Dual-energy computed tomography (DECT) involves acquisition of two simultaneous datasets through the use of two different X-ray energy spectra, which in turn allows material decomposition analysis [1, 2]. Decomposition analysis is achieved through spectral separation and by exploiting the differences in CT number or attenuation ratios of two materials at low and high tube potentials (e.g., 80 and 140 kVp) [1, 3]. Tissue characterization through spectral separation was first reported in the late 1970s and early 1980s showing potential use [1, 4]. However, momentum of rapid spread and clinical adoption of early dual-energy technology was stifled by inherent technical limitations, such as noise from low tube potentials and spatial misregistrations between low- and high-energy datasets due to prolonged and separate data acquisitions [1, 3, 4]. However, these limitations were largely mitigated with the subsequent advent of newer CT technologies allowing more rapid data acquisition. Availability of simultaneous (e.g., dual-layer detector DECTs) and near simultaneous (e.g., single source, rapid kilovoltage-switching DECT and dual-source DECT) acquisition technology sparked renewed interest in DECT applications with resultant increase in reports of clinical applications in the literature [1, 3–15].

DECT applications in abdominal imaging have been widely reported, including use in liver, pancreas, renal, and adrenal imaging [1, 3–17]. Despite this and continued advances in the technology, widespread adoption and integration into routine clinical practice has not occurred [1]. This is despite well-reported benefits of increased lesion conspicuity, reduction in iodinated contrast media use, and lower overall exam radiation dose [16, 18–21]. Historically, higher radiation dose and poorer image quality of DECT compared with conventional single-energy CT (SECT) were cited for lack of widespread use. However, recent advances have allowed DECT to have equivalent or reduced radiation dose [through improvements in scan acquisition technology and replacement of true unenhanced (TUE) images with virtual unenhanced (VUE) images] and comparable image quality to SECT (through the use of noise reduction algorithms) [22–28]. Currently, the fee-for-service payment model in some ways disincentivizes the use of DECT. DECT studies with VUE images replacing TUE images cannot be charged as CT examinations requested for without and with intravenous contrast even though the patient receives less total radiation by virtue of eliminating an acquisition (i.e., the TUE). On the contrary, in future value-based alternative payment models, DECT can show tremendous value through reduced total radiation dose and the ability to lower volumes of iodinated contrast due to increased tissue contrast gained from lower energy monochromatic images. Additional barriers for adoption include multiple decision points that exist when considering incorporating DECT in clinical practice ranging from patient and exam type selection for dual-energy scanning as well as questions regarding image reconstruction, routing, and interpretation (Fig. 1). Moreover, contribution of dual-energy scanning on increased CT technologists’ workload must also be considered. Although there are reports of prior experience with successful clinical implementation of DECT into practice, universally accepted workflow standardization does not exist [2]. Moreover, a roadmap is needed to serve as initial reference points for those practices considering adopting DECT into their practice.

Fig. 1
figure 1

Decision tree flowchart showing examples of points of decisions that must be made when considering DECT implementation. CT computed tomography, DECT dual-energy computed tomography, SECT single-energy computed tomography, VMI virtual monochromatic images, PACS picture archiving and communication system

Full size image

The purpose of this study was to therefore standardize patient selection and exam protocol workflow for DECT involving single and multiphasic abdominopelvic exams via a multi-institutional consensus.

Materials and methods

Conference participants and preparation

Twenty-two abdominal radiologists representing nine major academic institutions with expertise in and experience with DECT formed a working group. The purpose of this group was to discuss clinical and research applications of DECT as well as provide a forum for discussions related to DECT optimization and troubleshoot any challenges that may be encountered. The working group met regularly over live webinar sessions (GoToWebinar, Citrix Systems, LLC; Fort Lauderdale, FL). Prior to consensus gathering, two of the authors (B.P. and D.M.) presented a PowerPoint over a single live webinar session to the working group that included background material, including current DECT workflow and potential areas needing standardization. These authors also developed a standardized questionnaire (Qualtrics LLC, Provo, UT) with general questions on patient selection for DECT as well as specific questions on gastrointestinal and genitourinary CT protocol-specific questions (Fig. 2). The survey was reviewed and discussed by the working group over webinar sessions and subsequently revised based on those discussions. The final survey was then emailed to nine individuals (average number of years of DECT experience, 6.1 years; range 1–14 years) who were voluntarily selected to serve as faculty leads responsible for filling out the survey based on local consensus from discussions at their respective institutions.

Fig. 2
figure 2figure 2

Images show questions that were used in the web-based survey (Qualtrics)

Full size image

Consensus scoring

For questions specific to CT examination protocols, two scoring systems were utilized with five-point ordinal scales (Table 1). Survey participants rendered a score for each protocol question using an agreement scale (AS) and a utilization scale (US) (both out of 4 points). The agreement scoring system was chosen to confer the degree of indication or contraindication of dual-energy CT for a particular protocol. The utilization score was chosen to provide information on current practice standards and experience with dual energy for a particular exam protocol. A consensus for the use of DECT was considered when a score of ≥2.5 was achieved for a scale. A 2.5 threshold was chosen as it would represent a majority score (i.e., 2.5 out of 4 total points or 62.5%).

Table 1 Table showing the five-point ordinal scales used in the survey
Full size table

Data collection and discussions

Survey responses from the nine institutions were collected and the data were analyzed. Collective responses from the nine institutions were tabulated and anonymized and presented to the working group over two live webinar sessions. Comments and discussions with salient points for each question and their results were captured.

Results

Demographics

A total of 32 abdominal radiologists from nine major academic institutions participated in formulating survey responses. This included 22 members of the working group and additional abdominal radiologists at each local institution chosen at the discretion of the faculty lead at each institution. The average number of years in practice among the participants was 12.3 years (range 1–35) with an average number of years of experience with DECT of 4.6 years (range 1–14). Four institutions are equipped with rapid kV-switching single-source DECT (rsDECT) scanners, one institution with a dual-source DECT (dsDECT), and four institutions have both platforms.

Patient selection

Seven (78%) institutions perform DECT based on the examination protocol type. Five (56%) institutions exclude large patients either by total body weight (n = 2), transverse dimension based on the frontal scout radiograph (n = 2), or both (n = 1). Weight cutoffs used include 260 lbs (118 kg) (n = 2) and 280 lbs (127 kg) (n = 1). Transverse diameter cutoffs used include 38 cm (n = 1), 42 cm (n = 1), and 46 cm (n = 1).

Image reconstruction and general preferences

For reconstruction of virtual monochromatic images from the acquired DECT datasets, 7 (78%) institutions prefer and send to picture archiving and communication system 50 keV for the low energy level (i.e., high contrast) and 70 keV for medium energy level (i.e., low noise). Few institutions reconstruct high keV virtual monochromatic images (i.e., metal artifact reduction), with energies of 120 keV (n = 1) or 140 keV (n = 3). Other institutions either do not reconstruct high keV virtual monochromatic images, or use metal artifact reduction algorithms or iodine material density images.

Eight (89%) institutions send iodine material density and virtual unenhanced images to the picture archiving and communication system (PACS). Eight (89%) use independent, vendor-specific thin-client software for selective cases depending on the need for advanced visualization and/or quantification of materials. Common examples of use for the thin-client platforms include material quantification (e.g., iodine) (n = 8), image optimization (e.g., select keV for optimal CNR) (n = 6), and image color overlays (n = 6). For DECT examination interpretation, 5 (56%) institutions stated their radiologists prefer using the same approach/order in viewing different series independent of the type of examination. The most common approach was viewing DECT datasets in the following order: high keV images, low keV images, virtual unenhanced images, and iodine material density images.

Examination-specific protocols

Multiphasic liver examination

DECT for multiphasic (i.e., arterial, portal venous, and delayed venous phases) liver CT examinations is indicated (AS 3.67; US 3.44) in the arterial phase (AS 3.78; US 3.22) and participants are performing it (Fig. 3). DECT during both arterial and venous phases is not indicated nor being performed (AS 2.22; US 2.33). Although there was interest in performing DECT during the portal venous or delayed phases for better detection of washout of hepatocellular carcinomas (HCC), most centers are not performing it (AS 2.67; US 2.22). DECT during the venous phases only is not indicated and not performed (AS 1.67; US 0.78). Additionally, participants agreed that a true unenhanced examination in addition to the multiphasic liver DECT dataset was not indicated (AS 1.56; US 0.44) as the VUE images usually suffice.

Fig. 3
figure 3

Graph shows score results of survey for dual-energy CT performed for multiphasic liver exams

Full size image

Dual pancreas examination

Surveyed participants agreed that DECT for dual pancreas is indicated with institutions currently performing it (AS 3.78; US 3.44) during the arterial phase (AS 3.78; US 3.11) (Fig. 4). DECT during both phases, even for certain indications, is not indicated nor are centers performing it (AS 2.44; US 2.33). Additionally, DECT during the venous phase only is not indicated and surveyed institutions do not perform it (AS 2.00; US 1.78). A true unenhanced examination in addition to DECT datasets is not indicated nor performed (AS 1.56; US 1.44).

Fig. 4
figure 4

Graph shows score results of survey for dual-energy CT performed for multiphasic pancreas exams

Full size image

Renal cell carcinoma (RCC) examination

DECT for RCC examinations is indicated and being performed (AS 3.44; US 3.33) in the arterial phase (AS 3.33; US 2.78) (Fig. 5). Participants agreed that DECT is not indicated and is not being performed during both arterial and venous phases (AS 1.89; US 1.78) or during the venous phase only (AS 2.44; US 2.22). DECT plus a true unenhanced examination is indicated and being performed (AS 2.67; US 2.89) as VUE images were not felt to adequately replace TUE images.

Fig. 5
figure 5

Graph shows score results of survey for dual-energy CT performed for RCC exams

Full size image

CT urogram

DECT for CT urography is indicated and performed (AS 3.33; US 3.22) during the nephrographic phase (AS 3.11; US 2.89) (Fig. 6). Additionally, DECT plus a true unenhanced exam is indicated and being performed (AS 2.78; US 2.78). During live discussion, participants agreed that VUE cannot currently completely replace TUE due to limitations of VUE images for renal application (see Discussion section “DECT for Multiphasic GU Exams”). DECT during both the nephrographic and excretory phases is not indicated nor is it being performed (AS 1.78; US 1.78).

Fig. 6
figure 6

Graph shows score results of survey for dual-energy CT performed for CT urogram (CTU)

Full size image

Mesenteric ischemia/gastrointestinal bleeding exams

DECT for mesenteric ischemia and gastrointestinal bleeding is indicated (AS 3.33; US 3.00), particularly during the arterial phase (AS 3.11; US 3.11) (Fig. 7). Additionally, participants agreed that there would be value in performing DECT during both the arterial and venous phases (AS 2.89; US 2.79). Despite high utilization scores, most centers do not have much experience with performing DECT for this examination protocol. Surveyed participants agreed that VUE datasets can replace TUE examinations and that a TUE acquisition in addition to VUE dataset was not indicated (AS 1.78; 2.44).

Fig. 7
figure 7

Graph shows score results of survey for dual-energy CT performed for mesenteric ischemia and gastrointestinal bleeding exams

Full size image

Single-phase exams

DECT for routine single-phase (i.e., portal venous phase) exams for common indications such as abdominal pain or oncology follow-up is indicated (AS 3.00). Only 4 centers perform this while the remainder do not but would like to (US 3.11). During live webinar discussion, barriers that were cited for lack of adoption at respective institutions included the time required for image reconstruction and lack of uniform awareness of utility of DECT for single routine portal venous exams. For renal stone protocol, participants agreed that DECT is indicated (AS 3.00) for stone composition analysis depending on referring urologist preferences at particular institutions. Most centers would like to perform it (US 3.00) but do not based on their discussion with local urology referral practice who do not express desire for routine compositional analysis of stones.

Discussion

Although dual-energy CT applications in abdominal imaging have been well reported, widespread adoption into routine clinical practice has not occurred. Moreover, there continues to be significant variation in practice patterns and applications of the technology, which in turn can add to the challenge to a practice considering implementation of DECT. The results of this study could ease implementation of DECT in practice by providing basic guidelines that are based on common current practice patterns among large academic centers with experience. A summary of consensus statements derived from the results of our survey is listed in Table 2.

Table 2 A summary of consensus statements based on the results from the survey
Full size table

Patient and protocol selection

One of the challenges of DECT implementation is to determine which patients are appropriate for scanning. Lower energy monochromatic energy datasets provide higher contrast; however, there is a trade-off with associated higher noise as compared to the higher energy datasets [11, 15, 29, 30]. Larger body size can further degrade image quality due to associated increase in noise and likelihood of photon starvation [29, 31]. Studies have shown the impact of body sizes on the diagnostic performance of dual energy for various clinical applications [32–35]. dsDECT platforms have a limited reconstruction field of view associated with the high-voltage tube, with a maximum diameter of 26 cm for first generation and 33 cm for second-generation systems. Although central organs such as kidneys and pancreas may not be as affected with this system, it is critical to center the patient properly so that portions of the peripheral organs (such as, liver, small bowel, or colon) are not excluded from the reconstruction field of view [1, 2, 5, 11]. Large body sizes could make this positioning difficult. While the rsDECT has a full field of view (50 cm), the rapid switching between the high- and low-energy acquisition times results in relatively decreased X-ray flux making this system susceptible to noise with larger body sizes. Moreover, the lower tube potential cannot be altered and is limited to 80 kVp thereby limiting the ability to scan larger patients with acceptable image quality due to noise. The higher noise associated with the lower energy tube potential can be reduced through slowing the gantry rotation (increasing rotation time from 0.5 to 0.8 s) as well as through the use of iterative reconstruction techniques [1, 2, 5]. Because of the aforementioned effects of body size on image quality, many centers use a cutoff to exclude patients from dual-energy CT scanning. Our results showed that the practice patterns vary in terms of which parameters for cutoff (i.e., weight vs. transverse diameter from the initial scout image) are used. While the weight cutoffs were relatively similar among institutions that use a cutoff (260–280 lbs.), the lateral dimensions varied widely (38, 42, and 46 cm). This likely reflects the patient demographic makeup for each institution and the acceptable image noise level within the practice. There is a paucity of literature on which cutoff method is better to use (weight vs. lateral dimensions for dual energy) as well as what the optimal cutoff value should be. Guimarães et al reported that the optimal patient size cutoff to generate acceptable image quality with the low tube potential (80-kVp) ranged from 34.9 to 35.8 cm depending on the organ of interest being imaged using 14 × 1.2-mm detector configuration on a dual-source platform [31].

Another initial challenge for a practice considering DECT adoption is deciding how to determine which protocols will be performed with dual energy and which ones will not. A majority of the surveyed institutions incorporate DECT for specific exam protocols (i.e., dual pancreas CT) rather than incorporating it based on exam indication (e.g., pre-operative pancreatic cancer staging). This likely reflects the ease of implementation into workflow if protocols are designated for DECT scanning rather than sorting through indications to determine whether or not DECT scanning is appropriate for the clinical indication [2].

DECT for multiphasic GI exams

Our results showed that a majority of centers are performing dual-energy CT in the arterial phase for multiphasic liver exams. This practice pattern is well supported by studies showing increased lesion conspicuity of hypervascular tumors, such as HCC and hypervascular metastases, on lower energy virtual monochromatic energy datasets [17, 30]. In addition to HCC lesion detection, other DECT applications have been reported and include distinguishing bland from tumor thrombus with iodine quantification and detection of recurrence after trans-arterial chemoembolization using iodine material density images [36, 37]. Interestingly, our results also demonstrated growing interest in using dual-energy CT during the portal venous and/or delayed venous phases, in addition to the arterial phase, allowing potential improved detection of lesion washout in cases of HCC. Despite the interest, this was not routinely practiced at most centers, namely due to slight incremental increase in radiation dose on the single source platform as well as the lack of supporting data in the literature. At this time, the use of dual-energy CT during the venous phases for multiphasic liver exams is not a well-established practice pattern.

There was also a consensus among participants that dual-energy CT should be performed during the arterial phase for multiphasic pancreas examinations. Pancreatic parenchymal attenuation increases with lower tube potential, as expected, and this can improve lesion conspicuity of hypoattenuating tumors such as adenocarcinomas. Detecting small, less than 2 cm, pancreatic adenocarcinomas and isoattenuating pancreatic tumors, which occur in up to 11%–27% of cases, can pose a diagnostic challenge [3, 38]. Improved detection of small and isoattenuating pancreatic adenocarcinomas has been shown using lower energy monochromatic images which provide higher contrast between the tumor and normal parenchyma [3, 16]. Similarly, high-contrast lower energy monochromatic images can also aid in detecting small hypervascular pancreatic tumors, such as neuroendocrine [39] or renal cell carcinoma metastasis.

As one would expect, dual-energy CT could prove useful for the detection of gastrointestinal bleeding source or the detection of bowel ischemia. Our results showed most participants agreed that current dual-energy CT applications for bowel imaging could be useful in these scenarios. Despite this intuitive application, most centers that were surveyed do not have experience with implementing dual-energy CT scanning in these two clinical settings. To date, only two studies have demonstrated the role of dual-energy CT in detecting active gastrointestinal bleeding, both using a dual-source platform showing sensitivity for detection of active bleeding between 89%–94%. Both studies showed potential radiation dose savings of approximately 30% compared with conventional triple-phase scans (i.e., true unenhanced, arterial, and portal venous) and one study interestingly showed better image quality of the virtual unenhanced images as compared with true unenhanced images with no difference in CT numbers [21, 40]. Similarly, lower energy monochromatic images and iodine maps could be used to detect lack of bowel wall enhancement in cases of bowel ischemia. Animal models using rsDECT showed higher contrast-to-noise (CNR) ratios for ischemic bowel segments on reconstructed virtual monochromatic images as compared to the conventional polychromatic 120-kVp images, with 51-keV and iodine material density images having the highest CNRs [41]. Despite demonstration of maximum attenuation of bowel wall at low (i.e., 40-keV) energy monochromatic images and feasibility of dual-energy datasets to allow more confident detection of small bowel enhancement in one study, subjective noise was reported to be higher with monochromatic 70-keV images as compared with polychromatic images [42].

DECT for multiphasic GU examinations

As for the multiphasic GI exams, low-energy virtual monochromatic images reconstructed from the dual-energy dataset can, by way of increases contrast, increase the conspicuity of arterially enhancing renal masses, such as RCC. Low-energy virtual monochromatic images can also allow accurate pre-operative assessment of RCC and renal vasculature through increased iodine contrast. Our results reflect these applications with surveyed participants indicating dual-energy CT scanning commonly being performed for RCC protocols [43]. Iodine quantification can also be used to determine which subtype of renal cell carcinoma is present as well as confer the grade of the tumor [44].

For patients undergoing hematuria workup, a split-bolus technique with simultaneous acquisition of the nephrographic and excretory phases has been reported to carry a sensitivity for detection of upper tract urothelial carcinomas up to 99%–100%. Additionally, split-bolus dual-energy technique has the potential to reduce radiation dose by eliminating the TUE dataset and replacing it with the VUE dataset [45, 46]. Several studies have confirmed that VUE images can be comparable to TUE images in terms of renal lesion characterization and detection of contrast enhancement when comparing post-contrast images with VUE images [47–50]. However, VUE images remain limited in their use to due their susceptibility to noise artifacts as well as limitations in detecting small, less than 4-mm renal stones [51, 52]. Practice patterns across surveyed institutions support these data as there was a consensus that VUE cannot replace TUE for hematuria work-up at this time.

Renal lesion characterization can pose a diagnostic dilemma as often it can be difficult to distinguish benign hyperdense cysts from low-grade solid renal lesions due to high attenuation. Because of this, multiphasic renal studies are often subsequently obtained with a true unenhanced and nephrographic phase to determine if there is enhancement, defined as a change in attenuation between TUE images and nephrographic phase as an increase in attenuation of 10–20 Hounsfield Units (HU) [43, 53]. Additionally, small, less than 2 cm, simple cysts can show a spurious increase in attenuation, or pseudoenhancement, between the unenhanced and enhanced images largely attributed to beam hardening. Dual-energy reconstructed monoenergetic datasets simulate an ideal X-ray beam and are therefore not susceptible to beam hardening artifacts. Thus, DECT virtual monochromatic images can and have been shown to overcome pseudoenhancement and provide a solution to the aforementioned clinical dilemma when encountering pseudoenhancement on conventional multidetector CT [35]. Moreover, iodine material density images can allow easier detection of enhancement as voxels with iodine will be bright or color coded on iodine overlay images [11]. A quantitative approach using iodine material density images to determine enhancement may also be performed, with a threshold of 2 and 0.5 mg/mL for rsDECT and dsDECT, respectively [53].

DECT single phase

Our study showed that DECT for single-phase examinations, such as evaluation of acute abdominal pain or oncology follow-up is indicated though many were not performing it at their institution. Oncologic applications include lesion response to therapy, detection of recurrent or metastatic disease, and treatment planning [5]. Similarly, participants agreed that DECT for renal stone protocols can allow stone composition analysis. However, centers were only performing it based on the perceived value of composition analysis by their referring urologists. DECT can play a significant role in stratifying patient management through stone composition analysis. For example, DECT can allow discrimination of uric from non-uric acid stones as well as characterize those stones that may be resistant to extracorporeal shock wave lithotripsy [54, 55].

Limitations

Our study has several limitations that merit attention. First, the survey was limited to major academic institutions, and thus results and perceived issues with DECT adoption are biased toward those encountered in academic settings. Second, we also had a small sample size of nine institutions though individuals from each institution amounted to 32 total participants. Finally, our study was subject to inherent biases associated with survey-based works.

Conclusions

Although dual-energy CT applications in abdominal imaging have been well reported, there is a lag in widespread adoption. Despite the availability of literature, there is variability in DECT use and practice patterns across departments and imaging centers. This variability further adds to the challenge of those practices considering implementing DECT into their workflow. We herein provide a basic roadmap and guidance to those practices considering either adopting DECT or changing their current DECT workflow.