Quality of virtual-non-contrast phases derived from arterial and delayed phases of fast-kVp switching dual-energy CT in patients after endovascular aortic repair

Abdominal aortic aneurysms (AAAs) are a common health problem that affects up to approximately 7.6% of the male population [1]. Reduction of postoperative mortality has led to the adoption of endovascular aneurysm repair (EVAR) as the preferred method of treating AAAs [2]; however, EVAR patients need lifelong surveillance to diagnose possible life threating complications, such as stent-graft migration and endoleak [2, 3].

Computed tomography angiography (CTA) and color Doppler ultrasound are currently leading methods in the surveillance of patients with acute and chronic aortic disease, as well as those after EVAR procedure. One of the most promising methods of CTA data acquisition is dual-energy CTA (DECTA). DECTA allows acquisition of two simultaneous or nearly simultaneous datasets (depending on the scanning technique) acquired in different energies. Differences between attenuation levels in the acquired datasets enable differentiating materials and elements (e.g. iodine, calcium) in the scanned area and creating maps of their occurrence [4]. In the subsequent step, it enables subtraction of iodine and reconstruction of the so-called virtual non-contrast (VNC) phase. This may allow omission of the true non-contrast (TNC) acquisition and a substantial reduction in the radiation dose [5‐12]. Studies cited above confirmed the possibility of obtaining images similar in quality to TNC and the validity of using VNC in patients after EVAR. The use of iodine subtraction algorithms was, however, associated with differences between TNC and VNC images. In these studies, VNC was obtained only from the delayed phase of the study. It is, therefore, reasonable to compare the images of TNC and VNC obtained from the arterial phase (VNCa) and the delayed (VNCd) study. To our knowledge, only one study analyzing differences between TNC, VNCa, and VNCd in patients after EVAR has been conducted, but it was performed using a specific data acquisition method (dual-source scanner, Siemens Healthineers) [13].

The aim of the study was to compare TNC, VNCa, and VNCd obtained with the fast-kVp-Switching technique (GE Healthcare) by assessing the effect of image noise on the subjective image quality parameters and by performing measurements of CT numbers with the assessment of the signal-to-noise ratio (SNR). Moreover, the possible reduction of the radiation dose with the use of VNC was calculated.

The VNC images obtained in this study were characterized by a higher SNR compared to the TNC images. There were significant differences in the densities of vascular structures and endoleaks in the TNC, VNCa, and VNCd reconstructions, with significantly lower mean densities in VNCd. The above correlations are concordant with part of the literature [13, 15].The qualitative and quantitative analyses performed showed that VNCd images may be a better approximation of TNC images than VNCa images in the assessment of endoleaks, and their use enables a reduction of the radiation dose.

The presence of hyperdense calcifications within the aneurysm sac, subtracted in VNC, may lead to a false positive recognition of endoleaks. This phenomenon has already been described in the literature, and the degree of subtraction was noted as absent/mild [5, 6, 8] or was not assessed [9‐11]. In our analysis, the subjective degree of subtraction of calcifications was determined in most cases as minimal or intermediate, and in 4% of cases, the level of calcification subtraction was defined as significant. Statistical analyses did not show a significant correlation between image noise and the degree of subtraction of calcifications. Moreover, image noise did not significantly affect the subjective assessment of the quality of the images, as well as the level of acceptance of the images in the assessment of the presence of endoleak.

The analysis of the ROI measurements showed significant differences in the CT numbers of the structures undergoing intense contrast enhancement, depending on the phase of the study, from which VNC images were reconstructed. The VNCa images showed a significantly higher densities of ROI located within the aorta, iliac artery, and endoleak compared to the TNC and VNCd images. The phenomenon of statistically significant differences in the degree of radiation absorption between TNC and VNC within arteries has already been described in the literature [13, 15, 16]. There are also publications in which no statistically significant differences in the attenuation of abdominal aorta and other abdominal vessels was found between TNC and VNC images [6, 8]. The issue of differences in tissue imaging between TNC and VNC images acquired in DECT systems has been examined by several researchers, including Sauter et al. [17]. In their study, they found no significant differences between tissues, except for spongious bone, when comparing TNC and VNC reconstructions [17]. However, they did find differences above 15 HU in 19% of aortic measurements. Similar results were found in previously mentioned studies [13, 15, 16]. The authors of the study concluded that iodine subtraction algorithms depend on DECT data acquisition, and that dual layer DECT systems are more capable of precise iodine subtraction than other dual-energy systems.

The previously mentioned study by Lehti et al. [13] showed a tendency similar to the one obtained in our research—the occurrence of a higher degree of radiation absorption within arteries in VNCa images, with a similar degree of radiation absorption in muscles and adipose tissue. The phantom study performed by Toepker et al. [15] also showed differences in vessel densities in VNCa and VNCd, with higher mean values of radiation absorption in VNCa than in VNCd and TNC. Therefore, it can be assumed that the algorithms processing VNC images, in some cases, are unable to properly quantify iodine present in high concentrations in the examined structures. The phenomenon of a higher mean density of areas identified later as an endoleak in VNCa than in TNC may potentially be important in the process of identifying endoleaks, thus making their diagnosis difficult and reducing the diagnostic value of the examination if biphasic examination is used with VNCa reconstruction. On the other hand, since endoleaks in VNCd have lower densities than in TNC, this could potentially result in a better enhancement of the presence of endoleaks during the comparison of VNCd images and the delayed phase of the study. Further research should focus on the influence of the type of the dual-energy acquisition method and software version used on this phenomenon. A separate issue is the possibility of using VNC reconstructions in a variety of clinical applications, in which it is essential to assess differences in structure densities in the subsequent stages of the study (e.g. assessment of adrenal adenomas, liver and kidney lesions).

The DECT system used in our study had a major limitation, which was the need for prospective determination of the examination protocol, along with the risk of using a higher radiation dose. Dual-source and sequential DECT systems default to single-energy mode, which precludes acquiring spectral data and thus VNC reconstructions [18]. Fortunately, split-filter, multi-layer detectors DECT systems, as well as photon-counting detectors (PCD), are free from such limitations, allowing for retrospectively acquiring DE datasets and spectral reconstructions. In our opinion, such systems allow for broader and more frequent use of reconstructions such as VMI and material decomposition algorithms (VNC, virtual non-calcium). The use of these types of reconstructions significantly increases the diagnostic value of angiographic examinations and brings significant benefits to the patient [18‐20]. Moreover, frequent use of these tools allows for their expert use and application in particularly useful clinical applications.

The recently PCD CT systems, after the introduction of multi-row and DECT systems, are a significant advancement in medical CT scanners. PCD CT systems are able to preserve and quantify the energy of each photon and generate spectral reconstructions of wide range [21]. The 2022 publication by Decker et al. demonstrated the high diagnostic value of VNC phase reconstructed with PCD (VNCpc) in post-EVAR CT assessment [22]. VNCpc images derived from CT angiographies of the aorta showed potential as a substitute for TNC images in follow-up scans after EVAR. Expert readers rated 95% of VNCpc images as suitable for replacing TNC-series. VNCpc images demonstrated high image quality with complete aortic contrast removal and minimal erroneous subtraction of stent parts or calcifications. They also showed lower image noise, higher SNR, and smaller CT-value differences to TNC-series than conventional DECT-acquired VNC series. These findings are accompanied by results of another study performed in 2022 by Mergen et al. [23]. The authors found no significant difference in errors between arterial or portal venous VNC reconstructions (3.3 HU vs 3.5 HU, P = 0.16). Subjective image quality was rated lower in VNC images, but diagnostic quality was reached in 99–100% of patients. The study concluded that abdominal virtual noncontrast images from the arterial and portal venous phase of PCD CT yielded accurate CT attenuation and good image quality compared with true noncontrast images, although combined with higher noise values. The above-mentioned studies are supported by a few other, showing comparable results [24, 25]. However, there were also papers published, showing significant differences in CT numbers in TNC and VNCpc images [26]. Similar to DECT systems, these discrepancies in research results suggest the need for further work on VNC reconstruction algorithms, which likely reveal vendor- and software-specific differences. However, these promising results, combined with the ability of PCD systems to generate spectral and VNC reconstructions of each scan, allow us to believe that the widespread use and development of DECT and PCD systems will enable routine application of VNC reconstructions on a global scale, leading to a significant reduction in radiation doses received by patients during CT scans.

VNCa and VNCd images are characterized by a higher SNR compared to TNC images. There are significant differences in the densities of vascular structures and endoleaks in TNC, VNCa and VNCd reconstructions, with significantly lower mean densities in VNCd. No statistically significant correlation was found between image noise and the subjective image quality parameters and the degree of subtraction of calcifications. In conclusion, VNCd images are more suitable than VNCa images and sufficient as a replacement of TNC in the diagnosis of endoleaks with a possible 23.28% reduction of radiation dose.