| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 All authors: Department of Radiology, Innsbruck University Hospital, Anichstra. 35, 6020 Innsbruck, Austria.
Received September 10, 2001;
accepted after revision June 26, 2002.
Address correspondence to A. Mallouhi.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. In 16 patients, 16 renal artery stents were evaluated with MDCT renal angiography and digital subtraction angiography (DSA). CT data were evaluated using multiplanar volume reformations and the volume-rendering algorithm with three different volume-rendered parameter settings (low-to-high, high-to-low, and high—low—high opacity transfer functions: VRLH, VRHL, and VRVE, respectively). Targeted images of each stent were rendered in paraaxial and paracoronal planes and were interactively interpreted. The overall restenosis severity was measured on postprocessed paraaxial and paracoronal images and compared with that obtained on DSA using linear regression analysis. Image quality and lumen delineation on rendered images were also compared using Wilcoxon's signed rank test.
RESULTS. Eight restenoses were identified on DSA. Correlations between restenosis severity measured with DSA and those measured with MDCT were significant (p < 0.001). Volume rendering with VRHL allowed the best correlation with DSA (reviewer 1, r2 = 0.86; reviewer 2, r2 = 0.94) and was significantly better than multiplanar volume reformations (p = 0.028). Overall image quality was high with all rendering techniques and with no significant differences (p > 0.59, for all comparisons). Stent lumen was well delineated with volume-rendering modalities; however, VRHL was significantly better than VRLH (p = 0.033).
CONCLUSION. Volume-rendered MDCT angiography enabled high-quality three-dimensional reproducible evaluation of the patency of implanted renal artery stents. Volume rendering with VRHL achieved the best performance.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Single-detector CT angiography complemented with volume rendering as a postprocessing algorithm has emerged as a useful imaging tool for arterial stenosis [5, 6], particularly renal artery stenosis [7]. The improved scanning efficiency of the multidetector CT (MDCT) scanner contributes significantly to the performance of CT angiography and influences substantially the overall image quality of three-dimensional reconstructions [8, 9]. The purpose of our study was to evaluate the diagnostic potential of volume-rendered MDCT angiography for estimating the patency of renal artery stents by comparing three volume-rendering techniques with DSA and multiplanar volume reformations.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CT Angiography Acquisition and Reconstruction Parameters
MDCT was performed on a LightSpeed QX/i scanner (General Electric Medical
Systems, Milwaukee, WI) with a 0.8-sec gantry rotation period. During a single
breath-hold, CT angiograms were acquired in a superior-to-inferior direction
at the level of the renal arteries, covering 10-14 cm in the z-axis
using 1.25-mm collimation, a high-quality mode (pitch, 3), a 1.25-mm
reconstructed slice thickness, a reconstruction interval of 0.8 mm, and a
standard reconstruction kernel. The X-ray tube voltage was 140 kV, and the
current was 250 mA. Depending on the patient's weight, a total volume of
90-120 mL of nonionic contrast material was administered at a rate of 3
mL/sec. The scanning delay was determined using the Smart-Prep software
(General Electric Medical Systems) that continuously monitored the attenuation
values in the pulmonary trunk. When the attenuation increased to 180-200 H
after contrast material injection, MDCT was manually initiated under the
supervision of a radiologist who was experienced in the use and interpretation
of CT. The source images were then reconstructed using a 9.6-cm display field
of view; the region of interest captured the aorta, the renal artery stent,
and the proximal segment of the renal artery. The axial reconstructed images
were transferred to an independent workstation (Ultra 60; Sun Microsystems,
Mountain View, CA) running the Advantage Windows software (version 4.0;
General Electric Medical Systems). The CT data were evaluated using
multiplanar volume reformations and the volume-rendering algorithm with three
volume-rendered parameter settings.
Multiplanar volume reformation images were obtained in paraaxial and paracoronal planes at 90° intervals along the longitudinal axis of the renal artery stent. The window width and level settings were customized subjectively to allow clear visualization of the stent lumen.
Display parameters (window width and level, opacity, and brightness) for generating volume-rendered images of the stented renal arteries were defined by consensus between two of the authors after a preclinical study in which different models of volume-rendered images from CT angiography data sets were tested. Targeted volume-rendered images were generated using a thin slab (1.3-1.7 mm) positioned on the multiplanar volume reformation images through the longitudinal axis of the stent in the paraaxial and paracoronal planes at 90° intervals. Volume-rendered virtual endoscopic images were generated by manual camera movements along the longitudinal axis of the renal artery stent.
For targeted volume-rendered images, we used two opacity transfer functions. The first opacity transfer function (VRLH) was determined to maximize the visualization of high-attenuation materials (i.e., enhanced renal arteries and stent) using a low-to-high opacity curve type that incorporates high-attenuation voxels and excludes voxels with attenuation less than 100 H. An opacity value of 50% and a brightness value of 100% were assigned to the selected materials. Voxels with attenuation between 100 and 200 H were reflected at a linearly increasing opacity, and voxels between 200 and 2000 H were reflected with maximal opacity (50%). The second opacity transfer function (VRHL) was determined to maximize the visualization of a probably existing low-attenuation material (i.e., intimal hyperplasia) and to decrease the reflection of high-attenuation materials (i.e., stent and contrast materials) using a high-to-low opacity curve type that incorporates all voxels with attenuation between -50 and 2000 H. Three thresholds were defined to classify the different attenuation materials in the histogram, and a color was assigned for each threshold: the first threshold (range, -50 to 100 H; color, yellow) determines the paravascular fat tissue and, most important, the intimal lining; the second threshold (range, 100-300 H; color, red) determines the enhanced arterial lumen; and the third threshold (range, >300 H; color, blue) relates to the stent material. The threshold values were slightly adjusted to concur with source data obtained from different patients. An opacity value of 50% and a brightness value of 100% were assigned to the selected materials. Voxels with attenuation of less than 100 H were reflected with maximal opacity (50%), and voxels between 100 and 2000 H were reflected with a linearly decreasing opacity.
For volume-rendered virtual endoscopy (VRVE), we determined the opacity transfer function to maximize the visualization of low-attenuation voxels (<100 H) and high-attenuation voxels (>350 H), as well as rendering the vascular lumen invisible using a high—low—high opacity curve type.
The time required to perform diagnostic images of the stent lumen was approximately 1-2 min for multiplanar volume reformations and 2-4 min for each volume-rendering technique.
Image Analysis
MDCT data were reconstructed and interactively evaluated by two
radiologists independently who were experienced in three-dimensional
reconstructions and who were unaware of the DSA findings. Each data set was
evaluated by multiplanar volume reformations—VRLH,
VRHL, and VRVE—in a different randomized order and
on different occasions to prevent a consistent bias in the interpretation of
one image set based on a prior viewing of a different image set of the same
patient. Using digital calipers, we measured the diameter of the enhanced
lumen at its narrowest section and compared it with that of the stent on the
paraaxial and paracoronal images. The overall percentage of stenosis was
assessed and compared with that obtained on DSA. For further assessment, we
applied an ordinal 4-point scale to evaluate the impression of overall image
quality (1, very good; 2, good; 3, average; and 4, unsatisfactory) and
delineation of the lumen (1, well defined; 2, moderately defined; 3, vaguely
defined; and 4, lumen not identified).
DSA Technique and Image Analysis
Flush abdominal aortography was performed with an injection of 36 mL of
nonionic contrast material at 12 mL/sec using a 4-French pigtail catheter.
Images were obtained in 10-15° left or right anterior oblique projections.
Pressures were also measured in the renal artery and the abdominal aorta to
judge the significance of stenosis. Digital subtraction angiograms were
evaluated in consensus by two senior interventional radiologists who analyzed
each stented renal artery on the basis of intraarterial transstenotic blood
pressure gradients and the percentage of diameter reduction at the maximal
narrowing of the stent lumen compared with the stent diameter.
Statistical Analysis
As a first step, linear regression analysis was performed to investigate
the correlation between the percentage of in-stent stenosis obtained by each
rendering algorithm and that measured on DSA. The coefficient of determination
(r2) and the p value of the correlation were
determined by the analysis. Differences among the rendering techniques were
analyzed with a two-tailed paired Student's t test. In the second
step of the analysis, we evaluated the semi-quantitative data arising from
image quality and vascular delineation mode scores to assess the performance
of each rendering algorithm using Wilcoxon's signed rank test. All p
values less than 0.05 were regarded as significant. Finally, to compare
observer performance for stenosis severity measurements, without consensus, we
used the limits of agreement method
[10]. The kappa statistic
[11] was used for the
assessment of interobserver agreement on image quality and lumen
delineation.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Quantitative Image Analysis
On DSA, in-stent restenosis was confirmed in eight patients and ruled out
in eight. The volume-rendering techniques detected all restenoses with no
false-negative or false-positive visualizations of a significant restenosis
(
50%).
A linear regression analysis for both reviewers revealed a statistically significant correlation between stenosis severity measured on DSA and that assessed with volume-rendering techniques and multiplanar volume reformations. The highest correlation with DSA was achieved by means of volume rendering with VRHL (reviewer 1, r2 = 0.86; reviewer 2, r2 = 0.94). Correlation with DSA decreased to r2 = 0.79 and r2 = 0.87 by multiplanar volume reformations; r2 = 0.73 and r2 = 0.75 by VRLH; and r2 = 0.80 and r2 = 0.73 by VRVE for reviewer 1 and 2, respectively. In comparison with VRHL, restenosis quantification did not differ significantly using VRLH (p = 0.219 and p = 0.679, for reviewers 1 and 2, respectively) or VRVE (p = 0.048 and p = 0.096, for reviewers 1 and 2, respectively). On the contrary, multiplanar volume reformations differed significantly from VRHL (p = 0.028, for reviewers 1 and 2).
Interobserver agreement on the measurement of in-stent stenosis severity with multiplanar volume reformations and volume-rendering techniques was judged as good. The limits of agreement were -28%, 28% on multiplanar volume reformations; -30.3%, 27.7% on VRHL; -12.4%, 16.8% on VRLH; and -16.5%, 12.7% on VRVE.
Semiquantitative Image Analysis
Both reviewers considered the overall image quality on coronal and axial
volume-rendered images and multiplanar volume reformation images to be high,
inferred from a mean score ranging from 1.1 ± 0.3 on VRVE to
1.6 ± 0.6 on VRLH. No statistically significant differences
were found among the four rendering techniques (p > 0.59, for all
comparisons). In terms of vascular delineation, volume-rendered MDCT
angiography enabled well-defined delineation between the enhancing lumen and
the surrounding stent (Fig.
1A,1B,1C,1D)
and, in cases of an instent stenosis, between the intimal lining and the
enhancing residual lumen on one side and the stent on the other side (Figs.
2A,2B,2C,2D,2E,2F
and
3A,3B,3C,3D,3E).
The mean score ranged from 1.2 ± 0.5 on VRVE to 1.8 ±
0.5 on VRLH. The delineation of stent lumen with VRHL
was found to be significantly (p = 0.033 and p = 0.018, for
reviewers 1 and 2, respectively) better than that with VRLH.
Interobserver agreement was moderate to substantial for the assessment of
image quality (
= 0.58-0.67) and lumen delineation ( =
0.56-0.78).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
As an alternative, we assessed the usefulness of renal MDCT angiography with volume rendering for the detection of in-stent stenosis. The relative small diameter of the stent and the artifacts caused by its metallic material pose substantial challenges for CT angiography. With the implemented acquisition and reconstruction parameters, three-dimensional resolution was maximized and thus contributed a more comprehensive perceptibility to the CT angiographic algorithms. Specifically, the longitudinal resolution was substantially improved using the high-quality mode (pitch, 3) and a reconstruction interval of 0.8 mm. The application of the high-quality mode introduced an effective section thickness that is equal to the nominal section thickness (i.e., 1.25 mm) and resulted in minimizing the slice sensitivity profile while maintaining signal-to-noise ratio [16]. Furthermore, the retrospective targeted reconstruction of the field of view decreased the pixel size and slightly increased the axial resolution on the reconstructed data set in comparison with the original images [17].
Because the course of the renal artery is usually oblique to the axial plane, and in-stent stenosis is small and often eccentric, the axial source images, which are subject to partial volume averaging, may not be helpful in the detection or quantification of restenosis in the coronal plane (Fig. 2A). The volume-rendering technique has been found to be more accurate than other rendering techniques in quantifying vascular stenosis [5,6,7], whereas multiplanar reformations present direct images of structures that would otherwise be obscured. The application of targeted volume rendering combines the advantages of multiplanar reformations [18] and volume rendering [7], enabling the rendered images to overcome partial volume averaging on multiplanar volume reformation images. With the interactive rendering, the stent lumen can be displayed in paracoronal and paraaxial images or endoscopic images that reveal the maximal detail inherent in the data set.
Although no significant differences were found among the three volume-rendering techniques in the quantification of in-stent stenosis, image quality and vascular delineation on VRHL images were considered better than those on VRLH. VRLH opacity transfer function, commonly used to generate volume-rendered images of nonstented arteries, maximizes the reflection of high-attenuation materials and excludes low-attenuation voxels that represent paravascular fat tissue as well as the intimal hyperplasia in the stent lumen from the histogram. This transfer function results in rendering the in-stent stenosis invisible. A similar opacity transfer function has been used to generate volume-rendered CT angiographic images of stented carotid arteries [19]. In contrast, volume rendering with an inverted VRHL function implemented a wider histogram so that voxels with values between -50 and 100 H were reflected at highest opacity, whereas voxels between 100 and 2000 H were reflected with a linearly decreasing opacity. The former parameter contributed to the visual perception and improved delineation of the intimal lining, and the latter prevented severe over-estimation of the stent wall. VRVE provided three-dimensional internal images of the stent lumen with high image quality and vascular delineation, particularly of patent stents. However, VRVE tended to underestimate the restenosis (Table 1).
|
Subvolume volume-rendered images allowed the opportunity to display a direct image of the stent lumen without the necessity of eliminating overlying structures or the need for time-consuming image segmentation. However, the targeted volume-rendered images were limited in visualizing merely the stented segment rather than the entire course of the renal artery, which could be evaluated on conventional volume-rendered images. A further limitation is the subjective selection of display parameters that may influence the precision of the stenosis quantification and a decrease in repeated examinations. Because generating a reliable volume-rendered angiogram depends on the density of the stent material and the contrast attenuation, it can be achieved, in our experience, by interactive modification of the window level so that it concurs with the attenuation of source data. In fact, despite some differences between both reviewers that were related to the variations involved in the interactive reconstruction of volume-rendered images, the interobserver variability was slight. However, the clinical applicability and generalizability of the results were limited by the small population size.
Finally, the 9.6-cm display field of view reconstructed using the standard algorithm was subject to increased image noise. The latter could be reduced using a smoother reconstruction kernel, which, however, tends to decrease spatial resolution [17]. Because the stent lumen comprises small and high-contrast structures, we believe that preserving spatial resolution has a greater impact on image quality than decreasing image noise and improving contrast resolution.
After stent deployment, patients are required to undergo continual clinical and radiologic follow-up examinations. At our institution, follow-up, including MDCT angiography, is performed 6 months after intervention and yearly thereafter or when restenosis is suspected. Patients with impaired renal function are usually excluded from MDCT angiography follow-up. When significant restenosis (> 50%) is seen on MDCT angiography, DSA is used at reintervention.
In summary, the small number of patients in our study does not allow us to draw final conclusions about the usefulness of MDCT angiography for the quantification of in-stent stenosis. The results, however, reflect the feasibility of MDCT angiography in the evaluation of renal artery stent lumen and detection of restenosis and suggest that the integration of volume rendering, particularly with inverted opacity transfer function, augments its clinical potential as a noninvasive technique for the assessment of renal artery patency in patients treated with stent deployment.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. C. Josephs, H. A. Rowley, G. D. Rubin, and for Writing Group 3 Atherosclerotic Peripheral Vascular Disease Symposium II: Vascular Magnetic Resonance and Computed Tomographic Imaging Circulation, December 16, 2008; 118(25): 2837 - 2844. [Full Text] [PDF]
|
|
|
|
 |
|
 M. Rieger, A. Mallouhi, T. Tauscher, M. Lutz, and W. R. Jaschke Traumatic Arterial Injuries of the Extremities: Initial Evaluation with MDCT Angiography. Am. J. Roentgenol., March 1, 2006; 186(3): 656 - 664. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. T. Johnson and E. K. Fishman IV Contrast Selection for MDCT: Current Thoughts and Practice Am. J. Roentgenol., February 1, 2006; 186(2): 406 - 415. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Suzuki, S. Furui, and T. Kaminaga Accuracy of Automated CT Angiography Measurement of Vascular Diameter in Phantoms: Effect of Size of Display Field of View, Density of Contrast Medium, and Wall Thickness Am. J. Roentgenol., June 1, 2005; 184(6): 1940 - 1944. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Z. Morvay, E. Nagy, R. Bagi, G. Abraham, R. Sipka, and A. Palko Sonographic Follow-up After Visceral Artery Stenting J. Ultrasound Med., August 1, 2004; 23(8): 1057 - 1064. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Suzuki, S. Furui, T. Kaminaga, and T. Yamauchi Measurement of Vascular Diameter In Vitro by Automated Software for CT Angiography: Effects of Inner Diameter, Density of Contrast Medium, and Convolution Kernel Am. J. Roentgenol., May 1, 2004; 182(5): 1313 - 1317. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Sheth and E. K. Fishman Multi-Detector Row CT of the Kidneys and Urinary Tract: Techniques and Applications in the Diagnosis of Benign Diseases RadioGraphics, March 1, 2004; 24(2): e20 - e20. [Abstract] [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
