- Email: [email protected]
Comparison of Whole Heart Computed Tomography Scanners for Image Quality Lower Radiation Dosing in Coronary Computed Tomography Angiography: The CONVERGE Registry
ARTICLE IN PRESS
Original Investigation
Comparison of Whole Heart Computed Tomography Scanners for Image Quality Lower Radiation Dosing in Coronary Computed Tomography Angiography: The CONVERGE Registry Nirali 1XD X Patel, MD, 2XD X Dong 3XD X Li, MD, 4XD X PhD, Rine 5XD X Nakanishi, MD, 6XD X PhD, Badiha 7XD X FatimaD,8X X Daniele 9XD X Andreini, MD, 0X1D X Gianluca 1X D X Pontone, MD, 2X1D X Edoardo 3X1D X Conte, MD, 4X1D X Rachael 5X1D X O'Rourke, MBBS, 6X1D X FRANZCR, Eranthi 7X1D X Jayawardena, BS, 8X1D X Christian 9X1D X Hamilton-Craig, MBBS, 0X2D X PhD, Manojna 1X2D X Nimmagadda, MD, 2X D X Matthew 3X2D X J. Budoff, MD 4X2D X Rationale and Objectives: Novel technology in coronary computed tomographic angiography allows assessment of coronary artery disease with high image quality (IQ). There are currently two wide detector “whole heart” coverage scanners available, which avoid misregistration artifacts. However, there are no data directly comparing IQ between the two scanners. The aim of the current study is to investigate if IQ is different between the most current scanners of GE and Toshiba broad detector scanners. Materials and Methods: Prospective, observational, multicenter international cohort study comparing 236 consecutive patients who underwent coronary computed tomographic angiography using whole-heart scanners; 126 patients on scanner S1 (Aquilion ONE, Toshiba), and 110 patients on scanner S2 (Revolution CT, GE Healthcare). Hounsfield units were measured using regions of interest in the descending aorta at 6 points (cranial slice, level of the visualized first, second, third, and fourth spines, and the caudal slice). We also compared the coverage length (z-axis) of the full width field of view between a single rotation of the two scanners. Results: Evaluating mean CT attenuation values Hounsfield units through the scan range, are progressively reduced across the descending aorta in the S1 group, resulting in the larger difference of contrast brightness between the cranial and caudal slices compared to the S2 group (absolute difference: S2 13.0 § 4.4 vs S1 141.9 § 16.4, p < 0.0001; Percent difference: 19.3 § 2.1 vs ¡3.4 § 1.2, <0.0001). The standard deviation (SD) is similar at the cranial slice between the two scanners, however, the S1 group demonstrated higher SD-differential from cranial to caudal than S2 group. Median radiation exposure was significantly lower for the S2 scanner 1.50 § 0.75 mSv vs the S1 system 1.9 mSv (IQR 1.7 2.7 mSv) (p = 0.01). Z-axis coverage was larger for the S2 scanner 152.5 mm (244 slices £ 0.625 mm/slice) than 133 mm for S1 (266 slices £ 0.5 mm/slice). Conclusion: Although both “volume” scanners cover the whole heart z-axis with one beat, scans using the S1 scanner have a larger variability in attenuation values throughout the scan range, resulting in 20% increase in nonuniformity from cranial to caudal slice. Additionally, SD variation across the field of view, a metric of noise, is larger when using the S1 scanner vs the S2 scanner. These results indicate that the GE Revolution CT has more uniform contrast enhancement and more coverage, lower radiation and lower image noise compared to the current Toshiba Aquilion ONE system. Key Words: Coronary computed tomographic angiography; image quality; diagnosis; cardiac CT. © 2019 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.
Acad Radiol 2019; &:1 7 From the Department of Medicine, Los Angeles Biomedical Research Institute, Torrance, CA (N.P., D.L., R.N., B.F., E.J., M.N., M.J.B.); Department of Cardiovascular Medicine, Toho University Graduate School of Medicine, Tokyo, Japan (R.N.); Centro Cardiologico Monzino, IRCCS, Milan, Italy (D.A., G.P., E.C.); Department of Clinical Sciences and Community Health, Cardiovascular Section, University of Milan, Italy (D.A.); Department of Medical Imaging, The Prince Charles Hospital, Brisbane, Queensland Australia (R.O., C.H.-C.); University of Queensland, Brisbane, Queensland, Australia (R.O., C.H.-C.). Received November 15, 2018; revised December 31, 2018; accepted January 2, 2019. Funding: This study was funded by General Electric Inc. The company had no role in study design, conduct, or manuscript preparation. Dr Budoff receives grant support from the National Institutes of Health and General Electric. Address correspondence to: M.J.B. e-mail: [email protected] © 2019 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.acra.2019.01.002
1
ARTICLE IN PRESS PATEL ET AL
INTRODUCTION
C
oronary computed tomographic angiography (CTA) has emerged as a promising noninvasive method for the detection and exclusion of obstructive coronary artery disease due to the very high negative predictive values (1,2). Recent advancements in CT scanner technology have enabled even further improvement in both image quality and diagnostic accuracy. Some advancements include increased temporal resolution through the introduction of dual source and software based motion correction, high definition detectors, iterative reconstruction, and advanced work stations. (3). Additionally, two CT system manufacturers (Toshiba, GE) have released systems with larger volume of coverage in the z-axis that reduce misalignment and “stair-step artifacts” and improve myocardial and blood pool homogeneity by allowing for whole heart coverage without need for table movement. There have been studies demonstrating the improved image quality associated with these larger coverage scanners as compared to 64-slice CT technology (4). However, there has not, to our knowledge, been a direct comparison between the two wide coverage systems, specifically in the assessment of myocardial and blood pool homogeneity. The purpose of this study is to make such a direct comparison between the current wide coverage CTs in terms of myocardial and blood pool homogeneity.
MATERIAL AND METHODS This study prospectively consented and evaluated consecutive patients who underwent coronary CTA studies at multiple sites as part of the CONVERGE Registry. Sites included Los Angeles Biomedical Research Institute, Torrance, CA; Centro Cardiologico Monzino, IRCCS, Milan, Italy; The Prince Charles Hospital, Brisbane, Queensland Australia and Baptist Hospital of Miami Florida. Two consecutive patient cohorts from multiple institutions were evaluated, each cohort obtained concurrently from March 2017 through August 2017. The first cohort (S1, n = 126) images were acquired in one heart cycle with volumetric coverage of 16 cm (320 slice CT £ 0.5 mm) (Aquilion ONE, Toshiba) and a gantry rotation speed of 0.35 s/rotation. The second cohort (S2, n = 110) were acquired in one heart cycle with a volumetric coverage of 16 cm (256 slice CT £ 0.625 mm) (Revolution CT, GE Healthcare) and a gantry rotation speed of 0.28 s/rotation. Inclusion criteria were weight <300 lbs and no prior history of cardiac surgery or percutaneous coronary intervention since these procedures would interfere with the measurement of coronary atherosclerosis or stenosis. Exclusion criteria for this study included atrial fibrillation, chronic kidney disease (estimated glomerular filtration rate <60 mL/min/1.73 m2 within 30 days of the CT) or a history of intravenous contrast allergy (5,6). Coronary CTA Scan Protocol and Image Reconstruction
The clinical coronary CTA exams were performed with beta blockade for heart rate (HR) control as per clinical routine if 2
Academic Radiology, Vol &, No &&, && 2019
their resting HR was greater than 60 beats per minute (bpm). The protocol was for patients undergoing coronary CTA to be given 100 mg metoprolol orally 1 hour before the scan if the HR was greater than 60 bpm and if needed up to 20 mg intravenous metoprolol titrated to achieve HR <60 bpm. We used prospectively ECG-triggered coronary CTA technique for both scanners. Both scanners used a 20 cm field of view for acquisition of coronary CTA's. The contrast injection was performed using a power injector (Stellant; Medrad, Warrendale, Pennsylvania) through an antecubital vein at a rate of 5.0 mL/s. A triple bolus injection protocol was utilized. Patients were given an injection of 60 cc of contrast in the first phase, followed by 20 cc of contrast and 30 cc of saline in the second phase and completed by 50 cc of saline. Scan Acquisition Protocol
For both scanners, the field of view (z-axis) included the mid ascending aorta to the upper abdomen. No table movement was required due to the wide volumetric acquisition. The z-axis collimation was selected based on the scout images demonstrating the heart size. No patient required more than 16-cm z-axis coverage. Tube voltages used fixed at 120 kilovolt potential (kVp), to provide comparable radiation exposure. All acquisitions were prospectively ECG triggered. Each scan was done in a single-beat acquisition within one cardiac cycle, regardless of HR. Scanner S1 Acquisition (Aquilion ONE): rotation speed was 350 milliseconds, with no table motion. The mA settings depend on the body mass index (BMI) of the patient, and set by the radiographer as follows: BMI < 20 kg/m2, 120 kVp 300 mA; BMI 21 25 kg/m2, 120 kVp 350 mA; BMI 26 30 kg/m2, 120 kVp 400 mA; BMI > 31 kg/m2,120 kVp 500 mA. Scanner S2 Acquisition (GE Revolution): rotation speed was 280 milliseconds, with no table motion. Tube current ranged between 122 and 740 mA, based upon the BMI of the patient, automatically determined by the system. A medium field of view was selected for all patients. The scanner is equipped with an “autogating” capability, which automatically adjusts HR-dependent settings for triggered acquisition and gated reconstruction. Autogating was used to automatically acquire diastolic phases for lower HRs and both systolic and diastolic phases for higher HRs. Image Reconstruction
For both scanner types, data reconstruction was performed using thin reconstructions with intervals ranging at 60 80% of the R-R cycle, and most of data were reconstructed at 75% of the R-R cycle. Iterative reconstruction algorithms (ADIR for S1, ASIR-V for S2) level of 40% was used. For CT exams performed on scanner S1, adaptive motion correction was applied to minimize motion. For CT exams performed on scanner S2, motion correction software (“SnapShot Freeze;” GE Healthcare) was used for correcting motion artifacts in patients with higher HRs.
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
COMPARISON OF CT SCANNERS FOR CARDIAC CT
All images were transferred to a single vendor workstation to provide standardized postprocessing for both scanner acquisitions (AW 4.7, GE Healthcare). An SCCT level-3 trained CT cardiologist read all CT scans at a central reading center (Los Angeles Biomedical Research Institute at Harbor UCLA in Torrance, California). Evaluation of Image Quality
CT images were analyzed at the core CT reading center by trained, experienced readers blinded to participants' clinical information and type of scanner used. All CT reads were done in the central CT core lab at Los Angeles Biomedical Research Institute in Torrance CA. Both readers are radiologists with >20 years experience reading cardiovascular CT imaging. Signal intensity was derived from CT attenuation values measured in Hounsfield units (HU) from a region of interest. All quantitative measurements were performed on axial source coronary CTA images by placing a circular region of interest of 100 mm2 in the descending aorta and left ventricular blood pool. The mean and SD of the HU values were measured in the descending aorta at 6 points (cranial slice, level of the visualized first, second, third, and fourth spines, and level of caudal slice; Figs. 1 and 2). Coverage We first measured full coverage in the z-axis for each scanner. We evaluated the total number of slices demonstrating the entire field of view present to determine coverage length of the scanners. However, both scanners demonstrate a smaller concentric circle of imaging from the detectors most distant from the x-ray tube (cranial- and caudal-most slices), and slowly increase visualization of the structure, so we also measured the coverage length along the z-axis which enabled a full width field of view (excluding those images without full axial visualization). Both scanners have truncated cranial and caudal slices, not allowing for visualization of the full field of view, and these slices were excluded from measure and analysis related to coverage length for each scanner. Radiation Dose The effective radiation dose in both systems was calculated using the following formula: Effective radiation dose = Dose length product £ Conversion coefficient for the chest (k = 0.014 mSv/mGy cm).
Statistical Analysis
Analyses were performed using statistical software (SAS version 9.1; SAS Institute, Cary, North Carolina). A statistically significant difference was defined as a p value (two-tailed) less than 0.05. Continuous variables were expressed as mean § SD, or quartile ranges. A two-sided t test was applied when the distribution of data was of equal variance, and Welch Satterthwaite Students t test was used when unequal variance was found. Multivariable analysis was done to assess the subjects age, gender, and BMI affect the results between scanners and if so how much of the variance is based on these factors and how much is based on which scanner is used in determining uniformity of age, gender, and BMI were entered as covariates. Logistic regression model was performed to determine if age, gender, and BMI were predictors for the contrast enhancement in the z axis, image noise (SD variation in the z axis) and radiation dose. No statistically significant differences were found for interactions between scanner type and the covariates (BMI, gender, and age). RESULTS Patient Demographics
There were no statistical differences between gender, patient age, BMI, and HR between the scanner groups (Table 1). Radiation Dose
We found that the median radiation dose was significantly lower for S2 scanner at 1.50 § 0.75 millisieverts (mSv) compared to 1.9 mSv (interquartile range, 1.7 2.7 mSv) for the S1 scanner (p = 0.01). Quantitative Analysis
Although the mean HU in the S2 group is higher than that in the S1 group at the cranial slice (618.3 § 19.1 vs 431.5 § 13.4, p < 0.0001), HUs are progressively reduced across the descending aorta in the S1 group, resulting in the larger difference of contrast brightness between the cranial and caudal slices compared to the S1 group (Absolute difference: S2 13.0 § 4.4 vs S1 141.9 § 16.4, p < 0.0001; Percent difference: 19.3 § 2.1vs ¡3.4 § 1.2, p <0.0001; Fig 1a,b). The SD is similar at the cranial slice between the two scanners, however, the S1 group indicated higher noise (increased SD of HU) than S2 group at the other levels in the descending aorta (Table 2).
TABLE 1. Demographics Patient Characteristics
Total
Age (years), Range: 28 70 BMI (kg/m2) Gender (% male) Heart rate (bpm)
55 § 6.9 27.9 § 3.2 62% 57 § 6
S1 Scanner (median § S.D.)
S2 Scanner (median § S.D.)
p Value (S1 vs S2)
55 § 9.8 28.0 § 4.0 61% 56 § 8
56 § 8.4 27.8 § 3.9 62% 59 § 7
0.49 0.76 0.89 0.24
SD, standard deviation.
3
ARTICLE IN PRESS PATEL ET AL
Academic Radiology, Vol &, No &&, && 2019
Figure 1. (a) Distribution of mean HU across descending aorta for GE (Scanner 1) and Toshiba (Scanner 2), (b) Distribution of Standard deviation of mean HU across descending aorta for GE (Scanner 1) and Toshiba (Scanner 2). HU, Hounsfield units.
Coverage
DISCUSSION
Total coverage (allowing for at least the heart to be seen centrally) for the S2 scanner was 160 mm (256 slices at 0.625 mm/slice) as compared to 135.5 mm for S1 (542 slices £ 0.25 mm). Full field of view coverage was greater for the S2 scanner 152.5 mm (244 slices/rotation £ 0.625 mm/slice), compared to S1 scanner 133.25 mm (533 slices/rotation £ 0.25 mm/slice).
CTA started with single slice scanners (electron beam tomography) over 20 years ago (7 9). Over the last 20 years, coverage has progressively increased, with more detectors available. Currently, two computed tomography systems have whole heart coverage, with at least 10 cm of coverage per rotation, also known as wide multi detector computed tomography. In this
4
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
COMPARISON OF CT SCANNERS FOR CARDIAC CT
TABLE 2. Image Quality of cohorts for Scanner 1 and Scanner 2 at different levels of the cardiac study GE Revolution (n = 110)
Toshiba (n = 126)
Top slice Mean HU 431.5 § 13.4 618.3 § 19.1 SD HU 38.2 § 0.7 55.9 § 10.0 Aorta at the level of the first spine Mean HU 439.7 § 13.8 616.5 § 20.1 SD HU 36.3 § 0.9 42.4 § 1.2 Aorta at the level of the second spine Mean HU 441.8 § 13.2 561.5 § 18.4 SD HU 36.4 § 0.8 41.9 § 1.0 Aorta at the level of the third spine Mean HU 445.7 § 13.6 548.2 § 17.9 SD HU 37.3 § 1.9 49.8 § 5.7 Aorta at the level of the fourth spine Mean HU 448.1 § 14.1 537.1 § 17.0 SD HU 40.7 § 3.2 44.0 § 0.9 Bottom slice Mean HU 441.5 § 14.1 470.4 § 13.9 SD HU 43.0 § 0.8 46.6 § 0.9 Absolute difference (Top¡Bottom) Mean HU ¡13.0 § 4.4 141.9 § 16.4 SD HU ¡4.7 § 1.1 ¡7.6 § 1.1 Percent difference (Top¡Bottom) Mean HU ¡3.4 § 1.2 19.3 § 2.1 SD HU ¡16.5 § 3.1 ¡25.9§3.4
p Value
<0.0001 0.84 <0.0001 0.0001 <0.0001 0.0001 0.0001 <0.0001 0.0002 <0.0001 0.17 0.007 <0.0001 0.056 <0.0001 0.061
prospective, multicenter, observational cohort study, we compared image quality from consecutive coronary CTA scans performed on two “whole-heart” single-rotation scanner platforms; the GE Revolution, and the Toshiba Aquilion ONE. One of the key benefits of whole-heart imaging is achieving signal homogeneity from isophasic acquisition, which may have significant value in myocardial CT perfusion as well as other quantitative applications. Therefore, we sought to quantitate the uniformity throughout the cranio-caudal z-axis in clinical exams
of patients with suspected coronary artery disease, and to compare true z-axis coverage length and radiation exposures. The Revolution CT and the Toshiba scanner both have 160 mm coverage as quoted by the manufacturers. While the Revolution CT achieves 160 mm coverage in cardiac mode, the Aquilion One only allowed for reconstruction of 135.5 mm. Furthermore, when analyzing the portion of the effective coverage (the number of slices which contain diagnostic information across the full field of view), it was significantly greater on the Revolution than on the Toshiba system (152.5 mm vs 133.25 mm, respectively, p = 0.011). While both scanners covered the whole heart within a single rotation, scans on the Aquilion ONE have a larger variability of attenuation values throughout the scan range, resulting in 20% reduction in contrast brightness from cranial slice to caudal slice. SD variation is also larger on the Aquilion ONE than that the Revolution system, indicating the latter has more consistent contrast enhancement while maintaining less image. We also found that median radiation exposure on the Aquilion ONE scanner (S1) was higher than the Revolution scanner (S2). In this study, we evaluated coronary CT angiography images from two commercially available systems with wide-detector whole-heart coverage (quoted 16 cm scan range). Both systems allow acquisition of the entire heart without moving the table. This inherently removes misalignment artifacts, and has the promise to improve signal homogeneity due to isophasic acquisition, which has value in emerging applications such as myocardial CT perfusion imaging and atrial fibrillation acquisitions. However, the physics of wide coverage CT also presents unique challenges such as increased heel effect and scatter that confound the goal of isophasic signal homogeneity. Our study indicates that, as compared to the wide coverage Aquilion ONE system, the Revolution CT system yielded significant improvement in contrast uniformity and noise variation in the z-axis, and with significantly lower radiation exposure. In addition, the portion of the 16 cm z-coverage which allowed diagnostic visualization of the structures within the full field of view (excluding those slices with
Figure 2. Assessment of mean and SD of HU on the descending aorta. HU, Hounsfield units; SD, standard deviation.
5
ARTICLE IN PRESS PATEL ET AL
truncated images,) was significantly larger with the Revolution than the Aquilion ONE. The Revolution system contains a few unique characteristics that may explain these findings. The GE Revolution employs 3D collimator situated between patient and detector, which blocks scattered photons prior to reaching the detector bank, as distinct from other vendor systems which attempt to collimate the beam prior to reaching the patient. Additionally, the Revolution system has an integrated “Volume HD” reconstruction platform that contains an algorithm call Multi-Material Artifact Reduction. This algorithm performs beam hardening correction, but takes into account the spectrum variation across the field of view due to the heel effect. One possible explanation of the differences seen in consistency of images is the use of different detectors by each system. The Revolution system uses the “Gemstone” garnet-based detector, while “PUREVISION” detector used in Toshiba system is forged from a solid ceramic ingot. Compared to ceramic (Gd2O2S) scintillator, the garnet-based detector has 100-times faster decay time (faster rate of light being emitted after the detector is excited) and 75% shorter afterglow (time needed for the material to return to pre-excitation state, ready for the next excitation) (10). Overall, these properties translate into improved spatial resolution and reduced image noise due to increased data sampling per gantry rotation, which is conventionally limited by the properties of the ceramic material. A proprietary 3D collimator in GE scanner ensures contrast uniformity and minimizes scatter and beam hardening artifacts associated with wide coverage systems, which is supported by the findings of the present study. Along with such hardware innovations, advancements in spectral modeling, projection-based material decomposition, and noise modeling have led to increased uniformity across the full z-coverage, as demonstrated by our quantitative data (11 14). Limitations
The limitations of the study include the study design, which was a prospective multicenter observational registry that did not allow for each patient to be scanned twice (once on each scanner) due to logistics as well as concerns about radiation exposure. However the patient groups were well matched, and the postprocessing standardized for both scanner acquisitions. Each scanner uses different determinations of tube current based on subject size/BMI. While we carefully matched and controlled for age and BMI, there are intrinsic differences in the acquisition parameters and perhaps in in body morphology despite similar BMI that we could not control for. The goal was to identify, using clinical scan parameters used by each scanner to see if this results in differences in image quality. It should be noted that image quality is also affected by ECG triggering and other algorithms that may differ between vendors, but these variables are beyond the scope of this study (15 17). 6
Academic Radiology, Vol &, No &&, && 2019
CONCLUSION We identified significant improvements in contrast uniformity, noise variation, full-field scan range, and decreased radiation exposure using the Revolution CT compared to the Aquilion One system. The clinical implications of the results are significant. Lower radiation exposures is always clinically important, and increased scan length (coverage) allows for single rotation scanning of more patients, including bypass and aortic studies depending on height and scan length of the patient. Lower image noise will impact overall image quality and allow for better acquisition for postprocessing of such applications as fractional flow reserve by CT (18), plaque imaging, transluminal gradient assessment, pericoronary fat attenuation and other measures that require superb image quality. The revolution CT allows for improved quantitative uniformity of iodinated contrast across the whole z-axis scan range, which may be advantageous for applications such as myocardial CT perfusion imaging, whilst at the same time improving dose performance. DISCLOSURE Dr. Budoff receives grant support from General Electric. Dr Hamilton-Craig has received research support from SanofiGenzyme and Siemens. REFERENCES 1. Budoff MJ, Dowe D, Jollis JG, et al. Diagnostic performance of 64-multidetector row coronary computed tomographic angiography for evaluation of coronary artery stenosis in individuals without known coronary artery disease: results from the prospective multicenter ACCURACY (Assessment by Coronary Computed Tomographic Angiography of Individuals Undergoing Invasive Coronary Angiography) trial. J Am Coll Cardiol 2008; 52(21):1724–1732. 2. Wang H, Xu L, Fan Z, et al. Clinical evaluation of new automatic coronary-specific best cardiac phase selection algorithm for single-beat coronary CT angiography. PLoS One 2017; 12(2). 3. Alani A, Nakanishi R, Budoff MJ. Recent improvement in coronary computed tomography angiography diagnostic accuracy. Clin Cardiol 2014; 37(7):428–433. 4. Fujimoto S, Matsutani H, Kondo T, et al. Image quality and radiation dose stratified by patient heart rate for coronary 64- and 320-MDCT angiography. AJR Am J Roentgenol 2013; 200(4):765–770. 5. Korada SKC, Zhao D, Tibuakuu M, et al. Frailty and subclinical coronary atherosclerosis: the Multicenter AIDS Cohort Study (MACS). Atherosclerosis 2017; 266:240–247. 6. Schuetz GM, Schlattmann P, Achenbach S, et al. Individual patient data meta-analysis for the clinical assessment of coronary computed tomography angiography: protocol of the Collaborative Meta-Analysis of Cardiac CT (CoMe-CCT). Syst Rev 2013; 2(1):13. 7. Budoff MJ, Oudiz RJ, Zalace CP, et al. Intravenous three dimensional coronary angiography using contrast enhanced electron beam computed tomography. JACC 1997; 29(2):8003 8003. 8. Lu B, Dai RP, Jing BL, et al. Evaluation of coronary artery bypass graft patency using three-dimensional reconstruction and flow study on electron beam angiography. J Comput Assist Tomogr 2000; 24:663–670. 9. Budoff MJ, Lu B, Mao S. Gadolinium-enhanced three-dimensional electron beam coronary angiography. J Comput Assist Tomogr 2002; 26 (Nov-Dec(6)):879. 10. Jiang H, Vartuli J, Vess C. Gemstone—the ultimate scintillator for computed tomography Haochuan Jiang CT detectors. GE white paper; c 2009. 11. Min JK, S. R, Vass M, et al. High-definition multidetector computed tomography for evaluation of coronary artery stents: comparison to
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
standarddefinition64-detector row computed tomography. J Cardiovasc Comput Tomogr. 2009; 3(Jul-Aug(4)):246–251. 12. Nakanishi R, Matsumoto S, Alani A, et al. Diagnostic performance of transluminal attenuation gradient and fractional flow reserve by coronary computed tomographic angiography (FFRCT) compared to invasive FFR: a sub-group analysis from the DISCOVER-FLOW and DeFACTO studies. Int J Cardiovasc Imaging 2015; 31:1251–1259. 13. Gupta M, Kadakia J, Jug B, et al. Detection and quantification of myocardial perfusion defects by resting single-phase 64-slice cardiac computed tomography angiography compared with SPECT myocardial perfusion imaging. Coron Artery Dis. 2013; 24(4):290–297. 14. Cury RC, Kitt TM, Feaheny K, et al. A randomized, multicenter, multivendor study of myocardial perfusion imaging with regadenoson CT perfusion vs single photon emission CT. J Cardiovasc Comput Tomogr 2015; 9(2):103–112. e2.
COMPARISON OF CT SCANNERS FOR CARDIAC CT
15.
16.
17.
18.
Gopal A, Mao SS, Karlsberg D, et al. Radiation reduction with prospective ECG-triggering acquisition using 64-multidetector computed tomographic angiography. Int J Cardiovasc Imaging 2009; 25(4):405–416. Mao SS, Bakhsheshi H, Lu B, et al. Effect of ECG triggering on reproducibility of coronary artery calcium scoring. Radiology 2001; 220(Sep 3): 707–711. Gupta M, Hacioglu Y, Kadakia J, et al. Left ventricular volume: an optimal parameter to detect systolic dysfunction on prospectively triggered 64multidetector row computed tomography: another step towards reducing radiation exposure. Int J Cardiovasc Imaging 2011; 27:1015–1023. Nakanishi R, Matsumoto S, Alani A, et al. Diagnostic performance of transluminal attenuation gradient and fractional flow reserve by coronary computed tomographic angiography (FFRCT) compared to invasive FFR: a sub-group analysis from the DISCOVER-FLOW and DeFACTO studies. Int J Cardiovasc Imaging 2015; 31:1251–1259.
7
Original Investigation
Comparison of Whole Heart Computed Tomography Scanners for Image Quality Lower Radiation Dosing in Coronary Computed Tomography Angiography: The CONVERGE Registry Nirali 1XD X Patel, MD, 2XD X Dong 3XD X Li, MD, 4XD X PhD, Rine 5XD X Nakanishi, MD, 6XD X PhD, Badiha 7XD X FatimaD,8X X Daniele 9XD X Andreini, MD, 0X1D X Gianluca 1X D X Pontone, MD, 2X1D X Edoardo 3X1D X Conte, MD, 4X1D X Rachael 5X1D X O'Rourke, MBBS, 6X1D X FRANZCR, Eranthi 7X1D X Jayawardena, BS, 8X1D X Christian 9X1D X Hamilton-Craig, MBBS, 0X2D X PhD, Manojna 1X2D X Nimmagadda, MD, 2X D X Matthew 3X2D X J. Budoff, MD 4X2D X Rationale and Objectives: Novel technology in coronary computed tomographic angiography allows assessment of coronary artery disease with high image quality (IQ). There are currently two wide detector “whole heart” coverage scanners available, which avoid misregistration artifacts. However, there are no data directly comparing IQ between the two scanners. The aim of the current study is to investigate if IQ is different between the most current scanners of GE and Toshiba broad detector scanners. Materials and Methods: Prospective, observational, multicenter international cohort study comparing 236 consecutive patients who underwent coronary computed tomographic angiography using whole-heart scanners; 126 patients on scanner S1 (Aquilion ONE, Toshiba), and 110 patients on scanner S2 (Revolution CT, GE Healthcare). Hounsfield units were measured using regions of interest in the descending aorta at 6 points (cranial slice, level of the visualized first, second, third, and fourth spines, and the caudal slice). We also compared the coverage length (z-axis) of the full width field of view between a single rotation of the two scanners. Results: Evaluating mean CT attenuation values Hounsfield units through the scan range, are progressively reduced across the descending aorta in the S1 group, resulting in the larger difference of contrast brightness between the cranial and caudal slices compared to the S2 group (absolute difference: S2 13.0 § 4.4 vs S1 141.9 § 16.4, p < 0.0001; Percent difference: 19.3 § 2.1 vs ¡3.4 § 1.2, <0.0001). The standard deviation (SD) is similar at the cranial slice between the two scanners, however, the S1 group demonstrated higher SD-differential from cranial to caudal than S2 group. Median radiation exposure was significantly lower for the S2 scanner 1.50 § 0.75 mSv vs the S1 system 1.9 mSv (IQR 1.7 2.7 mSv) (p = 0.01). Z-axis coverage was larger for the S2 scanner 152.5 mm (244 slices £ 0.625 mm/slice) than 133 mm for S1 (266 slices £ 0.5 mm/slice). Conclusion: Although both “volume” scanners cover the whole heart z-axis with one beat, scans using the S1 scanner have a larger variability in attenuation values throughout the scan range, resulting in 20% increase in nonuniformity from cranial to caudal slice. Additionally, SD variation across the field of view, a metric of noise, is larger when using the S1 scanner vs the S2 scanner. These results indicate that the GE Revolution CT has more uniform contrast enhancement and more coverage, lower radiation and lower image noise compared to the current Toshiba Aquilion ONE system. Key Words: Coronary computed tomographic angiography; image quality; diagnosis; cardiac CT. © 2019 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.
Acad Radiol 2019; &:1 7 From the Department of Medicine, Los Angeles Biomedical Research Institute, Torrance, CA (N.P., D.L., R.N., B.F., E.J., M.N., M.J.B.); Department of Cardiovascular Medicine, Toho University Graduate School of Medicine, Tokyo, Japan (R.N.); Centro Cardiologico Monzino, IRCCS, Milan, Italy (D.A., G.P., E.C.); Department of Clinical Sciences and Community Health, Cardiovascular Section, University of Milan, Italy (D.A.); Department of Medical Imaging, The Prince Charles Hospital, Brisbane, Queensland Australia (R.O., C.H.-C.); University of Queensland, Brisbane, Queensland, Australia (R.O., C.H.-C.). Received November 15, 2018; revised December 31, 2018; accepted January 2, 2019. Funding: This study was funded by General Electric Inc. The company had no role in study design, conduct, or manuscript preparation. Dr Budoff receives grant support from the National Institutes of Health and General Electric. Address correspondence to: M.J.B. e-mail: [email protected] © 2019 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved. https://doi.org/10.1016/j.acra.2019.01.002
1
ARTICLE IN PRESS PATEL ET AL
INTRODUCTION
C
oronary computed tomographic angiography (CTA) has emerged as a promising noninvasive method for the detection and exclusion of obstructive coronary artery disease due to the very high negative predictive values (1,2). Recent advancements in CT scanner technology have enabled even further improvement in both image quality and diagnostic accuracy. Some advancements include increased temporal resolution through the introduction of dual source and software based motion correction, high definition detectors, iterative reconstruction, and advanced work stations. (3). Additionally, two CT system manufacturers (Toshiba, GE) have released systems with larger volume of coverage in the z-axis that reduce misalignment and “stair-step artifacts” and improve myocardial and blood pool homogeneity by allowing for whole heart coverage without need for table movement. There have been studies demonstrating the improved image quality associated with these larger coverage scanners as compared to 64-slice CT technology (4). However, there has not, to our knowledge, been a direct comparison between the two wide coverage systems, specifically in the assessment of myocardial and blood pool homogeneity. The purpose of this study is to make such a direct comparison between the current wide coverage CTs in terms of myocardial and blood pool homogeneity.
MATERIAL AND METHODS This study prospectively consented and evaluated consecutive patients who underwent coronary CTA studies at multiple sites as part of the CONVERGE Registry. Sites included Los Angeles Biomedical Research Institute, Torrance, CA; Centro Cardiologico Monzino, IRCCS, Milan, Italy; The Prince Charles Hospital, Brisbane, Queensland Australia and Baptist Hospital of Miami Florida. Two consecutive patient cohorts from multiple institutions were evaluated, each cohort obtained concurrently from March 2017 through August 2017. The first cohort (S1, n = 126) images were acquired in one heart cycle with volumetric coverage of 16 cm (320 slice CT £ 0.5 mm) (Aquilion ONE, Toshiba) and a gantry rotation speed of 0.35 s/rotation. The second cohort (S2, n = 110) were acquired in one heart cycle with a volumetric coverage of 16 cm (256 slice CT £ 0.625 mm) (Revolution CT, GE Healthcare) and a gantry rotation speed of 0.28 s/rotation. Inclusion criteria were weight <300 lbs and no prior history of cardiac surgery or percutaneous coronary intervention since these procedures would interfere with the measurement of coronary atherosclerosis or stenosis. Exclusion criteria for this study included atrial fibrillation, chronic kidney disease (estimated glomerular filtration rate <60 mL/min/1.73 m2 within 30 days of the CT) or a history of intravenous contrast allergy (5,6). Coronary CTA Scan Protocol and Image Reconstruction
The clinical coronary CTA exams were performed with beta blockade for heart rate (HR) control as per clinical routine if 2
Academic Radiology, Vol &, No &&, && 2019
their resting HR was greater than 60 beats per minute (bpm). The protocol was for patients undergoing coronary CTA to be given 100 mg metoprolol orally 1 hour before the scan if the HR was greater than 60 bpm and if needed up to 20 mg intravenous metoprolol titrated to achieve HR <60 bpm. We used prospectively ECG-triggered coronary CTA technique for both scanners. Both scanners used a 20 cm field of view for acquisition of coronary CTA's. The contrast injection was performed using a power injector (Stellant; Medrad, Warrendale, Pennsylvania) through an antecubital vein at a rate of 5.0 mL/s. A triple bolus injection protocol was utilized. Patients were given an injection of 60 cc of contrast in the first phase, followed by 20 cc of contrast and 30 cc of saline in the second phase and completed by 50 cc of saline. Scan Acquisition Protocol
For both scanners, the field of view (z-axis) included the mid ascending aorta to the upper abdomen. No table movement was required due to the wide volumetric acquisition. The z-axis collimation was selected based on the scout images demonstrating the heart size. No patient required more than 16-cm z-axis coverage. Tube voltages used fixed at 120 kilovolt potential (kVp), to provide comparable radiation exposure. All acquisitions were prospectively ECG triggered. Each scan was done in a single-beat acquisition within one cardiac cycle, regardless of HR. Scanner S1 Acquisition (Aquilion ONE): rotation speed was 350 milliseconds, with no table motion. The mA settings depend on the body mass index (BMI) of the patient, and set by the radiographer as follows: BMI < 20 kg/m2, 120 kVp 300 mA; BMI 21 25 kg/m2, 120 kVp 350 mA; BMI 26 30 kg/m2, 120 kVp 400 mA; BMI > 31 kg/m2,120 kVp 500 mA. Scanner S2 Acquisition (GE Revolution): rotation speed was 280 milliseconds, with no table motion. Tube current ranged between 122 and 740 mA, based upon the BMI of the patient, automatically determined by the system. A medium field of view was selected for all patients. The scanner is equipped with an “autogating” capability, which automatically adjusts HR-dependent settings for triggered acquisition and gated reconstruction. Autogating was used to automatically acquire diastolic phases for lower HRs and both systolic and diastolic phases for higher HRs. Image Reconstruction
For both scanner types, data reconstruction was performed using thin reconstructions with intervals ranging at 60 80% of the R-R cycle, and most of data were reconstructed at 75% of the R-R cycle. Iterative reconstruction algorithms (ADIR for S1, ASIR-V for S2) level of 40% was used. For CT exams performed on scanner S1, adaptive motion correction was applied to minimize motion. For CT exams performed on scanner S2, motion correction software (“SnapShot Freeze;” GE Healthcare) was used for correcting motion artifacts in patients with higher HRs.
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
COMPARISON OF CT SCANNERS FOR CARDIAC CT
All images were transferred to a single vendor workstation to provide standardized postprocessing for both scanner acquisitions (AW 4.7, GE Healthcare). An SCCT level-3 trained CT cardiologist read all CT scans at a central reading center (Los Angeles Biomedical Research Institute at Harbor UCLA in Torrance, California). Evaluation of Image Quality
CT images were analyzed at the core CT reading center by trained, experienced readers blinded to participants' clinical information and type of scanner used. All CT reads were done in the central CT core lab at Los Angeles Biomedical Research Institute in Torrance CA. Both readers are radiologists with >20 years experience reading cardiovascular CT imaging. Signal intensity was derived from CT attenuation values measured in Hounsfield units (HU) from a region of interest. All quantitative measurements were performed on axial source coronary CTA images by placing a circular region of interest of 100 mm2 in the descending aorta and left ventricular blood pool. The mean and SD of the HU values were measured in the descending aorta at 6 points (cranial slice, level of the visualized first, second, third, and fourth spines, and level of caudal slice; Figs. 1 and 2). Coverage We first measured full coverage in the z-axis for each scanner. We evaluated the total number of slices demonstrating the entire field of view present to determine coverage length of the scanners. However, both scanners demonstrate a smaller concentric circle of imaging from the detectors most distant from the x-ray tube (cranial- and caudal-most slices), and slowly increase visualization of the structure, so we also measured the coverage length along the z-axis which enabled a full width field of view (excluding those images without full axial visualization). Both scanners have truncated cranial and caudal slices, not allowing for visualization of the full field of view, and these slices were excluded from measure and analysis related to coverage length for each scanner. Radiation Dose The effective radiation dose in both systems was calculated using the following formula: Effective radiation dose = Dose length product £ Conversion coefficient for the chest (k = 0.014 mSv/mGy cm).
Statistical Analysis
Analyses were performed using statistical software (SAS version 9.1; SAS Institute, Cary, North Carolina). A statistically significant difference was defined as a p value (two-tailed) less than 0.05. Continuous variables were expressed as mean § SD, or quartile ranges. A two-sided t test was applied when the distribution of data was of equal variance, and Welch Satterthwaite Students t test was used when unequal variance was found. Multivariable analysis was done to assess the subjects age, gender, and BMI affect the results between scanners and if so how much of the variance is based on these factors and how much is based on which scanner is used in determining uniformity of age, gender, and BMI were entered as covariates. Logistic regression model was performed to determine if age, gender, and BMI were predictors for the contrast enhancement in the z axis, image noise (SD variation in the z axis) and radiation dose. No statistically significant differences were found for interactions between scanner type and the covariates (BMI, gender, and age). RESULTS Patient Demographics
There were no statistical differences between gender, patient age, BMI, and HR between the scanner groups (Table 1). Radiation Dose
We found that the median radiation dose was significantly lower for S2 scanner at 1.50 § 0.75 millisieverts (mSv) compared to 1.9 mSv (interquartile range, 1.7 2.7 mSv) for the S1 scanner (p = 0.01). Quantitative Analysis
Although the mean HU in the S2 group is higher than that in the S1 group at the cranial slice (618.3 § 19.1 vs 431.5 § 13.4, p < 0.0001), HUs are progressively reduced across the descending aorta in the S1 group, resulting in the larger difference of contrast brightness between the cranial and caudal slices compared to the S1 group (Absolute difference: S2 13.0 § 4.4 vs S1 141.9 § 16.4, p < 0.0001; Percent difference: 19.3 § 2.1vs ¡3.4 § 1.2, p <0.0001; Fig 1a,b). The SD is similar at the cranial slice between the two scanners, however, the S1 group indicated higher noise (increased SD of HU) than S2 group at the other levels in the descending aorta (Table 2).
TABLE 1. Demographics Patient Characteristics
Total
Age (years), Range: 28 70 BMI (kg/m2) Gender (% male) Heart rate (bpm)
55 § 6.9 27.9 § 3.2 62% 57 § 6
S1 Scanner (median § S.D.)
S2 Scanner (median § S.D.)
p Value (S1 vs S2)
55 § 9.8 28.0 § 4.0 61% 56 § 8
56 § 8.4 27.8 § 3.9 62% 59 § 7
0.49 0.76 0.89 0.24
SD, standard deviation.
3
ARTICLE IN PRESS PATEL ET AL
Academic Radiology, Vol &, No &&, && 2019
Figure 1. (a) Distribution of mean HU across descending aorta for GE (Scanner 1) and Toshiba (Scanner 2), (b) Distribution of Standard deviation of mean HU across descending aorta for GE (Scanner 1) and Toshiba (Scanner 2). HU, Hounsfield units.
Coverage
DISCUSSION
Total coverage (allowing for at least the heart to be seen centrally) for the S2 scanner was 160 mm (256 slices at 0.625 mm/slice) as compared to 135.5 mm for S1 (542 slices £ 0.25 mm). Full field of view coverage was greater for the S2 scanner 152.5 mm (244 slices/rotation £ 0.625 mm/slice), compared to S1 scanner 133.25 mm (533 slices/rotation £ 0.25 mm/slice).
CTA started with single slice scanners (electron beam tomography) over 20 years ago (7 9). Over the last 20 years, coverage has progressively increased, with more detectors available. Currently, two computed tomography systems have whole heart coverage, with at least 10 cm of coverage per rotation, also known as wide multi detector computed tomography. In this
4
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
COMPARISON OF CT SCANNERS FOR CARDIAC CT
TABLE 2. Image Quality of cohorts for Scanner 1 and Scanner 2 at different levels of the cardiac study GE Revolution (n = 110)
Toshiba (n = 126)
Top slice Mean HU 431.5 § 13.4 618.3 § 19.1 SD HU 38.2 § 0.7 55.9 § 10.0 Aorta at the level of the first spine Mean HU 439.7 § 13.8 616.5 § 20.1 SD HU 36.3 § 0.9 42.4 § 1.2 Aorta at the level of the second spine Mean HU 441.8 § 13.2 561.5 § 18.4 SD HU 36.4 § 0.8 41.9 § 1.0 Aorta at the level of the third spine Mean HU 445.7 § 13.6 548.2 § 17.9 SD HU 37.3 § 1.9 49.8 § 5.7 Aorta at the level of the fourth spine Mean HU 448.1 § 14.1 537.1 § 17.0 SD HU 40.7 § 3.2 44.0 § 0.9 Bottom slice Mean HU 441.5 § 14.1 470.4 § 13.9 SD HU 43.0 § 0.8 46.6 § 0.9 Absolute difference (Top¡Bottom) Mean HU ¡13.0 § 4.4 141.9 § 16.4 SD HU ¡4.7 § 1.1 ¡7.6 § 1.1 Percent difference (Top¡Bottom) Mean HU ¡3.4 § 1.2 19.3 § 2.1 SD HU ¡16.5 § 3.1 ¡25.9§3.4
p Value
<0.0001 0.84 <0.0001 0.0001 <0.0001 0.0001 0.0001 <0.0001 0.0002 <0.0001 0.17 0.007 <0.0001 0.056 <0.0001 0.061
prospective, multicenter, observational cohort study, we compared image quality from consecutive coronary CTA scans performed on two “whole-heart” single-rotation scanner platforms; the GE Revolution, and the Toshiba Aquilion ONE. One of the key benefits of whole-heart imaging is achieving signal homogeneity from isophasic acquisition, which may have significant value in myocardial CT perfusion as well as other quantitative applications. Therefore, we sought to quantitate the uniformity throughout the cranio-caudal z-axis in clinical exams
of patients with suspected coronary artery disease, and to compare true z-axis coverage length and radiation exposures. The Revolution CT and the Toshiba scanner both have 160 mm coverage as quoted by the manufacturers. While the Revolution CT achieves 160 mm coverage in cardiac mode, the Aquilion One only allowed for reconstruction of 135.5 mm. Furthermore, when analyzing the portion of the effective coverage (the number of slices which contain diagnostic information across the full field of view), it was significantly greater on the Revolution than on the Toshiba system (152.5 mm vs 133.25 mm, respectively, p = 0.011). While both scanners covered the whole heart within a single rotation, scans on the Aquilion ONE have a larger variability of attenuation values throughout the scan range, resulting in 20% reduction in contrast brightness from cranial slice to caudal slice. SD variation is also larger on the Aquilion ONE than that the Revolution system, indicating the latter has more consistent contrast enhancement while maintaining less image. We also found that median radiation exposure on the Aquilion ONE scanner (S1) was higher than the Revolution scanner (S2). In this study, we evaluated coronary CT angiography images from two commercially available systems with wide-detector whole-heart coverage (quoted 16 cm scan range). Both systems allow acquisition of the entire heart without moving the table. This inherently removes misalignment artifacts, and has the promise to improve signal homogeneity due to isophasic acquisition, which has value in emerging applications such as myocardial CT perfusion imaging and atrial fibrillation acquisitions. However, the physics of wide coverage CT also presents unique challenges such as increased heel effect and scatter that confound the goal of isophasic signal homogeneity. Our study indicates that, as compared to the wide coverage Aquilion ONE system, the Revolution CT system yielded significant improvement in contrast uniformity and noise variation in the z-axis, and with significantly lower radiation exposure. In addition, the portion of the 16 cm z-coverage which allowed diagnostic visualization of the structures within the full field of view (excluding those slices with
Figure 2. Assessment of mean and SD of HU on the descending aorta. HU, Hounsfield units; SD, standard deviation.
5
ARTICLE IN PRESS PATEL ET AL
truncated images,) was significantly larger with the Revolution than the Aquilion ONE. The Revolution system contains a few unique characteristics that may explain these findings. The GE Revolution employs 3D collimator situated between patient and detector, which blocks scattered photons prior to reaching the detector bank, as distinct from other vendor systems which attempt to collimate the beam prior to reaching the patient. Additionally, the Revolution system has an integrated “Volume HD” reconstruction platform that contains an algorithm call Multi-Material Artifact Reduction. This algorithm performs beam hardening correction, but takes into account the spectrum variation across the field of view due to the heel effect. One possible explanation of the differences seen in consistency of images is the use of different detectors by each system. The Revolution system uses the “Gemstone” garnet-based detector, while “PUREVISION” detector used in Toshiba system is forged from a solid ceramic ingot. Compared to ceramic (Gd2O2S) scintillator, the garnet-based detector has 100-times faster decay time (faster rate of light being emitted after the detector is excited) and 75% shorter afterglow (time needed for the material to return to pre-excitation state, ready for the next excitation) (10). Overall, these properties translate into improved spatial resolution and reduced image noise due to increased data sampling per gantry rotation, which is conventionally limited by the properties of the ceramic material. A proprietary 3D collimator in GE scanner ensures contrast uniformity and minimizes scatter and beam hardening artifacts associated with wide coverage systems, which is supported by the findings of the present study. Along with such hardware innovations, advancements in spectral modeling, projection-based material decomposition, and noise modeling have led to increased uniformity across the full z-coverage, as demonstrated by our quantitative data (11 14). Limitations
The limitations of the study include the study design, which was a prospective multicenter observational registry that did not allow for each patient to be scanned twice (once on each scanner) due to logistics as well as concerns about radiation exposure. However the patient groups were well matched, and the postprocessing standardized for both scanner acquisitions. Each scanner uses different determinations of tube current based on subject size/BMI. While we carefully matched and controlled for age and BMI, there are intrinsic differences in the acquisition parameters and perhaps in in body morphology despite similar BMI that we could not control for. The goal was to identify, using clinical scan parameters used by each scanner to see if this results in differences in image quality. It should be noted that image quality is also affected by ECG triggering and other algorithms that may differ between vendors, but these variables are beyond the scope of this study (15 17). 6
Academic Radiology, Vol &, No &&, && 2019
CONCLUSION We identified significant improvements in contrast uniformity, noise variation, full-field scan range, and decreased radiation exposure using the Revolution CT compared to the Aquilion One system. The clinical implications of the results are significant. Lower radiation exposures is always clinically important, and increased scan length (coverage) allows for single rotation scanning of more patients, including bypass and aortic studies depending on height and scan length of the patient. Lower image noise will impact overall image quality and allow for better acquisition for postprocessing of such applications as fractional flow reserve by CT (18), plaque imaging, transluminal gradient assessment, pericoronary fat attenuation and other measures that require superb image quality. The revolution CT allows for improved quantitative uniformity of iodinated contrast across the whole z-axis scan range, which may be advantageous for applications such as myocardial CT perfusion imaging, whilst at the same time improving dose performance. DISCLOSURE Dr. Budoff receives grant support from General Electric. Dr Hamilton-Craig has received research support from SanofiGenzyme and Siemens. REFERENCES 1. Budoff MJ, Dowe D, Jollis JG, et al. Diagnostic performance of 64-multidetector row coronary computed tomographic angiography for evaluation of coronary artery stenosis in individuals without known coronary artery disease: results from the prospective multicenter ACCURACY (Assessment by Coronary Computed Tomographic Angiography of Individuals Undergoing Invasive Coronary Angiography) trial. J Am Coll Cardiol 2008; 52(21):1724–1732. 2. Wang H, Xu L, Fan Z, et al. Clinical evaluation of new automatic coronary-specific best cardiac phase selection algorithm for single-beat coronary CT angiography. PLoS One 2017; 12(2). 3. Alani A, Nakanishi R, Budoff MJ. Recent improvement in coronary computed tomography angiography diagnostic accuracy. Clin Cardiol 2014; 37(7):428–433. 4. Fujimoto S, Matsutani H, Kondo T, et al. Image quality and radiation dose stratified by patient heart rate for coronary 64- and 320-MDCT angiography. AJR Am J Roentgenol 2013; 200(4):765–770. 5. Korada SKC, Zhao D, Tibuakuu M, et al. Frailty and subclinical coronary atherosclerosis: the Multicenter AIDS Cohort Study (MACS). Atherosclerosis 2017; 266:240–247. 6. Schuetz GM, Schlattmann P, Achenbach S, et al. Individual patient data meta-analysis for the clinical assessment of coronary computed tomography angiography: protocol of the Collaborative Meta-Analysis of Cardiac CT (CoMe-CCT). Syst Rev 2013; 2(1):13. 7. Budoff MJ, Oudiz RJ, Zalace CP, et al. Intravenous three dimensional coronary angiography using contrast enhanced electron beam computed tomography. JACC 1997; 29(2):8003 8003. 8. Lu B, Dai RP, Jing BL, et al. Evaluation of coronary artery bypass graft patency using three-dimensional reconstruction and flow study on electron beam angiography. J Comput Assist Tomogr 2000; 24:663–670. 9. Budoff MJ, Lu B, Mao S. Gadolinium-enhanced three-dimensional electron beam coronary angiography. J Comput Assist Tomogr 2002; 26 (Nov-Dec(6)):879. 10. Jiang H, Vartuli J, Vess C. Gemstone—the ultimate scintillator for computed tomography Haochuan Jiang CT detectors. GE white paper; c 2009. 11. Min JK, S. R, Vass M, et al. High-definition multidetector computed tomography for evaluation of coronary artery stents: comparison to
ARTICLE IN PRESS Academic Radiology, Vol &, No &&, && 2019
standarddefinition64-detector row computed tomography. J Cardiovasc Comput Tomogr. 2009; 3(Jul-Aug(4)):246–251. 12. Nakanishi R, Matsumoto S, Alani A, et al. Diagnostic performance of transluminal attenuation gradient and fractional flow reserve by coronary computed tomographic angiography (FFRCT) compared to invasive FFR: a sub-group analysis from the DISCOVER-FLOW and DeFACTO studies. Int J Cardiovasc Imaging 2015; 31:1251–1259. 13. Gupta M, Kadakia J, Jug B, et al. Detection and quantification of myocardial perfusion defects by resting single-phase 64-slice cardiac computed tomography angiography compared with SPECT myocardial perfusion imaging. Coron Artery Dis. 2013; 24(4):290–297. 14. Cury RC, Kitt TM, Feaheny K, et al. A randomized, multicenter, multivendor study of myocardial perfusion imaging with regadenoson CT perfusion vs single photon emission CT. J Cardiovasc Comput Tomogr 2015; 9(2):103–112. e2.
COMPARISON OF CT SCANNERS FOR CARDIAC CT
15.
16.
17.
18.
Gopal A, Mao SS, Karlsberg D, et al. Radiation reduction with prospective ECG-triggering acquisition using 64-multidetector computed tomographic angiography. Int J Cardiovasc Imaging 2009; 25(4):405–416. Mao SS, Bakhsheshi H, Lu B, et al. Effect of ECG triggering on reproducibility of coronary artery calcium scoring. Radiology 2001; 220(Sep 3): 707–711. Gupta M, Hacioglu Y, Kadakia J, et al. Left ventricular volume: an optimal parameter to detect systolic dysfunction on prospectively triggered 64multidetector row computed tomography: another step towards reducing radiation exposure. Int J Cardiovasc Imaging 2011; 27:1015–1023. Nakanishi R, Matsumoto S, Alani A, et al. Diagnostic performance of transluminal attenuation gradient and fractional flow reserve by coronary computed tomographic angiography (FFRCT) compared to invasive FFR: a sub-group analysis from the DISCOVER-FLOW and DeFACTO studies. Int J Cardiovasc Imaging 2015; 31:1251–1259.
7