Author + information
- Received July 26, 2011
- Revision received December 12, 2011
- Accepted December 22, 2011
- Published online August 1, 2012.
- Aaron So, PhD⁎,†,⁎ (, )
- Jiang Hsieh, PhD‡,
- Yasuhiro Imai, PhD‡,
- Suresh Narayanan, PhD‡,
- John Kramer, PhD‡,
- Karen Procknow, PhD‡,
- Sandeep Dutta, PhD‡,
- Jonathon Leipsic, MD§,
- James K. Min, MD∥,
- Troy LaBounty, MD∥ and
- Ting-Yim Lee, PhD⁎,†
- ↵⁎Reprint requests and correspondence:
Dr. Aaron So, Imaging Research Laboratories, Robarts Research Institute, 100 Perth Drive, London, Ontario N6A 5K8, Canada
Dual-energy computed tomography (DECT) has recently been introduced for clinical use. One potential application of DECT is myocardial perfusion imaging through the significant reduction of beam-hardening artifacts by using monochromatic image reconstruction; analysis of these images can improve the accuracy of quantitative measurement of myocardial perfusion. Single-source DECT enabled by rapid switching between the low and high tube potentials (kV) can minimize misregistration of the high and low kV projection datasets from cardiac motion. We have recently implemented prospective electrocardiography-triggering capability in our rapid kV-switching computed tomography (CT) scanner to reduce the high effective dose from a quantitative CT myocardial perfusion imaging study with DECT. Our initial investigation suggests that prospectively electrocardiography-triggered rapid kV-switching DECT can eliminate beam hardening and provide a more reproducible myocardial perfusion measurement compared with the traditional single-energy CT protocol.
The 2 most commonly used cardiac imaging techniques to evaluate patients with suspected coronary artery disease are single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) for functional assessment of stress-induced myocardial ischemia and coronary computed tomographic angiography (CTA) for anatomical evaluation of the degree of luminal narrowing in coronary arteries. SPECT evaluates myocardial hypoperfusion based on the amount of uptake of perfusion tracers relative to that of the normally perfused myocardium. The overall diagnostic accuracy of SPECT is confounded, however, by its qualitative nature, which tends to underestimate the true extent of a left main or triple-vessel coronary artery disease where myocardial perfusion (MP) is globally attenuated (“balanced” ischemia). Patients with left main or triple-vessel disease are at highest risk of future cardiac events and will receive the most benefit from revascularization. In contrast, coronary CTA has emerged as the noninvasive gold standard for the evaluation of coronary anatomy but provides little information as to the hemodynamic significance of the stenosis. It has been shown consistently that anatomic stenosis detected by using both coronary CTA and invasive coronary angiography only modestly correlates with attendant ischemia (1,2). Thus, quantitative computed tomography (CT) MPI could be an important tool to allow accurate stenosis assessment and ischemia evaluation with a single test. Nevertheless, the clinical use of CT MPI remains relatively limited to date. One of the obstacles for CT MPI to overcome is beam hardening (BH) arising from high-density iodinated contrast, which is intravenously injected for the study, into the heart chambers. Uncorrected BH can significantly affect the accuracy of MP measurement.
We herein report our initial experience with prospectively electrocardiography (ECG)-triggered rapid kV-switching dual-energy computed tomography (DECT) to minimize BH and reduce radiation dose in quantitative MPI studies.
BH in CT arises from the polychromatic (polyenergetic) nature of x-rays used in clinical CT scanners and the energy dependence of x-ray attenuation: low-energy x-rays are attenuated more than their high-energy counterparts. Depending on the path taken by the x-rays (e.g., through heart chambers filled with highly attenuating contrast or less-attenuating soft or lung tissue), the different projection views acquired around the heart in CT will have inconsistent magnitude of x-ray attenuation among them. CT image reconstruction by back-projecting these inconsistent projections will result in BH artifacts, which manifest as inaccuracies in the CT number of the reconstructed images. Although all CT images have BH correction for soft tissue during the initial reconstruction process, BH arising from the high-density materials, particularly iodinated contrast in the heart chambers in CT MPI, cannot be effectively minimized by such correction. In contrast-enhanced images of the heart, the BH effect typically results in dark bands in the inferior (basal) and anterior (apical) wall and bright bands in the septal and lateral wall in the 4-chamber tomographic plane (approximate horizontal long-axis view). These artifacts will lead to inaccurate assessment of MP using either a qualitative or quantitative approach.
BH Effect on Quantitative Measurement of MP
To understand the effect of BH on quantitative CT MPI, one must first be familiar with MP CT acquisitions and the processing method used. A bolus of iodinated contrast is injected intravenously while dynamic scanning with prospective ECG triggering at 1 scan every 1 to 2 heartbeats and, with breathhold, is initiated 3 to 5 s into the injection for 30 s to monitor the first pass of contrast through the heart. From the dynamic contrast enhanced (DCE) CT images, time-density curves (TDCs) of the myocardium and its supply arteries are measured. The arterial TDC is measured from either the left ventricular chamber or aorta instead of the coronary arteries to avoid underestimation of the arterial density owing to partial volume averaging. The myocardial TDC is dependent on the local blood flow and volume and the arterial TDC. A mathematical operation called deconvolution is applied to decouple the effect of the arterial TDC on the myocardial TDC to estimate absolute MP in units of milliliters per minute per gram of tissue (3,4). BH in DCE images leads to distortion of both arterial and myocardial TDCs; deconvolution of these distorted TDCs will lead to inaccurate estimation of absolute MP.
Post-reconstruction image-based method
Several algorithms have been proposed to minimize the BH effect in DCE images of the heart. We have shown in previous porcine studies that the BH effect on MP measurement can be significantly reduced by using a post-reconstruction image-based correction method (3). This method first applies Hounsfield unit (HU) thresholds to differentiate the low-attenuation soft tissue region from the high-attenuation region (bone and iodinated contrast) in a DCE image. Because conventional CT images have BH correction for water/soft tissue already applied, residual BH artifacts mainly arise from high-attenuating regions where the water correction fails to minimize the BH effect. In our method, the high-attenuating region in the thresholded image is forward-projected to retrieve the projections through the high-attenuating materials. The BH-induced errors in these projections are estimated and back-projected to form the error image, which is then subtracted from the original CT image to produce the BH-corrected image (1). The linear regression plots of MP measurements by CT perfusion are displayed without (Fig. 1A) and with (Fig. 1B) image-based BH correction against MP measurements by the gold standard radiolabeled microspheres technique in pigs with acute myocardial infarction (1). Despite a similar regression slope, the y-intercept of the non–BH-corrected plot was >2 times higher than that of the BH-corrected plot (68.3 vs. 29.8 ml/min/100 g), indicating that the BH-induced overestimation of MP can be significantly reduced after proper correction.
The image-based BH correction method requires a complete and accurate knowledge of the high-attenuating materials within the acquisition field of view. For the correction in CT MPI, it is also important to accurately estimate the concentration of iodine in the heart. Because the image-based correction relies on the initial “water-corrected” images, neither the segmentation of high- and low-attenuating materials within the field of view nor the BH correction is exact, leading to suboptimal correction. This is evident by the residual overestimation of MP with respect to the gold standard MP measurements after image-based BH correction (Fig. 1B).
Dual-energy computed tomography
DECT could be a more effective technique for correcting BH due to the ability to reconstruct monochromatic (monoenergetic) CT images (3). Because BH largely arises from the polychromatic nature of x-rays used in clinical CT scanners, monochromatic images therefore simulate images acquired with a monoenergetic x-ray beam and are free of BH artifact. Unlike conventional CT, where projections are acquired with a single x-ray energy spectrum (SECT), DECT simultaneously acquires 2 sets of projections using 2 different x-ray energy spectrums (the detail of acquisition techniques are discussed later). The basis of monochromatic imaging with DECT is briefly reviewed below.
Generation of Synthesized Monochromatic Images
In CT imaging, the attenuation of x-rays by a material is dominated by 2 interactions: absorption (photoelectric) and scatter (Compton), both of which are dependent on x-ray energy. One objective of DECT is to determine the unique contributions of these 2 effects for each material in the scanned object by taking measurements at 2 different x-ray energies (because 2 equations are required to solve for 2 unknown quantities). Once the absorption and scatter contributions along each projection path are determined, the scanned object can be represented by an “absorption” image and a “scatter” image instead of a single image of linear attenuation coefficients or CT numbers (as in the case of SECT). From the clinical point of view, however, it is difficult to interpret the resulting images because the 2 physics effects do not exhibit the usual relationships with clinical pathologies that exist with CT numbers. Instead, a pair of basis materials, water and iodine, that are commonly encountered in CT imaging and are vastly different in their atomic numbers (and hence their x-ray absorption and scatter characteristics) are usually used instead of the 2 physics effects for image representation.
The 2 sets of projections measured at the high and low energies in DECT are corrected for BH and transformed into 2 sets of equivalent-density projections of the basis materials (water and iodine). The tomographic reconstruction process is then applied to these projections to produce equivalent-density images of water and iodine. The “water equivalent-density” and “iodine equivalent-density” images have units given as grams per centimeters cubed instead of HU, and they depict the amount of water and iodine (or equivalent densities) present in the scanned object to produce the measured projections at the 2 energies. The process of transforming the measured projections into equivalent-density projections of 2 basis materials and reconstruction of the equivalent-density images of water and iodine is called material decomposition (MD). MD is a useful application of DECT because it may provide additional information of different tissue types for their discrimination from each other compared with conventional SECT. Finally, because the linear attenuation coefficients of water and iodine at any specific x-ray energy are known, monochromatic energy images in HU can be synthesized from the water and iodine equivalent-density images (5) (Fig. 2).
In addition to the projection-based MD technique discussed here, MD can also be achieved by using an image-based method. In this approach, the conventional 80 and 140 kV SECT images are acquired first; monochromatic images and water and iodine equivalent-density images are then generated by linear combinations of the 80 and 140 kV images with proper weightings of each kV image. Compared with the projection-based technique, the image-based approach is simpler because it does not require sophisticated processing of the projection data. However, image-based MD could be less accurate than the projection-based method because BH is already present in the low and high kV images, which would affect the accuracy of the derived equivalent-density images and monochromatic images.
DECT is currently implemented in 2 ways. One approach is to use a CT scanner with 2 independent x-ray tube/detector pairs offset by 90° (e.g., SOMATOM Definition Flash, Siemens Healthcare, Erlangen, Germany) with each operating at a different tube potential (kV) setting (usually 80/100 and 140 kV). The other approach is to use a CT scanner with a single x-ray tube capable of rapidly switching between 80 and 140 kV (e.g., Discovery CT750 HD, GE Healthcare, Waukesha, Wisconsin) (Figs. 3A and 3B). In this single-source CT system, the 2 projection sets corresponding to the low and high kV are acquired in an interlaced manner minimizing the misregistration error from cardiac motion. Furthermore, the scintillating material used as x-ray detectors has an ultrafast decay time to ensure no “ghosting” between projection views.
The Discovery CT750 HD scanner applies a set of fixed tube currents (ranging from 360 to 630 mA) as presets for scanning at different gantry speeds and patient sizes. Although the current single-source CT system cannot modulate the tube current at the same rate as the tube potential, different dwell times are used for 80 and 140 kV acquisitions to optimize the noise performance. The resulting effective dose for a non–ECG-gated quantitative MPI study of a 4-cm section of the heart ranges from 31 to 54 mSv. To reduce the high radiation dose, we have recently implemented prospective ECG triggering (Snapshot Pulse, GE Healthcare) with rapid kV-switching DECT, in which image acquisition is now triggered at every other heart beat (Fig. 3C). The gantry rotation speed is also 30% faster than before (0.35 s vs. 0.5 s per rotation) to further reduce the x-ray exposure time and radiation dose in each acquisition. These improvements can reduce the effective dose of a quantitative CT MPI study to between 11 and 23 mSv over 4 cm of coverage.
Quantitative Measurement of MP With Prospectively ECG-Triggered Rapid KV-Switching DECT
In this section, we present our initial investigation of quantitative CT MPI with a prospectively ECG-triggered rapid kV-switching DECT protocol in a normal (nonischemic) pig.
A 70-kg normal pig was scanned on a GE Healthcare Discovery CT750 HD scanner with the following scan protocol: 22 ECG-triggered dual-energy scan once every other heartbeat (to cover roughly 30 s) using 140 and 80 kV alternating at 0.2 ms intervals, 630 mA, and a 0.35 s rotation period starting at 4 s into an intravenous injection of contrast (Isovue 370, Bracco Diagnostics Inc., Princeton, New Jersey) at a rate of 4 ml·s-1 and a dosage of 0.7 ml·kg-1. 5-mm 140 kV DCE images and the corresponding monochromatic 70 keV DCE images were reconstructed with full 360° BH uncorrected and corrected projections, respectively. Both image sets were analyzed with the CT Perfusion software (GE Healthcare) to generate MP functional maps and the average map for comparison.
Figures 4 and 5 show the 140 kV and 70 keV average and MP maps, the corresponding arterial and myocardial TDCs and mean perfusion values in different regions of the left ventricular myocardium. In the 140 kV average map, rather than the expected uniform enhancement throughout the normal myocardium, BH induced severe hypo-enhancement in the apical wall and an increase in enhancement in the septal wall relative to the lateral wall (Fig. 4A); all were confirmed by the 140 kV myocardial TDCs (Fig. 4C). In contrast, the BH effect in the 70 keV average map was minimized, resulting in a more uniform enhancement throughout the normal myocardium during the first-pass of contrast (Fig. 4B). The 70 keV myocardial TDCs from different parts of the myocardium were more consistent with each other (Fig. 4D). The BH artifacts in 140 kV DCE CT images led to a heterogeneous MP map in a normal heart (Fig. 5A). In particular, MP in the septal wall was almost 2 times higher than that in the apical wall (Figs. 5C and 5D).
This pig study demonstrates that rapid kV-switching projection-based DECT may permit a more accurate MP measurement by eliminating the BH effect in contrast-enhanced CT images of the heart. The quantitative CT MPI study with the proposed DECT protocol reported here has an effective dose of 19 mSv for 4-cm coverage of the heart even though prospective ECG triggering is used. This is owing to the high tube current (630 mA) used for scanning and the large number of projections acquired per gantry rotation. Future research will focus on dose reduction in several areas. First, a goal will be to investigate whether iterative reconstruction (e.g., adaptive statistical iterative reconstruction , GE Healthcare) can reduce noise in projections when a much lower tube current (360 mA) is used for DECT acquisition. Initial experiences suggest adaptive statistical iterative reconstruction is effective in reducing noise in SECT images acquired with low tube current. Another goal will be to investigate the feasibility of reducing the number of projections acquired per gantry rotation with rapid kV-switching DECT without inducing artifacts in the synthesized monochromatic images. In the proposed rapid kV-switching DECT protocol, about 1,000 projections are acquired at 80 and 140 kV to match the number of projections acquired in standard SECT at either 80 or 140 kV. The effective dose is directly proportional to the number of projections acquired. It has been shown that decreasing the number of projections leads to streak artifacts in images reconstructed with filtered back-projection (7). Preliminary data suggest that such streak artifacts can be minimized by using a new image reconstruction algorithm (e.g., prior image–constrained compressed sensing) (7).
Lastly, MP measurement with the proposed DECT protocol should be validated against microspheres measurement to determine whether the accuracy of MP quantification is improved as a result of a more exact BH correction than that of the post-reconstruction image-based method (Fig. 1).
We demonstrated for the first time using a porcine model that prospectively ECG-triggered rapid kV-switching projection-based DECT may be a useful technique for the elimination of BH in CT MPI, which should facilitate more accurate measurements of MP with CT Perfusion. The clinical use of the technique for functional assessment of coronary artery disease is currently restricted by the high radiation dose, but the barrier could be overcome by the use of iterative reconstruction techniques to allow low-dose DECT acquisition without affecting image quality and the accuracy of perfusion measurements.
The authors thank Jennifer Hadway and Laura Morrison for their assistance with the animal study, and Dr. Xiaogang Chen for his assistance with computer programming.
This work was supported in part by the Canadian Institutes of Health Research, Canada Foundation for Innovation, Ontario Research Fund, Ontario Innovation Trust, the Research and Education Foundation of the Radiological Society of North America, and GE Healthcare. Dr. Leipsic is on the Speakers' Bureau and Advisory Board of GE Healthcare. Dr. Min is on the Speakers' Bureau and Medical Advisory Board of GE Healthcare; is a consultant to Edwards LifeSciences; and holds equity interest in TC3. Dr. Lee has received a research grant from and has a software licensing agreement for CT Perfusion with GE Healthcare. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- beam hardening
- computed tomography
- computed tomographic angiography
- dynamic contrast enhanced
- dual-energy computed tomography
- material decomposition
- myocardial perfusion
- myocardial perfusion imaging
- single-energy computed tomography
- single-photon emission computed tomography
- time-density curve
- Received July 26, 2011.
- Revision received December 12, 2011.
- Accepted December 22, 2011.
- American College of Cardiology Foundation
- van Werkhoven J.M.,
- Schuijf J.D.,
- Gaemperli O.,
- et al.
- Meijboom W.B.,
- Van Mieghem C.A.,
- van Pelt N.,
- et al.