unaffected liver parenchyma (dual-source: 108 HU, spectral detector: 2.1 mg/mL) was measurable (dual-source: p = 0.32, spectral detector: p = 0.15). Conclusion: Visceral-motion-related artifacts on the liver can be markedly reduced by viewing 120-kVp-like images for dual-source and twin-beam DECT scanners and iodine images for fast kV-switching and dual-layer spectral detector DECT scanners.


Introduction
Computed tomography (CT) is frequently used for oncologic staging and follow-up. Optimization of CT image quality is essential for accurate assessment of the liver, as it is commonly affected by metastases and primary tumors that may be small or subtle. Image artifacts at CT degrade image quality and may compromise diagnostic accuracy [1,2].
Different types of CT image artifacts are known to relate to CT data acquisition, image reconstruction, metallic implants, and patient motion [1][2][3][4][5]. In particular, motion-related artifacts caused by involuntary movement such as intestinal peristalsis or cardiac motion

Materials and Methods
In total, 131 of the 458 patients were included in a previous study [9]. This prior article dealt with visceral-motion-related artifact reduction with a dual-layer spectral detector DECT scanner only, whereas in this manuscript, we report on visceral-motion-related artifact reduction with different DECT scanner models, as DECT methodologies differ significantly between models.
This study is compliant with the Health Insurance Portability and Accountability Act. Need for informed consent was waived by the institutional review board. The National Institutes of Health and Philips Healthcare provided funding for our study as part of a research grant. The funders had no role in the study design, data collection, data analyses, data interpretation, manuscript writing, or in publishing the results.

Study Population
In this retrospective analysis, we evaluated all contrast-enhanced CT scans (arterial phase, venous phase, delayed phase) of the abdomen acquired as part of clinical routine on a dual-source CT scanner during the time period from 29 April 2021 to 28 May 2021, on a twin-beam CT scanner from 8 May 2021 to 28 May 2021, on a fast kV-switching CT scanner from 1 February 2018 to 5 March 2018, and on a dual-layer spectral detector CT scanner from 13 September 2017 to 1 April 2018. Patients <18 years-of-age were excluded. No antispasmodic medication was administered. Following the clinical routine standard operating procedure of our department, fasting prior to CT image acquisition was not required. CT scans with artifacts on the liver originating from oral contrast material or Diagnostics 2022, 12, 2155 3 of 13 metallic foreign materials were excluded. CT scans with missing source dual-energy data were excluded from artifact evaluation.

Image Analysis
Images were viewed on standard picture archiving and communication system (PACS) workstations. The commercially available dedicated postprocessing software Syngo. Via version VB30A (Siemens Healthineers, Forchheim, Germany) was used for the evaluation of dual-source and twin-beam images, GSI Volume Viewer version 13.0 (GE Healthcare, Chicago, IL, USA) for fast kV-switching images, and IntelliSpace Portal version 11.1 (Philips Healthcare, Cleveland, OH, USA) for dual-layer spectral detector images. Two independent radiologists (Reader A: S.G., 4 years of experience in abdominal radiology) (Reader B: K.V., 6 years of experience in abdominal radiology) analyzed axial images of the different DECT reconstructions. DECT images were reviewed on default window level settings for the presence or absence of visceral-motion-related artifacts on the liver. The organ where the visceral-motion-related artifacts originated from was recorded. An artifact was considered to be present if it was scored as visible by one or both readers. To assess inter-reader and intra-reader variability, all CT scans with visceral-motion-related artifacts were evaluated independently by Reader A and B, then 26 CT scans with visceral-motion-related artifacts were evaluated again four weeks after the first reading by Reader A.
Circular regions of interest (ROI) were manually placed in the brightest area of the artifact, darkest area of the artifact, and in neighboring unaffected liver parenchyma. ROImax was defined as bright, ROImin as dark, on low-keV images. Areas of inhomogeneity due to partial volume effect, vessels, or tissue borders were avoided. ROIs were propagated onto the axial 120-kVp(-like), low-keV, high-keV, VNC, and iodine images with identical ROI sizes in identical anatomical ROI locations. ROI measurements are presented in HU (120-kVp(-like), low-keV, high-keV, and VNC images) or iodine concentrations in mg/mL (iodine images). Comparisons of ROI measurements were performed between artifacts and neighboring liver parenchyma not affected by artifacts, which served as reference tissue.

Qualitative Visceral-Motion-Related Artifact Evaluation
Visceral-motion-related artifact severity on the liver in 120-kVp(-like), low-and high-keV, VNC, and iodine images was qualitatively evaluated on a five-point Likert scale [9]: 1 = Absence of visceral-motion-related artifact on the liver; 2 = Visible visceral-motionrelated artifact with no effect on diagnosis on the liver; 3 = Moderate visceral-motion-related artifact that may decrease confidence in diagnosing a 0.5 to 0.9 cm liver lesion; 4 = Distinct visceral-motion-related artifact that prevents the diagnosis of a 0.5 to 0.9 cm liver lesion, and that may decrease confidence in diagnosing a ≥ 1.0 cm liver lesion; 5 = Severe visceralmotion-related artifact that prevents the diagnosis of a ≥ 1.0 cm liver lesion [9].

Statistical Analysis
Continuous variables were expressed as mean ± standard deviation (SD), categorical variables as frequencies and percentages. The Shapiro-Wilk test was used for testing for normality.
The Wilcoxon signed-rank test was used to compare the depth of visceral-motionrelated artifact extension into the liver and qualitative artifact scores between 120-kVp(-like), low-keV, high-keV, VNC, and iodine images of each scanner, respectively. The paired ttest was performed to evaluate the relationship between ROI measurements (HU, iodine concentration) of visceral-motion-related artifacts and the neighboring liver parenchyma not affected by artifacts in 120-kVp(-like), low-keV, high-keV, VNC, and iodine images of each scanner, respectively.
In addition, p-values < 0.05 were considered to denote statistical significance. Statistical analysis was performed with the open-source software RStudio Version 1.4.1103 (RStudio Team (2020), RStudio: Integrated Development for R. RStudio, PBC, Boston, MA, USA).

Study Population
In total, we evaluated 615 contrast-enhanced CT scans of the abdomen in 458 patients (mean age: 61 ± 14 years, 331 men) (dual-source scanner: 127 scans in 90 consecutive patients, twin-beam scanner: 142 scans in 140 consecutive patients, fast kV-switching scanner: 126 scans in 97 consecutive patients, dual-layer spectral detector scanner: 220 scans in 131 consecutive patients). A total of 36/615 (6%) CT scans were excluded due to artifacts on the liver originating from metallic foreign materials or oral contrast material.

Quantitative Visceral-Motion-Related Artifact Evaluation
The inter-reader agreement in the quantitative assessment of depth of artifact extension into the liver was moderate (ICC = 0.56, p < 0.001), and of Hounsfield Unit and iodine concentration differences of visceral-motion-related artifacts from normal liver parenchyma was moderate (ICC = 0.64, p < 0.001). The intra-reader agreement in the assessment of depth of artifact extension was excellent (ICC = 0.97, p < 0.001), and of Hounsfield Unit and iodine concentration differences was excellent (ICC = 0.98, p < 0.001).

Discussion
We found that, in routine clinical DECT scans of the abdomen, visceral-motion-related artifacts on the liver could be minimized by viewing specific DECT image reconstructions, depending on the type of DECT scanner. For dual-source and twin-beam DECT scanners, visceral-motion-related artifacts were significantly less severe, both quantitatively and qualitatively, when the liver was viewed using 120-kVp-like images, but not iodine images. Conversely, for fast kV-switching and dual-layer spectral detector DECT Figure 4. Qualitative artifact scores (1 = none to 5 = severe) of visceral-motion-related artifacts on the liver in axial 120-kVp(-like), low-and high-keV, VNC, and iodine images acquired on a dual-source, twin-beam, fast kV-switching, and dual-layer spectral detector CT scanner.

Discussion
We found that, in routine clinical DECT scans of the abdomen, visceral-motion-related artifacts on the liver could be minimized by viewing specific DECT image reconstructions, depending on the type of DECT scanner. For dual-source and twin-beam DECT scanners, visceral-motion-related artifacts were significantly less severe, both quantitatively and qualitatively, when the liver was viewed using 120-kVp-like images, but not iodine images. Conversely, for fast kV-switching and dual-layer spectral detector DECT scanners, visceralmotion-related artifact on the liver were significantly less severe in iodine images, but not in 120-kVp(-like) images. Dual-source 120-kVp-like images and dual-layer spectral detector iodine images reduced visceral-motion-related artifacts to an extent that no significant difference in measurements of HU or iodine concentrations between artifacts and unaffected liver parenchyma was detectable.
These strikingly different findings for different DECT implementations likely result from different gantry geometry in different DECT scanner models. Dual-source and twinbeam models acquire the high-and low-kVp datasets at dissimilar gantry positions [19].
Consequently, visceral-motion-related artifacts appear in different positions on the highversus low-kVp datasets [29]. As the 120-kVp-like images are a blended mix of the high-and low-kVp datasets, bright and dark visceral-motion-related artifact components cancel each other out (see Figure 5). Conversely, for fast kV-switching scanners and dual-layer spectral detector scanners, there is close or perfect geometric alignment and temporal resolution of high-and low-kVp datasets [19,29]. Therefore, visceral-motion-related artifacts on these scanners have a very similar configuration of bright and dark components on the high-and low-kVp datasets. Blending high-and low-kVp datasets to reduce artifacts is not efficient in these scanner models. However, because visceral-motion-related artifacts affect the highand low-kVp datasets to a similar degree, and iodine images depict voxels with a relatively large decrease in HU between low-kVp and high-kVp images, visceral-motion-related artifacts are substantially reduced in iodine images [9,13,29]. scanners, visceral-motion-related artifact on the liver were significantly less severe in iodine images, but not in 120-kVp(-like) images. Dual-source 120-kVp-like images and duallayer spectral detector iodine images reduced visceral-motion-related artifacts to an extent that no significant difference in measurements of HU or iodine concentrations between artifacts and unaffected liver parenchyma was detectable. These strikingly different findings for different DECT implementations likely result from different gantry geometry in different DECT scanner models. Dual-source and twinbeam models acquire the high-and low-kVp datasets at dissimilar gantry positions [19]. Consequently, visceral-motion-related artifacts appear in different positions on the highversus low-kVp datasets [29]. As the 120-kVp-like images are a blended mix of the highand low-kVp datasets, bright and dark visceral-motion-related artifact components cancel each other out (see Figure 5). Conversely, for fast kV-switching scanners and dual-layer spectral detector scanners, there is close or perfect geometric alignment and temporal resolution of high-and low-kVp datasets [19,29]. Therefore, visceral-motion-related artifacts on these scanners have a very similar configuration of bright and dark components on the high-and low-kVp datasets. Blending high-and low-kVp datasets to reduce artifacts is not efficient in these scanner models. However, because visceral-motion-related artifacts affect the high-and low-kVp datasets to a similar degree, and iodine images depict voxels with a relatively large decrease in HU between low-kVp and high-kVp images, visceralmotion-related artifacts are substantially reduced in iodine images [9,13,29]. Figure 5. Visceral-motion-related artifact (arrows) appearance on images with (A,D) low X-ray photon energy and (B,E) high X-ray photon energy acquired with a dual-source (top row) and duallayer spectral detector (bottom row) DECT scanner. As hyperdense artifact components on (A) lowenergy dual-source images appear hypodense on (B) high-energy dual-source images, and vice versa, visceral-motion-related artifacts are substantially reduced on (C) blended 120-kVp-like dualsource images. As hyperdense artifact components on (D) low-energy dual-layer spectral detector images remain hyperdense on (E) high-energy dual-layer spectral detector images, and vice versa, Figure 5. Visceral-motion-related artifact (arrows) appearance on images with (A,D) low X-ray photon energy and (B,E) high X-ray photon energy acquired with a dual-source (top row) and dual-layer spectral detector (bottom row) DECT scanner. As hyperdense artifact components on (A) low-energy dual-source images appear hypodense on (B) high-energy dual-source images, and vice versa, visceral-motion-related artifacts are substantially reduced on (C) blended 120-kVp-like dual-source images. As hyperdense artifact components on (D) low-energy dual-layer spectral detector images remain hyperdense on (E) high-energy dual-layer spectral detector images, and vice versa, visceral-motion-related artifacts are substantially reduced on (F) iodine dual-layer spectral detector images.
Our results are in line with previous studies evaluating visceral-motion-related artifact reduction in DECT image reconstructions. Winklhofer et al. showed in DECT scans of a bowel peristalsis phantom model and 100 patients acquired on a fast kV-switching scanner that general visceral-motion-related artifacts were substantially reduced in iodine images compared to 70-keV, 120-keV, and VNC images (p < 0.001, each) [7]. However, this study did not compare different DECT scanner models and did not further investigate the effects of visceral-motion-related artifacts on the liver, other than frequency [7]. In a phantom study Obmann et al. assessed the influence of different DECT scanner models on the severity of visceral-motion-related artifacts in 120-kVp-like and iodine images using a qualitative three-point score [29]. It was shown that visceral-motion-related artifacts were perceived as least severe in iodine images for fast kV-switching and dual-layer spectral detector scanners, and in 120-kVp-like images for dual-source and twin-beam scanners [29]. Our study adds to the field by evaluating the influence of different DECT scanner models on the severity of visceral-motion-related artifacts on the liver in actual clinical DECT scans of 458 consecutive patients. In addition to qualitative five-point artifact scores based on oncologic liver evaluation, we also obtained quantitative measurements of visceral-motionrelated artifacts on the liver. Besides 120-kVp and iodine images, monoenergetic low-and high-keV images as well as VNC images were also assessed for the four commonly used DECT models.
Our study has limitations. As the appearances of the monoenergetic, VNC, and iodine images are very characteristic, effective blinding of the readers to the different image reconstructions was not possible. The four DECT techniques were not employed among the same patients, since routine clinical DECT scans were evaluated. In order to obtain the maximum effect of low-keV monoenergetic reconstructions on visceral-motion-related artifacts, monoenergetic 40-keV images were evaluated instead of more commonly used 50-keV images. Diagnostic accuracy of liver lesion detection was not assessed, as our study population did not contain a sufficient number of liver lesions masked by visceral-motionrelated artifacts.
In conclusion, visceral-motion-related artifacts on the liver can be substantially reduced by viewing 120-kVp-like images for dual-source and twin-beam DECT scanners, and iodine images for fast kV-switching and dual-layer spectral detector DECT scanners.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/diagnostics12092155/s1. Materials and Methods S: CT image acquisition. Results S: Study population. Figure S1: Depth of extension of visceral-motion-related artifacts into the liver parenchyma. Measurements of Reader A only. Figure S2: Depth of extension of visceral-motion-related artifacts into the liver parenchyma. Measurements of Reader B only. Figure S3: Qualitative artifact scores of Reader A only. Figure S4: Qualitative artifact scores of Reader B only. Table S1: Quantitative measurements of visceral-motion-related artifacts on the liver of Reader A only. Table S2: Quantitative measurements of visceral-motion-related artifacts on the liver of Reader B only. Funding: This study has received funding by a grant from the National Institutes of Health (Award Number: R01CA226868) and Philips Healthcare, Cleveland, OH, USA. The content is solely the responsibility of the authors, and does not necessarily represent the official views of the funders. The funders had no role in the study design, data collection, data analyses, data interpretation, manuscript writing, or in the decision to publish these results.

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of the University of California, San Francisco (IRB #: 12-10492 Reference #: 286873, Approval Date: 7 August 2020).
Informed Consent Statement: Patient consent was waived by the Institutional Review Board because the research is no more than minimal risk to subjects, could not practicably be conducted without the waiver, could not practicably be conducted without identifiable information, will not adversely affect rights and welfare of subjects with the waiver, and will provide subjects with additional pertinent information after participation, whenever it is appropriate. The waiver of informed consent applies to all subjects. Data Availability Statement: Data are available upon reasonable request. Requests should be sent to the corresponding author and are subject to approval.