| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Vascular and Interventional Radiology |
1 From the Institute of Clinical Radiology, Ludwig-Maximilians-University of Munich, Marchioninistrasse 15, 81377 Munich, Germany (S.O.S., J.R., C.H.W., H.J.M., T.W., O.D., M.F.R.); and Central Biostatistics Unit, German Cancer Research Center, Heidelberg, Germany (C.I.). Received October 20, 2003; revision requested January 12, 2004; final revision received June 20; accepted July 26. Address correspondence to S.O.S. (e-mail: stefan.schoenberg@med.uni-muenchen.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: The study was approved by the institutional review board, and consent was obtained from all patients. Forty-five patients (17 women, 28 men; mean age, 62.2 years) were evaluated for suspected renal artery stenosis. Three-dimensional gadolinium-enhanced MR angiograms were acquired with isotropic spatial resolution of 0.8 x 0.8 x 0.9 mm in 23-second breath-hold with an integrated parallel acquisition technique. In-plane diameter of stenosis was measured along vessel axis, and perpendicular diameter and area of stenosis were assessed in cross sections orthogonal to vessel axis, on multiplanar reformations. Interobserver agreement between two radiologists in measurements of in-plane and perpendicular diameters of stenosis and perpendicular area of stenosis was assessed with mean percentage of difference. In a subset of patients, degree of stenosis at MR angiography was compared with that at DSA (n = 20) and intravascular US (n = 11) by using Bland-Altman plots and correlation analyses.
RESULTS: Mean percentage of difference in stenosis measurement was reduced from 39.3% ± 78.4 (standard deviation) with use of in-plane views to 12.6% ± 9.5 with use of cross-sectional views (P < .05). Interobserver agreement for stenosis grading based on perpendicular area of stenosis was significantly better than that for stenosis grading based on in-plane diameter of stenosis (mean percentage of difference, 15.2% ± 24.2 vs 54.9% ± 186.9; P < .001). Measurements of perpendicular area of stenosis on MR angiograms correlated well with those on intravascular US images (r2 = 0.90).
CONCLUSION: Evaluation of cross-sectional images reconstructed from high-spatial-resolution three-dimensional gadolinium-enhanced MR renal angiographic data increases the accuracy of the technique and decreases interobserver variability.
© RSNA, 2005
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
A second limitation of 3D gadolinium-enhanced MR angiography is the high interobserver variability in morphologic grading of renal artery stenosis on the basis of the sole evaluation of coronal source images and maximum intensity projections. Investigators in two previous studies reported moderate to poor interobserver agreement, with 
values of less than 0.50, a result that points to another source of error in the grading of renal artery stenosis (3,4).
Meta-analyses have shown that the inclusion criteria for renal artery intervention are inconsistently defined, with the degree of stenosis varying from 30% to 70% (7,8). One important prerequisite for appropriate patient selection is a consistent cutoff criterion for the degree of stenosis. Along with other contributing factors, this might be a reason for the moderate results demonstrated in patient outcomes after intervention: Only one-third of patients showed improvement in blood pressure or renal function (7,9). Researchers in recent cardiology studies about coronary artery interventions clearly demonstrated that the hemodynamic significance of atherosclerotic stenoses, because of their eccentricity, is more accurately determined with intravascular ultrasonography (US), which can be used to assess the true reduction in vessel area at the stenosis site (10). This third limitation of 3D gadolinium-enhanced MR angiography has also fueled discussions about the overall clinical usefulness of this technique (11).
In principle, 3D gadolinium-enhanced MR angiography has potential for assessment of the vessel area at the stenosis site, because a 3D data set is acquired that can be reformatted in any plane at a workstation. Multiplanar reformation, however, requires 3D data sets with isotropic submillimeter spatial resolution in all three dimensions to avoid geometric distortions and ensure accurate delineation of the vessel area, while spatial resolution is limited by the breath-hold time for acquisition of the 3D data set. Most investigators, therefore, have used data sets with anisotropic spatial resolution and relatively thick sections of more than 1.5 mm in the anteroposterior direction. With the incorporation of parallel acquisition techniques, spatial resolution can be substantially improved by a factor of two to three without prolonging the data acquisition (12).
Thus, the purpose of our study was to retrospectively compare renal 3D gadolinium-enhanced MR angiography, performed with an integrated parallel acquisition technique (iPAT) for high isotropic spatial resolution, with selective DSA and intravascular US for accuracy of diameter and area measurements in renal artery stenosis.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Inclusion criteria were adopted from the international guidelines for the work-up of patients with renal artery stenosis that were published by Rundback et al (13). Because of unilateral occlusion of the renal artery in four patients and unilateral nephrectomy in one patient, 85 renal arteries were assessed with 3D gadolinium-enhanced MR angiography.
Of the 45 patients, 20 underwent selective DSA. Inclusion criteria for DSA were (a) that the patient was a candidate for intervention for treatment of hemodynamically significant (>50%) renal artery stenosis (9) (n = 11) with percutaneous transluminal angioplasty and stent placement or (b) that the referring clinician lacked confidence in the results of MR angiography and required confirmation with DSA (n = 9). In patients who were candidates for percutaneous transluminal angioplasty, only the renal artery that would be subject to intervention was imaged with selective DSA. In the remaining nine patients, both renal arteries were examined by obtaining a midstream aortogram at the level of the renal arteries. Since one of these patients underwent nephrectomy, 17 renal arteries were imaged with this method. Thus, a total of 28 renal arteries were available for the comparison of DSA images with MR angiograms. In the remaining patients, DSA could not be performed, either because the patient was not a candidate for intervention according to the published criteria (because of previous severe reaction to intravenously administered iodinated contrast media, glomerular filtration rate < 30 mL/min, or severe renal atrophy with a kidney size < 7 cm; n = 10) or because the referring physician had accepted the result of MR angiography without further invasive work-up of the patient (n = 15) and had initiated medical treatment of stenosis according to the general regimen reported by investigators in the Dutch Renal Artery Stenosis Intervention Cooperative study group (14).
In all 11 patients undergoing intervention with percutaneous transluminal angioplasty and stent placement, intravascular US of the stenosed artery was performed. One patient had three stenosed renal arteries, and the other 10 patients had one stenosed artery each, for a total of 13 intravascular US data sets available for comparison.
Intravascular US and DSA were performed in the same session in patients who were to undergo intervention. The mean interval between MR angiography and DSA was 17.2 days ± 24.3 (range, 30 days prior to 61 days after), and that between MR angiography and intravascular US was 7.6 days ± 6.9 (range, 1 day prior to 20 days after).
As all studies were considered clinically necessary and were requested by the referring physician, all patients gave oral and written consent according to the local institutional review board standards for clinically indicated MR imaging, DSA, and intravascular US. The study was approved by the institutional review board.
High-Spatial-Resolution MR Angiography
Three-dimensional gadolinium-enhanced MR angiography was performed with a 1.5-T MR system (Magnetom Sonata Maestro Class; Siemens Medical Solutions, Erlangen, Germany) with maximum gradient strength of 40 mT/m (minimum rise time, 200 µsec) and eight receiver channels. For signal reception, a recently introduced dedicated 12-element array coil system was used that consists of one anterior and one posterior flexible coil, each with a set of six receiver elements. Outside elements on each side were combined to fit the limit of eight receiver channels. Parallel imaging was performed with a generalized autocalibrating partially parallel acquisitions (GRAPPA) algorithm based on autocalibrating simultaneous acquisition of spatial harmonics (15). Contrary to that with other algorithms for parallel imaging (eg, sensitivity encoding), correction to the parallel data acquired with this algorithm is performed in the k-space domain and not in the image domain. This algorithm therefore is less susceptible to artifacts that are propagated at the center of the image by aliasing of tissue outside the field of view (FOV) (Fig 1). The necessary coil profiles are determined during 3D gadolinium-enhanced MR angiography by acquiring additional reference lines in the center of k-space, a process referred to as autocalibration. The GRAPPA parameters were set to an iPAT acceleration factor of two, with 24 reference lines (12 additionally acquired autocalibration signal lines). These settings allowed the acquisition of a full 3D data set with an isotropic spatial resolution of 0.8 x 0.8 x 0.9 mm in a 23-second breath hold by using the following parameters: 3.79/1.39; FOV (readout direction), 400 mm; FOV (phase-encoding direction), 87%; number of sections per slab, 80; section thickness, 0.9 mm; frequency encoding steps, 512; flip angle, 25°; and bandwidth, 350 Hz per pixel. Because of the integrated reference acquisition, a linear k-space acquisition mode had to be chosen for the 3D gadolinium-enhanced MR angiographic sequence. The phase-encoding direction was set left to right, and the section-encoding direction was set anterior to posterior. An oblique coronal acquisition plane was chosen for the 3D slab along the course of the abdominal aorta. No additional saturation pulses were applied for signal suppression outside the FOV.
|
R. In addition, the SNR of iPAT images is generally further reduced by the so-called geometry factor g, the value of which depends on the chosen coil configuration. If spatial resolution is increased at the same time, the overall SNR is further decreased. For the chosen settings (acceleration factor of two, 512 x 512 matrix), an overall decrease in SNR to about 50% of the original value has to be considered. To regain SNR, we used as a contrast agent the gadolinium chelate gadobutrol (Gadovist; Schering, Berlin, Germany), which contains 1.0 mol/L gadolinium and for which the reported SNR and contrast-to-noise ratio are higher than those for the standard 0.5-mol gadolinium chelates. Gadobutrol was injected at a dose of 0.125 mmol per kilogram of body weight and at a rate of 2 mL/sec. Arrival time of the contrast agent at the level of the renal arteries was determined with a test bolus of 1 mL gadobutrol, as previously described (16). Since the depth of inspiration is not absolutely identical between multiple breath holds, and to avoid artifacts from inconsistencies of subtraction at the vessel margins, subtraction from a precontrast image was not used.
All gadolinium-enhanced MR angiographic studies were of diagnostic quality, without substantial artifacts from motion or inadequate bolus timing. The assessment of image quality was limited to the main stem (proximal, middle, and distal segments) of the renal artery. In particular, no severe aliasing artifacts were observed at the center of the image with use of the GRAPPA algorithm, despite the fact that tissue located more than 2 cm outside the lateral kidney margin was excluded from the FOV (Fig 1). Slight ring artifacts occurred only in patients with massive obesity and did not affect the image interpretation.
DSA Examinations
All DSA examinations were performed by a radiologist (T.W.) with 10 years of experience in performing DSA. First, a transfemorally inserted 4-F pigtail catheter was positioned at the expected level of the renal artery origins, between the 12th thoracic and first lumbar vertebrae. Anteroposterior DSA was performed with a C-arm system (Multistar T.O.P.; Siemens Medical Solutions) and with an injection of 25 mL of the contrast medium iohexol (Omnipaque; GE Healthcare Bio-Sciences, Little Chalfont, England) at a rate of 16 mL/sec. Additional oblique views were obtained at an angle of 18°–25° when necessary because of tortuous vessels or the overlap of another vessel over the renal artery. If stenosis was suspected on the basis of these views, the renal artery was also selectively catheterized with a 4-F renal double-curve catheter or Cobra catheter, with manual injection of an additional 5–10 mL of iohexol. Hemodynamically significant stenosis was defined as stenosis of at least 50% of the vessel diameter, which mathematically corresponds to at least 75% of the vessel area.
Intravascular US Examinations and Image Interpretation
Intravascular US (ClearView Ultra; Boston Scientific, Maple Grove, Minn) of the renal artery was performed by using a 0.014-inch pressure guidewire (PressureWire; RADI Medical Systems, Uppsala, Sweden) for insertion of a 40-MHz single-element ("monorail") 2.7-F intravascular US catheter (Atlantis SR; Boston Scientific). The intravascular US catheter was advanced through the guidewire and past the stenosed segment. The US catheter was connected to a motorized pullback system with a pullback speed of 1 mm/sec. The complete drawback was videotaped. All intravascular US examinations were performed by a cardiologist with 5 years of experience in performing intravascular US. Data analysis was immediately performed with off-line viewing of the videotape. The intravascular US data were independently evaluated by a fellow in vascular and interventional radiology (C.H.W.) who was blinded to the results of DSA and MR angiography. According to the standard method of coronary vessel analysis, the lumen area within the maximally stenosed segment (minimal lumen area) and the lumen area of a nearby distal undiseased reference segment (reference lumen area) were determined (17). The area of stenosis could then be calculated from the ratio of the minimal lumen area to the reference lumen area. Since the intravascular US examination procedure entails a minimal risk of renal artery dissection, only the 13 renal arteries (in 11 patients) with a significant (at least 50%) stenosis amenable to treatment with dilation and stent placement were examined with intravascular US.
Analysis of MR Angiograms and DSA Images
The 3D MR angiographic data sets and the DSA images were independently evaluated by two board-certified attending radiologists with special training in vascular and interventional radiology (S.O.S., J.R.), each of whom had 6 years of experience in acquiring and reading MR angiograms and DSA images. Both were blinded to the results of intravascular US and had no clinical information about patients, since the data sets were presented in an anonymous and randomized fashion and were identifiable only by a four-digit random number. All data sets were reviewed electronically at a dedicated 3D postprocessing workstation to allow electronic measurements, windowing, and 3D reformation.
Aliasing artifacts that appeared at the center of the MR images because of the use of iPAT were rated on a three-point scale as absent, mild and not affecting diagnostic evaluation, or severe and limiting diagnostic assessment.
Analysis of the 3D MR angiographic data sets included viewing of various oblique reformations along the vessel axis to assess the degree of in-plane diameter of stenosis (Fig 2a). If the stenotic vessel lumen was not completely visualized in a single oblique plane, a targeted thin-section maximum intensity projection with a maximum thickness of 5 mm was used. Measurement of the in-plane vessel diameter at the stenosis site (Dis) and in-plane diameter in a normal vessel segment (Din) distal to the stenosis (in a segment without dilatation) was performed electronically. The in-plane diameter of stenosis (DSi) was calculated as a percentage on the basis of in-plane diameter measurements, according to the generally used North American Symptomatic Carotid Endarterectomy Trial definition criteria (18), as follows: DSi = (1 – Dis/Din) · 100.
|
|
To account for eccentric renal artery stenosis, the maximum diameter in the cross sections was chosen for the measurement of perpendicular diameter both in the normal vessel and at the site of stenosis.
In addition, the perpendicular area of the stenosed vessel (ASp) was determined, also as a percentage, from electronic measurements of the vessel area in cross sections at the site of maximum stenosis (Aps) and in a normal vessel segment (Apn), as follows: ASp = (1 – Aps/Apn) · 100.
If no stenosis was identified on the images, the two readers were advised to perform the measurements in two proximal renal artery segments 1 cm apart. To avoid a confounding effect, the diameters of accessory renal arteries were excluded from the analysis. The mean reading time for each reader, including postprocessing (multiplanar reformations), was 20 minutes.
On DSA images, the degree of stenosis could be measured in plane only. Measurements of the in-plane diameter at the stenosis site (DSAis) and the in-plane normal vessel diameter (DSAin) distal to the stenosis, in a segment without poststenotic dilatation, were performed electronically by using the digital data sets. The in-plane diameter of stenosis (DSAi) was calculated as a percentage according to the generally used North American Symptomatic Carotid Endarterectomy Trial definition criteria, as follows (18): DSAi = (1 – DSAis/DSAin) · 100.
In addition to the degree of stenosis, the location and type of stenosis were evaluated and noted by both readers independently. The location of stenosis was classified as ostial and/or proximal for a stenosis located within 10 mm of the ostium of the renal artery or as intermediate and/or distal for a stenosis located farther downstream, in the middle or distal segment. The stenosis type was defined according to the apparent cause of stenosis (either atherosclerosis or fibromuscular dysplasia). In cases of fibromuscular dysplasia, the number and location of strictures were noted.
Statistical Analysis
Statistical analyses were performed by using software (SPSS, version 11.5.1 for Windows; SPSS, Chicago, Ill).
Differences in age and weight between women and men were assessed by means of an unpaired Wilcoxon rank sum test.
For comparison of MR angiography with DSA, 20 patients and 28 renal arteries were evaluated. For comparison of MR angiography with intravascular US, 11 patients and 13 renal arteries were evaluated. Comparisons of MR angiographic, DSA, and intravascular US measurements were performed by means of correlation analysis. Bland-Altman plots were prepared, and the correlation coefficient (r2) and 95% confidence interval were calculated (19). Differences between the percentage of stenosis measured at MR angiography and that measured at DSA were calculated with the following equation: D = (MRAs – DSAs)/DSAs · 100, where D is the percentage of difference, MRAs is the percentage of stenosis as measured on MR angiograms, and DSAs is the percentage of stenosis as measured on DSA images.
Differences between the two readers with regard to the percentage of stenosis were calculated as follows: D = (FRs – SRs)/SRs · 100, where FRs is the percentage of stenosis as measured by reader 1 and SRs is the percentage of stenosis as measured by reader 2.
Because the results of Kolmogorov-Smirnov and Shapiro-Wilk tests for normally distributed data negated the null hypothesis, nonparametric tests were used for the subsequent analysis. Differences between MR angiography and DSA with regard to the assessment of percentage of stenosis on the basis of diameter measurements were analyzed for statistical significance by means of a paired Wilcoxon signed rank test. All values were calculated as the mean ± standard deviation. Differences between the two readers in the percentage of stenosis calculated on the basis of diameter measurements (both in-plane and cross-sectional), as well as in the calculated area of stenosis, were also assessed with a paired Wilcoxon signed rank test. A P value of less than .05 was considered to indicate a significant difference. Nonparametric methods were used because they require few assumptions about the distribution of data. In particular, the classic assumption of normal distribution is not required. Nevertheless, the nonparametric procedures used for the analysis are only slightly less efficient than the corresponding normal sampling theory procedures when the underlying distribution is normal, and they may be more efficient when the distribution is not normal (20,21).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Comparison of MR Angiography and DSA
Results of the statistical analysis for agreement between measurements of diameter and area of stenosis obtained with MR angiography and those obtained with DSA are shown in Table 1 and Figure 3. Measurements of the diameter of stenosis in cross sections perpendicular to the vessel axis revealed a significantly higher accuracy for MR angiography compared with DSA than did measurements of the in-plane diameter of stenosis (P < .05). The mean difference between the perpendicular diameter of stenosis measured at MR angiography and the in-plane diameter of stenosis measured at DSA was only 12.6% ± 9.5 (reader 1), compared with a difference of 39.3% ± 78.4 for the in-plane diameter of stenosis measured with MR angiography versus the in-plane diameter measured with DSA. The correlation between the perpendicular diameter of stenosis at MR angiography and the in-plane diameter at DSA was significantly better (r2 = 0.89, P < .05) than that between the in-plane diameter at MR angiography and the in-plane diameter at DSA. Figure 4 shows a patient with a low-grade stenosis in whom the accuracy of stenosis grading improved with use of the perpendicular (versus the in-plane) diameter measurement at MR angiography in comparison with the in-plane diameter measurement at DSA.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
It is widely believed that 3D gadolinium-enhanced MR angiography results in overestimation of the degree of renal artery stenosis on the basis of measurements of the in-plane diameter of stenosis on targeted subvolume maximum intensity projections (22,23). The Bland-Altman plots from this study, however, show that the effect on estimation of renal artery stenosis is not one-sided. Instead, both substantial over- and underestimation occur, particularly with degrees of stenosis of approximately 50%. For stenoses of less than 20% or more than 80%, the agreement between 3D MR angiography and DSA is substantially better. This observation leads one to conclude that the measurement of in-plane stenosis diameter with 3D gadolinium-enhanced MR angiography alone is an acceptable technique only for ruling out a renal artery stenosis or for confirming a high-grade stenosis and that the results of this technique for assessment of stenoses of 40%–60% are almost random. It is therefore concluded that any stenosis determined to be within the range of 40%–60% should be confirmed with DSA if 3D gadolinium-enhanced MR angiography is performed with limited spatial resolution that allows only the determination of the in-plane diameter of stenosis.
On the other hand, the accuracy of 3D gadolinium-enhanced MR angiography for stenosis grading is significantly improved if the diameter of stenosis is measured perpendicular to the vessel axis, in cross sections, particularly for stenoses of 40%–60%, as demonstrated by the Bland-Altman plots. This effect can be explained mainly by two facts: the typical tortuous course of the renal artery, which induces geometric distortions, and the eccentricity of atherosclerotic plaques (24). From the ostium, the renal arteries typically take an oblique course in all three dimensions, with an initial curve in the anterior direction followed by a more cranial and posterior course. Atherosclerotic changes tend to occur in the proximal segment, predominantly in the anterior or posterior wall, a location that usually results in a more oval or irregularly shaped residual lumen rather than a round one (24). The determination of the perpendicular diameter of stenosis is much less influenced by these two effects than is that of the in-plane diameter of stenosis, since measurements of the former are performed in exactly perpendicular cross sections for both normal and stenotic segments. In addition, the largest oblique diameter was chosen in both segments to eliminate the effect of an eccentric stenosis.
The results are even more pronounced when one considers the interobserver variability for the grading of renal artery stenosis. Interobserver variability is as important for a diagnostic modality as is numeric accuracy of the measurements (4). The Bland-Altman plots show an almost rhombus-shaped distribution of the measurement differences between readers 1 and 2, revealing maximum discrepancy in measurement of stenoses of 40%–60%. Again, this discrepancy is substantially reduced when the diameter of stenosis is measured in a plane perpendicular to the vessel axis. However, interobserver agreement is further improved in this particular range if the perpendicular area of stenosis is measured. This can be explained by the fact that it is technically more demanding for readers independently to identify the largest oblique diameter within cross sections than to draw a complete contour along the vessel margin for measurement of the vessel area. On the other hand, the correlation of the perpendicular area of stenosis on MR angiograms with the in-plane diameter of stenosis on DSA images is worse than that of the perpendicular diameter of stenosis on MR angiograms, since DSA is a projection technique that allows measurements of in-plane stenosis diameter only.
Investigators in the Renal Artery Diagnostic Imaging Study in Hypertension also came to the conclusion that 3D gadolinium-enhanced MR angiography has an acceptable specificity but performs poorly with regard to the exact grading of renal artery stenosis (6). In that study, however, measurements of the perpendicular diameter and perpendicular area of stenosis were not obtained because of the lower spatial resolution achieved with use of 3D gadolinium-enhanced MR angiography. To our knowledge, measurements of the perpendicular diameter and perpendicular area of stenosis with 3D MR angiography, and the correlation of these parameters with measurements obtained with DSA and intravascular US, respectively, have not previously been performed. This is most likely related to the fact that 3D gadolinium-enhanced MR angiography with isotropic submillimeter spatial resolution has not been feasible within a tolerable breath-hold time of less than 25 seconds. Integrated parallel acquisition techniques appear to hold promise for high-spatial-resolution contrast-enhanced MR imaging of the renal arteries. With current 1.5-T MR systems, an acceleration factor of two is routinely possible and enables the near doubling of spatial resolution for a given acquisition time. The use of the GRAPPA algorithm for parallel imaging allows tailoring of the FOV to the region of interest without the propagation of major aliasing artifacts at the center of the image. With use of the GRAPPA algorithm, artifacts are substantially reduced in comparison with those that result from the use of sensitivity encoding. This reduction is related to the fact that the signal from the aliased tissue is more uniformly distributed throughout the entire image when the GRAPPA algorithm is used (25).
In this in vivo study in renal arteries, measurements of perpendicular diameter of stenosis at MR angiography were validated in comparison with the reference standard of selective DSA, and measurements of perpendicular area of stenosis at MR angiography were validated in comparison with intravascular US. To our knowledge, no previous comparative studies between the area of stenosis at 3D gadolinium-enhanced MR angiography and that at intravascular US have been performed for the renal arteries. A good correlation was found between measurements of perpendicular area of stenosis in renal arteries at MR angiography and those at intravascular US, with a tendency toward underestimation of perpendicular area of stenosis at MR angiography. This result is supported by data from comparisons of MR angiography with intravascular US in other vessel territories, such as the superficial femoral artery (26). It is possible that the technique of measuring the diameter and area of stenosis in vessel cross-sections on reformatted images from high-resolution MR angiographic data sets will gain wide use for grading of renal artery stenosis in the near future, since it is the only method that allows assessment of the true reduction of the luminal vessel area at the stenosis site (24). This method is already used in cardiology practice at some institutions to assess the hemodynamic significance of coronary artery stenoses. Commonly, a 75% area of stenosis—which, theoretically, corresponds to a 50% stenosis diameter—is considered hemodynamically significant (27). It is evident from this study that the criterion of a 75% perpendicular area of stenosis measured at 3D gadolinium-enhanced MR angiography is a much more reliable criterion than is a 50% stenosis diameter measured in plane. For the latter measurement, the mean percentage of difference between the two readers in our study was 55% (range within 1 standard deviation, –36% to +45%), while for the perpendicular area measurement it was only 15% (range within 1 standard deviation, –20% to +18%).
Motion of the renal artery during 3D gadolinium-enhanced MR angiographic acquisition remains an unsolved problem for the accurate assessment of renal artery stenosis. This motion occurs even in the presence of completely suspended breathing, as a result of additional diaphragmatic motion. Investigators in one study demonstrated that these effects are much more severe in the distal part of the renal artery than in the proximal segments (28). Those authors claimed that shortening of the total acquisition time would reduce these effects at the cost of a further decrease in SNR. Until this problem is solved, the performance of 3D gadolinium-enhanced MR angiography for grading of distal segmental renal artery stenosis remains uncertain. Therefore, it has to be concluded that, particularly in patients with fibromuscular dysplasia of the distal and intrarenal vessel segments, verification with DSA is still required, since these patients show a great benefit even from dilation of distal stenoses (29). Investigators in the Renal Artery Diagnostic Imaging Study in Hypertension obtained poor results with 3D gadolinium-enhanced MR angiography for the assessment of fibromuscular dysplasia, with accuracies of less than 60%. In our patient group, only three stenoses were found downstream of the ostial and proximal segments of the renal artery. This number is too limited to allow assessment of the accuracy of measurements of perpendicular diameter and perpendicular area of distal stenoses.
In our study, interobserver agreement for DSA was excellent, with less than 10% deviation between the two readers for the exact degree of renal artery stenosis. It must be pointed out, however, that these results are not uniformly supported by the literature. Investigators in two large studies found only moderate interobserver agreement for DSA. In addition, there are reports of renal artery stenoses that were underestimated or missed at DSA (30–32). The reason for the good interobserver agreement in this study might be the fact that, in all cases of a suspected renal artery stenosis, the overview angiograms were completed by using selective catheterization of both renal arteries, and images were obtained at a minimum of two different angles. This is particularly important for vessels with a tortuous course. In addition, all images were reviewed by two experienced attending radiologists with special training in vascular and interventional radiology.
One limitation of our trial is the small number of cases, which allows only modest conclusions about the overall accuracy of measurements of perpendicular area of stenosis. However, the new method of renal artery intravascular US is still technically challenging and was available only in patients with clearly treatable renal artery stenosis, since the procedure involves a minimal risk of renal artery dissection. In addition, the narrow 95% confidence interval for the correlation between intravascular US and the area of stenosis at MR angiography underlines the validity of the results. No statistical adjustment for multiple testing was carried out, however, which might impose a limitation on the interpretation of all statistically significant results.
In summary, it has been demonstrated that measurements of diameter and area of stenosis on cross-sectional images derived from high-spatial-resolution 3D gadolinium-enhanced MR angiographic data sets improve accuracy and interobserver agreement in the grading of proximal renal artery stenosis and correlate well with measurements at DSA and intravascular US. In addition, the cutoff value for a hemodynamically significant stenosis can be better defined with measurement of the area of stenosis. Therefore, consideration should be given to the use of this type of image analysis on a routine basis as a replacement for current methods of measuring in-plane diameter of stenosis, if the spatial resolution of the 3D gadolinium-enhanced MR angiograms permits. Nevertheless, the role of 3D gadolinium-enhanced MR angiography for assessment of distal renal artery stenosis remains unclear.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, S.O.S., J.R., M.F.R.; study concepts, S.O.S., J.R., O.D., T.W.; study design, S.O.S., J.R., O.D., C.H.W.; literature research, S.O.S., J.R., H.J.M., O.D., C.I.; clinical studies, S.O.S., J.R., H.J.M., T.W., C.H.W., M.F.R.; data acquisition, S.O.S., J.R., H.J.M., T.W., C.H.W., M.F.R.; data analysis/interpretation, S.O.S., J.R., O.D., C.H.W., H.J.M., C.I.; statistical analysis, S.O.S., J.R., H.J.M., C.I.; manuscript preparation, O.D., T.W., C.H.W., S.O.S., J.R., C.I.; manuscript definition of intellectual content, S.O.S., J.R., M.F.R., O.D.; manuscript editing, S.O.S., J.R., O.D., H.J.M., M.F.R.; manuscript revision/review, T.W., C.H.W., C.I.; manuscript final version approval, S.O.S., J.R., M.F.R., C.I.
S.O.S. and J.R. contributed equally to this work.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
), correlation between measurements at MR angiography and intravascular US is excellent; for reader 2 (
), correlation is somewhat weaker because of substantial underestimation of a single high-grade stenosis at intravascular US. In general, area of stenosis was slightly underestimated with 3D gadolinium-enhanced MR angiography. (b) Bland-Altman plot shows a narrow range of variation between measurements of area of stenosis at intravascular US (IVUS) and those at 3D gadolinium-enhanced MR angiography (MRA).
