Spectral CT quantification stability and accuracy for pediatric patients: A phantom study

Abstract Background Spectral computed tomography (spectral CT) provides access to clinically relevant measures of endogenous and exogenous materials in patients. For pediatric patients, current spectral CT applications include lesion characterization, quantitative vascular imaging, assessments of tumor response to treatment, and more. Objective The aim of this study is a comprehensive investigation of the accuracy and stability of spectral quantifications from a spectral detector‐based CT system with respect to different patient sizes and radiation dose levels relevant for the pediatric population. Materials and methods A spectral CT phantom with tissue‐mimicking materials and iodine concentrations relevant for pediatric imaging was scanned on a spectral detector CT system using a standard pediatric abdominal protocol at 100%, 67%, 33% and 10% of the nominal radiation dose level. Different pediatric patient sizes were simulated using supplemental 3D‐printed extension rings. Virtual mono‐energetic, iodine density, effective atomic number, and electron density results were analyzed for stability with respect to radiation dose and patient size. Results Compared to conventional CT imaging, a pronounced improvement in the stability of attenuation measurements across patient size was observed when using virtual mono‐energetic images. Iodine densities were within 0.1 mg/ml, effective atomic numbers were within 0.26 atomic numbers and electron density quantifications were within ±1.0% of their respective nominal values. Relative to the nominal dose clinical protocol, differences in attenuation of all tissue‐mimicking materials were maintained below 1.6 HU for a 33% dose reduction, below 2.7 HU for a 67% dose reduction and below 3.7 HU for a 90% dose reduction, for all virtual mono‐energetic energies equal to or greater than 50 keV. Iodine, and effective atomic number quantifications were stable to within 0.1 mg/ml and 0.06 atomic numbers, respectively, across all measured dose levels. Conclusion Spectral CT provides accurate and stable material quantification with respect to radiation dose reduction (up to 90%) and differing pediatric patient size. The observed consistency is an important step towards quantitative pediatric imaging at low radiation exposure levels.


| INTRODUCTION
Spectral computed tomography (spectral CT) provides quantitative information that enhances conspicuity of disease and tissue characterization capabilities. The quantitative capabilities of spectral CT arise from access to elemental composition information. This is achieved through measurements of the energy-dependent materialspecific x-ray attenuation at different energies during a single CT scan. 1 Dual-Energy Computed Tomography (DECT), the first realization of spectral CT, has been investigated since the 1970s and became a commercially available clinical tool more than a decade ago. 2 DECT provides a variety of spectral results enabling quantitative measurement of endogenous and exogenous materials within the patient.
Spectral results include virtual mono-energetic images that, unlike conventional CT images, provide well-defined attenuation quantifications that enhance conspicuity and tissue characterization capabilities 3,4 while reducing metal and beam-hardening artifacts. 3,5 Other spectral results provide material-specific quantifications that enable density assessment and virtual attenuation suppression of various clinically relevant materials, e.g. iodine 6,7 and calcium 8 or electron density (ED) or effective atomic number (Zeff) estimations that enable improved material differentiation. 6 Recommended applications of spectral CT for children currently include vascular imaging, bowel imaging, assessment of perfused lung blood volume, detection and characterization of lesions in solid organs, tumor treatment response evaluations, metal artifact reduction, and renal calculi composition assessments. 9 All these applications rely on accurate attenuation or material quantifications, and their clinical benefit improves with increased consistency. For example, accurate iodine maps are used to generate perfused blood volume images that map iodine distributions in the lung parenchyma and allow both visualization and quantification of perfusion defects, and energy-dependent x-ray attenuation profiles are utilized for classifying different types of common stones that are found in children based on their predominant materials.
Different technologies can be used to implement the concept of acquiring paired attenuation measurements from two different energy spectra. Current commercial approaches to spectral CT include the rapid kVp-switching method, 10 the dual-source method, 2 the split-beam method, 11 the spin-spin method 1 and the dual-layer detector method, 12 which we utilized in our study. While sourcebased approaches achieve spectral separation by variations in the radiation spectra that pass through the patient, implemented either by changing the accelerating tube voltage or by different x-ray filtrations, the only commercially available detector-based spectral CT system owes its spectral separation capabilities to a dedicated duallayered detector. The unique detector design consists of a horizontal configuration of an Yttrium-based top layer that is more sensitive to low-energy photons and serves as filtration for the Gadolinium oxysulfide-based bottom layer, which mostly detects high-energy photons. Spectral information with sufficient energy separation is acquired with perfect spatial and temporal alignment at every 120 or 140 kVp scan, obviating the need to prospectively modify scan parameters in order to enable a spectral CT acquisition mode. 13 While the proven value of spectral CT is continuously rising, 14 there are only a few studies focused on spectral CT applications for pediatric imaging. 9,15,16 For this unique population, imaging applications concentrated on a different domain of clinical questions compared to those of adult patients is of interest. In addition, research of different technical aspects that are related to image quality are of importance, specifically, with respect to the smaller sizes of pediatric patients, and, perhaps more importantly, their sensitivity to ionizing radiation and the obligation to use lower doses. 17 Any application of CT for pediatric patients must include a risk-benefit analysis weighing the risk of exposure to ionizing radiation with the benefit of the information obtained from the scan. This is because young children are more sensitive to the effects of radiation than adults. [18][19][20] There are very few studies that include phantom measurements dedicated for pediatric patient spectral CT imaging. 21 Of these, the ability to perform dose-neutral spectral CT imaging was the primary concern, 22,23 rather than the resulting spectral accuracy. Comprehensive studies that evaluate the quantification capabilities of spectral CT while focusing on pediatric imaging are therefore still necessary for the adoption of spectral applications for this unique population.
The aim of this study is to conduct a thorough investigation of the accuracy and stability of spectral CT quantitative measurements with respect to different patient sizes and radiation dose levels that are relevant to pediatric patients.

2.A | Phantoms
In order to assess the accuracy of various spectral results as a function of patient size and radiation dose, a custom spectral CT phantom was used (Tissue Equivalent Materials & CTIodine ® , QRM GmbH, Moehrendorf, Germany). The phantom, shown in Fig. 1, contains a 10 cm diameter insert that can hold up to eight different tissue-mimicking or iodine density rods at a time. The insert and 1 cm diameter material rods are all 10 cm in length. The insert is made of a solid water equivalent plastic with X-ray attenuation properties similar to those of (liquid) water. Due to the importance of iodine in spectral CT imaging, we included four iodine rods at different concentrations: 0.5, 2, 5, and 10 mg/ml. The remaining rods were made of liver, adipose, 100 mg/ml hydroxyapatite (HA) and 400 mg/ml HA tissue-mimicking materials (HA simulates bone).
Reported waist circumference measurements in infants and children 24,25 were converted into approximate body diameters (through division by π) to determine median patient sizes of newborns, 1.5-2 yr-old, and 9-yr-old patients and select the phantom sizes (10-20 cm diameter) utilized in this study. That said, the large variability in waist circumferences at these early ages imply that, for example, a 20 cm diameter phantom size corresponds to the 85th percentile of 5-yr-old female patients as well as the 10th percentile of 13-yr-old male patients such that imaging of the selected SHAPIRA ET AL.

2.B | Scanner and acquisition protocol
The experiments were performed on a commercially available spectral detector (dual-layer) CT system (IQon Spectral CT, Philips Healthcare, Eindhoven, The Netherlands). The system is based on a conventional multi-detector CT system with rotation times down to 0.27 s, a 700 mm bore and a 120 kW generator allowing tube voltages of 80, 100, 120, and 140 kVp. The system is equipped with a 4 cm dual-layer spectral detector where the top layer is more sensitive to low-energy photons and serves as filtration for the bottom layer, which mostly detects high-energy photons. Detector materials and thicknesses were designed to result in equal noise levels for scans of typical body sizes. Spectral information with sufficient energy separation is acquired at every 120 or 140 kVp scan, eliminating the need to prospectively modify scan parameters in order to enable a spectral CT acquisition mode. 13 Owing to the detector-based spectral separation design of the system, data are acquired with perfectly aligned spatial and temporal registration of the two spectra in the projection domain. This provides the opportunity to accurately account for beam-hardening effects and to achieve highly efficient noise reduction with dedicated projection-based material decomposition and spectral denoising algorithms. 12 67, 33 and 10 mAs. These correspond to CTDI vol dose levels of 9, 6, 3, and 0.9 mGy, or effective dose levels of 0.18, 0.12, 0.06, and 0.018 mSv per every 10 cm scanned, when using the appropriate conversion (k) factor of 0.02 for abdomen and pelvis of 5-yr-old pediatric patients. 28 Reconstruction of each acquisition was performed using the clinical standard body kernel ("B" on the IQon scanner), a reconstruction field of view (FOV) diameter of 250 mm, 3 mm slice thicknesses and increments, and a matrix size of 512 × 512 pixels. Each reconstruction resulted in a series of spectral base images (SBI) which contain all the information that is required to generate all available spectral results. From these SBIs, VMIs of various energies as well as iodine density, Zeff, and ED images were generated.

2.C | Nominal values calculations
The nominal attenuation at various VMI energies [HU(E)] for the four tissue-mimicking rods were calculated based on the material composition (given as normalized elemental mass fractions, w i ) and physical density (ρ phys ) of each rod, which were provided by the phantom manufacturer, according to: where μ i E ð Þ and μ water E ð Þ are the mass attenuation coefficients (expressed in cm 2 /g) at photon energy Eof element i and water, respectively, and the sum is over all elements in the elemental composition. Values for μ i E ð Þ of the various elements within the elemental compositions of the rods and for water were obtained from publicly available data provided by the National Institute of Standards and Technology (NIST). 29 Similarly, nominal effective atomic number (Zeff) values for the four tissue-mimicking rods were calculated according to: where Z i is the atomic number of each element in the rod composi- Finally, the nominal electron density (ED) values were calculated according to: where ED water is the electron density of water (3:343 Â 10 23 m À3 ) and

2.D | Evaluation matrices
Analysis was based on mean and standard deviation measurements copied to all of the slices that correspond to the specific phantom configuration. All voxels that satisfy  Similarly, our analysis includes a size dependency of an additional set of VMI results, whose energy level (67 keV) was selected based on their closest quantification values to those of the conventional result. Figure 3 presents the size dependency of the conventional images compared to that of the 67 keV VMI spectral results. Here too, the vertical trends that appear in Fig. 3(a) are emphasized in

| DISCUSSION
Quantitative medical imaging is receiving greater acknowledgment from clinicians and healthcare providers due to its rapidly increasing clinical value. 34,35 Utilization of spectral CT applications for pediatric patients is still in its early adoption stages, with a pressing lack of quantitative evaluations to determine the spectral capabilities for this unique patient population. The dose-neutrality of dual-energy CT, compared to conventional CT, 23,36 provides an opportunity to improve patient care also for the pediatric population. For example, the use of quantitative measurements could be very helpful for In our study, we evaluated quantification accuracy and stability with respect to patient size and radiation dose of spectral results from a spectral detector CT scanner. CT solution should be independent of acquisition parameters such as rotation time, collimation, and pitch. This is due to the simultaneous acquisition of signals from the two spectra, which is unique to the detector-based DECT approach enables. Notably, the pediatric protocol that we utilized in this study was already configured with the fastest rotation time available on the system (0.27 s per rotation), extremely important for the pediatric population. Finally, our study included only phantoms and did not involve any clinical data. Moreover, the phantoms that were used in our study were not anthropomorphic phantoms but rather a material quantification phantom, designed for spectral quantification evaluation purposes and equipped with two extension rings that simulate different pediatric patient sizes. The authors are unaware of commercially available pediatric-sized anthropomorphic phantoms that are adequate for evaluating DECT material quantifications. For our purposes of quantifying the accuracy, consistency and stability of spectral results, a phantom study is the only viable solution for material quantifications with known ground truths and for performing dose dependency evaluations without introducing risk to a vulnerable patient population. However, while phantoms are important as an initial evaluation step, the benefits of the stability and consistencies of spectral results will require validation and further assessments in dedicated clinical studies and with respect to specific clinical applications for pediatric patients.

| CONCLUSION
Spectral results from a spectral detector DECT scanner provide accurate and consistent material characterizations with respect to different pediatric patient sizes and radiation dose reduction of up to 90%. The stability of the material quantifications such as mono-energetic attenuation estimations, as well as iodine density, effective atomic number and electron density quantifications enables an important step towards quantitative CT imaging for children of various ages with low exposure to ionizing radiation.

ACKNOWLEDG MENTS
Research reported in this publication was supported by the University of Pennsylvania Research Foundation (URF) and Philips Healthcare. In addition, we acknowledge Michael Geagan for his help with 3D-printing of the extension rings.

AUTHOR CONTRI BUTION STATEMENT
NS performed the measurements and participated in the analysis and in writing the manuscript, KM participated in the analysis and in writing the manuscript, PBN participated in the analysis and in writing the manuscript.

CONFLI CTS OF INTEREST
The authors declare no conflict of interest.