| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Cardiac Imaging |
1 From the Departments of Radiology (J.B., S.D., R.B.) and Cardiology (J.P., F.E.R.), Gasthuisberg University Hospital, Leuven, Belgium; and Department of Radiology, Mayo Clinic, Jacksonville, Fla (R.K.). Received October 29, 2001; revision requested December 10; final revision received July 17, 2002; accepted August 13. Address correspondence to J.B., Department of Radiology, University Hospitals, Catholic University of Leuven, Herestraat 49, B-3000 Leuven, Belgium (e-mail: jan.bogaert@uz.kuleuven.ac.be).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Twenty-one patients underwent 3D navigator MR coronary angiography immediately before catheterization. Two observers independently graded image quality on a scale from 1 (unreadable) to 5 (excellent), quantified coronary artery visualization, and evaluated the presence of significant (ie, >50% narrowing) stenoses. 
statistics were used to assess interobserver agreement, and receiver operating characteristic (ROC) analysis was used to assess stenosis detection.
RESULTS: For two of 21 patients, MR coronary angiogram quality was insufficient for analysis (mean score < 2). For the remaining 19 patients, the mean image quality scores assigned by observers 1 and 2 were 3.3 ± 1.0 (SD) and 3.2 ± 0.9, respectively. A mean of 71% of all coronary artery segments were visible at MR coronary angiography, and there was 91% agreement between the observers ( = 0.78). Observers 1 and 2 detected significant stenoses (n = 29) at MR coronary angiography with sensitivities of 44.4% and 55.5%, respectively; specificities of 95.1% and 83.7%, respectively; and 80% agreement ( = 0.35). Areas under the ROC curve were 0.817 and 0.795 for observers 1 and 2, respectively.
CONCLUSION: Large portions of the coronary arteries can be visualized with MR coronary angiography. Imaging results are not consistently reliable, however. The examination is premature for routine clinical assessment of significant coronary artery stenosis owing to low sensitivity and large observer variability.
© RSNA, 2003
Index terms: Coronary angiography, 54.1242, 54.1244 • Coronary vessels, MR, 54.121412, 54.121416, 54.12142 • Coronary vessels, stenosis or obstruction, 54.76 • Magnetic resonance (MR), vascular studies, 54.121412, 54.121416, 54.12142
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Our aim in the present study was to examine the value of a commercially available 3D real-time navigator MR coronary angiographic examination in the detection of significant coronary artery stenoses, with conventional coronary angiography as the standard of reference.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MR Coronary Angiography
All MR coronary angiographic examinations were performed with a 1.5-T whole-body MR unit (Gyroscan ACS-NT; Philips Medical Systems, Best, the Netherlands) with Powertrak 6000 (Philips Medical Systems) gradients (23 mT/m, 220-µsec rise time), a cardiac software patch, and a five-element phased-array cardiac coil (Synergy Coil; Philips Medical Systems). An extended description of the MR coronary angiography technique used in this study has been published elsewhere (14,16,17). Patients were placed in the supine position, and the electrocardiographic leads were placed on the anterior left hemithorax.
The MR coronary angiography protocol included four different image measurements, all of which were performed during free breathing: two-dimensional gradient-echo localizer imaging, 3D turbo- field-echo (TFE) echo-planar localizer imaging (to obtain a scout image), and two 3D TFE submillimeter MR coronary angiographic sequences (for the left and right coronary artery systems). For coronary artery localization and navigator positioning at the right hemidiaphragm, two localizer sequences were performed. The first localizer sequence was an electrocardiographically triggered, free-breathing, multisection two-dimensional segmented gradient-echo (11/2.4 [repetition time msec/echo time msec], 256 x 128 matrix, 450-mm field of view) acquisition of nine transverse, nine coronal, and nine sagittal interleaved images of the thorax (acquisition duration, 1 minute). On these images, the navigator was positioned at the dome of the right hemidiaphragm, and the localized transverse 3D volume for the subsequent scout image was planned (Fig 1).
|
For enhancement of the contrast between blood and myocardium, a T2-weighted prepulse preparation (18) preceded the navigator and imaging parts of the examination (Fig 2). This preparation, as well as the following two imaging sequences, was performed during the middle phase of diastole. The third and fourth sequences were used to perform 3D submillimeter TFE MR angiography of the coronary arteries—that is, one sequence for the left and one sequence for the right coronary artery system. The 3D-segmented k-space TFE sequences (7.1/1.97, 30° flip angle) started with the described flow-insensitive T2-weighted prepulse for contrast enhancement followed by a localized anterior saturation prepulse, the navigator, the spectrally selective fat-saturation pulse, and finally the TFE image acquisitions (Fig 2). A bandwidth of 135 Hz per pixel was used. A field of view of 360 mm and a matrix of 512 x 360 yielded an in-plane acquisition spatial resolution of 0.7 x 1.0 mm. The TFE mode consisted of 38 shots, with a low to high fill in of the k space. A 30-mm-thick slab was obtained and yielded 20 1.5-mm-thick sections.
|
The navigator used in the second to fourth sequences (ie, 3D TFE echo-planar scout sequence and two 3D TFE submillimeter sequences) is a vertical two-dimensional selective real-time navigator through the dome of the right hemidiaphragm with a gating window at end expiration of 5 mm (14,17). Thus, shots are only accepted when the diaphragm is within the gating window at end expiration. The measurement efficiency is defined as the number of shots accepted divided by the total number of heart beats to complete an imaging examination and is indicative of the total acquisition time per measurement. The measurement efficiency and total time per MR coronary angiography measurement were measured. In the present study, the navigator was manually started during the downslope (ie, expiration) of the breathing curve. Furthermore, a local shim volume was used to optimize the magnetic field homogeneity of the heart.
Submillimeter MR coronary angiograms through the right and left coronary artery systems were obtained in a random order to reduce or prevent the influence of the duration of the examination (eg, patient movements due to prolonged lying) on the quality of images of the left or right coronary artery system.
Conventional Coronary Angiography
All patients underwent selective conventional coronary artery angiography by means of the Judkins technique within 24 hours after undergoing MR coronary angiography. The images were interpreted by consensus between two cardiologists with 26 (J.P.) and 14 years of experience and who were not involved in the MR coronary angiographic examinations and were blinded with regard to the MR imaging data at the time of the readings. The severity of a coronary artery stenosis was expressed as the percentage reduction in the luminal diameter, as determined by using the quantitative coronary analysis method (19). Only hemodynamically significant lesions (ie, with 
50% reduction in luminal diameter) in the four coronary arteries (ie, RCA, LM, LAD, LCX) were considered. For exact localization of the lesions and comparison of the conventional and MR coronary angiograms, the distance of the lesion along the coronary artery to the origin of the coronary artery was determined.
MR Angiogram Analysis
The MR coronary angiograms were transferred to a commercially available workstation (Easy Vision; Philips Medical Systems) and analyzed in a stepwise manner. The native MR coronary angiograms were independently evaluated by two readers with 10 (J.B.) and 12 (F.E.R.) years of experience in cardiac MR imaging. Consensus was not used to resolve interpretation disagreements. These readers were not involved in performing the MR coronary angiographic examinations, nor were they aware of the conventional coronary angiography results.
First, image quality was qualitatively evaluated on a per-coronary-artery basis by using a five-point grading system similar to that previously used by Duerinckx and Urman (6) to score two-dimensional MR coronary angiograms: 5 meant excellent quality; 4, good quality; 3, moderate quality; 2, poor quality; and 1, nondiagnostic or unreadable. This grading system was based on the signal-to-noise ratio, contrast-to-noise ratio, motion artifacts, and visualization of side branches on each image. For instance, a grade of 2 was assigned when a curvilinear structure was seen in the presumed position of the coronary artery, without clear depiction of the vessel anatomy, whereas a grade of 5 was assigned when large portions of the coronary arteries were visible as sharply defined structures with visible branch vessels. Patients with a mean image quality grade lower than 2 for MR coronary angiographic depiction of the coronary artery tree were not included in the study analysis.
Next, the length and diameter of each coronary artery (ie, CA, LM, LAD, and LCX) were measured. By using electronic calipers, we measured the diameter of each coronary artery proximally—that is, 1 cm beyond the origin.
Finally, the two readers evaluated the MR coronary angiograms for the presence of significant coronary artery lesions—specifically, stenoses of 50% or greater narrowing. With the exception of the LM, each coronary artery was divided into proximal (<3 cm), middle (3–6 cm), and distal (>6 cm) segments, which were abbreviated 1, 2, and 3, respectively. We preferred this approach, in contrast to that involving the use of branch vessels to divide the coronary arteries into segments that was applied in some previous studies (6,15), because branch vessels are not always well visualized, especially on MR coronary angiograms with moderate or poor image quality, and because this approach is less observer dependent—that is, it is less influenced by, for example, the misinterpretation of branch vessel patency status. Thus, for each patient, the coronary artery tree was divided into 10 segments: LM, RCA1, RCA2, RCA3, LAD1, LAD2, LAD3, LCX1, LCX2, and LCX3.
Coronary artery segments that were not visible at MR coronary angiography were excluded from analysis. The remaining visible coronary artery segments were then judged to be normal (ie, no stenosis) or positive for stenosis or occlusion. To each evaluated coronary artery segment, a confidence score from 1 to 5 was assigned: 1 meant definitely normal; 2, minor irregularities but probably normal; 3, possibly a stenotic lesion; 4, probably a stenotic lesion; and 5, definitely a stenotic lesion. These confidence levels were used for receiver operating characteristic (ROC) analysis.
In calculating the sensitivity, specificity, accuracy, and positive and negative predictive values of MR coronary angiography, we considered a segment to be normal if it was given a score of 1 or 2 and positive for stenosis or occlusion if it was given a score of 3, 4, or 5. The distance of the stenotic lesion to the origin measured at conventional coronary angiography was used to localize the lesion in a coronary artery segment at MR coronary angiography. To adjust for bias from learning as the images were read, the second observer read the MR coronary angiograms in reverse of the order in which the first observer interpreted the images. Measurement efficiency and total imaging time per coronary artery measurement were measured.
Statistical Analysis
The mean values of vessel length, vessel diameter, and image quality score ± the SDs were calculated. A paired Student t test was used to compare the measurement efficiency between the left and right coronary artery systems. P < .05 was considered to indicate a statistically significant difference. ROC analysis also was performed to assess lesion detection in coronary artery segments (with nonvisualized segments excluded). ROC analysis was performed by using a maximum-likelihood algorithm (20). The ROC curves were constructed on a per-segment rather than per-patient basis. For these ROC analyses, a degree-of-confidence decision task was used. Confidence scores of 1–5 were used to define four cutoffs points at which sensitivity and specificity values were calculated, and ROC curves were plotted for both readers (21). statistic analysis was used to measure agreement between the two readers, which was defined as the agreement beyond chance divided by the amount of agreement possible beyond chance (22).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
|
= 0.78). The readers disagreed regarding the visualization of 17 (9%) of 186 coronary artery segments. As shown in Table 1, the more distal the coronary artery segment was, the lower the number of visible segments was. One hundred percent of the LM and RCA1 segments were visible. More than 80% of the LAD1, LAD2, LCX1, and RCA2 coronary segments were visible. The percentages for visualization of LAD3, RCA3, and LCX2 were much lower, and LCX3 was not depicted. The mean lengths and diameters of the coronary arteries are listed in Table 2. The LM had a mean length of 11 mm but a wide range of lengths: Some patients had a very short (ie, 2-mm) LM with a very proximal bifurcation, whereas others had a very long (ie, 19-mm) LM. The lengths of the other coronary arteries corresponded well to the values of coronary artery segment visibility presented in Table 1.
|
Image Quality of MR Coronary Angiograms
The mean overall image quality (Table 3) grade for MR coronary angiograms of the coronary artery tree was 3.3 ± 1.0 for reader 1 and 3.2 ± 0.9 for reader 2. These values indicate that the two readers independently assigned similar grades of overall image quality. The values also indicate that the quality of images of the different coronary arteries, with the exception of that for the LM, was judged to be equivalent. Second, on a 1–5 grade scale, an average grade of just higher than 3 indicates moderate overall image quality, with an image quality range of excellent (Fig 3) to poor or even unreadable.
|
|
|
|
= 0.35) between the readers regarding the presence of significant coronary artery stenoses.
|
|
|
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
The overall image quality for the remaining 19 patients was moderate, with close agreement between readers but a wide range of values. Image quality was unpredictable: There were strong variations among patients and between the right and left coronary artery systems. Some patients who were considered good candidates for MR coronary angiography had poor image quality, whereas some who were considered poor candidates had good image quality. Potential explanations for these results are irregularities in heart rhythm or breathing pattern, the patient’s health status (eg, lung emphysema in elderly patients), and/or technical issues such as suitability of the navigator technique for patients. In the present study, although the length of time for the MR coronary angiographic examination was not analyzed statistically, we found no evidence that it affected image quality or that the use of double-oblique imaging planes—in example, for imaging the RCA—negatively affected image quality.
As in previous studies (13,15), the proximal parts of the coronary artery tree were best visualized. This good visualization was owing to the larger size of the proximal parts of the coronary arteries. In addition, the described real-time navigator technique is probably optimized to depict this part of the coronary artery tree. In general, more blurring was noticed in the distal right atrioventricular groove of the RCA. Moreover, only small portions of LCX1 were visible, and the more distal portions of this vessel generally were hardly visible, even in those patients with excellent image quality. The reasons for the poorer visualization of these parts of the coronary artery system are not well understood and may be related to navigator problems, the length of time for image acquisition, complex coronary artery motion patterns, or, in cases of imaging the LCX, the distance to the surface coil. The diameters of the coronary arteries at MR coronary angiography corresponded well with published data (5,6,23).
The main goal of performing MR coronary angiography, aside from the detection of coronary artery anomalies or coronary aneurysms in Kawasaki disease (24,25), is the visualization of coronary atheromatosis. For analysis in the present study, the coronary artery tree depicted at MR coronary angiography was divided into 10 segments, each of which was evaluated for the presence of significant coronary artery stenosis. Although a mean percentage of visualization of 71% of all segments might be disappointing, in the present study, the segments encompassed a much longer part of the coronary arteries as compared with the segments evaluated in most previously published studies (16). In addition, there was excellent agreement, greater than 90%, between the two readers with regard to the visualization of segments, and the value was high.
In most studies (26), a 50% reduction in luminal diameter is used to define a hemodynamically significant lesion. In the present study, two of 29 lesions were missed at MR coronary angiography owing to nonvisualizaton of the coronary artery segment. Of the 27 lesions in segments that were visible at MR coronary angiography, only 12 and 15 were correctly diagnosed by readers 1 and 2, respectively; thus, sensitivity was low. Inclusion of the lesions that were in segments that were not visible at MR coronary angiography would have led to a further decrease in sensitivity. These sensitivities are much lower than some of those in previously published studies involving the use of the older two-dimensional MR coronary angiography approach (5) (90%) or 3D MR coronary angiography with a retrospective rather than real-time navigator, a much longer acquisition window (178 msec vs 65 msec) (27), and much lower spatial resolution (1.2 x 2.3 mm vs 0.7 x 1.0 mm) (82%) (15).
The reasons for the lower sensitivities in the present study are unclear. Neither reader, both of whom are well trained in cardiac MR imaging, was involved in performing the MR coronary angiographic examinations or aware of the coronary catheterization (ie, conventional angiography) results. Additionally, all visible coronary artery segments were included in the analysis.
To determine the real value of MR coronary angiography in the detection of coronary artery stenoses, we compared MR coronary angiograms with conventional coronary angiograms on a segmental (ie, proximal, middle, and distal segments) rather than per-coronary-artery or per-patient basis. The latter comparisons yield much higher sensitivities at the expense of specificity. Therefore, we added ROC curves, which are a more efficient way to demonstrate the relationship between sensitivity and specificity. The closer an ROC curve is to the upper left-hand corner of the graph, the more accurate MR coronary angiography is. Poor image quality contributes to low sensitivity values. Furthermore, even with use of optimized MR coronary angiography techniques with a submillimeter in-plane spatial resolution, some lesions may be difficult to detect.
Because the contrast at MR coronary angiography is mainly dependent on T2 relaxation rather than true coronary vessel blood flow, a proximal hyperintense thrombus or occlusion with filling of the coronary artery by collateral vessels can perfectly mask the presence of a coronary artery abnormality. Contrast material–enhanced MR coronary angiography may be appealing, but the clinical role of this examination is not yet established (28–30). The specificity and negative predictive values in the present study, however, were much better than the sensitivity values.
The level of agreement between the two readers regarding the presence or absence of coronary artery stenosis appeared to be good (80%). However, the value, a better expression of the reliability of measurements because it takes into account the chance of agreement rather than the pure observed agreement, was low (0.35). The closer the value is to 1, the better the real agreement between readers is. Thus, in the present study, the low value was indicative of poor agreement between the readers: They disagreed on the presence or absence of coronary artery stenosis in 25 segments. Moreover, the difference in the course of ROC curves between readers is indicative of different observation behavior between readers; this means that reader 1 scored better at less strict confidence thresholds, whereas the opposite was true with regard to the scoring behavior of reader 2. Improved MR coronary angiogram image quality is expected to lead to improved ROC curves (ie, closer to 1) and better interobserver agreement.
Several limitations need to be mentioned: First, the number of patients included in our study was relatively limited. In a published multicenter trial in which a similar MR coronary angiography technique was used to examine 109 patients from seven different centers, sensitivities that were higher than those in this study but specificities that were lower and accuracies that were slightly lower were reported (16). Moreover, in that study, only the very proximal portions of the coronary arteries were evaluated and the selection criteria for patient inclusion were not well defined.
Second, we evaluated coronary artery stenosis detection on a per-coronary-artery-segment basis but assessed image quality on a per-coronary-artery basis. Although we could not determine the precise relationship between image quality and coronary artery stenosis assessment at MR coronary angiography, in general, the MR coronary angiograms with poor image quality incorrectly depicted the patency status of a higher number of coronary segments than did the MR coronary angiograms with good or excellent image quality. This close relationship has been described by Duerinckx and Urman (6).
Third, the coronary artery segment approach has the advantage that it enables more objective comparisons between readers (eg, with no misinterpretations of segment patency status due to confusion regarding side branches). The length of the segments (ie, 3 cm) was sufficient to minimize the risk of the two readers localizing lesions in different segments. However, lesions crossing segments may be located in different segments by different readers.
Fourth, only major coronary arteries, not the branches, were included in our analysis. Seventy-one percent of the segments of the major coronary arteries depicted at MR coronary angiography were useful for analysis; this percentage of visualization is probably inadequate for routine clinical practice. When we consider the proximal segments only, this visualization increases to almost 100%.
Fifth, the degree of stenosis depicted on conventional coronary angiograms was assessed by using quantitative coronary artery analysis; however, visual grading was used to assess MR coronary angiographic depiction. At present, the submillimeter spatial resolution at MR coronary angiography is suboptimal for exact determination of the degree of arterial stenosis, and this is probably another factor that contributes to the poor sensitivity. For instance, one cannot differentiate a stenosis of just less than 50% from a stenosis of just greater than 50%.
Sixth, a factor that hinders the incorporation of MR coronary angiography into a combined cardiac MR imaging protocol to study other facets, such as myocardial function, perfusion, or late enhancement, is the long total imaging time, which exceeds 15 minutes per MR coronary angiography measurement. Finally, MR coronary angiography yields images of the vessel lumina only; no information on the characteristics of the atherosclerotic plaques is generated. The role of black-blood MR coronary angiography is being investigated (8,31).
In conclusion, the results of the present study of 3D MR coronary angiography with a real-time navigator and submillimeter spatial resolution show that this examination is premature for the detection of significant coronary artery stenoses in the clinical setting. Image quality results are not consistently reliable, and, thus, the overall image quality is inadequate in some patients, and the distal coronary artery segments—namely, the distal segments of the RCA and LAD (ie, >6 cm from origin) and the middle and distal parts of the LCX (ie, >3 cm from origin)—may not be visualized. At present, the sensitivity of commercially available MR coronary angiographic systems for the detection of hemodynamically significant stenoses is too low, mainly because of poor image quality, and there is much observer variability. However, although this technique has poor sensitivity, it has relatively good specificity and thus is potentially useful for the exclusion of significant coronary artery disease in certain patient groups (eg, before valve replacement).
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, J.B.; study concepts, R.K.; study design, S.D., F.E.R.; literature research, J.B., R.B.; clinical studies, R.K., S.D., J.P.; data acquisition, R.B., J.P.; data analysis/interpretation, R.K., J.B., F.E.R., J.P.; statistical analysis, R.K., J.B.; manuscript preparation, J.B.; manuscript definition of intellectual content, J.B., R.K.; manuscript editing, J.B., F.E.R.; manuscript revision/review, R.K.; manuscript final version approval, J.B., F.E.R.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  H. Sakuma, Y. Ichikawa, N. Suzawa, T. Hirano, K. Makino, N. Koyama, M. Van Cauteren, and K. Takeda Assessment of Coronary Arteries with Total Study Time of Less than 30 Minutes by Using Whole-Heart Coronary MR Angiography Radiology, October 1, 2005; 237(1): 316 - 321. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N M C So, W W M Lam, D Li, A K Y Chan, J E Sanderson, and C Metreweli Magnetic resonance coronary angiography with 3D TrueFISP: breath-hold versus respiratory gated imaging Br. J. Radiol., February 1, 2005; 78(926): 116 - 121. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. L. Gerber, E. Coche, A. Pasquet, E. Ketelslegers, D. Vancraeynest, C. Grandin, B. E. Van Beers, and J.-L. J. Vanoverschelde Coronary Artery Stenosis: Direct Comparison of Four-Section Multi-Detector Row CT and 3D Navigator MR Imaging for Detection--Initial Results Radiology, January 1, 2005; 234(1): 98 - 108. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. G. Danias, A. Roussakis, and J. P.A. Ioannidis Diagnostic performance of coronary magnetic resonance angiography as compared against conventional x-ray angiography: A meta-analysis J. Am. Coll. Cardiol., November 2, 2004; 44(9): 1867 - 1876. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
U. J. Schoepf, C. R. Becker, B. M. Ohnesorge, and E. K. Yucel CT of Coronary Artery Disease Radiology, July 1, 2004; 232(1): 18 - 37. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. A. Obuchowski How Many Observers Are Needed in Clinical Studies of Medical Imaging? Am. J. Roentgenol., April 1, 2004; 182(4): 867 - 869. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
= reader 1, 
= reader 2. Straight line represents the line of chance.