Multi-energy CT imaging for large patients using dual-source photon-counting detector CT

Abstract

Multi-energy CT imaging of large patients with conventional dual-energy (DE)-CT using an energy-integrating-detector (EID) is challenging due to photon starvation-induced image artifacts, especially in lower tube potential (80–100 kV) images. Here, we performed phantom experiments to investigate the performance of DECT for morbidly obese patients, using an iodine and water material decomposition task as an example, on an emulated dual-source (DS)-photon-counting-detector (PCD)-CT, and compared its performance with a clinical DS-EID-CT. An abdominal CT phantom with iodine inserts of different concentrations was wrapped with tissue-equivalent gel layers to emulate a large patient (50 cm lateral size). The phantom was scanned on a research whole-body single-source (SS)-PCD-CT (140 kV tube potential), a DS-PCD-CT (100/Sn140 kV; Sn140 indicates 140 kV with Sn filter), and a clinical DS-EID-CT (100/Sn140 kV) with the same radiation dose. Phantom scans were repeated five times on each system. The DS-PCD-CT acquisition was emulated by scanning twice on the SS-PCD-CT using different tube potentials. The multi-energy CT images acquired on each system were then reconstructed, and iodine- and water-specific images were generated using material decomposition. The root-mean-square-error (RMSE) between true and measured iodine concentrations were calculated for each system and compared. The images acquired on the DS-EID-CT showed severe artifacts, including ringing, reduced uniformity, and photon starvation artifacts, especially for low-energy images. These were largely reduced in DS-PCD-CT images. The CT number difference that was measured using regions-of-interest across field-of-view were reduced from 20.3 ± 0.9 (DS-EID-CT) to 2.5 ± 0.4 HU on DS-PCD-CT, showing improved image uniformity using DS-PCD-CT. Iodine RMSE was reduced from 3.42 ± 0.03 mg ml⁻¹ (SS-PCD-CT) and 2.90 ± 0.03 mg ml⁻¹ (DS-EID-CT) to 2.39 ± 0.05 mg ml⁻¹ using DS-PCD-CT. DS-PCD-CT out-performed a clinical DS-EID-CT for iodine and water-based material decomposition on phantom emulating obese patients by reducing image artifacts and improving iodine quantification (RMSE reduced by 20%). With DS-PCD-CT, multi-energy CT can be performed on large patients that cannot be accommodated with current DECT.

1. Introduction

Multi-energy CT has demonstrated its unique value in clinical practice by enabling various applications such as iodine quantification and virtual non-contrast imaging (Graser et al 2009b, Chandarana et al 2011, Marin et al 2014, Mccollough et al 2015, Fulwadhva et al 2016). Conventional CT systems typically utilize an energy-integrating detector (EID). Due to the energy integrating nature of this type of detector technology, the signal generated with EID is proportional to the energy of all received photons, while the energy information of each individual photon is lost. Consequently, a multi-energy CT system employing EID usually acquires multi-energy data by scanning with different tube potentials in order to create sufficient energy separation between different x-ray spectra (Mccollough et al 2015). Various multi-energy CT implementations have been developed, such as fast kV switching (Kalender et al 1986, Xu et al 2009) and the dual-source dual-energy approach (Flohr et al 2006). Other techniques that require single kV acquisition were also developed, such as split-beam filtration (Almeida et al 2017) and dual-layer detectors (Carmi et al 2005).

Despite the unique benefits in various clinical applications, it remains a challenge to perform multi-energy CT acquisition on morbidly obese patients (Graser et al 2009a, Modica et al 2011, Patino et al 2016). This is in part because the high attenuation of very large patients can lead to various image artifacts, such as photon starvation, beam hardening, ring and streak artifacts, as well as image non-uniformity. These artifacts substantially impact CT number accuracy, and consequently degrade multi-energy performance (e.g. accuracy of material quantification). Image artifacts are especially significant for images acquired with lower tube potentials such as 80 kV or 100 kV (Guimaraes et al 2010). In addition, due to substantial x-ray attenuation caused by large patients, the incident photon flux into detectors is relatively low. Hence, electronic noise may have a larger contribution to the overall noise compared to images from a small or medium-sized patient. Although a dual-layer detector may have more photons in the low energy data set, the spectral separation between the low and high energy data still suffer from considerable overlap due to the beam hardening effect, especially for very large patients (Atwi et al 2019). For these reasons, dual-energy CT application is typically not suitable for large patients.

Compared to the EID, photon-counting detectors (PCD) can resolve the energy of incident photons and are therefore capable of multi-energy imaging with a single kV scan. On a PCD-CT, a set of energy thresholds (typically 2 to 8) can be selected and then used to generate a series of energy bin data sets, with each bin data set containing photons of energy between two nearby thresholds. Recently, PCDs have been introduced to human CT imaging (Kappler et al 2014, Gutjahr et al 2016a, Yu et al 2016c, Leng et al 2018). A series of studies have demonstrated that PCD-CT has many advantages over the conventional EID-CT, such as improved iodine contrast due to the removal of energy weighting in signal detection (Gutjahr et al 2016a), capability of radiation dose-efficient high-resolution imaging (Leng et al 2018), reduced electronic noise and improved CT number stability in low-radiation-dose scenarios (Yu et al 2016a, Symons et al 2017). However, the energy discriminating capability of PCD suffers from non-ideal physical effects, such as charge sharing, K-escape, and pulse pileup, which can reduce the energy separation between acquired energy bin data, and therefore compromise multi-energy imaging performance (Kim et al 2011, Taguchi and Iwanczyk 2013, Koenig et al 2014). Currently, whole-body prototype PCD-CT systems have been developed by clinical CT system providers and are undergoing clinical evaluation. Several studies performed on these systems have demonstrated that PCD-CT could yield improved imaging performance compared to conventional EID-CT for various clinical applications such as temporal bone, sinus, kidney stone, and lung imaging (Symons et al 2017, Marcus et al 2018, Bartlett et al 2019, Leng et al 2019, Rajendran et al 2020).

By combining PCD technology with a conventional multi-energy imaging technique, i.e. the dual-source approach, there is potential to further improve energy separation among the acquired multi-energy data sets (Faby et al 2015, Tao et al 2019a, 2019b). The use of different tube potentials for each source can create an initial separation between high and low energy spectra. Additional filtration such as a tin filter can also be applied to the high tube voltage scans to further improve energy separation, mirroring conventional dual-source dual-energy CT (Primak et al 2010). With dual-source (DS)-PCD-CT, more energy bin data can be acquired. For a PCD system allowing two energy bin acquisition, a total of four energy bin data sets can be acquired with the DS-PCD approach. In addition, PCDs have better resistance to electronic noise as well as better CT number stability at low photon flux regimes which may be more beneficial for large patient imaging (Yu et al 2016a, Symons et al 2017).

In this work, we investigate the feasibility of multi-energy CT imaging on large sized patients using an emulated dual-source (DS)-PCD-CT. Phantom experiments were performed on a research whole-body PCD-CT system to evaluate image artifact and investigate the performance of iodine and water based material decomposition. The results were further compared with that of a clinical EID-based dual-source dual-energy (DSDE) CT system to demonstrate the advantage of DS-PCD-CT.

2. Materials and methods

2.1. Research whole-body PCD-CT

Phantom experiments were performed on a research whole-body PCD-CT system (SOMATOM CounT, Siemens Healthcare GmbH, Erlangen, Germany) (Kappler et al 2014, Gutjahr et al 2016). The system has been previously evaluated and technical details of this system can be found in recent publications (Yu et al 2016c, Leng et al 2018). This is a single-source (SS) PCD-CT system model based on a second-generation DS-EID-CT (SOMATOM Definition Flash, Siemens Healthcare GmbH), with the detector of the second sub-system replaced with a PCD unit. The EID subsystem and PCD subsystems are positioned about 90° apart from each other (referred to as Tube A and Tube B systems, respectively), and operate independently. The x-ray tube of the PCD sub-system is identical to that of the conventional dual-source dual-energy CT of the same model, with the capability of including an additional 0.4 mm tin filtration to harden the x-ray beam. The tube potentials available on the PCD subsystem (i.e. 80, 100, 120 and 140 kV) are the same as that on the tube A system of the second-generation dual-source CT. The PCD subsystem allows for high photon flux with a tube current up to 550 mA at 140 kV, sufficient for most whole-body acquisitions. The PCD sub-system has a field-of-view (FOV) of 27.5 cm. The limited FOV compared to the 50.0 cm FOV of a typical CT system is due to the prototype nature of this system but it is not a fundamental limitation of PCD technology (Yu et al 2016c). To scan objects larger than the FOV of the PCD, a low-radiation-dose data completion scan using the EID sub-system can be performed to prevent truncation artifact in the PCD images (Yu et al 2016b).

The research PCD subsystem provides four data acquisition modes by combining detector pixels and using different parings of energy thresholds, which are Macro mode, Chess mode, Sharp mode, and ultra-high-resolution (UHR) mode. The Macro, Sharp and UHR mode allow two energy thresholds and are capable of generating two energy bin data sets. The main differences between these modes are the longitudinal collimation and detector readout pixel size. The Sharp and UHR modes provide narrower longitudinal collimation (12/8 mm for Sharp/UHR vs. 16 mm for Macro/Chess) and smaller detector pixel size (0.25 mm for Sharp/UHR vs. 0.5 mm for Macro at isocenter), which are designed for high resolution imaging applications. The Chess mode is capable of generating four energy bin data sets, and is primary used for applications involving multi-contrast imaging. Details of this system were described in previous publications (Yu et al 2016c, Leng et al 2018).

All the experiments in the current work were performed using the Macro mode with two configurable energy thresholds (i.e. low-energy threshold (TL) and high-energy threshold (TH)), 0.5 mm detector size at isocenter, and 32 × 0.5 = 16 mm longitudinal collimation. Macro mode was selected in this study because our main goal was to evaluate the iodine quantification performance for abdomen CT applications, and compare it with that of conventional EID-CT. Macro mode provides detector pixel size that are closer to that of EID-CT (0.6 mm detector size and 38.4 mm collimation).

2.2. Experiments

Numerical simulations were performed to evaluate the spectral separation of the multi-energy CT data sets with SS-PCD-CT and DS-PCD-CT using vendor-provided simulation tools (Li et al 2017). The energy spectra were obtained after attenuation by 50 cm water to simulate a scan of a large sized patient, and then compared with that of a clinical DS-EID system.

For phantom experiments, an abdominal phantom (QRM, Moehrendorf, Germany) with extension rings was scanned on the PCD-CT system. The phantom has a lateral dimension of 40 cm, and was further wrapped in soft tissue-equivalent gel layers to emulate a very large patient of 50 × 40 cm (left-right × anterior-posterior) dimension. Figure 1 shows the CT image of this phantom. A series of iodine inserts of 2, 5, 10, and 15 mg ml⁻¹ iodine (Gammex Inc. Middleton, WI) were also scanned together with the abdominal phantom.

Figure 1.

CT image showing a cross-section of the phantom used in this work. It consists of an abdominal phantom with 40 × 30 cm dimension which was wrapped in soft tissue-equivalent gel layers to emulate a very large patient of 50 × 40 cm dimension. A series of iodine inserts of 2, 5, 10, and 15 mg ml⁻¹ iodine (Gammex Inc. Middleton, WI) as well as solid water inserts were included in the central space of the phantom. The dashed circles show the regions-of-interest (ROI) used for artifact assessment.

The phantom was scanned on a clinical DS-EID-CT system (SOMATOM Definition Flash, Siemens Healthcare GmbH), the same platform that the PCD-CT was built on. On the DS-EID-CT, the phantom was scanned using a liver VNC protocol with 100 kV and 140 kV (with 0.4 mm tin filtration) for each subsystem; a longitudinal collimation of 32 × 0.6 mm, spiral pitch = 0.4, and a total radiation dose of CTDI_vol = 58.24 mGy. The CTDI_vol was determined based on the system’s dual-energy VNC base protocol (tube A/B potential = 100/Sn140 kV, CareDose quality reference mAs = 230/178 mAs for tube A/B). The protocol, including spiral pitch, was modified to increase the total radiation dose according to the large patient size while keeping the tube power under the system limit. Note that a 100 kV tube potential instead of 80 kV was used for the lower kV acquisition to reduce image artifacts.

Next, the same phantom was scanned on a PCD-CT system, assuming a SS-PCD architecture with 140 kV tube potential, TL/TH = 25/75 keV, a longitudinal collimation of 32 × 0.5 mm, spiral pitch = 0.4, and radiation dose of CTDI_vol = 58.43 mGy, similar to that of the EID scan. The use of 140 kV tube potential and energy thresholds of 25/75 keV provide a balance of photon flux in the low and high energy bins compared to 120 kV tube potential and other energy thresholds, and provide better iodine quantification performance based on our experience, which were therefore adopted in this work (Leng et al 2017).

Finally, to emulate a DS-PCD architecture, the phantom was scanned twice on the same PCD-CT using different tube potentials of 100 and 140 kV with energy threshold of TL/TH = 25/70 keV for 100 kV and 25/75 keV for 140 kV (0.4 pitch). A 0.4 mm tin filtration was included during the 140 kV scan. The energy thresholds were empirically determined. The use of 70 keV for the 100 kV tube, instead of 75 keV as used for Sn140 kV tube, was to include more photons for the high energy bin. The total radiation dose (CTDI_vol = 58.48 mGy) of the DS-PCD scan was the same as that of the DS-EID and SS-PCD scans. The radiation dose partitioning between the two individual sub-systems was the same as that of the DS-EID-CT (radiation dose ratio between low/high energy tubes = 1:0.67). The phantom was scanned on each system (SS-PCD, DS-PCD and DS-EID-CT) and was repeated five times.

2.3. Data processing

All images acquired with each system architecture were reconstructed using vendor-provided reconstruction tools with the same quantitative medium-smooth D30 kernel (image matrix = 512, FOV = 27.5 cm, slice thickness = 5 mm) (Stierstorfer et al 2004). The image reconstruction platform for PCD-CT is provided by the system vendor and provides matching performance in terms of spatial resolution compared to the conventional clinical kernel. For SS-PCD-CT, two energy bin data sets can be generated (i.e. bin 1 and bin 2), while for DS-PCD-CT, a total of four energy bin data set can be obtained. For DS-EID-CT, the energy low and energy high data sets are generated. The energy bin images (PCD-CTs) and energy low/high images (EID-CT) were then used to perform iodine and water based material decomposition using an image domain, least-squares-based approach (Yu et al 2018). No denoising was performed in order to evaluate the performance of raw output images using different system architectures. The iodine quantification RMSE, which combines estimation bias and noise, was then calculated using images acquired on SS-PCD-CT, DS-EID-CT, and DS-PCD-CT for comparison. Statistical comparison of iodine quantification RMSE between DS-EID-CT and DS-PCD-CT was performed based on a (one-sided) student t-test with the null hypothesis that the iodine RMSE of DS-PCD-CT is no lower than that of DS-EID-CT. Since the material decomposition techniques available on commercial CT systems usually employ denoising techniques, the RMSE values reported here are generally higher compared to those reported by system vendors. The reconstructed images from all systems were also visually examined for image artifacts.

To quantitatively evaluate CT image uniformity and artifact on DS-EID-CT and DS-PCD-CT systems, the CT numbers of regions-of-interest (ROI) at different locations of the phantom (ROIs shown in figure 1) were measured from the low (100 kV) and high energy (Sn140 kV) images acquired on DS-EID-CT. The maximal CT number difference among different ROIs was then calculated for each of the five repeated phantom scans. For comparison, the ROI CT numbers were also measured from the threshold-low images acquired with the same tube potentials (100 kV and Sn140 kV) on the DS-PCD-CT. The maximal CT number difference among different ROIs in the DS-PCD-CT images was then calculated and compared with that of DS-EID-CT. Statistical comparison of the ROI CT number differences between DS-EID-CT and DS-PCD-CT systems was then performed based on a one-sided student t-test, with the null hypothesis that the ROI CT number difference in DS-PCD-CT images is no lower than that of DS-EID-CT. Note that rejecting the null hypothesis suggests that the DS-PCD-CT has lower CT number difference among ROIs than DS-EID-CT, hence reduced image artifact and better uniformity.

3. Results

Figure 2(a) shows the energy spectra of the low and high energy channels from DS-EID-CT obtained using numerical simulations. The energy spectra of energy bin data sets obtained from SS-PCD and DS-PCD acquisitions are shown in figures 2(b)–(c), respectively. Note that the spectral responses of the EID and PCD system shown in figure 2 account for both the distribution of input x-ray photon flux and the response function of the detector which is different between EID and PCD. Also note that PCD can eliminate the non-uniform energy weighting that occurs on EID, due to its ability to count each individual photon, which explains the difference of detector response between EID and PCD (Taguchi and Iwanczyk 2013, Leng et al 2019). The mean energy levels of different energy bin data sets are also shown. Due to an imperfect PCD response, the difference in mean energy levels between bin 1 and bin 2 of SS-PCD (104.9 keV—83.8 keV = 21.1 keV) is smaller compared to the DS-EID scan (104.8 keV—75.7 keV = 29.1 keV). The mean energy levels of each energy bin for the DS-PCD scan were 70.8/83.6 keV for bin 1/2 from the 100 kV acquisition, and 94/107.7 keV for bin 1/2 from the Sn 140 kV acquisition. The energy difference between the lowest and highest energy bin was 107.7 keV—70.8 keV = 36.7 keV, higher than that of the SS-PCD-CT DS-EID-CT. Therefore, a DS-PCD-CT approach can provide energy spectra with wider spread compared to the other two systems.

Figure 2.

(a) Normalized energy spectra of low and high energy channels of the dual-source (DS)-EID acquired with 100 kV and 140 kV with tin filtration (Sn 140 kV). The mean energy levels of the low and high energy channels are 75.7 keV and 104.8 keV. (b) Energy spectra of the energy bin data sets acquired on a single-source (SS) PCD with 140 kV. The mean energy levels of bin 1 and 2 data are 83.8/104.9 keV. (c) Energy spectra of the energy bin data sets acquired on a DS-PCD with 100/Sn140 kV tube potentials. The mean energy levels of bin 1 and bin 2 acquired with 100 and Sn 140 kV are 70.8/83.6 keV and 94/107.7 keV, respectively.

Figure 3 shows low energy and high energy CT images acquired on DS-EID-CT compared to the energy threshold-low (TL) images acquired on DS-PCD-CT with the same total radiation dose. The low-energy image acquired on DS-EID-CT exhibits substantial image artifacts, including shading and ringing, and consequently, the image uniformity was compromised. In contrast, the images acquired on the PCD-CT show reduced image artifacts and improved uniformity, especially for the low-energy acquisition (100 kV). The high energy images are more immune to these artifacts for both systems. Table 1 shows the CT number measurements using different regions-of-interest (ROI) across the image. The EID-CT images show a substantial CT number variation among different ROIs, especially for low-energy acquisition. The largest difference between the ROIs was observed to be larger than 20 HU (ROIs 3 vs. 4). On the other hand, the ROI measurements among different ROIs in DS-PCD images were more uniform, with the largest difference in low kV acquisition around 2.5 HU. The CT number differences among ROIs were calculated from five repeated phantom scans. The differences were found to be 20.3 ± 0.9 HU for DS-EID-CT and 2.5 ± 0.4 HU for DS-PCD-CT. Additionally, the t-test results show that the ROI CT number difference was significantly lower for DS-PCD-CT compared to that of DS-EID-CT (p < 0.05), showing the better image uniformity of DS-PCD-CT.

Figure 3.

Examples of CT images acquired on DS-EID-CT and DS-PCD-CT with the same tube potential and matched radiation dose. (a), (b) Low and high energy images acquired with 100 kV and 140 kV (with Sn filtration) tube potential on the DS-EID-CT. (c), (d) Threshold-low images acquired with low and high tube potential of 100 kV and 140 kV (with Sn filtration) on the DS-PCD-CT. Note the image artifacts in the low-energy image (a) acquired on EID-CT, such as ringing and shading (arrows). The bony structure is also blurred (arrow head). These artifacts are reduced in the PCD image (c).

Table 1.

Artifact assessment using the mean CT number values measured from different ROIs within the image plus/minus the standard deviation of the mean CT numbers over five repeated scans. The ROI locations are marked in figure 1.

	ROI 1 (HU)	ROI 2 (HU)	ROI 3 (HU)	ROI 4 (HU)
EID (100 kV)	13.0 ± 0.3	25.5 ± 0.3	26.9 ± 0.7	6.6 ± 0.3
EID (Sn 140 kV)	38.4 ± 0.3	42.1 ± 0.2	37.9 ± 0.2	38.6 ± 0.4
PCD (100 kV)	23.1 ± 0.1	22.3 ± 0.3	21.5 ± 0.1	20.6 ± 0.3
PCD (Sn 140 kV)	33.0 ± 0.4	33.3 ± 0.3	32.5 ± 0.3	31.0 ± 0.3

Figure 4 shows the iodine- and water-specific images generated using DS-EID-CT images and DS-PCD-CT data sets. Image artifacts from the source CT images can propagate into material-specific images as seen in the iodine image (see arrows in figure 4(a)). The iodine-specific images generated using the DS-PCD images, on the other hand, are more uniform and suffer less from image artifacts. The iodine image generated from the DS-EID images (figure 4(a)) also appears to be patchy and blurred compared to the PCD results (figure 4(c)). Note that the bone structure in figure 4(b) is blurred (arrow head) compared to 4(d).

Figure 4.

Examples of iodine (a), (c) and water-specific images (b), (d) generated using images acquired on DS-EID-CT (a), (b) and DS-PCD-CT (c), (d). Artifacts, including shading and streaking (see arrows) as well as blurring of bone (arrow head), appeared in the results of DS-EID-CT (a), (b). These effects are reduced in the DS-PCD-CT results (c), (d).

Iodine quantification using images acquired on DS-EID-CT, SS-PCD-CT and DS-PCD-CT with matched total radiation dose were measured from iodine inserts of 2, 5, 10, and 15 mg ml⁻¹ concentrations. The iodine quantification RMSE values were 2.90 ± 0.03, 3.42 ± 0.03, and 2.39 ± 0.05 mg ml⁻¹ for DS-EID-CT, SS-PCD-CT and DS-PCD-CT, respectively. The SS-PCD-CT has a higher RMSE than DS-EID-CT due to the non-ideal energy response of PCD-CT, which compromises the energy separation between the two energy bin spectra, as shown in figure 2(b) vs. 2(a). The DS-PCD has the lowest iodine quantification RMSE. The iodine RMSE of DS-EID-CT was significantly higher than that of the DS-PCD-CT (p < 0.05), showing the improvement using DS-PCD-CT. Bin 2 of the low energy tube (i.e. 100 kV) acquisition only represents a small portion of photons; therefore it may not contribute significantly to the outcome of material decomposition. The iodine RMSE without that data was 2.40 ± 0.05 mg ml⁻¹, a marginal increase from that of using bins 1 and 2 of low energy tube.

4. Discussion

In this work, we performed numerical simulations and phantom studies to investigate the combination of conventional dual-source dual-energy CT method with PCD technology to conduct dual-energy CT imaging for large patients, and compare the performance of the resultant DS-PCD system architecture with conventional DS-EID CT as well as SS-PCD CT. Our results demonstrate that a DS-PCD approach can reduce image artifacts as compared to DS-EID-CT, especially for lower energy images. This can be attributed to the reduced effect of electronic noise of PCD and its improved CT number stability at low photon flux. The reduced image artifact using PCD is especially relevant since the severe artifacts observed on very large patients typically prohibit the use of dual-energy CT on this population. Note that we used 100 kV for the lower kV scan on the DS-EID-CT, which provides better resistance to image artifact. The use of lower tube potential such as 80 kV would suffer from more severe artifacts. Although we focused on comparison with a dual-source based dual-energy EID-CT, other EID-based dual-energy technologies, such as fast-kV switching and dual-layer detectors, also suffer similar limitations for imaging large patients. Our results demonstrated the feasibility of multi-energy CT imaging for large patients using DS-PCD-CT.

As a result of imperfect detector response of PCD, iodine quantification performance using SS-PCD data was worse compared to the DS-EID-CT. As shown in the simulation, the energy separation between the two energy bins are lower compared to that between the low and high energy channels of DS-EID. This resulted in a higher RMSE for SS-PCD (3.4 mg ml⁻¹ for SS-PCD and 2.9 mg ml⁻¹ for DS-EID). The combination of DS acquisition and PCD technique allows broader energy spectra as well as more energy bin data to be generated, which was shown to benefit iodine quantification by reducing the RMSE to 2.4 mg ml⁻¹. In this work, we use the same radiation dose for each sub-system as the DS-EID based on existing clinical protocols. Further optimization of radiation dose partitioning, as well as energy threshold setup for each PCD sub-system may further improve decomposition performance. In addition, image denoising is typically performed prior to or coupled with material decomposition in practice, which may have further reduced the RMSE while retaining the inherent benefits from PCD for artifact reduction. In this study, we used conventional filtered back projection (FBP) reconstruction and least-square-based material decomposition algorithm to generate the material specific images, which are linear algorithms. This was done to compare the performance of different systems without bringing in confounding factors such as iterative image reconstruction and image denoising. The benefit of the proposed system can be further improved using advanced image processing and reconstruction methods which can enhance image quality and results of material decomposition. These include iterative image reconstruction and multi-spectral reconstruction utilizing regularizations that force sharing structural information between different energy bins (Sawatzky et al 2014, Geyer et al 2015, Xi et al 2015, Wang et al 2016, Kazantsev et al 2018, Wu et al 2018).

In this study, we used a body shaped phantom (i.e. abdomen QRM phantom) with internal structures and presence of air gap between inserts, rather than a homogeneous phantom, as used in previous works evaluating the iodine quantification on clinical DS-EID-CT (Jacobsen et al 2018). The phantom used in this study allowed us to emulate a very large patient that may be encountered in clinical practice, compared to the phantom used by Jacobsen et al which emulates a medium to large size patient. Although image artifact can arise as a consequence of the phantom configuration compared to a homogeneous phantom, it is our intension to compare the severity of these artifacts between images acquired on different systems.

This study was performed on a prototype PCD-CT system that is currently available for whole-body imaging. With further development of PCD technology in the future, the performance of both SS-PCD-CT and DS-PCD-CT may be improved. In addition, the use of spectral domain correction on multi-energy CT data may also improve the performance of material decomposition (Lee et al 2017).

We focused on comparing with the DS-PCD-CT with Siemens Somatom Definition Flash system in this study because the PCD-CT is modeled based on the same system platform. This allowed a direct comparison between the performances of different detector technologies, i.e. PCD vs EID. The performance of multi-energy CT can be improved on the newer Siemens Somatom Definition Force system as compared to Flash systems, which uses a different x-ray tube and generator, as well as a higher kV with a thicker Sn filter (0.6 mm Sn filter on Force vs. 0.4 mm Sn on Flash). In theory, the same set of hardware as that used on Force system may also improve the performance of PCD and DS-PCD.

One limitation of this study is that the effect of cross-scattering between two sub-systems is not considered in the phantom experiment. This is due to the unavailability of a true DS-PCD-CT system. However, correction techniques can be applied to effectively suppress cross-scattering effects, as done in current DS dual-energy CT (Bruder et al 2008).

5. Conclusion

In this work, we investigated the feasibility of multi-energy CT imaging for large patients using DS-PCD-CT, and compared its performance with conventional DS dual-energy CT. The results of our phantom study demonstrated that the DS-PCD-CT can reduce image artifact compared to the DS-EID-CT, and improve iodine quantification RMSE by 20%. With DS-PCD-CT, multi-energy CT can be reliably performed on large patients that cannot be accommodated with current DECT technology.