Radiation dose efficiency of multi-energy photon-counting-detector CT for dual-contrast imaging

Abstract

Compared to traditional multi-scan single-energy CT (SECT), one potential advantage of single-scan multi-energy CT (MECT) proposed for simultaneous imaging of multiple contrast agents is the radiation dose reduction. This phantom study aims to rigorously evaluate whether the radiation dose can truly be reduced in a single-scan MECT protocol (MECT_1s) in biphasic liver imaging with iodine and gadolinium, and small bowel imaging with iodine and bismuth, compared to traditional two-scan SECT protocols (SECT_2s). For MECT_1s, mixed iodine/gadolinium samples were prepared corresponding to late arterial/portal-venous phase for biphasic liver imaging. Mixed iodine/bismuth samples were prepared representing the arterial/enteric enhancement for small bowel imaging. For SECT_2s, separate contrast samples were prepared to mimic separate scans in arterial/venous phase and arterial/enteric enhancement. Samples were placed in a 35 cm wide water phantom and scanned by a research whole-body photon-counting-detector-CT (PCD-CT) system (‘chess’ mode). MECT images were acquired with optimized kV/threshold settings for each imaging task, and SECT images were acquired at 120 kV. Total CTDIvol was matched for the two protocols. Image-based three-material decomposition was employed in MECT_1s to determine the basis material concentration values, which were converted to CT numbers at 120 kV (i.e. virtual SECT images) to compare with the SECT images directly acquired with SECT_2s. The noise difference between the SECT and the virtual SECT images was compared to evaluate the dose efficiency of MECT_1s. Compared to SECT_2s, MECT_1s was not dose efficient for both imaging tasks. The amount of noise increase is highly task dependent, with noise increased by 203%/278% and 110%/82% in virtual SECT images for iodine/gadolinium and iodine/bismuth quantifications, respectively, corresponding to dose increase by 819%/1328% and 340%/230% in MECT_1s to achieve the same image noise level. MECT with the current PCD-CT technique requires higher radiation dose than SECT to achieve the same image quality.

1. Introduction

With the advance in photon-counting-detector (PCD) based and other energy-integrating-detector (EID) based multi-energy CT (MECT) techniques, the potential clinical benefits of a single-scan MECT with two or more contrast agents have been demonstrated in simulation, phantom, and animal studies (Anderson et al 2010, Taguchi and Iwanczyk 2013, Muenzel et al 2016, 2017a, 2017b, Cormode et al 2017, Symons et al 2017a, 2017b, Dangelmaier et al 2018, Panta et al 2018, Ren et al 2018a, 2018b, Stayman and Tilley 2018, Yu et al 2018). Using a single MECT scan and basis material decomposition, distributions of different contrast agents can be separated and quantified, which correspond to different phases acquired by multiple scans in traditional single-energy CT (SECT). A combination of clinical iodine- and gadolinium-based agents have been used for this purpose, where the two agents are injected sequentially with a certain time interval in between, to enable biphasic data acquisition from one single MECT scan. Simultaneous imaging of iodine and gadolinium could potentially be used in biphasic liver imaging (Muenzel et al 2016, 2017a), biphasic kidney imaging (Symons et al 2017a), CT colonography (Muenzel et al 2017b), or CT angiography (Symons et al 2017b). Another promising pair involves iodine and bismuth, which are administered via intravascular and enteric routes, respectively. Simultaneous imaging of these two contrast agents could potentially be used in CT enterography to differentiate arterial enhancement in vascularized bowel wall and enteric enhancement in the intestinal lumen from a single MECT scan (Symons et al 2017a, Yu et al 2018). Potential clinical applications of other combinations of two or more contrast agents such as iodine/gold, iodine/gadolinium/bismuth, and iodine/gadolinium/gold have also been explored in phantom and animal models (Cormode et al 2017, Symons et al 2017a, Panta et al 2018, Stayman and Tilley 2018).

Compared to traditional multi-scan SECT, two potential advantages of performing a single MECT scan to simultaneously image multiple contrast agents have been claimed (Muenzel et al 2016, 2017a, Cormode et al 2017, Symons et al 2017a, 2017b). The first potential advantage is the perfect or near perfect image co-registration among different phases that can be obtained from one single MECT scan. Traditional SECT scans at multiple phases may suffer from motions of visceral organs caused by bowel peristalsis and depths of breathing among different scans. The second potential advantage is the radiation dose reduction by allowing omission of some of the scans. While the first potential advantage seems obvious, the second one has not been thoroughly investigated.

The purpose of this work, therefore, is to perform a systematic and quantitative phantom study to compare SECT images acquired with traditional two-scan SECT protocol (SECT_2s) and virtual SECT images generated from material-decomposed images with single-scan MECT protocol (MECT_1s), with matched total radiation dose levels for the two protocols. The comparison will be performed for two clinical imaging tasks that may potentially benefit from simultaneous imaging of two contrast agents (referred to as dual-contrast imaging): biphasic liver imaging using iodine and gadolinium, and small bowel imaging using iodine and bismuth since these two applications received great attention recently and are representative of common dual-contrast MECT applications (Qu et al 2012, Muenzel et al 2016, 2017a, Symons et al 2017a, Yu et al 2018).

2. Materials and methods

2.1. Scan protocols

2.1.1. Biphasic liver imaging

Two protocols for biphasic liver imaging, SECT_2s and MECT_1s, are illustrated in figures 1(a) and (b), respectively. In SECT_2s, the current clinical protocol, the iodine intravenous contrast is injected at time T₀, followed by two SECT scans performed at time T₁ and T₂ to capture the iodine contrast enhancement at late arterial phase and portal-venous phase, respectively. The time delays (t₁ and t₂) are determined by the enhancement phases of interest. As labelled in figure 1(a), the delay time (t₁) and the additional delay time (t₂) are required for a bolus of contrast material (iodine) to enter the arterial structures and transit further to the venous structures of the liver. In MECT_1s, gadolinium and iodine intravenous contrasts are injected at T₀ and $T_1^{\prime }$, respectively. One single MECT scan is performed at time T₂ to simultaneously image the iodine and gadolinium. Similarly as in SECT_2s, the delay time (t₁) is required for the contrast material of iodine injected at $T_1^{\prime }$ to enter the arterial structures of the liver. A total delay time (t₁ + t₂) is required for contrast material of gadolinium injected at T₀ to reach to the venous structures of the liver. By this design, the timing of the iodine enhancement corresponds to the late arterial phase and the timing of the gadolinium enhancement corresponds to the portal-venous phase. For comparison, the total radiation dose (CTDIvol) of SECT_2s and MECT_1s is matched as 2 × D₀ mGy.

Figure 1.

Scan protocols for biphasic liver imaging with (a) two SECT scans for iodine imaging and (b) one MECT scan for simultaneous iodine and gadolinium imaging; the total radiation dose is matched for comparison.

2.1.2. Small bowel imaging

For small bowel imaging, a SECT_2s protocol is designed as a reference for comparing the dose efficiency of MECT, where the iodine intravenous contrast and bismuth oral contrast are administered to the patients through two separate procedures (figures 2(a) and (b)). Iodine contrast is injected intravenously at time $T_1^{\prime }$ and a SECT scan is performed at time $T_1^{″}$ to image the arterial enhancement (figure 2(a)). Bismuth contrast is administered orally at time $T_1^{\prime }$ and one SECT is performed at time $T_1^{″}$ to image the enteric enhancement (figure 2(b)). Both time delays ($t_1^{\prime }$ and $t_2^{\prime }$) are determined by the contrast enhancement of iodine or bismuth. In the MECT_1s protocol (figure 2(c)), bismuth contrast is administered orally at $T_0^{″}$ and iodine contrast is injected intravenously at $T_0^{\prime }$. One MECT scan is performed at time $T_1^{″}$ to simultaneously image the iodine, representing the arterial enhancement, and the bismuth, representing the enteric enhancement. For comparison, the total radiation dose is matched for SECT_2s and MECT_1s as 2 × D₀ mGy.

Figure 2.

Scan protocols for small bowel imaging with (a) one SECT scan for iodine imaging, (b) one SECT scan for bismuth imaging, and (c) one MECT scan for simultaneous imaging of both iodine and bismuth; the total radiation dose is matched for comparison.

2.2. Sample preparation and experimental phantom design

For the biphasic liver imaging task in SECT_2s, two sets of iodine samples with multiple concentrations (5.0, 10.0 and 15.0 mg cc⁻¹) were prepared using commercially available iodine-based Iohexol (Omnipaque 350, GE Healthcare, Princeton, NJ), with the enhancement values corresponding to the late arterial (figure 3(a)) and portal-venous (figure 3(b)) phases, respectively. Note that these two sets of iodine samples were not mixed and were scanned separately, to mimic separate scans in the arterial/venous phases. In MECT_1s, a set of mixed iodine and gadolinium samples was prepared using iodine-based Iohexol and gadolinium-based gadopentetate dimeglumine (Gadavist, Bayer Healthcare, Whippany, NJ), as shown in figure 3(c), with the iodine enhancement corresponding to the late arterial phase and the gadolinium enhancement to the portal-venous phase. The concentrations of iodine samples prepared for MECT_1s were identical to those prepared for SECT_2s, namely, 5.0, 10.0 and 15.0 mg cc⁻¹. The concentrations of gadolinium samples were determined as 3.3, 6.7 and 10.0 mg cc⁻¹, in an effort to match the enhancement of iodine samples at a 120 kV SECT scan.

Figure 3.

Phantom layouts for testing the two tasks: biphasic liver imaging (a)–(c) and small bowel imaging (d)–(f). (a) Iodine samples mimicking contrast enhancement at late arterial phase in SECT_2s, (b) iodine samples mimicking contrast enhancement at portal-venous phase in SECT_2s, and (c) mixed iodine and gadolinium samples for contrast enhancement at the two phases in MECT_1s; (d) iodine samples mimicking arterial enhancement in SECT_2s, (e) bismuth samples mimicking enteric enhancement in SECT_2s, and (f) mixed iodine and bismuth samples for both arterial and enteric enhancements in MECT_1s (phantom lateral dimension: 35 cm).

For the small bowel imaging task in SECT_2s, one set of iodine samples (5.0, 10.0 and 15.0 mg cc⁻¹) and one set of bismuth samples (10.0, 15.0 and 20.0 mg cc⁻¹) were prepared using iodine-based Iohexol and bismuth subsalicylate (Pepto-Bismol, Proctor&Gamble, Cincinnati, Ohio) to mimic the arterial (figure 3(d)) and enteric (figure 3(e)) enhancements, respectively. The concentration values for bismuth samples were determined to provide about 200–500 HU enteric enhancement in small bowel lumen. In MECT_1s, a set of mixed iodine and bismuth samples was prepared with the iodine representing the arterial enhancement and the bismuth representing the enteric enhancement (figure 3(f)). Note that the concentrations of iodine and bismuth samples prepared for MECT_1s were identical to those prepared for SECT_2s for comparison.

Samples were contained in plastic cylindrical tubes (diameter: 2.54 cm), which were placed in a 35 cm wide water tank to mimic the attenuation of a medium-size patient. Phantom layouts with labelled material types and concentrations are illustrated in figure 3.

2.3. PCD-CT system and data acquisition

Phantom experiments were performed on a research whole-body PCD-CT system (SOMATOM CounT, Siemens Healthineers), which was designed based on the same platform as the second-generation dual-source, dual-energy CT system (SOMATOM Definition Flash, Siemens Healthineers), with the second EID-based system replaced by a PCD-based system. The system therefore, consists of two independent subsystems, namely the EID subsystem with a full field of view (FOV) of 50 cm and the PCD subsystem with a FOV of 27.5 cm. When the maximum dimension of the imaged object (e.g. 35 cm in the current study) exceeds the designated FOV in the PCD subsystem, a separate, low-dose data completion scan using the EID subsystem is performed to estimate the missing PCD projections and complete the required PCD data (Yu et al 2016a). Thus, truncation artifacts in the reconstructed PCD images are minimized.

The macro pixels of the PCD are of size 0.9 × 0.9 mm², which corresponds to 0.5 × 0.5 mm² at the isocenter of the PCD subsystem. Each macro pixel is grouped by 4-by-4 subpixels, and each subpixel contains two photon-counting circuits allowing two different energy thresholds. Multiple data acquisition modes are available on the PCD-CT system with different arrangement of the energy thresholds for each subpixel (Leng et al 2018). In the current work, experiments were performed using a ‘chess’ mode, which provides the capability of multi-energy imaging with four energy bins. To achieve this, the 16 subpixels in each macro pixel are interlaced, where the thresholds of half subpixels are set to two energy thresholds (threshold low 1, or TL₁, and threshold high 1, or TH₁), and the other half subpixels are set to two different energy thresholds (threshold low 2, or TL₂, and threshold high 2, or TH₂), with 0 < TL₁ < TL₂ < TH₁ < TH₂ < Ep, where Ep denotes the peak tube potential. Eight sets of projection data are generated in the ‘chess’ mode including four sets of threshold data ([TL₁ Ep], [TL₂ Ep], [TH₁ Ep], and [TH₂ Ep]) and four sets of bin data ([TL₁ TL₂], [TL₂ TH₁], [TH₁ TH₂], and [TH₂ Ep]). Note that each 8-subpixel grouping of the detector pixels only uses half of the x-ray photons incident on a macro pixel, and the ‘chess’ mode has 50% detector efficiency during data acquisition (Yu et al 2016b). To compensate for this, the radiation dose level was doubled in the current study with respect to the routine dose level determined by an initial 120 kV SECT scan with quality reference effective mAs set as 200 and automatic exposure control enabled in the EID subsystem (Michalak et al 2016). More details regarding the research whole-body PCD-CT system can be found elsewhere (Yu et al 2016a, Li et al 2017, Zhou et al 2017, Leng et al 2018, Tao et al 2018, Zhou et al 2018).

MECT_1s phantoms containing iodine/gadolinium samples (figure 3(c)) and iodine/bismuth samples (figure 3(f)) were scanned at 80 kV with the four energy thresholds at 25, 35, 50, and 55 keV, and 140 kV with 25, 50, 75, and 90 keV, respectively. The kV and energy threshold settings for each imaging task were determined based on both literature studies (Tao et al 2019a, Symons et al 2017a, Tao et al 2019b) and additional phantom experiments by going through a range of kV and energy threshold settings, to yield optimal or near optimal material decomposition performance. Specifically, a lower (i.e. 80 kV) and a higher (i.e. 140 kV) x-ray tube potential are desired to allocate more photons near the k-edge of gadolinium and bismuth, respectively. The threshold of 25 keV in both imaging tasks was used to maximally reject electronic noise without excessively sacrificing the photon statistics. The 50 keV in iodine/gadolinium imaging and 90 keV in iodine/bismuth imaging were used to capture the k-edge of gadolinium and bismuth, respectively. The other two thresholds, 35 and 55 keV in iodine/gadolinium imaging, and 50 and 75 keV in iodine/bismuth imaging, were determined to better utilize the k-edges via balancing the number of photons within adjacent energy bins. PCD-CT x-ray spectra along with the linear attenuation coefficients of the contrast materials for the two imaging tasks are plotted in figure 4.

Figure 4.

PCD-CT x-ray spectra for (a) biphasic liver imaging with iodine and gadolinium and (b) small bowel imaging with iodine and bismuth; the linear attenuation curves (LACs) derived from pure iodine, gadolinium, and bismuth were also plotted.

SECT_2s phantoms designed for biphasic liver imaging (figures 3(a) and (b)) were scanned at 120 kV with four energy thresholds at 25, 35, 50, and 55 keV, and those designed for small bowel imaging (figures 3(e) and (f)) were scanned at 120 kV with 25, 50, 75, and 90 keV. The [TL₁ Ep] image, namely[25 120] keV, which was formed by all the incident x-ray photons, was used to represent the SECT image acquired at 120 kV in SECT_2s. Other three energy thresholds, namely TL₂, TH₁, and TH₂ (i.e. 35, 50, and 55 keV, or 50, 75, and 90 keV) were arbitrarily selected and not used to form SECT images. Here 120 kV was selected according to the abdominal SECT protocols in our institute. SECT_2s phantoms designed for biphasic liver imaging were also scanned at 80 kV with four energy thresholds at 25, 35, 50, and 55 keV, and those designed for small bowel imaging were also scanned at 140 kV with four thresholds at 25, 50, 75, and 90 keV, both with identical kV and threshold settings as in the MECT_1s scans as references.

Note that the SECT images were acquired with the PCD-CT subsystem rather than the EID subsystem to provide a direct and fair comparison between SECT_2s and MECT_1s by avoiding any systematic differences between EID subsystem and PCD subsystem. As a matter of fact, the noise performance of the PCD subsystem ([TL₁ Ep] image) was comparable to that of the EID subsystem when the noise measurement is normalized by the spatial resolution (Yu et al 2016b). Information about data acquisition geometry, radiation dose, and image reconstruction in SECT_2s and MECT_1s are summarized in table 1. Note that the total CTDIvol was matched for the two protocols.

Table 1.

Data acquisition geometry, radiation dose, and image reconstruction in SECT_2s and MECT_1s.

Imaging protocol	SECT_2s		MECT_1s
Imaging task	Biphasic liver imaging	Small bowel imaging	Biphasic liver imaging	Small bowel imaging
Contrast material(s)	Late arterial: iodine	Arterial: iodine	Late arterial: iodine	Arterial: iodine
	Portal-venous: iodine	Enteric: bismuth	Portal-venous: gadolinium	Enteric: bismuth
Tube potential (kV)	80, 120	140, 120	80	140
Threshold settings (keV)	25, 35, 50, 55	25, 50, 75, 90	25, 35, 50, 55	25, 50, 75, 90
Pitch			0.6
Rotation time (s)			1.0
Collimation (mm)			32 × 0.5
Scan/Recon FOV (mm)			275/275
CTDIvol (mGy)	23.0 + 23.0a		46.0
Reconstruction Kernel			D30

^aThe routine dose level determined by an initial 120 kV scan was 11.5 mGy.

2.4. MECT basis material decomposition

The MECT_1s scan was followed by a generic image-based material decomposition to determine the basis material concentrations at each pixel. The assumption of volume conservation is included as an additional physical constraint (Yu et al 2018, Ren et al 2019a), as given by

$$\{\begin{array}{c}\mu \left(E_1\right)=\sum _{m=1}^M{\left(\frac{\mu }{\rho }\right)}_m\left(E_1\right)\cdot {\rho }_m\\ \mu \left(E_2\right)=\sum _{m=1}^M{\left(\frac{\mu }{\rho }\right)}_m\left(E_2\right)\cdot {\rho }_m\\ ⋮\\ \mu \left(E_N\right)=\sum _{m=1}^M{\left(\frac{\mu }{\rho }\right)}_m\left(E_N\right)\cdot {\rho }_m\\ 1=\sum _{m=1}^M\frac{{\rho }_m}{{\rho }_{m0}}\end{array},$$ (1)

where μ (E_i), i = 1, 2, …, N represent the effective linear attenuation coefficient measured at ith energy (E) bin; ${\left(\frac{\mu }{\rho }\right)}_m\left(E_i\right)$, m = 1, 2,…,M, i = 1, 2, …, N represent the mass attenuation coefficient of the mth basis material at ith energy bin, determined beforehand by using a calibration procedure; N is the number of energy bins (N = 4 for the current study); M is the number of basis materials (M = 3 in the current study); the last $\text{row }1=\sum _{m=1}^M\frac{{\rho }_m}{{\rho }_{m0}}$ refers to volume conservation with ρ_m and ρ_m0 as the concentration of basis material m in the mixture and density in its pure form, respectively.

Specifically, the basis material concentration ρ_m at each pixel in the above linear equation system was solved using a generalized least squares optimization method, with the cost function given by

$$\overrightarrow{\rho }=\underset{\overrightarrow{\rho }}{\text{arg min}}{(\overrightarrow{\mu }-A\overrightarrow{\rho })}^T{V}^{-1}(\overrightarrow{\mu }-A\overrightarrow{\rho }),$$ (2)

where $\overrightarrow{\mu }=\left[\begin{array}{c}\mu \left(E_1\right)\\ ⋮\\ \mu \left(E_N\right)\\ 1\end{array}\right],\overrightarrow{\rho }=\left(\begin{array}{c}{\rho }_1\\ ⋮\\ {\rho }_m\end{array}\right)$, and $A=\left[\begin{array}{cccc}{\left(\frac{\mu }{\rho }\right)}_1\left(E_1\right)& {\left(\frac{\mu }{\rho }\right)}_2\left(E_1\right)& \cdots & {\left(\frac{\mu }{\rho }\right)}_M\left(E_1\right)\\ ⋮& \ddots & & ⋮\\ {\left(\frac{\mu }{\rho }\right)}_1\left(E_N\right)& {\left(\frac{\mu }{\rho }\right)}_2\left(E_N\right)& \cdots & {\left(\frac{\mu }{\rho }\right)}_M\left(E_N\right)\\ \frac{1}{{\rho }_{10}}& \frac{1}{{\rho }_{20}}& \cdots & \frac{1}{{\rho }_{M0}}\end{array}\right]$ denote energy bin measurements, basis material concentrations, and coefficient matrix, respectively; V represents the variance-covariance matrix defining the variance V_nn(n ⩽ 4) in energy bin image μ (E_n) and covariance V_nn′ (n, n^′ ⩽ 4) between energy images μ (E_n) and μ (E_n′). The variance-covariance matrix V(n, n′ ⩽ 4) can be calculated using an identical ROI with P pixels in all the energy bin images, as given by Faby et al (2015)

$$V_{nn}=\frac{1}{P}\sum _{p\in ROI}{\left[{\mu }_p\left(E_n\right)-\overline{\mu \left(E_n\right)}\right]}^2,n⩽4$$ (3)

$$V_{n{n}^{\prime }}=\frac{1}{P}\sum _{p\in \text{ROl}}\left[{\mu }_p\left(E_n\right)-\overline{\mu \left(E_n\right)}\right]\left[{\mu }_p\left(E_{n^{\prime }}\right)-\overline{\mu \left(E_{n^{\prime }}\right)}\right],n,n^{\prime }⩽4$$ (4)

where $\overline{\mu \left(E_n\right)}$ and $\overline{\mu \left(E_n\right)}$ are the mean values in the chosen ROI in energy images μ (E_n) and μ (E_n′). Note that V_5j = V_j5 = 0, j = 1, 2, 3, 4, and 5, since the last element of $\overrightarrow{\mu }$ is equal to a constant of 1. Thus the inverse of V is calculated using the Moore-Penrose pseudoinverse.

2.5. Virtual SECT (vSECT) Images in MECT_1s

To compare with the SECT images acquired directly with the SECT_2s, the concentration values in each contrast-specific image determined from the material decomposition process needed to be converted to CT numbers at 120 kV. In this study, the converted SECT images at 120 kV from MECT are referred hereafter to as Virtual SECT (vSECT) images. In biphasic liver imaging, the vSECT image for late arterial phase [μ_vSE,LA (E_{120 kV})] was a combination of iodine-specific image (ρ_I) and water image (ρ_w), which is equivalent to the mixed image subtracted by the gadolinium-specific image, and the vSECT image for portal-venous phase [μ_vSE,PV (E_{120 kV})] was a combination of gadolinium-specific image (ρ_Gd) and water image (ρ_w), which is equivalent to the mixed image subtracted by the iodine-specific image. The two vSECT images are given by

$${\mu }_{\text{vSE},\text{LA}}\left(E_{120\text{ kV}}\right)={\left(\frac{\mu }{\rho }\right)}_I\left(E_{120\text{ kV}}\right)\cdot {\rho }_I+{\left(\frac{\mu }{\rho }\right)}_w\left(E_{120\text{ kV}}\right)\cdot {\rho }_w,$$ (5)

$${\mu }_{\text{vSE},\text{PV}}\left(E_{120\text{ kV}}\right)={\left(\frac{\mu }{\rho }\right)}_{Gd}\left(E_{120\text{ kV}}\right)\cdot {\rho }_I+{\left(\frac{\mu }{\rho }\right)}_w\left(E_{120\text{ kV}}\right)\cdot {\rho }_w,$$ (6)

where ${\left(\frac{\mu }{\rho }\right)}_I\left(E_{120\text{ kV}}\right)$, ${\left(\frac{\mu }{\rho }\right)}_{Gd}\left(E_{120\text{ kV}}\right)$, and ${\left(\frac{\mu }{\rho }\right)}_w\left(E_{120\text{ kV}}\right)$ are the effective mass attenuation coefficients for the three basis materials at 120 kV; E_{120 kV} denotes the effective beam energy of the broad 120 kV spectrum.

Similarly, in small bowel imaging, the vSECT image for arterial enhancement [μ_vSE,AE (E_{120 kV})] was a combination of the iodine-specific image (${\rho }_I^{\prime }$) and water image (${\rho }_W^{\prime }$), and the vSECT image for enteric enhancement [μ_vSE,EE (E_{120 kV})] is a combination of the bismuth-specific image (ρ_Bi) and water image (${\rho }_W^{\prime }$), which are given by

$${\mu }_{\text{vSE},\text{AE}}\left(E_{120\text{ kV}}\right)={\left(\frac{\mu }{\rho }\right)}_I\left(E_{120\text{ kV}}\right)\cdot {\rho }_I^{\prime }+{\left(\frac{\mu }{\rho }\right)}_w\left(E_{120\text{ kV}}\right)\cdot {\rho }_w^{\prime },$$ (7)

$${\mu }_{\text{vSE},\text{EE}}\left(E_{120\text{ kV}}\right)={\left(\frac{\mu }{\rho }\right)}_{Bi}\left(E_{120\text{ kV}}\right)\cdot {\rho }_{Bi}+{\left(\frac{\mu }{\rho }\right)}_w\left(E_{120\text{ kV}}\right)\cdot {\rho }_w^{\prime },$$ (8)

where ${\left(\frac{\mu }{\rho }\right)}_{Bi}\left(E_{120\text{ kV}}\right)$ denotes the mass attenuation coefficient for bismuth at 120 kV. The difference in noise between SECT and vSECT (generated from MECT_1s) images at matched radiation dose was calculated as:

$$\mathrm{\Delta }\sigma =\frac{\sigma \left[{\mu }_{v\text{SE},P_i}\left(E_{120\text{ kV}}\right)\right]-\sigma \left[{\mu }_{SE,P_i}\left(E_{120\text{ kV}}\right)\right]}{\sigma \left[{\mu }_{\text{SE},P_i}\left(E_{120\text{ kV}}\right)\right]}\times 100\%,$$ (9)

where σ [μ_vSE,Pi (E_{120 kV})] and σ [μ_SE,Pi (E_{120 kV})] are the noise (standard deviation) on vSECT and SECT images for different phases/enhancements (i.e. i = LA, PV, AE, or EE). Note that direct comparison of noise in SECT and vSECT images is equivalent to compare the contrast-to-noise ratio (CNR), because the contrast values in SECT and vSECT images for small bowel imaging were matched (identical samples in two protocols at the same x-ray tube voltage of 120 kV), while those for biphasic liver imaging were also matched at 120 kV through purposely selecting the concentration values between iodine and gadolinium samples. It is thus sufficient to compare the SECT/vSECT image noise as a measure of CNR.

The dose difference could also be determined based on the noises in SECT and vSECT images, as if the same image noise was targeted for the two protocols (Leng et al 2016):

$$\mathrm{\Delta }D=\frac{{\sigma }^2\left[{\mu }_{\text{vSE},P_i}\left(E_{120\text{ kV}}\right)\right]-{\sigma }^2\left[{\mu }_{\text{SE},P_i}\left(E_{120\text{ kV}}\right)\right]}{{\sigma }^2\left[{\mu }_{\text{SE},P_i}\left(E_{120\text{ kV}}\right)\right]}\times 100\%.$$ (10)

3. Results

3.1. Biphasic Liver Imaging

Figures 5(a)–(c) depicts the basis material images of iodine, gadolinium, and water determined using MECT data with the three-material decomposition algorithm. Note that the significant noise reduction in water image, compared to iodine- and gadolinium-specific images was due to the inclusion of the volume conservation, as detailed in our previous study (Ren et al 2019a). The concentration values of iodine and gadolinium samples were measured in iodine-specific and gadolinium-specific images from eight circular ROIs (noted in red, figures 5(a) and (b)), and plotted against the nominal values (figures 5(d) and (e)). Strong linear correlations between measured and nominal concentrations were found for both iodine and gadolinium samples (R² ⩾ 0.99, 1.00 ⩽ slope ⩽ 1.05, and −0.13 ⩽ offset ⩽ 0.11 mg cc⁻¹).

Figure 5.

Biphasic liver imaging using MECT scan with three-material decomposition algorithm: (a) iodine-specific image, (b) gadolinium-specific image, and (c) water image; linearity analysis between measured and nominal concentrations: (d) iodine samples and (e) gadolinium samples.

Figure 6(a) shows the SECT images (1st column) and the vSECT images (2nd column) both at 120 kV for two vascular phases. Compared with the SECT images, one can visually observe the noise increases in vSECT images. To quantitatively compare all SECT/vSECT images, mean CT numbers (and standard deviations) were measured in five sample areas (labeled in figure 6(a)) and summarized in table 2. The measured mean CT numbers were comparable between SECT and vSECT images both at 120 kV for each vascular phase since the sample concentrations were either identical (iodine in SECT_2s versus iodine in MECT_1s) or purposely matched (iodine in SECT_2s versus gadolinium in MECT_1s) at 120 kV; the average mean CT number difference was less than 7%. The noise levels (standard deviations) on SECT and vSECT at 120 kV were also summarized in figure 6(b) for comparison, indicating an increased noise level by 203% and 278% for late arterial phase and portal-venous phase, respectively, in the vSECT images. Again, thanks to the matched contrast enhancement between SECT_2s and MECT_1s at 120 kV, the comparison of noise is equivalent to that of CNR. To achieve the same image noise, 819%–1328% more radiation dose is needed for biphasic liver imaging with MECT_1s. Note that in table 2, larger CT numbers were measured for all ROIs in SECT images at 80 kV, compared to SECT/vSECT images at 120 kV, due to stronger attenuations of both iodine and gadolinium at a lower energy beam. The average noise levels across all samples in two phases were calculated as 24.4 ± 1.5 HU and 20.6 ± 1.3 HU, for SECT 80 kV and 120 kV images, with a relative difference about 18.1% (p < 0.001). The noise levels measured on the threshold low images acquired using the PCD-CT system for different tube potential (e.g. 80 kV and 120 kV) with equal total radiation dose level were consistent with a previous study (Gutjahr et al 2016).

Figure 6.

(a) SECT images (1st column) acquired from SECT_2s and vSECT images (2nd column) generated from MECT_1s for biphasic liver imaging; (b) noise level comparison between SECT and vSECT images.

Table 2.

Summary of contrast/noise in SECT/vSECT images for biphasic liver imaging (unit: HU).

		Sample 1	Sample 2	Sample 3	Sample 4	Sample 5
Late arterial phase	SECT (80kV)	352.5 ± 23.6	186.1 ± 22.8	551.0 ± 26.8	389.2 ± 27.2	174.6 ± 24.1
	SECT (120kV)	224.2 ± 21.9	114.1 ± 21.6	343.7 ± 20.6	244.5 ± 20.2	109.1 ± 17.1
	vSECT (120kV)a	242.6 ± 57.0	124.6 ± 62.2	331.8 ± 66.2	257.1 ± 61.3	122.5 ± 60.8
Portal-venous phase	SECT (80kV)	551.1 ± 25.4	393.8 ± 22.9	176.5 ± 22.9	351.1 ± 24.0	186.2 ± 23.8
	SECT (120kV)	350.7 ± 21.0	250.7 ± 21.1	109.7 ± 21.1	222.2 ± 20.8	113.6 ± 20.8
	vSECT (120kV)a	346.2 ± 78.9	238.2 ± 72.0	108.4 ± 72.0	208.8 ± 79.7	92.0 ± 78.6

^avSECT images at 120 kV were converted from 80 kV PCD acquisition.

3.2. Small bowel imaging

Figures 7(a)–(c) show the basis material images of iodine, bismuth, and water determined using the MECT data with the three-material decomposition algorithm. Again, thanks to the incorporation of volume conservation, significant noise reduction in water image was achieved in comparison with the iodine- and bismuth-specific images (Ren et al 2019a). The concentration values of iodine and bismuth samples were measured in iodine-specific and bismuth-specific images from eight circular ROIs (noted in red, figures 7(a) and (b)), and plotted against the nominal values (figures 7(d) and (e)). Strong linear correlations between measured and nominal concentrations were found for both iodine and bismuth samples (R² ⩾ 0.99, 1.00 ⩽ slope ⩽ 1.02, and −0.11 ⩽ offset ⩽ −0.03 mg cc⁻¹).

Figure 7.

Small bowel imaging using MECT scan with three-material decomposition algorithm: (a) iodine-specific image, (b) bismuth-specific image, and (c) water image; linearity analysis between measured and nominal concentrations: (d) iodine samples and (e) bismuth samples.

Figure 8(a) shows the SECT images (1st column) and the vSECT images (2nd column) both at 120 kV for arterial and enteric enhancements. Compared with the SECT images, one can visually observe the noise increases in vSECT images. To quantitatively compare all SECT/vSECT images, mean CT numbers (and standard deviations) were measured in five sample areas (labeled in figure 8(a)) and summarized in table 3. The measured mean CT numbers were comparable between SECT and vSECT images both at 120 kV since the sample concentrations were identical in SECT_2s and MECT_1s; the average mean CT number difference was less than 3%. The noise levels (standard deviations) on SECT and vSECT at 120 kV were also summarized in figure 8(b) for comparison, indicating an increased noise level by 110% and 82% for arterial and enteric enhancement, respectively, in the vSECT images. Due to the matched contrast enhancement between SECT_2s and MECT_1s at 120 kV, the comparison of noise is equivalent to that of CNR. To achieve the same image noise, 230%–340% more radiation dose is needed for small bowel imaging with MECT_1s. Note that in table 3, slightly smaller CT numbers were measured for all ROIs in SECT images at 140 kV, compared to SECT/vSECT images at 120 kV due to slightly less attenuations of both iodine and bismuth at a slightly higher energy beam. The average noise levels across all samples in two phases were calculated as 19.7 ± 1.7 HU and 19.9 ± 1.4 HU, for SECT 140 kV and 120 kV images, with a relative difference about 1.1% (p > 0.05). The noise levels measured on the threshold low images acquired using the PCD-CT system for different tube potential (e.g. 140 kV and 120 kV) with equal total radiation dose level were consistent with a previous study (Gutjahr et al 2016).

Figure 8.

(a) SECT images (1st column) and vSECT images (2nd column) for small bowel imaging; (b) noise level comparison between SECT and vSECT images.

Table 3.

Summary of contrast/noise in SECT/vSECT images for small bowel imaging (unit: HU).

		Sample 1	Sample 2	Sample 3	Sample 4	Sample 5
Arterial enhancement	SECT (140kV)	191.8 ± 18.4	97.7 ± 20.4	298.6 ± 18.6	214.3 ± 16.9	96.3 ± 18.1
	SECT (120kV)	223.2 ± 17.8	115.5 ± 19.8	347.2 ± 20.9	247.7 ± 18.4	108.9 ± 20.6
	vSECT (120kV)a	230.0 ± 44.1	113.1 ± 39.0	351.5 ± 40.9	260.4 ± 37.6	113.0 ± 42.7
Enteric enhancement	SECT (140kV)	464.2 ± 22.0	344.9 ± 22.3	226.6 ± 21.0	338.4 ± 20.1	230.1 ± 19.4
	SECT (120kV)	466.2 ± 19.9	349.0 ± 18.7	227.3 ± 22.8	343.5 ± 19.7	233.3 ± 20.8
	vSECT (120kV)a	462.7 ± 39.0	353.8 ± 37.6	222.8 ± 35.8	328.3 ± 37.8	221.1 ± 35.2

^avSECT images at 120 kV were converted from 140 kV PCD acquisition.

4. Discussion

Quantitative dual-contrast imaging using MECT has been actively explored in phantom and animal studies (Anderson et al 2010, Taguchi and Iwanczyk 2013, Muenzel et al 2016, 2017a, Cormode et al 2017, Symons et al 2017a, 2017b, Dangelmaier et al 2018, Panta et al 2018, Ren et al 2018a, 2018b, Stayman and Tilley 2018, Yu et al 2018), many of which claimed that by reducing the number of scans, radiation dose can be reduced using single-scan MECT compared to multi-scan SECT (Muenzel et al 2016, 2017a, Cormode et al 2017, Symons et al 2017a, 2017b). Using a research whole-body PCD-CT system, we investigated the dose efficiency of MECT for two potential clinical imaging tasks that may benefit from simultaneous dual-contrast imaging: biphasic liver imaging with iodine and gadolinium, and small bowel imaging with iodine and bismuth, in a phantom study. It was found that when the total radiation dose is matched between the two protocols, SECT_2s and MECT_1s, the vSECT images generated from MECT_1s had much higher noise compared with the SECT images acquired with SECT_2s at the same beam energy. These results provided strong evidence that MECT for simultaneous dual-contrast imaging is dose inefficient compared with SECT.

One important finding in this study is that the dose efficiency of MECT dual-contrast imaging is highly dependent on the diagnostic task and the contrast materials involved. The noise in vSECT images is increased by 203%/278% and 110%/82% for iodine/gadolinium and iodine/bismuth quantifications, respectively. These correspond to dose increases by 819%/1328% and 230%/340% in MECT, compared to SECT, to achieve the same image noise for biphasic liver imaging and small bowel imaging, respectively.

This difference can be explained by the atomic number and k-edge energy difference between the first and the second contrast materials. The first contrast material in both tasks is iodine, which has an atomic number of 53 and a k-edge energy of 33.2 keV. The atomic number of gadolinium is 64 and k-edge is 50.2 keV, which are much closer to those of iodine compared with bismuth (atomic number: 83; k-edge: 90.5 keV) to iodine. The further apart in atomic number, the better the materials’ spectral distinction and noise properties of the material decomposition (Kelcz et al 1979). Therefore, the material decomposition for iodine and gadolinum magnifies image noise much higher than for iodine and bismuth.

To make a fair comparison between MECT and SECT, we used the optimal x-ray tube voltage and threshold settings for each dual-contrast pair on the current PCD-CT system. For the iodine/gadolinum pair, the optimal tube potential and threshold settings were 80 kV with the four energy thresholds at 25, 35, 50, and 55 keV. For the iodine/bismuth pair, they were 140 kV with 25, 50, 75, and 90 keV. In general, compared to gadolinium (50.2 keV) in biphasic liver imaging, the k-edge of bismuth (90.5 keV) in small bowel imaging can be better utilized, in that reasonable amounts of x-ray photons are distributed below and above the bismuth’s k-edge, as demonstrated in figure 4. As a result, the dose-efficiency of MECT with iodine and bismuth is much better than with iodine and gadolinium.

The dose inefficiency of dual-contrast imaging using MECT can be partially attributed to the severe spectral distortions on the current PCD-CT platform caused by many physical non-idealities (e.g. K-escape, pulse pile-up, charge sharing, etc), as shown in figure 4 (Shikhaliev et al 2009, Taguchi and Iwanczyk 2013, Ren et al 2018b). Though distorted, a fair amount of x-ray photons are still allocated below and above the k-edges of gadolinium and bismuth used in biphasic liver imaging and small bowel imaging, respectively. Therefore, the contrast materials can still be separated and quantified. Corrections for these non-ideal effects are ongoing research topics in both correction algorithms and PCD technologies. Evaluation of the radiation dose efficiency with corrected PCD responses warrants future studies.

A generic image-based material decomposition method was used without incorporating any denoising algorithms either before or within the material decomposition process. This was by design to evaluate the intrinsic dose efficiency of the MECT system for simultaneous imaging of two contrast agents. Incorporating a denoising algorithm into the material decomposition process may reduce image noise in the vSECT images, but these images do not reflect the fundamental properties of the MECT material decomposition due to the non-linear operations, particularly the use of the regularization term in the objective function of iterative reconstruction (Tao et al 2018, Tivnan et al 2019, Yao et al 2019a, 2019b). Denoising algorithms using adaptive filters and redundant information in the images and iterative reconstruction methods can also be applied to SECT images (Li et al 2014). To have a fair comparison, no denoising algorithms were applied to either the SECT or MECT images.

A 35 cm wide water tank was used in the current study to mimic the attenuation of a medium-size patient. The impact of phantom size on dose efficiency of dual-contrast imaging has not been investigated, but is predictable. For biphasic liver imaging with iodine and gadolinium, a smaller phantom size (lateral dimension < 30 cm) has less beam hardening effect, thus more x-ray photons could be allocated below gadolinium’s k-edge for the same kV and energy threshold settings (i.e. 80 kV with 25, 35, 50, and 55 keV), thus providing better material decompositon performance. For a larger phantom size (lateral dimension > 45 cm), beam hardening effect is strong and furthermore, a higher kV such as 100 kV or even 120 kV instead of 80 kV needs to be used, which significantly degrades material decompositon performance of iodine and gadolinium (Tao et al 2019b). For small bowel imaging with iodine and bismuth, however, the impact of phantom size and associated beam hardening effect may not be as signifcant as in biphasic liver imaging with iodine and gadolinium, as long as the kV and threshold settings in PCD-CT could be properly determined to capature the k-edge of bismuth.

In the single-scan MECT protocol, a virtual non-contrast image can be generated along with the contrast-material specific images. According to the biphasic liver CT protocols (multi-scan SECT protocol) in our institute, however, the non-contrast scan is optional. This is why we only doubled the radiation dose in MECT_1s relative to that in SECT_2s for image quality comparison and dose efficiency evaluation. In the case of prescribing a non-contrast scan, the radiation dose in MECT_1s should be tripled for a fair comparison to SECT_2s (strictly speaking, the SECT protocol now refers to as SECT_3s: three-scan SECT protocol) and the noise increase and dose efficiency could be predicted based on the current results. Specifically in the current study, the total radiation dose for MECT_1s should be increased from 46 mGy to 69 mGy, corresponding to dose increase by 513% and 852% (819% and 1328% for doubled dose) in biphasic liver imaging, and 193% and 120% (340% and 230% for doubled dose) in small bowel imaging. The above calculations indicate that the assumption of prescribing a non-contrast scan would definitely improve the dose efficiency for both imaging tasks, but not change the conclusion made in the current study.

One potential disadvantage of applying single-scan MECT protocol, particularly in biphasic liver imaging with iodine and gadolinium, is the doubled dose of contrast, compared to the multi-scan SECT protocol (Muenzel et al 2017b, Symons et al 2017a). The increased contrast dose level may be a concern due to contrast toxicity. Gadolinium has not been proven to be a viable CT contrast yet, given the fact that higher gadolinium contrast dose is required compared to that used in MRI scan, in which however, the use of gadolinium has caused controversies due to retention (Layne et al 2018). The potential adverse effects of injecting iodine and gadolinium consecutively in the same imaging session at CT-specific contrast doses is currently unknown. A thorough investigation is necessary to assess whether the benefits of such a dual-contrast approach outweigh the risks for biphasic liver imaging in patients.

The findings in this study clarified that the reduction in radiation dose cannot be rendered as a potential benefit of dual-contrast imaging using MECT. To achieve the same target image quality, SECT with multiple scans may have a better dose efficiency. However, MECT-based dual-contrast imaging provides certain advantages over multi-scan SECT. A notable benefit is the perfect or near perfect co-registration between different enhancement phases captured in a single MECT acquisition, consequently improving the spatial fidelity of material maps pertinent to dual-contrast applications. The dose efficiency of simultaneous imaging of two contrast agents were evaluated for the above two imaging tasks using dual-energy CT (DECT) in a phantom study, also demonstrating radiation dose inefficiency compared to the SECT protocol (Ren et al 2019b).

5. Conclusion

Quantitative dual-contrast imaging using a research whole-body PCD-CT system was demonstrated for two potential clinical imaging tasks: biphasic liver imaging with iodine and gadolinium, and small bowel imaging with iodine and bismuth. When the total radiation dose level is matched to that in a SECT protocol, however, MECT for dual-contrast imaging resulted in higher image noise than SECT in both imaging tasks. The dose efficiency of MECT for dual-contrast imaging is task and contrast-material dependent, with iodine/bismuth imaging better than iodine/gadolinium imaging, but both applications are dose-inefficient than its corresponding two-scan SECT imaging due to the substantial increase in image noise.