Image quality evaluation of dual-layer spectral CT in comparison to single-layer CT in a reduced-dose setting

The introduction of dual-layer detector technology in computed tomography (CT) enabled the acquisition of spectral data for all performed scans without the need of an additional CT x-ray tube or additional acquisitions. Dual-layer spectral CT (DLCT) acquisitions allow material decomposition (virtual non-contrast, iodine-only imaging, and effective atomic numbers) as well as the calculation of virtual monoenergetic images. Several clinical studies have already been performed showing the advantages of DLCT for head CT for imaging intracerebral lesions and hemorrhage, for thoracic CT, for vertebral CT for differentiating bone lesions, and for abdominal CT angiographies for improved delineation of visceral arteries [1‐5]. However, for the image acquisition of such data, a tube potential of either 140 kVp or 120 kVp is necessary to allow for spectral decomposition under the exploitation of the energy-specific x-ray absorption of different materials. In contrast to changes in tube current, changes in tube potential have a non-linear effect on radiation dose: in comparison to 80 kVp, the x-ray tube output (i.e., air kerma or exposure) is 1.5 times higher for 100 kVp, 2.5 times higher for 120 kVp, and 3.4 times higher for 140 kVp [6].

While patient radiation exposure has long been a topic of interest in CT, as CT accounts for 49–66% of overall patient radiation exposure, this interest has recently led to new regulations in the European Union via the EURATOM directive, with the deadline for implementation into federal law in 2018 [7, 8]. Based on surveys, diagnostic reference levels (DRLs) at the 75th percentile of dose distribution have been defined as the limits of appropriate practice for CT acquisitions. In our institution, acquisition protocols for the thorax, abdomen, and pelvis have been previously optimized to reduce patient radiation exposure beyond these legal requirements. This was achieved, in part, by reducing tube potential from 120 to 100 kVp for single-layer CT (SLCT), resulting in an average patient dose 60% below the federal DRL for the thorax, abdomen, and pelvis, which is a DLP of 1000 mGy·cm and a CTDI_vol of 13 mGy.

With the installation of a novel DLCT, vendor’s recommended settings were applied including a tube potential of 120 kVp (acquisition protocol DLCT_high), which led to an increase in radiation exposure in comparison to previous reduced-dose SLCT protocols (acquisition protocol SLCT_low), albeit still remaining below the applicable diagnostic reference levels. After initial experience with DLCT_high, the tube current was reduced (acquisition protocol DLCT_low) to match the CTDI_vol of the reduced-dose protocol previously performed on the SLCT scanner at 100 kVp (SLCT_low).

The aim of the study was to quantitatively and qualitatively compare the image quality and radiation exposure of the SLCT acquisition protocol with reduced dose, SLCT_low (41% of the national DRL), and the DLCT acquisition protocol with reduced dose, DLCT_low (44% of the DRL). Thus, by comparing DLCT_low and SLCT_low, we want to aid in the deployment of protocols with reduced dose in DLCT with image quality comparable to previous low-tube potential SLCT_low, while allowing to benefit from the spectral information in a clinical setting.

The necessity to increase the tube potential to at least 120 kVp for spectral imaging with current DLCT results in an increase of patient radiation exposure, compared with conventional single-layer acquisitions performed at 100 kVp. As a compensation tube current-time product might be adapted to achieve equivalent CTDI_vol to standard-, reduced-, or low-dose SLCT protocols. The results in this study show that patient radiation exposure can be restored to levels close to that of reduced-dose SLCT with comparable image quality by adjusting automated exposure control (i.e., DRI). Image quality was comparable between DLCT_low and SLCT_low despite the changes in tube potential and the resulting changes in tube current-time product. Thus, a notable dose reduction due to adjusted automatic exposure control can be achieved in DLCT protocols compared to standard protocols, but further reductions might be possible by fine-tuning the lower threshold for tube current-time product.

Previously, Van Ommen et al [15] had already shown that DLCT can be used for a broad range of applications without increasing patient radiation exposure compared to normal-dose SLCT. However, neither quantitative nor qualitative image quality was evaluated in this study. Another aspect of our study was the adaptation of DLCT to match previous low-dose SLCT_low with DLCT_low, which shows lower radiation doses than the current literature [15, 16]. Moreover, the results show that even further dose reduction could be possible, especially in the thorax region where the automated exposure control reached the lower threshold of tube current-time product set in the acquisition protocol. According to the presented results, further dose reduction might be achieved by reducing the lower limit of the tube current from 65 to 55 mAs. According to Nagayama et al [17], DLCT showed a lower radiation dose than SLCT at equal tube potential of 120 kVp. However, with regard to image quality of conventional DLCT images, which is currently still the most frequently used reconstruction in clinical settings, no comparison was made with SLCT. Conventional CT images of dose equivalent DLCT and SLCT were already compared in a phantom study showing higher CNR but higher noise for DLCT compared to SLCT for 80–140 kVp, which was partly in agreement with the results presented here [18]. It should also be noted that despite matched CTDI_vol between SLCT_low and DLCT_low, as configured in the acquisition protocol (Table 1), the resulting CTDI_vol for DLCT_low was higher than that for the SLCT_low (Table 2), which could be connected to the lower limit for the tube current mentioned above.

In general, the aspect of patient radiation exposure in spectral imaging has already been discussed in previous studies. No radiation dose increase is necessary for dual-source, dual-energy scans without compromises in image quality of the thorax and abdomen [19‐21]. Rapid voltage switching dual-energy acquisition has shown ambiguous results. Singh et al showed dose equivalence to dual-source, dual-energy acquisition but with inferior image quality whereas other authors state that rapid voltage switching acquisition results in higher patient’s radiation [22, 23].

One limitation of the study is that the patients included in the quantitative evaluation were not scanned with all acquisition protocols, as 6 out of 31 patients were only scanned with two of the three protocols. As the quantitative evaluation did not rely on pairwise comparisons, but used an unbalanced design instead, it was possible to include patients where one of the acquisitions was missing as well as to include multiple acquisitions of the same patient. This was done to increase the statistical power of the results, as the overall number of patients included in this study was relatively low, which is another limitation. Because of the limited number of patients, the transferability of the obtained results to clinical practice may be limited as well. One of the reasons for the limited number of patients in this study is the restriction to a single acquisition protocol, which, even though it is the most frequently used acquisition protocol at that CT scanner, limits the overall number of eligible patients and also limits the transferability of the obtained results to other acquisition protocols. However, comparing different protocols at once would have led to an unnecessary complexity of the evaluation, as additional variables would have to be considered. This is also true for the inclusion of arterial-phase imaging, which is more susceptible to changes in patient habitus, cardiac output, and underlying disease (e.g., differences in enhancement patterns for different tumors, as evident in pancreatic carcinoma [24]). While the results achieved in this study are specific for the acquisition protocol and body region, and obtained from a limited number of patients, and reductions in patient radiation exposure may be larger or smaller for other protocols or patient collectives, which has to be the subject of further evaluation in the future, we do expect that the results are nonetheless fairly generalizable, as we chose to evaluate a single, widely applicable acquisition protocol that makes up the bulk of acquisitions at the evaluated CT scanner. Even though there is no clear consensus on the definition of low dose, with an average patient dose of 60% below the applicable DRL, the evaluated acquisition protocol in this study qualifies as “reduced dose” [25].

Another limitation of this study and the DLCT acquisition protocols with increased tube potential might be the inferior conspicuity of vessel structures at higher tube potential, as mean photon energies closer to the k-edge of iodine provide improved image contrast [26]. However, with the additional possibility of spectral post-processing provided by DLCT, vessel delineation might be improved with virtual monochromatic images at lower virtual photon energies [27]. Furthermore, only conventional images of DLCT were analyzed in this study, even though spectral imaging was the main reason for the tube potential increase. Nonetheless, the number of potential spectral post-processing applications is very large and the quantification of the potential benefit from the additional spectral information for clinical routine is beyond the scope of this study.

Regarding the results, the inter-reader agreement for qualitative analysis was found to be moderate. However, this might be a consequence of the forced choice method used for evaluation combined with the low number of available choice (3 images). In consequence, a disagreement of ranking for one patient always consists of at least two further mismatches, resulting in a moderate Cohen’s kappa. As all CT images were of diagnostic quality, the forced choice method was able to provide a side-by-side comparison, allowing to identify even small differences in image quality that might otherwise remain hidden with more conventional methods, like rating image quality on a Likert scale. Moreover, the quantitative analysis found differences in the delineation of vasculature as shown by the differences in CT numbers measured in the vessels. These differences did not lead to significant differences in subjective image quality, as delineation of vasculature is but one aspect of overall image quality. While Singh et al or Tabari et al [22, 28] suggest to link the evaluation of image quality to the clinical task of lesion detection, this approach was not applicable to this study as not all patients had lesions in all three evaluated body regions. In consequence, future studies may evaluate different aspects of image quality to provide a more complete evaluation for different clinical tasks.

Overall, image quality of DLCT_low was comparable to SLCT_low and a low-dose acquisition was possible using dual-layer technique, which has the additional benefit of providing spectral information. With DLCT_low, CTDI_vol could be reduced by 28% compared to DLCT_high, but remained above that of SLCT_low. Further reduction of patient radiation exposure for the thorax seems possible.