Assessment of quantification accuracy and image quality of a full‐body dual‐layer spectral CT system

Abstract The performance of a recently introduced spectral computed tomography system based on a dual‐layer detector has been investigated. A semi‐anthropomorphic abdomen phantom for CT performance evaluation was imaged on the dual‐layer spectral CT at different radiation exposure levels (CTDI vol of 10 mGy, 20 mGy and 30 mGy). The phantom was equipped with specific low‐contrast and tissue‐equivalent inserts including water‐, adipose‐, muscle‐, liver‐, bone‐like materials and a variation in iodine concentrations. Additionally, the phantom size was varied using different extension rings to simulate different patient sizes. Contrast‐to‐noise (CNR) ratio over the range of available virtual mono‐energetic images (VMI) and the quantitative accuracy of VMI Hounsfield Units (HU), effective‐Z maps and iodine concentrations have been evaluated. Central and peripheral locations in the field‐of‐view have been examined. For all evaluated imaging tasks the results are within the calculated theoretical range of the tissue‐equivalent inserts. Especially at low energies, the CNR in VMIs could be boosted by up to 330% with respect to conventional images using iDose/spectral reconstructions at level 0. The mean bias found in effective‐Z maps and iodine concentrations averaged over all exposure levels and phantom sizes was 1.9% (eff. Z) and 3.4% (iodine). Only small variations were observed with increasing phantom size (+3%) while the bias was nearly independent of the exposure level (±0.2%). Therefore, dual‐layer detector based CT offers high quantitative accuracy of spectral images over the complete field‐of‐view without any compromise in radiation dose or diagnostic image quality.


| INTRODUCTION
Computed tomography (CT) is widely used in diagnostic imaging.
Many of today's design considerations in state-of-the-art CT systems provide a reduction in radiation exposure and enhance contrast-todose efficiency while increasing the perceived contrast-to-noiseratio. Among these developments, advanced iterative reconstruction techniques are a very promising tool. [1][2][3][4] However, despite these advances, extracting accurate information from CT images like object size and composition is still a challenge. Over the last years, spectral imaging methods using different dual-energy approaches (kVp switching, 5 dual x-ray sources 6 ) have attracted increased attention in research and clinical practice. Besides CT, also interventional 2D imaging techniques are foreseen to benefit from spectral imaging thechniques. 7,8 Those methods enable the quantification of object composition by exploiting measurements of the material-and energy-dependent x-ray attenuation of various materials using a low and high energy spectrum. Despite dual-energy CT having been proposed shortly after the invention of CT itself, 9,10 clinically relevant systems have only been available for the recent years. Physically, the linear x-ray attenuation coefficient l that is reconstructed in each voxel of a conventional CT is a function of the x-ray energy E, the material composition in the voxel represented by the involved atomic numbers Z and the mass density q in the voxel. A parametrization that is commonly used in diagnostic imaging expresses the total linear attenuation as the sum of photoelectric absorption and Compton effect 9 lðE; Z; qÞ ¼ a ph ðZÞf ph ðEÞ þ a C ðZÞf C ðEÞ À Á Á q (1) In the above equation, f ph E ð Þ and f C E ð Þ are the energy-dependent spectral basis functions for the photoelectric absorption and Compton effect while a ph Z ð Þ and a C Z ð Þ provide material-specific weighting factors for both contributions. Therefore, the reconstructed HU-values in conventional CT can be the same for two different types of materials.
A very common and descriptive example for this is the discrimination of iodinated blood and calcified plaques in contrast-enhanced CT angiography. Despite a large difference in the atomic number of these two materials, the reconstructed HU values may be identical depending on the local mass density or concentration of dissolved material.
Using dual-energy spectral CT, this limitation can be overcome by measuring the attenuation at two distinct energy levels E L and E H .
Substituting these energies into (eq. 1) forms a set of two equations in two unknown variables and can be numerically solved for the underlying material composition prior to or following the CT reconstruction. Such effective energy levels can in principle be obtained by measuring the quality of the x-ray beam using known calibration objects. Although this approach is only exactly valid in case of monochromatic energies E L and E H , it can be extended to the case of polychromatic spectra such as those used in CT scanners.
With previously existing dual-energy CT, a range of applications have been developed. 11,12 Virtual mono-energetic imaging 13 offers images with an appearance as if they were acquired with a mono-energetic x-ray beam, thereby reducing beam hardening artifacts, 14 increasing image contrast between lesions and healthy parenchyma 15 and potentially reducing radiation exposure. 16 Furthermore, dualenergy based material separation can be utilized to quantify contrast medium uptake and to obtain virtual noncontrast material-enhanced images. 17 E.g., iodine density maps 18 are applied in CTA, lesion characterization, 19 and lung perfusion assessment. 20 Besides these applications, many other clinical applications of material-specific imaging have been identified in recent years. [21][22][23][24] In future and with the introduction of fully spectral CT systems it can be expected to define additional clinical applications, for example in the area of oncological imaging of therapy response.
The scope of this study is to characterize a novel type spectral CT system which is based on a dual-layer detector. Potential advantages of dual-layer CT include complete spatial and temporal regis-

2.A | Dual-layer spectral CT
The experiments were carried out on a commercially available duallayer spectral CT (IQon spectral CT, Philips Healthcare, Cleveland, OH, USA). This novel scanner acquires spectral data per default at each CT scan, exploiting a dedicated dual-layer detector concept. [25][26][27] In contrast, other current dual-energy CT systems can only acquire spectral information with preselected protocols. Figure 1(a) shows a picture of the dual-layer spectral CT system along with a schematic drawing of the detector principle in Fig. 1 of the x-ray beam with the patient and the detector properties to facilitate material decomposition. 30 By design of the system, the spectral data are acquired simultaneously and perfectly registered in the projection domain. This enables the use of projection-based material decomposition which allows for efficient correction of beam-hardening artifacts and highly efficient noise reduction in the reconstructed images. 11,25,26,31 Along with the spectral image data, conventional polychromatic images are always obtained by summing up the low and high energy channels in the projection space. Exploiting the dual-layer concept, the scanner offers dual-energy data acquisitions per default in every CT scan. Therefore, there is no need for selecting a specific dualenergy protocol prior to an examination and all data can be processed with any kind of spectral analysis offered by the system in retrospective.

2.B | Description of phantoms
To assess the image quality, a semi-anthropomorphic abdomen phan-  A central 100 mm borehole in the abdomen phantom allows the installation of task-specific inserts used to assess the performance of a CT system with respect to special questions (see Fig. 3).
To study the quantitative accuracy of HU values and CNR in VMIs, the iodine-water decomposition as well as effective-Z values, a custom-made dual-energy phantom from the same manufacturer was used which provides four additional slots for selectable inserts.
Available materials are listed in Table 1.

2.C | Data acquisition and image reconstruction
The image acquisition and reconstruction was carried out using standard clinical protocols. An abdominal protocol was employed with the details given in Table 2. Images were obtained using varying 3. Images of the custom dualenergy phantom insert with 100 mm diameter (a) and the specific configuration used to assess the CNR in VMIs (b). The background of the insert consists of waterequivalent material (app. 0 HU at 120 kVp). The dual-energy phantom features different rods made from tissue surrogate material as well as varying concentrations of iodine and Ca-Hydroxyapatite. Fillable inserts are additionally provided for customized applications.
T A B L E 1 Insert rods for the customized dual-energy phantom. Along with a description of the tissue equivalents, the table lists the approximate CT numbers for each material at 120 kVp tube voltage. The iodine concentrations are embedded in solid water-equivalent plastic rods.

Material equivalent insert
Nominal elemental composition (%) where q mat is the mass density of the material of interest, the index i identifies each chemical element in the feature a i is the associated mass fraction and l E ð Þ=q ð Þ i the corresponding mass attenuation coefficient tabulated e.g., in the XCOM database. 33 Finally, l E ð Þ H2O describes the attenuation of pure water at energy level E.
Using the spectral software, energy-dependent attenuation plots were generated for each selected ROI in the energy range between The determined mean HU are mostly independent on the radiation dose which is again underlining the good quantitative performance of the spectral reconstruction algorithm. F I G . 5. Measured contrast-to-noise (CNR) curves of the medium-sized phantom (350 9 250 mm 2 ) at various energy levels. (a) is the conventional CT image acquired using 120 kVp while (b) depicts the same slice at a 50 keV VMI derived from the spectral data. The curves in (c-e) show the energy-dependent CNR and noise values in the virtual mono-energetic images (VMI) for CTDI vol ranging from 10 mGy to 30 mGy. The left column of plots in (c-e) gives the energy-dependent behavior of image noise in terms of the HU standard deviation measured in each ROI. From these curves it can be seen that the noise is kept at a constant level which is lower than the conventional reference value for energy levels greater than 50 keV. Since the observed difference in HU values between two materials typically increases toward lower energies, increased CNR can be observed for all contrasts compared to the conventional image. The gain in CNR is most evident for materials with a larger difference in spectral behavior, yielding a more than threefold increase compared to the conventional reference image. Therefore, the increased CNR can mostly be attributed to the increased HU difference in the VMIs.

3.C | Quantitative assessment of iodine concentrations in spectral images:
The quantitative measurement of iodine concentrations is summarized in Fig. 7.  VMIs. In clinical routine higher levels of iterative processing will also be of interest to further reduce radiation exposure and improve image quality. The choice of the optimal reconstruction parameters is typically highly task-dependent as features of very different contrast and size are investigated in various diseases. Therefore, subsequent clinical studies will be needed to further evaluate the benefits of VMIs in the diagnosis of various diseases.
The quantification of iodine concentrations generally showed good agreement between nominal and measured values. Until now, the accuracy of iodine concentration measurements using dualenergy CT has been reported to decrease with larger patient sizes. 43,44 However, the size-dependent increase in iodine concentration bias [c.f. Fig. 8(d)] observed with dual-layer CT is considerably smaller than the values reported so far in work related to other dual-energy techniques. The iodine quantification results presented in this study generally seem slightly more accurate than previously reported in literature on dual-layer CT. 45,46 However, iodine quantification in the dual-layer scanner is optimized for clinical applications with mixtures of iodine with blood or tissue. 45 50 where the differences in iodine uptake clearly exceed the quantification bias. Especially the latter case in Ref. [50] demonstrates the potential for contrast medium dose reduction as the studies were performed with administration of 1.5 ml contrast agent per kg body weight. Reducing this amount to a third would in theory still result in a difference of 1.3 mg/ml iodine and would therefore be safe to diagnose.
The translation away from dual-energy concepts toward spectral detector-based solutions results in a paradigm shift in the clinical routine, because dual-layer spectral CT now allows to always acquire spectral information over the full field-of-view. This will allow to define additional diagnostic application for spectral imaging. Furthermore, this is a first step toward detector-based spectral imaging, which will be followed by spectral photon-counting CT (SPCCT) in the future. 30,51,52 This future clinical technology promises to overcome major drawback of current scanners with quantitative imaging, material specific (k-edge) imaging, and a new level of diagnostic EHN ET AL.
| 215 image quality in combination with significant reduction in radiation exposure. With respect to this development, dual-layer spectral CT allows now to start working on the additional diagnostic benefit for current and future systems.
In conclusion, we report on experimental evaluation of a fullbody dual-layer spectral CT. Our results demonstrate the clinical possibilities when fully utilizing the potentials of spectral imaging.
High quantitative accuracy of the various spectral results could be demonstrated in this study. In the near future, dual-layer spectral CT will enhance the capabilities of CT diagnostics in a wide range of patient groups and diseases.

ACKNOWLEDGMENTS
We acknowledge financial support through the European Research

CONF LICT OF I NTEREST
The authors declare no conflicts of interest.