European Journal of Radiology 83 (2014) 1715–1722

Contents lists available at ScienceDirect

European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad

Abdominal CT: An intra-individual comparison between virtual monochromatic spectral and polychromatic 120-kVp images obtained during the same examination Yoshitake Yamada a , Masahiro Jinzaki a, * , Takahiro Hosokawa a , Yutaka Tanami a , Takayuki Abe b , Sachio Kuribayashi a a b

Department of Diagnostic Radiology, Keio University School of Medicine, 35 Shinanomachi, Shinjuku-ku, Tokyo 160-8582, Japan Center for Clinical Research, Keio University School of Medicine, 35 Shinanomachi, Shinjuku-ku, Tokyo 160-8582, Japan

A R T I C L E I N F O

A B S T R A C T

Article history: Received 23 April 2014 Received in revised form 6 June 2014 Accepted 8 June 2014

Objectives: To compare quantitative and subjective image quality between virtual monochromatic spectral (VMS) and conventional polychromatic 120-kVp imaging performed during the same abdominal computed tomography (CT) examination. Materials and methods: Our institutional review board approved this prospective study; each participant provided written informed consent. 51 patients underwent sequential fast kVp-switching dual-energy (80/140 kVp, volume CT dose index: 12.7 mGy) and single-energy (120-kVp, 12.7 mGy) abdominal enhanced CT over an 8 cm scan length with a random acquisition order and a 4.3-s interval. VMS images with filtered back projection (VMS-FBP) and adaptive statistical iterative reconstruction (so-called hybrid IR) (VMS-ASIR) (at 70 keV), as well as 120-kVp images with FBP (120-kVp-FBP) and ASIR (120-kVp-ASIR), were generated from dual-energy and single-energy CT data, respectively. The objective image noises, signal-to-noise ratios and contrast-to-noise ratios of the liver, kidney, pancreas, spleen, portal vein and aorta, and the lesion-to-liver and lesion-to-kidney contrast-to-noise ratios were measured. Two radiologists independently and blindly assessed the subjective image quality. The results were analyzed using the paired t-test, Wilcoxon signed rank sum test and mixed-effects model with Bonferroni correction. Results: VMS-ASIR images were superior to 120-kVp-FBP, 120-kVp-ASIR and VMS-FBP images for all the quantitative assessments and the subjective overall image quality (all P < 0.001), while VMS-FBP images were superior to 120-kVp-FBP and 120-kVp-ASIR images (all P < 0.004). Conclusions: VMS images at 70 keV have a higher image quality than 120-kVp images, regardless of the application of hybrid IR. Hybrid IR can further improve the image quality of VMS imaging. ã 2014 Elsevier Ireland Ltd. All rights reserved.

Keywords: Multidetector computed tomography Abdomen Noise Signal-to-noise ratio

1. Introduction Polychromatic 120-kVp X-rays have been widely used as the standard acquisition condition in computed tomography (CT) ever since the introduction of CT in clinical diagnostics. The recently developed fast kVp-switching dual-energy CT enables almost simultaneous dual-energy CT data acquisition with a single tube

Abbreviations: ASIR, adaptive statistical iterative reconstruction; FBP, filtered back projection; VMS, virtual monochromatic spectral. * Corresponding author. Tel.: +81 3 3353 1977; fax: +81 3 3225 5715. E-mail addresses: [email protected] (Y. Yamada), [email protected] (M. Jinzaki), [email protected] (T. Hosokawa), [email protected] (Y. Tanami), [email protected] (T. Abe), [email protected] (S. Kuribayashi). http://dx.doi.org/10.1016/j.ejrad.2014.06.004 0720-048X/ ã 2014 Elsevier Ireland Ltd. All rights reserved.

and a single detector [1]. In this type of scanning, because two different types of kVp data are obtained for almost the same projection angle, projection-based beam-hardening correction with two materials (water and iodine) can be performed. This rigorous beam-hardening correction enables the acquisition of accurate material density images of the two basis materials during the projection-based material decomposition processing [2–6]. Thus, virtual monochromatic spectral (VMS) images are generated from a pair of accurate material density images and mass attenuation coefficients [1]. VMS images depict how the imaged object would look if the X-ray source produced a monochromatic beam [1,5]. Compared with the widely used polychromatic imaging, VMS images reconstructed with the more accurate beam-hardening correction provide improved linearity of the CT attenuation, a lower image noise and an improved contrast-to-

1716

Y. Yamada et al. / European Journal of Radiology 83 (2014) 1715–1722

noise ratio (CNR) [1,7–10]. Although VMS imaging at approximately 70 keV reportedly yielded lower image noise and higher CNR than 120-kVp CT imaging in a phantom study [7], to the best of our knowledge, no clinical study to date has compared the image quality between VMS and 120-kVp images obtained simultaneously in the same patients during routine abdominal CT examinations. Furthermore, no detailed clinical study has assessed the effectiveness of the iterative reconstruction (IR) technique for improving the image quality of VMS imaging. The purpose of this study was to compare the quantitative and subjective image quality between VMS and conventional polychromatic 120-kVp imaging performed during the same abdominal CT examination. 2. Materials and methods 2.1. Patients This prospective study was conducted with the approval of our institutional review board, and written informed consent was obtained from all of the patients. From April 2011 to June 2011, 63 consecutive patients who met the following inclusion criteria for the study were recruited: age greater than 55 years and scheduled for a single-phase contrast-enhanced abdominal CT as part of clinical standard-of-care on a specific scanner to avoid variations in the postcontrast timing of various abdominal CT protocols and to minimize the radiation dose. Patients were excluded if they were younger than 55 years old, were pregnant, potentially pregnant or lactating, had any contraindication to iodinated contrast material, such as a previous history of anaphylactoid reaction, had renal failure (serum creatinine level >2.0 mg/dL [177 mmol/L]) or did not provide written informed consent. Nine of the 63 patients were excluded from the study because of a previous history of anaphylactoid reaction to iodinated contrast material (n = 1), renal failure (n = 3) or refusal to provide informed consent (n = 5). To enable an understanding of the evaluation system and to improve the interobserver agreement in the subjective image quality study, the radiologists were trained using images obtained from the first three patients, which were subsequently eliminated from the analyses, and were blinded with regard to the image acquisition mode. Images from a total of 51 patients (36 men, 15 women; age range, 59–88 years; mean age, 72.9  6.9 years) were finally included in the analysis. The clinical indications for routine standard-of-care CT in the subjects of this study were staging or restaging of known malignancy (n = 45), suspected malignancy (n = 3), abdominal pain (n = 2), and further evaluation of an abnormality detected on an abdominal radiograph (n = 1). Thirty hepatic cysts (maximal diameter, 12.3  9.9 [range, 3–35] mm) in 12 patients, 14 hepatic metastases (maximal diameter, 17.2  13.6 [range, 7–60] mm) in 2 patients, and 45 renal cysts (maximal diameter, 15.7  10.0 [range, 3–44] mm) in 29 patients were identified in the research images described below by interpretation of both standard-of-care and research images. Proof of hepatic and renal cysts was based on the lesions showing no change in size compared to the previous or subsequent CT examinations (follow-up periods of 597.8  344.8 [range, 199– 1189] days for hepatic cysts and 754.5  335.9 [range, 133–1498] days for renal cysts). Proof of hepatic metastasis was based on the lesions showing increase in size compared to the previous CT examinations (follow-up period, 133 and 232 days). No other types of lesions were identified in the liver or kidney. Therefore, the hepatic lesions included hepatic cysts and metastases, and all of the renal lesions were cysts. The heights and weights of the patients were measured, and the body mass index (BMI) was calculated (weight in kilograms divided by height squared in meters).

2.2. Imaging protocol First, a clinically indicated standard-of-care portal-dominant phase contrast-enhanced abdominal CT examination was performed using the Discovery CT750HD scanner (GE Healthcare, Waukesha, Wisconsin, USA) 70 s after the injection of iohexol (Omnipaque 300; Daiichi-Sankyo, Japan) at a dose of 2.0 mL/kg using a power injector at a rate of 2 mL/s through the median cubital vein. Subsequently, 90 s after the administration of the contrast medium, sequential fast kVp-switching dual-energy (80/ 140 kVp) and single-energy (120 kVp) enhanced abdominal CT (or sequential single- and dual-energy CT) were performed for research purposes during a single breath-hold over an 8-cm scan length at the lower level of the liver (below the porta hepatis) including a portion of the kidneys, pancreas and spleen, using a random acquisition order to avoid contrast enhancement bias caused by a delay in scanning after the start of the injection. The interval between the beginnings of the two research scans was 4.3 s, which was the minimum setting. The other scanning parameters for fast kVp-switching between 80 and 140-kVp were as follows: tube current, 630 mA; detector collimation 0.625  64 mm; rotation speed, 0.5 s; pitch factor, 1.375; CT dose index volume, 12.7 mGy. The other scanning parameters for 120kVp single-energy CT were as follows: tube current, 495 mA; detector collimation 0.625  64 mm; rotation speed, 0.5 s; pitch factor, 1.375; and CT dose index volume, 12.7 mGy (the same as those for dual-energy CT). We therefore had two raw-data files of the same size for all of the patients. The effective dose estimate for the two research scans was 5.3 mSv, which was determined based on the DLP measurements and the previously reported appropriate normalized coefficients for abdominal CT (0.015 mSv/(mGy cm)) [11]. 2.3. Image reconstruction A series of contiguous 2.5-mm-thick conventional VMS images with filtered back projection (VMS-FBP) and VMS images with adaptive statistical IR (so-called hybrid IR) (VMS-ASIR) at 70 keV were generated from the dual-energy data using the standard kernel. The choice of 70 keV was based on previous phantom and clinical studies, which showed that VMS images at approximately 70 keV yielded the lowest image noise and the highest CNR among 101 sets of VMS images in the range of 40–140 keV at 1-keV intervals [7,10] and also showed that the CT numbers of the VMS images at approximately 70 keV were equal to those of the 120kVp CT images [7]. The series of contiguous 2.5-mm-thick 120-kVp images with FBP (120-kVp-FBP) and ASIR (120-kVp-ASIR) were generated from the single-energy CT data using the standard kernel. We used a 40% ASIR, which means that 40% of the ASIR image was blended with the FBP image, for both the VMS-ASIR and 120-kVp-ASIR images. The 40% ASIR was chosen based on the results of previous studies [12–16]. Thus, we obtained 204 sets of 2.5-mm-thick images, that is, 51 image sets each in the 4 image series: VMS-FBP, VMS-ASIR, 120-kVp-FBP and 120-kVp-ASIR. At the present time, the model-based iterative reconstruction (MBIR) technique cannot be applied to VMS images. Each image data set was coded, the patient information was removed, and the sets were randomized to enable double-blind evaluation. 2.4. Objective measurements The CT numbers (Hounsfield units [HU]) and objective image noise (i.e., standard deviation [SD] of the CT numbers) were measured for all 204 image sets using an independent workstation (Advantage workstation 4.5; GE Healthcare, Waukesha, Wisconsin, USA). Two board-certified radiologists with 7 and 22 years of

Y. Yamada et al. / European Journal of Radiology 83 (2014) 1715–1722

experience in interpreting abdominal CT placed, by consensus, a circular or ovoid region of interest (ROI) in the liver, kidney (renal cortex), pancreas, spleen, portal vein, abdominal aorta, inferior vena cava (IVC), paraspinal muscle and abdominal fat in the anterior abdominal wall at the level of the origin or proximal of the superior mesenteric artery, and in each hepatic and renal lesion. A constant ROI size of approximately 1.0 cm2 was used as much as possible at all sites. The size, shape and position of the ROIs were kept constant among the four protocols by applying a copy-and-paste function at the workstation. The signal-to-noise ratio (SNR) was then calculated by dividing the CT numbers by the corresponding image noise. The CNRs of the liver, kidney, pancreas, spleen, portal vein and aorta were calculated using the following formulae [17–19]: CNRLiver, Kidney, Pancreas, Spleen = (ROILiver, Kidney, Pancreas, Spleen ROIFat)/SDBackground, CNRPortal vein = (ROIPortal vein ROILiver)/SDBackground, CNRAorta = (ROIAorta ROIMuscle)/SDBackground, where ROILiver, ROIKidney, ROIPancreas, ROISpleen, ROIPortal vein, ROIAorta and ROIMuscle denote the CT numbers of the liver, kidney, pancreas, spleen, portal vein, aorta, and muscle respectively, and SDBackground denotes the image noise of fat [17–19]. The lesion-toliver and lesion-to-kidney CNRs were calculated using the following formula [10,20,21]: CNRlesion-to-liver, lesion-to-kidney = |ROIlesion ROILiver, kidney|/SDliver, kidney, where ROIlesion denotes the CT numbers of hepatic or renal lesions, and SDliver and SDkidney denote the image noises of the liver and kidney, respectively.

1717

2.6. Statistical analysis Variables were expressed as the mean  SD. A paired t-test was used to compare the CT numbers, objective image noises, SNRs and organ CNRs among the image reconstruction protocols. A general linear mixed-effects model containing the image reconstruction protocol as a fixed effect and the patient as a random effect was used to compare the lesion-to-liver and lesion-to-kidney CNRs among the protocols. Because only two patients had hepatic metastases, a subgroup analysis for these patients was not performed. The differences in the scores for subjective image quality were tested using the Wilcoxon signed rank sum test. The Bonferroni correction was used for all multiple comparisons. The interobserver agreement was evaluated using the percentage agreement and k statistics between the two radiologists for the subjective image quality scores. The significance level for all tests was 5% (two-sided). All data were analyzed using a commercially available software program (SPSS version 21; IBM SPSS, Armonk, New York, USA). 3. Results 3.1. Patient characteristics The mean height, weight and BMI of the 51 patients were 161.6  8.1 cm (range, 140.0–180.0 cm), 59.0  9.9 kg (range, 40.0– 81.0 kg) and 22.5  2.9 kg/m2 (range, 16.1–29.8 kg/m2), respectively. 3.2. Objective image quality

2.5. Subjective image quality All the image data sets were evaluated in a randomized manner using an independent workstation (Advantage workstation 4.5) to assess the subjective image quality. Two boardcertified radiologists with 6 and 12 years of experience in interpreting abdominal CT images independently evaluated the image quality in all 204 image sets in a blinded and randomized manner. Subjective image quality was assessed in terms of the subjective image noise on a 5-point scale; artifacts (including helical or windmill, streak and beam-hardening artifacts), on a 4point scale; sharpness with respect to the vascular wall (including the intrahepatic vessel wall) and the borders of organs, on a 5point scale; overall image quality, on a 5-point scale, as described in Table 1. The averages of the two individual scores were used for the image quality analysis. The radiologists evaluated the abdominal images using the default preselected abdominal window setting (window width, 400 HU; window level, 40 HU); they were, however, allowed to adjust the window level and width as desired.

The detailed objective image quality measurements are summarized in Tables 2 and 3. No significant differences in the CT numbers for any of the solid organs were observed among the 120-kVp-FBP, 120-kVp-ASIR and VMS-FBP images, while the CT number for the aorta on the VMS-FBP images was higher than the values on the 120kVp-FBP or 120-kVp-ASIR images. The CT numbers for all the regions were lower on the VMS-ASIR images than those on the VMS-FBP images; however, the differences were less than 1 HU. The CT numbers for the fat and portal vein were significantly different between the 120-kVp-ASIR images and the 120-kVp-FBP images; however, the differences were also less than 1 HU (Table 2). The VMS-ASIR images had a significantly lower objective image noise and significantly higher SNR and organ CNR in all the regions, and also had significantly higher lesion-to-liver and lesion-to-kidney CNRs (all P < 0.0001) compared to the 120-kVp-FBP, 120-kVp-ASIR, and VMS-FBP images (Tables 2 and 3). Compared with the 120-kVpFBP and 120-kVp-ASIR images, the VMS-FBP images had a significantly lower objective image noise and a significantly higher SNR and organ CNR in all the regions, and also had significantly higher lesion-to-liver and lesion-to-kidney CNRs (all P  0.0003).

Table 1 Grading scale for subjective image quality. Image quality Grading scale

Noise

Artifacts

Sharpness

Overall image quality

5

No or minimal artifacts

Sharpest

Superior

4

Minimum or no image noise Less than average noise Average image noise Above average noise

1

Unacceptable image noise

Not applicable

Better than average Average Poorer than average Blurry

Above average

3 2

Artifacts occupying a part of the body, but not interfering with diagnostic decision making Artifacts occupying the entire body, but diagnosis still possible Artifacts affecting diagnostic information

Average Suboptimal Unacceptable

1718

Y. Yamada et al. / European Journal of Radiology 83 (2014) 1715–1722

Table 2 CT numbers, objective image noises, signal-to-noise ratios and contrast-to-noise ratios in 120-kVp-FBP, 120-kVp-ASIR, VMS-FBP and VMS-ASIR images. Pair-wise comparison; P -value (unadjusted)a

Protocol 120-kVp-FBP

120-kVp-ASIR

VMS-FBP (70 keV)

VMS-ASIR (70 keV)

120-kVp-FBP and 120-kVp-ASIR

120-kVp-FBP and VMS-FBP

120-kVp-FBP and VMS-ASIR

120-kVp-ASIR and VMS-FBP

120-kVp-ASIR and VMS-ASIR

VMS-FBP and VMS-ASIR

(HU) 116.2  12.7 185.4  25.2 92.2  15.9 120.1  14.1 166.0  23.3 146.0  19.2 132.5  25.2 64.7  7.6 -100.5  17.1

116.2  12.7 185.2  25.4 92.1  15.9 120.1  14.1 165.7  23.3 145.9  19.2 132.3  25.2 64.7  7.5 -100.4  17.1

114.7  10.7 183.2  24.1 92.3  14.9 120.5  13.1 167.2  22.9 151.9  21.8 132.9  25.7 64.6  6.4 -101.4  18.2

114.6  10.7 182.8  24.1 92.0  14.9 120.4  13.1 166.4  22.7 151.6  21.7 132.5  25.6 64.3  6.4 -101.0  18.2

0.7147 0.0545 0.0564 0.3037 0.0001* 0.6872 0.0293 0.7933