| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Experimental Studies |
1 From the Department of Radiology, Duke University Medical Center, DUMC 3808, Durham, NC 27710 (D.T.B., E.M.M., E.K.P.); and Department of Radiology, University of Maryland Medical Center, Baltimore, Md (T.R.F.). Received May 15, 2007; revision requested July 23; revision received August 23; accepted September 19; final version accepted November 15. Address correspondence to D.T.B. (e-mail: daniel.boll{at}duke.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCE IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
Materials and Methods: Four coronary stents (2, 3, 4, and 5 mm) were examined at 64-section dual-energy multidetector CT by using a dual-detector "double-decker" imager with stacked high- and low-energy detector arrays, 0.67-mm section thickness, and 32 x 0.625-mm collimation. Simultaneous high- and low-energy data sets were acquired at 80 and 140 kVp and at 400 mAs. Cardiac motion was simulated in a moving anthropomorphic heart phantom. Stents were imaged longitudinally and axially with the phantom at rest and with it in motion. Use of an enhancement algorithm based on high- and low-energy absorption profiles was proposed. Stent lumen depiction and stent mesh delineation were quantified in terms of contrast-to-noise ratio (CNR) and kurtosis (
), respectively. Image quality was analyzed at univariate general linear model analysis in which peak voltage and data enhancement algorithm were dependent factors and stent orientation and cardiac motion were independent factors.
Results: Analysis of CNR and revealed an interdependency between CNR and and stent diameter: The CNR and of smaller stents increased significantly when these stents were imaged at lower peak voltages in the axial plane and with the enhancement algorithm applied to the 80-kVp data sets (P < .001). The CNR and of larger stents increased significantly when these stents were imaged at higher peak voltages in the longitudinal plane, and imaging of these stents benefited from the enhancement algorithm being applied to the 140-kVp data sets (P < .001).
Conclusion: Dual-energy multidetector CT performed with optimized acquisition parameters and alternative image postprocessing led to enhanced coronary stent lumen depiction to an extent beyond that achieved with single-energy multidetector CT.
© RSNA, 2008
Supplemental material: http://radiology.rsnajnls.org/cgi/content/full/2473070849/DC1
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCE IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
Sequentially allocated multidetector CT detector arrays have relied on x-ray tubes to generate a quantitatively homogeneous emission of x-ray photons along the entire detector width. However, x-ray tubes are also designed to produce photons that differ qualitatively, so different x-ray spectra and subsequently different contrast material– and tissue-specific absorption profiles can be created. With dual-energy imaging, x-ray beams of defined energies can be tuned to the characteristic absorption profiles of various tissues or contrast materials. Therefore, this technology can enable enhancement of the inherent contrast differences of tissues or materials. Dualenergy imaging has successfully evolved from two-dimensional chest imaging (10) to monochromatic cross-sectional CT (11,12) to coronary fluoroscopy (13) and more recently to coronary dual-energy multidetector CT (14).
We hypothesized that the use of dual-energy multidetector CT performed with optimized acquisition parameters such as x-ray tube voltage and alternative postprocessing techniques would lead to enhanced depiction of coronary stent lumens beyond the extent achievable with single-energy multidetector CT. Thus, the purpose of our study was to prospectively evaluate, by optimizing image acquisition and introducing alternative image postprocessing techniques, dual-energy multidetector CT for depiction of the lumens of coronary stents placed in a moving anthropomorphic heart phantom.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCE IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
|
The coronary vessel platform with implanted stents attached to the moving heart phantom was immersed in oil, which had a mean attenuation of approximately –70 HU to simulate epicardial fat. To account for the effect of the surrounding chest wall, a standardized circular Plexiglas layer (CIRS, Norfolk, Va) surrounded the experimental setup (14).
Dual-Energy Multidetector CT Data Acquisition and Reconstruction
Dual-energy data were acquired by using a 64-section multidetector CT imager with a recently introduced dual-detector design: two detector arrays of 32 equally spaced detector elements stacked in two layers (Brilliance 64 MultiEnergy; Philips Medical Systems, Cleveland, Ohio). This "double-decker" approach enables dual-energy imaging to be performed with only one x-ray tube: The x-ray spectrum is split into a low-energy component L0, which is detected by the upper detector layer, and a high-energy component L1, which is registered by the lower detector layer (18).
Similar to the data acquired with single-energy multidetector CT systems, the data acquired by each detector layer in the dual-energy multidetector CT system undergo preprocessing, according to the size and composition of the scanned object, to correct for attenuation variations due to beam-hardening effects. Data preprocessing with the dual-layer multidetector CT system consisted of individual polychromatic corrections for the L0 and L1 detectors. Each of these corrections is based on attenuation maps that contain for every pixel in the image a value that characterizes the average beam-hardening effects on that pixel. This value is calculated by adding all of the projections through that pixel (18). Finally, L0 and L1 detectors are calibrated so that Lc = L0 + L1, where Lc constitutes the combined absorption profiles of the low- and high-energy detectors and is equivalent to the absorption profile of a single-energy multidetector CT system. In contrast to a dual-tube dual-detector approach, dual-energy multidetector CT can be performed while the x-ray tube and the low- and high-energy detector arrays are in the same positions.
The helical scanning protocol involved the use of a 0.67-mm nominal section thickness, a detector pitch of 0.24, a gantry rotation period of 0.42 second, a collimation of 32 x 0.625 mm, and a matrix size of 512 x 512 pixels. Two dual-energy multidetector CT image series, each containing L0, L1, and Lc data sets, were acquired at 400 mAs: one with an x-ray peak voltage setting of 80 kVp and the other with 140 kVp. The coronary stents were imaged longitudinally and axially. While an electrocardiographic signal was being continuously supplied, imaging was performed with the cardiac phantom at rest and with it moving at a fixed heart rate of 75 beats per minute.
To reconstruct image data sets, we used a two-dimensional reconstruction kernel consisting of an edge-enhancing algorithm, as well as a noise-reduction filter that emphasized the contrast differences of linear anatomic structures (19). Images were retrospectively reconstructed at the 75% interval of the R-R cycle (20).
Dual-Energy Multidetector CT Data Postprocessing
To reduce the degrees of blooming artifacts and image noise and to enhance coronary stent lumen depiction, we applied a dual-energy enhancement algorithm (Le) to the originally acquired Digital Imaging and Communications in Medicine data sets. The proposed alternative Le consisted of three theoretic arms realized in a pixel-by-pixel approach (21): (a) A normalization image data set was calculated by performing a weighted multiplication of the acquired L0 and L1 data sets to reduce overall image noise; (b) a differential image data set was computed by subtracting L1 from L0 to generate a map of contrast differences due to dual-energy absorption effects; and (c) a weighted multiplication of the normalization image data set and the inverse of the differential image data set was performed. Thus, the Le was applied according to the equation Le = (L0 · L1) · (L1 – L0)–1. Dual-energy data postprocessing was performed by using ImageJ, version 1.36b, software (National Institutes of Health, Bethesda, Md).
Dual-Energy Multidetector CT Analysis
Dual-energy multidetector CT coronary stent analysis included assessment of stent lumen depiction and stent mesh delineation. Contrast-to-noise ratios (CNRs) for evaluating stent lumen depiction, expressed in Hounsfield units, were calculated by using the equation CNR = (HUM – HUL)/SDN, where HUM and HUL are the attenuations of the stent mesh and the stent lumen, respectively, and SDN is the standard deviation of the noise level.
We measured axial-plane stent lumen attenuation by centrally placing circular regions of interest with areas (in order of increasing stent diameter) of 1, 7, 19, and 38 mm2 in the lumen of the four analyzed coronary stents. We measured axial-plane stent mesh attenuation similarly by placing doughnut-shaped regions of interest measuring 2.5, 5.5, 8.5, and 11.5 mm2. We measured longitudinal stent lumen attenuation by centrally placing rectangular regions of interest measuring 2.5, 7.5, 12.5, and 17.5 mm2 in the lumen, and we measured longitudinal stent mesh attenuation by placing two rectangular regions of interest on opposing mesh grafts covering a combined area of 10 mm2 for each stent. We assessed standard deviations of noise levels by placing 100-mm2 regions of interest in artifact-free areas of the simulated epicardial fat.
Owing to the small size of the analyzed coronary stents, stent mesh delineation was quantitatively analyzed by defining a dissecting line that originated and terminated on opposing sides of the stent and crossed the center of the lumen in the axial and longitudinal stent planes. Attenuation values of pixels along the dissecting line were transferred onto a profile plot, and the kurtosis () of the resulting curve was calculated according to the following equation:
|
is the average of all attenuation values on the dissecting line.
Is a dimensionless mathematical parameter that subsumes the difference in maximal and minimal pixel attenuation values, as well as the ascending and descending slope of the measured profile plot. Values of less than 0 are termed platykurtic values and represent a shallow curve, as observed with substantial blooming artifacts that obscure the stent lumen. Values of greater than 0 are termed leptokurtic values and describe steep and well marginated bell-shaped curves, such as those representing small blooming artifacts that allow clear depiction of the stent lumen and clear delineation of the stent mesh.
All CNR and measurements were performed three times each, and the results were subsequently averaged. All placements of regions of interest and dissecting lines to assess CNR and , respectively, were performed by one radiologist (D.T.B.), who had more than 7 years experience interpreting gated vascular multidetector CT images. Dual-energy data were analyzed by using computer software (Brilliance Workspace, version 2.0 B2; Philips Medical Systems).
Statistical Analyses
We performed statistical analyses of CNR and by independently analyzing the originally acquired L0, L1, and Lc data sets and the postprocessed Le image data sets. First, we evaluated the originally acquired L0, L1, and Lc data sets at general linear model analysis by successively defining CNR and as dependent variables. To assess the effect of peak voltage (80 kVp or 140 kVp) on CNR and , peak voltage was designated a fixed factor; additional influencing factors such as stent orientation (axial or longitudinal) and cardiac activity (static or dynamic) were defined as independent factors or covariates. A balanced full factorial model with additional analysis of covariate interactions was chosen. Bonferroni post hoc analysis was performed for each dependent variable (CNR and ) individually. Corrections and tests were used for the fixed factor (peak voltage) only.
Second, by successively defining CNR and as dependent variables, we evaluated the postprocessed Le data sets at general linear model analysis to assess whether data enhancement, as compared with use of the originally acquired data sets, would lead to significant variations in stent lumen depiction and stent mesh delineation. To differentiate the effects of low-energy, high-energy, and combined-energy data acquisitions on postprocessed data sets, we defined acquisition technique (L0, L1, or Lc) as a fixed factor. Additional influencing factors such as peak voltage, stent orientation, and cardiac activity were designated independent factors or covariates. A balanced full factorial model with additional analysis of covariate interactions was chosen. Bonferroni post hoc analysis was performed for each dependent variable (CNR and ) individually. Corrections and tests were used for the fixed factors (L0, L1, or Lc) only. Statistical analyses were performed by using SPSS, version 13.0, software (SPSS, Chicago, Ill); P 
.05 was considered to indicate a significant difference.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCE IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
in Originally Acquired L0, L1, and Lc Data Sets
|
|
with stent mesh delineation quantified as a function of peak voltage revealed that when the smaller (2- and 3-mm) stents were imaged, a decrease in peak voltage from 140 to 80 kVp led to a significant increase in (P < .001). Stent orientation was found to be a significant factor of increased in the axial scanning plane (P < .001). However, the increase in was less striking in the dynamic image data sets than in the corresponding static image data sets (P < .001) (Fig 2). Larger (4- and 5-mm) coronary stents had an increase in (P < .001) when they were imaged at the higher peak voltage of 140 kVp, and imaging of these stents in the longitudinal orientation was favored (P < .001) (Fig 2).
|
increase and stent diameter: Smaller stents showed significant increases in CNR and when they were imaged at 80 kVp in the axial plane. However, larger stents had a significant increase in CNR and when they were imaged at 140 kVp—in particular, was enhanced at imaging in the longitudinal stent orientation (Figs 3, 4).
|
|
in Postprocessed Le Data Sets
|
Statistical analysis of the 3-, 4-, and 5-mm stents revealed that the CNR increase in the Le data sets acquired at 140 kVp was significantly larger than the CNR increase in the corresponding Le data sets acquired at 80 kVp (P < .001) (Fig 5). There was no significant effect on the overall CNR change in the Le data as a function of either stent orientation (P > .99) or cardiac phantom motion (P > .99).
Use of the proposed Le, as compared with use of the originally acquired L0, L1, and Lc data sets, facilitated a significant increase in . Stent diameter and enhancement factor due to Le had an inversely proportional relationship, which reached significant levels for all assessed coronary stents (P < .001) (Fig 6).
|
in the axial orientation at the lower x-ray tube peak voltage of 80 kVp (P < .001). Imaging of the larger stents benefited from the higher voltage of 140 kVp, especially in the longitudinal orientation (P < .001). Cardiac activity did not significantly affect overall variations in the Le data sets (P > .99).
Assessment of dual-energy multidetector CT postprocessing revealed that imaging of the smaller stents benefited substantially when the Le was applied to the originally acquired 80-kVp data sets. However, the larger stents showed a substantial increase in CNR and when the Le was applied to the originally acquired 140-kVp data sets (Figs 3, 4).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCE IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
factors, which are used to quantify depiction of the stent lumen and delineation of the stent mesh, respectively. Furthermore, dual-energy spectral analysis with a newly proposed algorithm of simultaneously acquired data led to significantly reduced blooming artifacts and image noise beyond the capabilities of single-energy multidetector CT technology alone. Our study findings emphasize the importance of x-ray tube voltage by showing that for smaller coronary stents, lumen depiction and mesh delineation were enhanced when a lower x-ray tube voltage of 80 kVp was applied, while imaging of the larger stents benefited from a higher x-ray tube voltage—140 kVp. In prior studies of x-ray tube voltage optimization in multidetector CT aortography, lower x-ray tube voltages caused increased attenuation of intravascular contrast material; however, greater image noise had to be accepted (23). Similarly, a reduction in voltage during imaging of the smaller stent lumens and meshes, as compared with higher voltage settings, led to an increase in the attenuation of intraluminal contrast material. However, for larger stents composed of more stent mesh material, higher x-ray tube voltages were required to overcome image noise and blooming artifacts and thereby enhance stent lumen depiction and stent mesh delineation.
The results of prior dual-energy examinations of ex vivo coronary artery specimens have suggested that advanced image postprocessing techniques have greater potential to enable visualization of the differences in contrast material– and tissue-specific absorption profiles than do individual analyses of the originally acquired and preprocessed low- and high-energy data sets (14). The Le proposed in our study was focused on two factors for increasing overall image quality: reduced image noise and enhanced image contrast. Rather than being based on the acquired nonfiltered raw data sets, the proposed enhancement algorithm was based on the individually adjusted and mutually synchronized low- and high-energy data sets: L0 and L1, respectively. By creating a normalization data set of the same anatomic area with differing x-ray tube peak voltages and resulting peak voltage–dependent noise levels, we reduced noise substantially with a moderate loss of tissue contrast (24). The differentiation data set was used to generate a map of contrast differences due to absorption spectra based on the low- and high-energy absorption profiles (14). Subsequent weighted multiplication enabled these two effects to be combined in one image data set to result in reduced image noise and enhanced image contrast.
Stent orientation is a factor based primarily on the patient's coronary artery anatomy. In multidetector CT, the anatomic arrangement of the coronary arteries translates into a longitudinal alignment of the proximal and wider coronary arterial segments and an axial orientation of the coronary segments that are smaller and more distal to the initial scanning plane (25). In simulating the human anatomy by varying the imaging plane in our study, we were able to show that larger stents implanted into larger and therefore more proximally located coronary segments can be visualized best in the longitudinal scanning plane. In contrast, the smaller and therefore more peripherally located coronary stents were best visualized in the axial scanning plane.
The cardiac motion simulated by the moving anthropomorphic heart phantom caused moderate to minor decreases in CNR and . Electrocardiographically gated multidetector CT image reconstruction was based on algorithms that combine partial scan reconstructions, which were optimized for high temporal resolution, with low-pitch multisection spiral weighting to image the cardiac phantom with longitudinal redundancy during multiple heart cycles. However, the fact that image quality was not degraded to the extent that stent lumen depiction and mesh delineation were not possible in any of the analyzed coronary stents reflects the substantial improvement in current stent-dedicated partial scan reconstruction algorithms.
Our study had limitations: First, the experimental setup simulating the chest wall, pericardial fat, coronary arteries, and cardiac motion represented an idealized anatomic environment. Second, only patent coronary artery stents without intraluminal narrowing or obstruction were examined. However, we designed our study as a pilot investigation representing a preliminary effort to understand whether dual-energy multidetector CT may be useful for minimally invasive coronary stent assessment. Finally, with use of one longitudinal and one axial imaging plane, only the most extreme orientations of the coronary anatomy—as opposed to the more realistic anatomic courses of tortuous coronary arteries—were replicated.
In conclusion, our study results show that dual-energy multidetector CT performed with optimized x-ray tube voltage settings and alternative image postprocessing techniques facilitated enhanced coronary stent lumen depiction beyond that achievable with single-energy multidetector CT. In particular, the larger coronary stents were visualized best in the longitudinal scanning plane when the Le was applied to the data sets acquired at 140 kVp, whereas depiction of the smaller coronary stents benefited from scanning in the axial plane with the Le applied to the data sets acquired at 80 kVp. Our study results also emphasize the importance of x-ray tube voltage as a factor for increasing the CNR. However, varying the x-ray tube voltage during image acquisition remains beyond the capability of current multidetector CT technology.
Practical application: The findings of this experimental pilot study suggest that as a practical application, minimally invasive dual-energy multidetector coronary CT performed to ensure stent patency might represent a valuable alternative to more invasive imaging techniques, such as conventional angiography, for follow-up imaging in patients after intracoronary stent implantation for treatment of coronary artery stenosis.
|
ADVANCE IN KNOWLEDGE |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCE IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
|
IMPLICATION FOR PATIENT CARE |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCE IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
|
FOOTNOTES |
|---|
Abbreviations: CNR = contrast-to-noise ratio • L0 = low-energy component of x-ray spectrum • L1 = high-energy component of x-ray spectrum • Lc = combined absorption profiles of low- and high-energy detectors • Le = dual-energy enhancement algorithm
Author contributions: Guarantors of integrity of entire study, all authors; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, all authors; experimental studies, all authors; statistical analysis, all authors; and manuscript editing, all authors
Authors stated no financial relationship to disclose.
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCE IN KNOWLEDGEIMPLICATION FOR PATIENT CAREReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  L.-J. Zhang, Y.-E Zhao, S.-Y. Wu, B. M. Yeh, C.-S. Zhou, X.-B. Hu, Q.-J. Hu, and G.-M. Lu Pulmonary Embolism Detection with Dual-Energy CT: Experimental Study of Dual-Source CT in Rabbits Radiology, July 1, 2009; 252(1): 61 - 70. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
