Monoenergetic extrapolation of cardiac dual energy CT for artifact reduction

Abstract

Background

Monoenergetic extrapolation of cardiac dual-energy computed tomography (DECT) could be useful in artifact reduction in clinical practice.

Purpose

To evaluate the potential of monoenergetic extrapolation of cardiac DECT data for reducing artifacts from metal and high iodine contrast concentration.

Material and Methods

With IRB approval and in HIPAA compliance, 35 patients (22 men, 61 ± 12 years) underwent cardiac DECT with dual-source CT (100 kVp and 140 kVp). Contrast material injection protocols were adapted to the patient’s weight using non-ionic low-osmolar 370 mgI/mL iopromide. Datasets were transferred to a stand-alone workstation and dedicated monoenergetic analysis software was used for postprocessing. Reconstructions with the following five photon energies were generated: 40 keV, 60 keV, 80 keV, 100 keV, and 120 keV. Artifact severity was graded on a 5-point Likert scale (0, massive artifact; 5, absence of artifact). The size of artifact and image noise (expressed as HU) in anatomic structures adjacent to the artifact were measured. Quantitative and subjective image quality was compared using Friedman and Wilcoxon tests.

Results

We observed artifacts arising from densely concentrated contrast material in the superior vena cava (SVC) in 18 patients, from sternal wires in 14, from bypass clips in eight, and from coronary artery stents in seven. Artifact size in monoenergetic reconstructions from 40 to 120 keV decreased from 21.3 to 19 mm for the SVC (P < 0.001), from 8.4 to 2.6 mm for sternal wires (P < 0.001), from 6.4 to 2.2 mm for bypass clips (P < 0.001), and from 5.9 to 2.7 mm for stents (P < 0.001), respectively. The quality score changed from 0.2 to 3.8 for the SVC (P < 0.001), from 0.1 to 4 for sternal wires (P < 0.001), from 0 to 3.9 for bypass clips (P < 0.001), and from 0 to 3.9 for stents (P < 0.001). Lowest noise in adjacent structures was found at 80 keV for the SVC (39.1 HU), for sternal wires (33.3), for bypass clips (26.9), and for stents (33.9).

Conclusion

A significant reduction of high-attenuation artifacts can be achieved by use of higher monoenergetic energy levels with cardiac DECT. However, image noise in anatomic structures affected by artifacts is lowest at 80 keV, which suggests an evaluation approach that makes use of multiple energy levels for a complete diagnosis.

Keywords

Dual energy CT (DECT)artifact monoenergetic extrapolation

Introduction

A dual-energy computed tomography (DECT) postprocessing method was developed in the 1980s to reduce beam-hardening artifacts in CT image reconstructions (1,2). However, this DECT technique has not been widely implemented in the clinical setting because it can confer reduced spatial resolution, unstable CT attenuation, and insufficient tube current at low tube voltages (3).

DECT is a long-recognized technology with clinical applications arising from its ability to provide material-specific and energy-specific information (3,4). These applications include but are not limited to analysis of renal stone composition, detection and quantification of tissue lipid and iron content, and optimization of enhancement with iodine-based contrast materials (5 –9). In dual-energy techniques, X-ray sources are operated at two different tube potentials, generating photons at two different energy levels. Characteristic changes in attenuation over the range of photon energies allow differentiation of materials with different atomic numbers.

Few experiences in cardiac DECT have been published (10 –14) and, to our knowledge, none has investigated the reduction of artifacts from metallic hardware or high iodine contrast concentration in cardiac CT.

The purpose of our investigation was to evaluate the potential of monoenergetic extrapolation of cardiac DECT data for reducing artifacts from metal and high iodine contrast concentration.

Material and Methods

Patients

Our Institutional Review Board approved the study. The study was conducted in HIPAA compliance. Image data of 35 symptomatic, consecutive patients with suspected CAD who had been referred for dual-source CT-based cardiac DECT were analyzed. Patients with a contraindication to iodinated and/or beta-blocking drugs and with a compromised renal function defined as creatinine of ≥1.2 mg/dL were excluded. Among the 35 patients, 13 were women and 22 were men. The mean age was 61 ± 12 years and the mean body mass index (BMI) was 31 ± 9 kg/m². All demographic data of our population are shown in Table 1.

Table 1.

Demographic data of study population.

Age (years)	61 ± 12
Height (cm)	166 ± 18
Weight (kg)	84 ± 17
BMI (kg/m²)	31 ± 9
Diabetes (n)	12
Dyslipidemia (n)	30
Smoke (n)	20
Family history of CAD (n)	23

Age, height, weight, and BMI values are expressed as mean ± standard deviation.

BMI, body mass index; CAD, coronary artery disease.

CT protocol

All examinations were performed on a second generation dual-source CT system (Somatom Definition Flash; Siemens Medical Solutions, Forchheim, Germany). DECT scan was performed after scout with the following parameters: tube voltages, 100 kVp and 140 kVp; gantry rotation time, 280 ms; collimation, 0.6 mm; reconstructed section width, 0.75 mm; and reconstruction increment, 0.3 mm. All 35 patients were scanned using commercially available automated anatomical tube current modulation software (CAREDose4D, Siemens Medical Solutions). Electrocardiography (ECG)-synchronization was used. Scan protocols used retrospective ECG gating in cases of irregular heart rate, heart rates >80 bpm, or when functional assessment was desired. Otherwise, prospective ECG triggering was used with scan acquisition during diastole (70% RR) for slower heart rates (<70 bpm) and during systole (40% RR) for faster heart rates. For contrast medium enhancement, bolus tracking was used with automated study commencement when the attenuation reached 100 HU in a region of interest (ROI) within the ascending aorta with a 6-s delay. A triphasic contrast medium injection protocol was used: undiluted contrast agent was followed by 30 mL of a 30%/70% contrast medium/saline mixture and a 40-mL saline flush, all administered at a flow rate of 5 mL/s. Contrast material injection protocols were adapted to each patient’s weight, using 1 mL per kg of nonionic low-osmolar 370 mgI/mL iopromide (Ultravist, Bayer Healthcare, Berlin, Germany). The contrast was injected via a power-injector (Stellant D, Medrad, Indianola, PA, USA).

Image reconstructions and quantitative and subjective image quality analysis

Datasets were transferred to a stand-alone workstation (Syngo MMWP VE 36 A, Siemens Medical Solutions) and dedicated monoenergetic analysis software was used for postprocessing. Reconstructions with the following five photon energies were generated: 40 keV, 60 keV, 80 keV, 100 keV, and 120 keV.

Artifact severity was graded on a 5-point Likert scale by an observer with 4 years of experience in cardiac CT. A score of 0 was defined as massive artifacts influencing the evaluation of adjacent tissue in an area >10 mm² around the artifact and/or not allowing the distinction of artifact from blood pool; a score of 1 was defined as submassive artifacts, producing an image alteration in an area of 5 mm² around the artifact and/or partially inhibiting the distinction of artifact from blood pool; a score of 2 as moderate artifacts producing an image alteration in an area <5 mm² around the artifact and/or partially inhibiting the distinction of artifact from blood pool; a score of 3 was defined as mild artifacts influencing the evaluation of adjacent soft tissue but allowing the distinction of artifact from blood pool; a score of 4 as mild artifacts not influencing the evaluation of adjacent soft tissue; finally, a score of 5 was defined as absence of artifacts and clear visualization of adjacent tissue. The size of artifacts (largest diameter), expressed in mm, was measured. Image noise was measured as the standard deviation (SD) of Hounsfield units (HU) within a ROI placed in anatomic structures adjacent to the artifacts. We analyzed artifacts produced by sternal wires, stents, surgical clips, and densely concentrated contrast material in the superior vena cava (SVC).

Statistical analysis

Quantitative and subjective image quality scores were compared between different monoenergetic reconstructions using Friedman and Wilcoxon tests. Continuous variables were expressed as mean ± SD and categorical variables as frequencies or percentages. Friedman test was used to compare overall keV groups for artifact size, subjective artifact severity, and image noise. The non-parametric Wilcoxon signed rank test was used to compare artifact size, subjective artifact severity, and image noise between pairs of different keV reconstructions: 40 keV vs. 60, 60 vs. 80, 80 vs. 100, and 100 vs. 120. All P values were two-sided with a value <0.05 considered statistically significant. Statistical analyses were performed using commercially available software (SPSS release 17; SPSS Inc., Chicago, IL, USA).

Results

We observed artifacts arising from densely concentrated contrast material in the SVC in 18 patients (Fig. 1), from sternal wires in 14 patients (Fig. 2), from bypass clips in eight patients (Fig. 3), and from coronary artery stents in seven patients (Fig. 4). Artifact size in monoenergetic reconstructions (40 keV, 60 keV, 80 keV, 100 keV, and 120 keV) decreased from 8.4 ± 1.8 mm (mean ± SD) to 2.6 ± 0.8 mm for sternal wires (P < 0.001), from 21.3 ± 2.6 mm to 17.4 ± 1.9 mm for the SVC (P < 0.001), from 5.9 ± 1.1 mm to 2.7 ± 1.1 mm for stents (P < 0.001) and from 6.4 ± 2.2 mm to 2.2 ± 1.1 mm for bypass clips (P < 0.001). The quality score changed, from 0.1 ± 0.3 to 4 ± 0 for sternal wires (P < 0.001), from 0.2 ± 0.5 to 3.8 ± 0.4 for the SVC (P < 0.001), from 0 ± 0 to 3.9 ± 0.4 for stents (P < 0.001), and from 0 ± 0 to 3.9 ± 0.4 for bypass clips (P < 0.001). Lowest artifact size was found at 120 keV for sternal wires (2.6 ± 0.8 mm) and bypass clips (2.2 ± 1.1 mm); artifact extension was equal at 100 keV and 120 keV for the SVC (17.4 mm) and stents (2.7 mm). The highest quality score was found at 120 keV for sternal wires (4 ± 0), stents (3.9 ± 0.4), bypass clips (3.9 ± 0.4), and the SVC (3.8 ± 0.4, equal to 100 keV). The lowest noise in adjacent structures was found at 80 keV for the SVC (39.1 HU-SD), sternal wires (33.3 HU-SD), bypass clips (26.9 HU-SD), and stents (33.9 HU-SD). All data are shown in Table 2 and Fig. 5. P values for comparisons of different keV reconstructions are presented in Table 3.

Fig. 1.

Artifact arising from densely concentrated contrast material in the superior vena cava at different keV: (a) 40, (b) 60, (c) 80, (d) 100, and (e) 120. For all images, window level was 200 and width 600.

Fig. 2.

Artifact arising from sternal wires and bypass clips at different keV: (a) 40, (b) 60, (c) 80, (d) 100, and (e) 120. For all images, window level was 200 and width 600.

Fig. 3.

Artifact arising from bypass clips at different keV: (a) 40, (b) 60, (c) 80, (d) 100, and (e) 120. For all images, window level was 200 and width 600.

Fig. 4.

Artifact arising from a stent at different keV: (a, f) 40, (b, g) 60, (c, h) 80, (d, i) 100, and (e, l) 120. For all images, window level was 200 and width 600.

Fig. 5.

Graphs of artifact size, quality score, and noise at different kiloelectron voltage (keV). HU, Hounsfield units; SD, standard deviation; SVC, superior vena cava.

Table 2.

Artifact size, quality score, and noise at different keV.

keV 40	keV 60	keV 80	keV 100	keV 120	P
Artifact size (mm)
Sternal wires	8.4 ± 1.8	5.6 ± 1.8	4.0 ± 0.9	3.1 ± 0.9	2.6 ± 0.8	<0.001
SVC	21.3 ± 2.6	19 ± 2.0	17.7 ± 2.3	17.4 ± 1.9	17.4 ± 1.9	<0.001
Stents	5.9 ± 1.1	4.6 ± 0.8	3.6 ± 1.0	2.7 ± 1.1	2.7 ± 1.1	<0.001
Clips	6.4 ± 2.2	3.6 ± 1.3	2.8 ± 1.2	2.6 ± 1.1	2.2 ± 1.1	<0.001
Quality score
Sternal wires	0.1 ± 0.3	2.1 ± 0.8	3.3 ± 0.6	3.7 ± 0.5	4 ± 0	<0.001
SVC	0.2 ± 0.5	1.7 ± 0.7	3.3 ± 0.7	3.8 ± 0.4	3.8 ± 0.4	<0.001
Stents	0 ± 0	1.4 ± 0.5	2.9 ± 0.7	3.7 ± 0.5	3.9 ± 0.4	<0.001
Clips	0 ± 0	1.9 ± 0.8	3.5 ± 0.8	3.8 ± 0.5	3.9 ± 0.4	<0.001
Noise (HU-SD)
Sternal wires	223.1	74.5	33.3	44.7	43.9	<0.001
SVC	349.6	99.3	39.1	44.7	57.4	<0.001
Stents	261.9	93.8	33.9	35.6	63.7	<0.001
Clips	210	85.6	26.9	42.3	44.1	<0.001

All values are expressed as mean ± standard deviation (SD).

All P values are calculated with Friedman test.

Table 3.

All P values for comparisons of different keV reconstructions.

keV 40 vs. 60	keV 60 vs. 80	keV 80 vs. 100	keV 100 vs. 120
Artifact size (mm)
Sternal wires	0.001	0.001	0.005	1
SVC	0.001	0.008	0.249	0.941
Stents	0.005	0.003	0.166	0.655
Clips	0.017	0.117	0.102	0.149
Quality score
Sternal wires	0.001	0.001	0.014	0.046
SVC	<0.001	<0.001	0.002	1
Stents	0.003	0.002	0.014	1
Clips	0.011	0.017	0.317	0.317
Noise (HU-SD)
Sternal wires	0.001	0.001	0.258	0.925
SVC	<0.001	<0.001	0.097	0.033
Stents	0.012	0.012	0.575	0.017
Clips	0.018	0.017	0.017	0.160

All P values are calculated with Wilcoxon test. HU, Hounsfield units; SD, standard deviation; SVC, superior vena cava.

Discussion

In this study, we provided initial evidence that high-attenuation artifacts can be substantially reduced using higher monoenergetic energy levels based on cardiac DECT. The results indicate a significantly improved diagnostic assessment of the structures surrounding sternal wires, stents, surgical clips, and high contrast concentrations in the SVC. We also showed that 80 keV may serve as an effective energy setting to reduce image noise in anatomic structures located near artifacts.

DECT acquisitions, which are increasingly used in radiological imaging, may provide an attractive option to mitigate metal artifacts, in particular for small cardiac structures such as stents in coronary arteries or bypass grafts.

Artifact size was lowest, for all structures analyzed, at 120 keV; however, no significant differences were found when 100 keV was compared to 120 keV. Comparing 80 keV to 100 keV, we found that 100 keV provided a significant reduction of artifact size, but only for sternal wires. Thus, we can speculate that 80 keV is the optimal monoenergetic reconstruction to reduce artifact size from SVC, stents, and bypass clips, but 100 keV is better than other keV reconstructions for sternal wire artifacts.

Quality score analysis showed the best result at 120 keV; however, no significant differences were found comparing 120 keV with 100 keV for any structure analyzed. For sternal wires, SVC, and stents, there was a significant reduction of quality score comparing 100 keV with 80 keV, but not for bypass clips. The quality score of bypass clips did not change significantly when photon energy was changed from 80 to 100 keV or 100 to 120 keV.

The noise was lowest at 80 keV but only bypass clips demonstrated a significant difference in going from 80 keV to 100 keV. Therefore, with regard to image noise, the best energy level to analyze structures near bypass clips is at 80 keV, although it is advantageous to choose 80 or 100 keV for the other artifacts.

Okayama et al. published their results of the influence of effective energy on CT number by material and tissue type in monoenergetic cardiac reconstructions at cardiac DECT (14). Another paper by Okayama et al., in which they optimized the energy levels of DECT, showed that image quality of coronary angiography could be improved by optimizing the energy level for individual patients (13). However, in these papers there is no mention of artifact reduction. More extensive investigations of artifact reduction with DECT have been published regarding orthopedic devices in phantoms (15,16) and patients (16 –19). In particular, there is an agreement about the reduction of metal artifacts with high keV. The mean optimal value shown in these papers was 120 keV (min 95 keV and max 150 keV), which is in accordance with our data. Lewis et al. confirmed the reduction of metal artifact size with increasing keV in a phantom study. Moreover they showed that at 80 keV, the SD of background attenuation is lowest (15). These data are comparable to our lowest value of image noise at 80 keV. This result confirms that artifact reduction is not correlated with contrast-to-noise ratio, evidenced by the fact that the decreased signal of the intravascular contrast observed at higher keV can significantly reduce the corresponding contrast-to-noise ratio when compared to lower keV datasets, regardless of whether or not the metallic artifact is reduced.

Our results should be read in light of our limitations. First, we did not investigate intra- or inter-reader reproducibility of our measurements. Nevertheless, the artifact evaluation was mainly performed using quantitative parameters such as length and image noise, which are less prone to subjective variability. Second, we performed our analysis for only five different keVs; however, this limited selection is well established in literature as the most effective (16 –18).

In conclusion, a significant reduction of high-attenuation artifacts can be achieved through the use of higher monoenergetic energy levels based on cardiac DECT. However, image noise in anatomic structures affected by artifacts is lowest at 80 keV, which suggests an evaluation approach that makes use of multiple energy levels for complete diagnosis. Monoenergetic reconstruction series from cardiac DECT can significantly reduce high-attenuation artifacts and may make this a suitable approach for patients with known metallic hardware.

Footnotes

Funding

UJS has received research grants from and is a member of the speakers’ bureau for Bayer Pharma AG, Berlin, Germany; UJS is consultant for and/or received research support from Bayer, Bracco, GE, Medrad, and Siemens.

References

Hemmingsson

Jung

Ytterbergh

. Dual energy computed tomography: simulated monoenergetic and material-selective imaging. J Comput Assist Tomogr 1986; 10: 490–499.

Fletcher

Takahashi

Hartman

. Dual-energy and dual-source CT: is there a role in the abdomen and pelvis. Radiol Clin N Am 2009; 47: 41–57.

Johnson

Krauss

Sedlmair

. Material differentiation by dual energy CT: initial experience. Eur Radiol 2007; 17: 1510–1517.

Kalender

Perman

Vetter

. Evaluation of a prototype dual-energy computed tomographic apparatus. I. Phantom studies. Med Phys 1986; 13: 334–339.

Coursey

Nelson

Boll

. Dual-energy multidetector CT: how does it work, what can it tell us, and when can we use it in abdominopelvic imaging? Radiographics 2010; 30: 1037–1055.

Yeh

Shepherd

Wang

. Dual-energy and low-kVp CT in the abdomen. Am J Roentgenol 2009; 193: 47–54.

Ruzsics

Schwarz

Schoepf

. Comparison of dual-energy computed tomography of the heart with single photon emission computed tomography for assessment of coronary artery stenosis and of the myocardial blood supply. Am J Cardiol 2009; 104: 318–326.

Choi

Song

. Myocardial perfusion imaging using adenosine-induced stress dual-energy computed tomography of the heart: comparison with cardiac magnetic resonance imaging and conventional coronary angiography. Eur Radiol 2011; 21: 26–35.

Arnoldi

Lee

Ruzsics

. CT detection of myocardial blood volume deficits: dual-energy CT compared with single-energy CT spectra. J Cardiovasc Comput Tomogr 2011; 5: 421–429.

10.

Zhang

Peng

. Dual source dual energy CT of acute myocardial infraction: correlation with histopathological findings in a canine model. Invest Radiol 2010; 45: 290–297.

11.

Lee

Imai

. Quantitative myocardial perfusion imaging using rapid kVp switch dual-energy CT: preliminary experience. J Cardiovasc Comput Tomogr 2011; 5: 430–442.

12.

Ruzsics

Lee

Zwerner

. Dual-energy CT of the heart for diagnosing coronary artery stenosis and myocardial ischemia-initial experience. Eur Radiol 2008; 18: 2414–2424.

13.

Okayama

Seno

Soeda

. Optimization of energy level for coronary angiography with dual-energy and dual-source computed tomography. Int J Cardiovasc Imaging 2012; 28: 901–909.

14.

Okayama

Soeda

Takami

. The influence of effective energy on computed tomography number depends on tissue characteristics in monoenergetic cardiac imaging. Radiol Res Pract 2012; 2012: 150980–150980.

15.

Lewis

Reid

Toms

. Reducing the effects of metal artefact using high keV monoenergetic reconstruction of dual energy CT (DECT) in hip replacements. Skeletal Radiol 2013; 42: 275–282.

16.

Meinel

Bischoff

Zhang

. Metal artifact reduction by dual-energy computed tomography using energetic extrapolation: a systematically optimized protocol. Invest Radiol 2012; 47: 406–414.

17.

Zhou

Zhao

Luo

. Monoenergetic imaging of dual-energy CT reduces artifacts from implanted metal orthopedic devices in patients with factures. Acad Radiol 2011; 18: 1252–1257.

18.

Guggenberger

Winklhofer

Osterhoff

. Metallic artefact reduction with monoenergetic dual-energy CT: systematic ex vivo evaluation of posterior spinal fusion implants from various vendors and different spine levels. Eur Radiol 2012; 22: 2357–2364.

19.

Bamberg

Dierks

Nikolaou

. Metal artifact reduction by dual energy computed tomography using monoenergetic extrapolation. Eur Radiol 2011; 21: 1424–1429.