| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From Medical Clinic I (H.P.K.) and the Dept of Medical Statistics (N.S.H.), University Hospital Aachen, Pauwelstrasse 30, 52057 Aachen, Germany; and Depts of Cardiology (T.S.P., A.M.B., A.C.v.R.) and Clinical Physics and Informatics (M.B.M.H.), Vrÿe University Medical Center, Amsterdam, the Netherlands. Supported by Netherlands Heart Foundation grant 2001.158. H.P.K. supported in part by grants from the Faculty of Medicine of the Rheinisch-Westfälische Technische Hochschule, Aachen, Germany, and by the Grimmke-Stiftung, Düsseldorf, Germany. Received Sept 11, 2002; revision requested Nov 7; final revision received May 12, 2003; accepted June 16. Address correspondence to H.P.K. (e-mail: hkuehl@ukaachen.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
= 0.84), while agreement was poor for the transmural extent of hyperenhancement ( = 0.32), which was attributed to the blurred appearance of the three-dimensional MR images. Findings with the one–breath-hold three-dimensional MR sequence allow assessment of nonviable myocardium with good agreement with those with the multiple–breath-hold two-dimensional MR sequence. © RSNA, 2003
Index terms: Heart, MR, 511.121419, 511.12143 • Myocardium, infarction, 511.771 • Myocardium, MR, 511.121419, 511.12143
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
The evaluation of myocardial hyperenhancement is clinically important. With the currently recommended two-dimensional (2D) MR imaging sequence, high-spatial-resolution images are acquired that allow assessment of the transmural distribution of hyperenhancement (6). This approach necessitates multiple breath holds to generate a set of MR images that cover the left ventricle, however, which is time consuming and may be difficult for patients. A method that enables rapid data acquisition within one breath hold would allow these limitations to be overcome. Rapid three-dimensional (3D) acquisition of a complete data volume is possible within one breath hold with a state-of-the-art MR imager. Thus, the purpose of our study was to compare a one–breath-hold 3D inversion-recovery gradient-echo MR pulse sequence, which was optimized for the detection of hyperenhancement by using a nonselective inversion-recovery prepulse, with a standard multiple–breath-hold MR sequence with a 2D inversion-recovery gradient-echo MR imaging technique for the detection of hyperenhancement in patients with chronic coronary artery disease.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
MR Imaging Protocols
MR images were acquired with the patient in the supine position with a 1.5-T whole-body MR system (Magnetom Sonata; Siemens, Erlangen, Germany) with a four-element phased-array cardiac coil for signal reception. A belt was strapped around the upper abdomen of each patient to monitor respiratory motion, and electrocardiography was performed to gate the acquisition and to monitor the patient’s condition. Scout images were acquired in long- and short-axis orientation for planning of the final long-axis and double-oblique short-axis views. For the assessment of global and regional left ventricular function, electrocardiographically gated cine images were obtained with a segmented true fast imaging with steady-state precession, or FISP, sequence (repetition time msec/echo time msec = 3.2/1.2, 5-mm section thickness, 35-msec temporal resolution) in three long-axis views (two-, three-, and four-chamber views) and seven to 11 short-axis views 1 cm apart to cover the whole left ventricle in repeated breath holds. Thereafter, gadopentetate dimeglumine (0.2 mmol per kilogram of body weight) was administered intravenously. Fifteen minutes after injection, contrast-enhanced MR images were acquired in the same orientation used for the cine short-axis images.
The parameters of the pulse sequences we investigated are summarized in Table 1. For 3D data acquisition, a segmented inversion-recovery gradient-echo sequence with central-ordered k space was used (standard mode on the MR imager). Image acquisition was triggered to every heartbeat and gated to middiastole. The acquisition window in the cardiac cycle was 293 msec. The field of view was set to 360 x 293 x 100 mm, and the matrix was 256 x 208 x 12, which resulted in a spatial resolution of 1.4 x 1.4 x 8.3 mm. Partial Fourier imaging was applied with 75% coverage of k space in both phase- and section-encoding directions. Section thickness was 8.3 mm during data acquisition and was interpolated to 5 mm by means of zero filling in the z direction, which resulted in 20 contiguous sections that covered a volume of 10 cm3. The inversion time was set separately in each patient to null the signal of normal myocardium after contrast material administration and was typically between 200 and 230 msec. Depending on the heart rate, acquisition duration was typically 14–20 seconds (20 heartbeats).
|
Data acquisition was performed with the standard 2D multiple–breath-hold MR imaging sequence and the 3D one–breath-hold sequence in each patient. The mean time between the two sequences was 9 minutes ± 3, with the 3D sequence performed first in most of the patients. Imaging was successful in all patients.
Analysis Design
Data were analyzed with both a quantitative approach and a qualitative approach. The MR images were first previewed on a personal computer workstation with commercial software (Radworks, version 5.0; Applicare Medical Imaging, Zeist, the Netherlands). Twenty MR images were generated with the 3D pulse sequence, but only 10 images were generated with the 2D sequence; therefore, images acquired with the 3D sequence had to be matched apriori to the images acquired with the 2D pulse sequence by a person not involved in data analysis (A.M.B.). Since the patients were not removed from the magnet bore during imaging, the images were matched according to anatomic landmarks and the section position indicated on the image. This was necessary because differences of up to 1 cm occurred between the spatial position on images acquired with the one–breath-hold 3D sequence and that on images acquired with the multiple–breath-hold 2D sequence, as a result of the variable position of the diaphragm during breath holding. Thus, of the images available from the 3D data set, only half were analyzed.
To minimize possible bias, each image was analyzed in a randomized and blinded fashion. For this purpose, all patient information and imaging parameters were removed from the original images, which were then stored in new digital files. Each image was randomly assigned a number that was used for patient and modality identification after data analysis. The observers (one observer for the quantitative analysis [H.P.K.] and two independent observers for the qualitative analysis [H.K.P., T.S.P.], both cardiologists with 1 year of experience in cardiac MR imaging) were thus blinded to patient information and the imaging sequence.
Quantitative Analysis
To assess agreement between the two sequences for the determination of total myocardial area and contrast-enhanced area, endocardial and epicardial borders and hyperenhanced areas were outlined manually in each image separately for each sequence. Hyperenhancement was defined as an area of high signal intensity that was 3 SDs or more greater than the signal intensity of nonenhanced myocardium. If differentiation between hyperenhanced subendocardium and the blood pool proved difficult, the subendocardial edge of the corresponding cine image was used as a reference. In addition, the contrast-to-noise ratio (CNR) for enhanced (E) myocardium versus that for nonenhanced (N) myocardium in each data set was assessed according to the following formula: CNR = (SIE - SIN)/noise, where SI is signal intensity. Noise was determined as the SD of signal intensity determined from a circular region of interest (10 cm2) placed in the field of view outside the patient’s body.
Qualitative Analysis
To assess agreement between the two sequences for the detection of hyperenhancement and spatial extent of hyperenhancement, segmental analysis was performed. For this purpose, each short-axis section in each data set of each patient was divided into eight equidistant segments angulated 45° apart starting from the anterior insertion of the right ventricular wall into the left ventricular myocardium by using an overlay (Fig 1). Each segment was evaluated with a scoring system for the presence or absence of delayed contrast enhancement: 1, no hyperenhancement present; 2, hyperenhancement present. In addition, each segment was evaluated for the distribution of hyperenhancement throughout the myocardial wall (transmural extent of hyperenhancement) and the distribution of hyperenhancement throughout the width of the segment (segmental width of hyperenhancement) by using the same scoring system: 1, no transmural extent or partial segmental width of hyperenhancement; 2, full transmural extent or full segmental width of hyperenhancement (Fig 2). Observer agreement was determined separately for the 2D and 3D sequences for the presence and spatial extent of hyperenhancement by using the same scoring criteria.
|
|
|
Agreement for qualitative variables was assessed by means of examination, or , statistics. Because there were multiple ratings per patient, a one-examination value was calculated first for each patient. After the one-examination values were obtained, a common or overall examination value was calculated according to the method described by Fleiss (13). In addition, the individual examination values were indicated and plotted against accepted benchmarks on the basis of the following examination values: < 0.21, poor agreement; = 0.21–0.40, fair; = 0.41–0.60, moderate; = 0.61–0.80, good; and > 0.80 = excellent (14). Observer agreement was assessed in the same way. Differences with a P value of less than .05 (two-tailed) were considered statistically significant.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
|
|
Qualitative Analysis
In one patient without hyperenhancement, complete agreement precluded calculation of a coefficient. Thus, 661 segments were considered in nine patients. Results are summarized in Table 2. Agreement for the detection of hyperenhancement was excellent, with a common examination value of = 0.84. For the evaluation of transmural extent and segmental width of hyperenhancement, 236 segments were considered in the nine patients. Agreement for the transmural extent of hyperenhancement was fair (common = 0.32). In half of the patients, the examination value was below the threshold for moderate agreement ( = 0.41), and results for only one patient demonstrated good agreement between the two techniques. The distribution of hyperenhancement was scored as transmural more often with the 3D MR sequence than with the 2D MR sequence, which demonstrates a trend for overestimation of the transmurality of hyperenhancement. Agreement on the segmental width of hyperenhancement was good (common = 0.62).
|
> 0.81 in the majority of subjects for both the 2D and 3D sequences (Table 3). Complete agreement was obtained for the patient without hyperenhancement (no examination value calculated), as well as in one and two patients evaluated with the 2D and 3D sequences, respectively ( = 1.0 for each). For the transmural extent of hyperenhancement, agreement was good for the 2D sequence. A similar range of examination values was found for the 3D sequence, although with fewer examination values greater than 0.80. For the segmental width of hyperenhancement, the distribution of examination values was similar for the 2D and 3D sequences.
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Advantages of the 3D Sequence
In the current approach for the detection of contrast-enhanced myocardium, a 2D inversion-recovery gradient-echo sequence is applied during acquisition of one image per breath hold during 16 heartbeats (6). Thus, repeated breath holds are required to cover the entire left ventricle for a comprehensive evaluation of the viability status of a given heart. This procedure is time consuming and difficult for patients with advanced heart failure. The advantage of the 3D inversion-recovery gradient-echo sequence is that it allows rapid and complete data acquisition that covers the left ventricle within one breath hold, albeit at the expense of a slightly longer breath hold. Moreover, the CNR was found to be threefold larger with the 3D sequence than that with the 2D sequence. To examine whether this was related to the physics of the sequences or to the systematic time difference between acquisition of the two data sets, we calculated the theoretic ratio in CNR for the two sequences. Differences in voxel size, receiver bandwidth, 2D versus 3D imaging, and partial Fourier imaging explain higher CNR by a factor of 2.0 for the 3D technique.
To exploit the effect of differences in the repetition time, number of segments, and k-space acquisition order within the cardiac cycle, a simulation of the MR signal was performed. The signal strength at the center of k space was determined with the formulas for a spoiled gradient-echo technique with inversion preparation (15). The effect of radiofrequency prepulses was taken into consideration. For normal myocardium at 18 minutes after injection of 0.2 mmol/kg gadopentetate dimeglumine, a T1 of 378 msec was assumed, and that for infarcted myocardium was 224 msec. These values were estimated by using an interpolated value of the blood concentrations of gadopentetate dimeglumine of 0.86 mmol/L, from Weinmann et al (16), and by assuming partition coefficients of 0.36 and 0.83 mL/g for normal and infarcted myocardium, respectively (17,18). Differences in k-space order (linear for the 2D technique and central for the 3D technique) contributed to additional improvement in CNR by a factor of 1.7 for the 3D sequence relative to the 2D sequence. Overall, estimations with the theoretic calculations showed improvement in CNR by a factor of 3.4 for the 3D sequence relative to the 2D sequence. These findings are in agreement with the observed improvement in CNR of 3.1 ± 1.6.
The effect of the systematic difference in acquisition time with the two techniques was also simulated. According to the data of Weinmann et al (16), the blood concentration of gadolinium-based contrast medium varied from 0.92 to 0.74 mmol/L at 15 and 24 minutes after administration. With the same assumption, the estimated time delay would cause an additional 12% increase in CNR with the 3D sequence relative to the 2D sequence. Thus, the increase in CNR observed with the 3D sequence is mainly related to the sequence parameters.
Limitations of the 3D Sequence
A disadvantage of the 3D sequence is the use of every heartbeat for data acquisition, which makes the contrast in the images and image quality more sensitive to heart-rate variability. This may substantially influence image quality in patients with rhythm disturbances such as atrial fibrillation or frequent premature beats. Moreoever, the effective spatial resolution of the 3D acquisition is less than that of the 2D sequence and is also less than the nominal resolution of the sequence itself. This may explain the differences between the 2D and 3D sequences for the detection of transmural extent of hyperenhancement.
Potential sources that might contribute to the lower effective resolution of the 3D sequence compared with that of the 2D sequence include differences in nominal resolution, cardiac motion, and artifacts in k space. First, nominal resolution in the section-encoding direction is less with the 3D sequence (1.4 x 1.4 x 8.3 mm) than that with the 2D sequence (1.3 x 1.6 x 5 mm). Second, cardiac motion is expected to result in more blurring with the 3D sequence as a result of a slightly larger acquisition window within the cardiac cycle (293 msec with the 3D sequence vs 245 msec with the 2D sequence), especially in patients with higher heart rates. The difference in effective acquisition window size is even larger when the central part of k space is considered (because of differences in 2D k space and 3D k space). Third, partial Fourier imaging with the 3D sequence may contribute to some image blurring. Finally, the central k-space order for the 3D acquisition resulted in a clear reduction in contrast at the edges of k space, which resulted in low pass filtering and loss of detail.
To overcome these limitations, possible improvements in the 3D sequence include acquisition of thinner sections by using more sections to cover the left ventricle, reduction of the acquisition window, and acquisition of full k space. The trade-off, however, would be increased acquisition duration, which is limited by acceptable breath-hold duration. For complete visualization of the left ventricle, this would necessitate acquisition of two stacks of images during one breath hold. Another option would be reduction of the flip angle in the 3D sequence to limit the low pass filtering.
Limitations of the Study
Patients with atrial fibrillation were excluded from the current study because the variable RR interval may adversely affect image quality, as well as the quality of nulling of the nonenhanced myocardium with the 3D sequence. The time interval between acquisition of the 2D and 3D images was rather large. Since the 3D sequence was applied first in most of the patients, one could argue that contrast material washout at the time of data acquisition with the 2D sequence may have changed the size and shape of the contrast-enhanced areas (19). Since we only included patients with chronic infarcts, however, we do not expect to see the differences in the wash-in and washout kinetics of the hyperenhanced regions that were reported in the setting of acute infarctions (19). Moreover, since imaging was started relatively late (15 minutes) after contrast material administration, a near steady-state situation can be expected. This is supported by findings in a recent report that the size of healed infarcts measured at contrast-enhanced MR imaging does not change between 10 and 30 minutes after contrast material administration (20).
The 2D and 3D data were not acquired in a randomized order. This is an important drawback that could have introduced a bias. The size of the hyperenhanced area, however, is not affected by imaging time in this patient population. In addition, at each imaging time point for each sequence, an "optimal" inversion time was sought that provided optimal contrast between enhanced and nonenhanced myocardium. Thus, we do not believe that this limitation might have adversely influenced our results. A complete blinding of the observers was not possible because images acquired with the 3D sequences could be identified as such. Thus, a certain amount of observer bias cannot be excluded in the analysis. The transmurality of delayed contrast enhancement was evaluated dichotomously as either transmural or nontransmural. A more gradual approach would not have been possible with the 3D sequence. For the determination of observer agreement on the transmural extent and segmental width of hyperenhancement, only segments scored as hyperenhanced by both observers were considered. Thus, examination values might have been overestimated because we ignored segments that were scored differently by both observers.
Clinical Implications
For the clinical application of contrast-enhanced MR imaging, rapid screening for the presence of hyperenhancement is important to reduce imaging time and increase acceptance of the method. With the stepwise image acquisition approach of the 2D sequence, approximately 10–15 minutes of imaging time is needed for complete visualization of the left ventricle. The 3D approach allows complete image acquisition within one breath hold. Small areas of delayed contrast enhancement, which can be found in patients with coronary artery disease (21), are readily detected with the 3D sequence with no loss of information compared with the 2D sequence (Fig 2). In patients with depressed left ventricular function of nonischemic origin, such as idiopathic dilated cardiomyopathy, with no hyperenhancement (22), contrast enhancement can be rapidly excluded with this method. Decisions on myocardial revascularization should be made cautiously with the 3D sequence, however, because overestimation of the transmural extent of hyperenhancement could lead to withholding of revascularization strategies in patients who might benefit from restoration of myocardial blood flow. Thus, the 3D acquisition approach is a rapid screening tool to assess the presence and extent of hyperenhancement. Selected views with the 2D sequence could then be added to assess the exact transmural distribution of hyperenhancement. With this approach, overall imaging time is reduced compared with that for full left ventricular coverage with the 2D sequence.
In conclusion, the results of this study demonstrate that the 3D inversion-recovery gradient-echo MR imaging sequence allows rapid evaluation of patients with chronic ischemic heart disease for the presence of hyperenhancement, with high agreement compared with results with the standard 2D inversion-recovery gradient-echo MR imaging sequence. The transmural extent of hyperenhancement, however, is overestimated with the 3D sequence. The overestimation may be related to lower effective spatial resolution as a result of intrinsic physical factors of the pulse sequence and acquisition scheme.
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, H.P.K., A.C.v.R.; study concepts, H.P.K., M.B.M.H.; study design, H.P.K., T.S.P., A.M.B.; literature research, H.P.K.; clinical studies, H.P.K., A.M.B.; data acquisition, H.P.K., A.M.B.; data analysis/interpretation, H.P.K., M.B.M.H.; statistical analysis, H.P.K., N.S.H.; manuscript preparation, H.P.K., M.B.M.H.; manuscript definition of intellectual content, H.P.K., M.B.M.H., A.M.B.; manuscript editing, H.P.K.; manuscript revision/review, M.B.M.H., A.C.v.R., N.S.H.; manuscript final version approval, A.C.v.R., M.B.M.H.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. U. Bauner, O. Muehling, D. Theisen, C. Hayes, B. J. Wintersperger, M. F. Reiser, and A. M. Huber Assessment of Myocardial Viability with 3D MRI at 3 T Am. J. Roentgenol., June 1, 2009; 192(6): 1645 - 1650. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. C. M. van den Bosch, J. J. M. Westenberg, J. C. Post, G. Yo, J. Verwoerd, L. J. M. Kroft, and A. de Roos Free-Breathing MRI for the Assessment of Myocardial Infarction: Clinical Validation Am. J. Roentgenol., June 1, 2009; 192(6): W277 - W281. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. K. Attili and P. N. Cascade CT and MRI of Coronary Artery Disease:Evidence-Based Review. Am. J. Roentgenol., December 1, 2006; 187(6 Suppl): S483 - S499. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Dewey, M. Laule, M. Taupitz, N. Kaufels, B. Hamm, and D. Kivelitz Myocardial Viability: Assessment with Three-dimensional MR Imaging in Pigs and Patients Radiology, June 1, 2006; 239(3): 703 - 709. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
