| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Experimental Studies |
1 From the Laboratory of Cardiac Energetics, National Heart, Lung and Blood Institute, National Institutes of Health, Department of Health and Human Services, Bldg 10, Rm B1D416, MSC 1061, 10 Center Dr, Bethesda, MD 20892-1061. Received April 11, 2003; revision requested June 30; final revision received January 6, 2004; accepted January 21. Supported by the Intramural Program of the National Heart, Lung and Blood Institute. Address correspondence to A.E.A. (e-mail: araia@nih.gov).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Coronary artery occlusion or intracoronary adenosine infusion was successfully performed in 16 beagles; both procedures were performed simultaneously in one animal. MBF was assessed at microsphere analysis. First-pass myocardial perfusion MR imaging was performed during a dual-bolus administration of gadopentetate dimeglumine (0.0025 mmol/kg followed by 0.10 mmol/kg). The absolute myocardial perfusion at MR imaging was calculated by using Fermi function deconvolution methods. Qualitative, semiquantitative, and absolute myocardial perfusion MR imaging measurements were compared with microsphere MBF measurements by using paired t tests, linear correlation, and Bland-Altman analysis.
RESULTS: Fully quantitative (ie, absolute) analysis of MBF at MR imaging correlated with microsphere MBF measurement (r = 0.95, P < .001) across the full range of blood flow rates encountered (from 0 to >5.0 mL/min/g). Similar close correlations were observed in endocardial and epicardial segments (representing approximately 0.85 g of the myocardium). With modest increases in MBF, qualitative measurements plateaued in the hyperemic zones. Semiquantitative measurements did not correlate with MBF as well (r = 0.69–0.89); they plateaued around 3.0 mL/min/g.
CONCLUSION: Dual-bolus MR imaging enabled accurate measurement of absolute epicardial and endocardial perfusion across a wide range of blood flow rates (0 to >5.0 mL/min/g). Use of qualitative MR imaging measures such as the contrast enhancement ratio led to substantially underestimated hyperemic blood flow measurements.
© RSNA, 2004
Index terms: Animals • Experimental study • Heart, MR, 511.121412, 511.12143, 511.12144, 511.12145 • Heart, perfusion • Magnetic resonance (MR), perfusion study, 511.12144
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Unlike perfusion methods such as those involving the use of sequestered microspheres or radionuclide tracers, first-pass techniques must accurately depict the delivery and transit of contrast agent through the vasculature and the interstitium of the myocardium. A number of methods to analyze first-pass perfusion at MR imaging have been described. These methods range from very simple techniques, such as those involving the ratio of the peak-to-baseline signal intensity in the myocardium, to complex models that incorporate multiple physiologic variables and yield absolute measurements of MBF that are expressed in milliliters per minute per gram of tissue (3–6).
Quantitative methods of perfusion measurement rely on the accurate depiction of the arterial input function. For practical considerations, the input function is usually characterized in the left ventricle (LV) or in the ascending aortic root. Because the overall contrast agent concentration is much higher in the LV than in the myocardium (where the distribution is primarily in the vasculature and in the extracellular space), to avoid distortions of the LV cavity signal intensity, determination of the input function requires contrast agent concentrations well below the optimal concentration needed for myocardial enhancement. The purpose of this study was to compare fluorescent microsphere measurements of MBF with qualitative, semiquantitative, and fully quantitative measurements of first-pass perfusion at MR imaging.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
At time zero, either the left anterior descending artery was totally occluded or intracoronary infusion of adenosine was performed at 20 mcg/kg/min. Perfusion MR imaging was then performed. Immediately thereafter, the animal was withdrawn from the magnet and approximately 5 million fluorescent-labeled microspheres 15 µm in diameter (IMT Laboratories, Irvine, Calif) were injected into the left atrial catheter with simultaneous reference sampling through the femoral catheter. The delay between gadopentetate dimeglumine (Magnevist; Berlex Laboratories, Wayne, NJ) administration and microsphere injection was less than 10 minutes. The animals were then sacrificed.
Perfusion Measurement at Dual-Bolus MR Imaging
To calculate the arterial input function from the LV blood pool and at the same time improve the signal-to-noise ratio for myocardial enhancement, dual-bolus perfusion MR imaging with peripheral vein injection of gadopentetate dimeglumine was performed (Fig 1). The first bolus consisted of a low concentration of gadopentetate dimeglumine, 0.0025 mmol/kg injected at 5 mL/sec, and was used to determine the LV blood pool input function. This low concentration resulted in reduced T2* artifacts, preserved the linearity between gadopentetate dimeglumine concentration and LV cavity signal intensity, and allowed the second bolus to be injected without simultaneous marked enhancement of the myocardium. Concentrations of 0.005–0.01 mmol/kg were tested to ensure that the gadopentetate dimeglumine concentration was linear with the signal intensity and generated negligible myocardial contrast (<5% more than baseline).
|
MR imaging was performed by using a 1.5-T system (Signa 1.5-T CV/i; GE Medical Systems, Milwaukee, Wis) with a phased-array knee coil. First-order shimming was performed on a volume containing the LV by using stimulated-echo spectroscopy (7). Saturation-recovery gradient-echo MR imaging with a segmented echo-planar readout (8) was performed during the dual-bolus intravenous injection of gadopentetate dimeglumine. The following imaging parameters were used: 7.6/1.6 (repetition time msec/echo time msec), 15° flip angle, 240 x 180-mm field of view, 8-mm section thickness, 128 x 72 matrix, 70°–90° saturation flip angle, saturation-recovery time of 10 msec, and echo train length of four.
Short-axis image sections (3,4) were acquired in each R-R interval during 60 cardiac cycles. The acquisition window for each image was 137 msec. The section spacing was adjusted so that representative apical, midventricular, and basal short-axis sections were acquired. The linearity of this contrast agent concentration with signal intensity has been verified (9). The MR images were displayed on standard computers by using the Windows NT operating system (Microsoft, Redmond, Wash) and custom written programs (written by D.W.R.) in Interactive Display Language (Research Systems, Boulder, Colo).
Pathologic Analyses
Pathologic analyses were performed by two authors (T.F.C., S.L.L.). Each heart was placed in 10% formalin for 48 hours and then cut into six 4-mm short-axis slices. The slices were paired to coincide with the MR perfusion image sections, which were 8 mm in thickness. Each pair of heart slices was divided into eight radial segments. Each of the eight segments was subdivided into epicardial and endocardial regions—for a total of 16 segments per slice—and weighed to the nearest 0.01 g. Microsphere estimates of MBF were calculated by using conventional formulas, in which the number of spheres per segment was substituted for the degree of radioactivity per segment (10). The MBF values for each heart slice were plotted circumferentially. Intervention zones (ie, occluded or adenosine-infused segments) were readily identified from these plots.
Analysis of Perfusion at MR Imaging
Each short-axis perfusion MR image was divided radially into eight epicardial and eight endocardial segments per section to match the heart segments for microsphere analysis and an LV region of interest. The segments were automatically divided after the user manually traced the epicardial and endocardial borders on the images. For each image section, a segment was selected by an observer (T.F.C., S.L.L.) to represent the intervention (ie, occlusion or adenosine-induced stress) zone and a second segment was chosen to represent the control zone. The selection of the region of interest was based on anatomic considerations. The intervention zone was anteroseptal and anterior, whereas the control zone was typically inferior or lateral.
Time–signal intensity curves were generated by using the mean signal intensity of the region of interest. For each experiment, six time–signal intensity curves—for the intervention zone and control zone in each of the three sections—were generated. The baseline of each curve was set at zero. The time–signal intensity curves for each zone were subsequently averaged for all animals to generate three summary curves representing the adenosine, occlusion, and control zones (Fig 2).
|
CER method.—The CER was calculated as follows:
|
Myocardial-to-LV upslope index method.—The initial upslope of the LV and myocardial time–signal intensity curves was calculated from consecutive acquisition frames (more than three for the LV, more than four for the myocardium). The ratio of myocardial-to-LV upslope was then calculated for each region of interest. This method controls to some extent for variations in the arterial input function.
Upslope integral ratio method.—The area under the curve of each baseline-adjusted myocardial time–signal intensity curve was measured, as described by Klocke et al (11), to obtain a perfusion index relative to the control region. This method allows one to adjust for different timings and upslopes of enhancement but with the assumption of a constant input function.
Absolute flow quantification method.—The most complex of the four measures involves the use of Fermi function deconvolution to determine an absolute myocardial perfusion value, as previously described (6,12,13). Figures 1 and 3 show how this method is adapted for dual-bolus MR imaging.
|
Statistical Analyses
Data are presented as means ± standard deviations. Unpaired t tests were used to compare continuous variables according to grouping variable. The paired t test was used to compare flow values between techniques. One-way analysis of variance was used to compare repeated measures of continuous variables according to myocardial zone, and post hoc unpaired t tests were used to compare subsets within the analysis of variance. P < .05 was considered to indicate significance. Simple linear regression analysis was used to compare the MR imaging perfusion values obtained with each MR imaging parameter measured.
To illustrate the nonlinearity between microsphere-measured blood flow and qualitative or semiquantitative MR imaging measures of blood flow, we plotted a line of identity forced through zero and the average measured value for points representing normal microsphere-measured blood flow (0.8–1.0 mL/min/g). We also plotted the linear correlation determined from points representing microsphere-measured flow greater than 0.8 mL/min/g to illustrate the tendency of several parameters to yield underestimated vasodilated blood flow values. Given that multiple heart segments from each animal were used, generalized estimating equations were completed. Bland-Altman plots (14) were generated to complement the simple plots of linear correlations.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The microsphere-measured blood flow values for the occlusion, control, and adenosine zones in each heart slice were significantly different (P < .001) and are summarized in the Table. The sample sizes resulted in a power of greater than 0.90, with an 
of .05 for all comparisons in the Table, except those involving endocardial perfusion (power of 0.75). For the control zones, endocardial flow was significantly higher (P < .001) than epicardial flow, with a 22% gradient between these areas. The pattern was reversed in segments in which hyperemia was induced with intracoronary infusion of adenosine: The flow rate in the epicardial segments was 26% higher than that in the endocardial segments (P < .001). The occlusion segments showed no transmural gradient in flow.
|
MR Imaging Analysis of First-Pass Perfusion
CER.—The correlation between CER and microsphere-measured MBF is shown in Figure 4, A. For ischemic to normal perfusion, the overall linear correlation was significant (r = 0.75, P < .001) when all segments were considered. However, there was a relative plateau in the association between these two variables at blood flow greater than 1.0 mL/min/g. Use of contrast enhancement measures alone led to underestimated hyperemic MBF values, as evidenced by the solid line in Figure 4, A. Findings were similar when epicardial and endocardial regions were analyzed separately.
|
Upslope integral.—The upslope integral (Fig 4, C) proved to be the most linear semiquantitative index of MBF (r = 0.89, P < .001 for all segments). The plateau effect during hyperemia was least severe with use of this measure. The linearity between microsphere-measured MBF and this measure was preserved at blood flow of up to 3 mL/min/g.
The Bland-Altman plots for the three qualitative and semiquantitative methods demonstrated a systematic underestimation of MBF at higher blood flow rates (Fig 4, bottom plot graphs).
Absolute flow quantification.—The summarized estimates of absolute perfusion obtained at MR imaging and the overall correlations between these estimates and the microsphere measurements are shown in Figures 5 and 6 and in the Table. Microsphere-measured MBF values closely correlated with transmural quantitative MR imaging estimates of MBF (MR imaging estimate = [0.95 · microsphere measurement] + 0.10; r = 0.95, P < .001), with 95% confidence limits for the correlation (r) of 0.88–1.01. Note that the agreement remained close when the analysis was performed separately in endocardial and epicardial regions (Fig 5).
|
|
There was a slight but consistent overestimation of the MBF rate (Table) in the control segments. The overestimation was significant in the transmural and epicardial regions and almost significant (P = .07) in the endocardial regions (Fig 6). The mean measurements matched closely. Quantitative MR imaging-derived values had a transmural pattern of distribution similar to that of microsphere measurements (Table). The MR imaging and microsphere values had similar intramural patterns of MBF. In the control zones, endocardial flow exceeded epicardial flow. In the adenosine zones, the endocardial flow–to–epicardial flow ratio was reversed. In the occlusion zones, the gradient was eliminated (Table).
Summary Time–Signal Intensity Curves
Summary time–signal intensity curves for each intervention are shown in Figure 2. There was a distinct change in curve morphology—including changes in slope and amplitude—with each of the three MBF states. Note that the amplitude of peak signal intensity with adenosine infusion did not increase 2.5- to fivefold above that with the control intervention as the MBF did with adenosine infusion. With adenosine infusion, there was also an overshoot (ie, conformation change in time–signal intensity curve) over the steady-state level observed at the end of the image acquisition.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The upslope integral (11) technique was the most effective semiquantitative method that we studied. The myocardial-to-LV upslope ratio appeared to plateau at MBF rates of greater than 3.0 mL/min/g, but radionuclide techniques have proved to be effective with similar or lower plateau thresholds (2). When the upslope integral is used to compare regions in the heart, it yields only a relative perfusion index and knowledge of a normal blood flow zone is required. Thus, there may be clinical limitations to the use of this parameter in the setting of multivessel disease. In addition, a relative perfusion method might not enable the detection of an inadequate vasodilator effect, which would be readily apparent with use of a measure of absolute MBF. Similar concerns regarding the upslope of myocardial enhancement might be raised; however, we observed a moderate correlation between blood flow and upslope of myocardial enhancement, in agreement with the findings of Kraitchman et al (16).
Theoretically, the use of fully quantitative perfusion analysis helps to avoid these problems, and the value measured with this method represents a potentially useful clinical index. Thus, the described fully quantitative MR imaging methods may have important clinical benefits compared with existing relative perfusion techniques.
Fermi function deconvolution is a method of analyzing how tissue-agent interactions transform LV contrast agent time–signal intensity characteristics to the contrast agent time–signal intensity characteristics observed in the myocardium (6,12,13). This method works well with imperfect signal-to-noise ratio data, as are obtained with first-pass perfusion MR imaging, and yields absolute MBF values. Our study data indicate that Fermi function deconvolution is effective in modeling myocardial perfusion, despite the numerous theoretical concerns (15–21).
Our dual-bolus MR imaging approach is an extended version of the method proposed by Jerosch-Herold et al (6), which allows the use of higher concentrations of contrast agent to improve the signal-to-noise ratio in the myocardium. This approach allowed us to estimate the transmural distribution of absolute myocardial perfusion in relatively small regions of the myocardium (approximately 0.85 g). In addition, fully quantitative analysis revealed no plateau effect across all ranges of clinically important MBF.
The theoretic reasons that the described analysis methods perform with variable degrees of success are worth considering. Fully quantitative deconvolution analysis works well because tissue-agent interactions are considered a system with this method. One derives a transfer function that "transforms" the arterial "input" time–signal intensity curve into the "output" observed in the myocardium (12,13). With the upslope integral, one does not try to determine the tissue-agent system itself by way of the system’s transfer function, but rather one compares the system’s output between normal and intervention zones. Had the LV cavity input been an instantaneous bolus, the upslope integral would have enabled an effective comparison of the cumulative probabilities that such a bolus had perfused the tissue by a certain amount of time. Ignoring the true arterial bolus profile can introduce errors.
The myocardial-to-LV upslope ratio method is a simplified approach to characterizing the tissue-agent system. The upslope from the LV cavity is a partial description of how the input function approaches an instantaneous bolus. The myocardial upslope is an index of mean transit time via the slope of the output, as observed in the myocardium. Upslope parameters describe part of the perfusion system but do not help determine the system’s transfer function; thus, these parameters are prone to errors.
The CER is a measure of the accumulation of contrast material but not of the rate of contrast material delivery or clearance. This was particularly evident in the adenosine zones, where peak enhancement was only modestly increased over that in the control zones despite a four to fivefold increase in MBF. The conformation change in the time–signal intensity curve (ie, overshoot) in regions of hyperemia were adequately modeled by using the deconvolution method but resulted in underestimations of perfusion with the semiquantitative methods, as has been previously described (5). This may explain why the CER index, which is entirely signal intensity dependent, performed worst in our comparisons. Thus, assessment of the kinetics of contrast material passage through the myocardium adds substantially to the measurement of perfusion as compared with assessment of signal intensity alone.
Schwitter and colleagues (3) used a measure of upslope enhancement that correlated well with nitrogen 13–ammonia positron emission tomography (PET) and coronary angiography measurements, with use of a low coronary flow reserve cutoff value of 1.65. However, with a coronary stenosis threshold of 50%, the sensitivity based on vascular territory decreased to nearly 80%, and specificity was about 70% (3). These results may indicate that the upslope index plateaus at higher blood flow rates, and this plateau could lead to the underestimation of degrees of intermediate stenosis. The quantitative perfusion seen at PET was not compared with the MR imaging upslope ratio, however. To date, clinical examinations of perfusion have performed optimally when the myocardial time–signal intensity curve was quantified. In addition, semiquantitative estimates of coronary blood flow reserve have correlated closely with percentages of luminal epicardial vessel stenosis measured at quantitative coronary angiography (4).
There were limitations to our study. The added complexity of the dual-bolus MR imaging examinations needs to be justified. Why did we not simply perform deconvolution by using a single bolus of gadopentetate dimeglumine as others have reportedly done? The described dual-bolus method allowed us to compare all analysis methods equally in terms of signal intensity. Single-bolus MR imaging examinations are limited to the use of low concentrations of gadolinium-based contrast material to avoid distortion of the LV cavity signal intensity. Because the myocardial signal-to-noise ratio is proportional to the contrast material concentration (9), the low concentration of gadolinium-based contrast material impairs the ability to measure time–signal intensity curves in the myocardium and is suboptimal for visual analysis.
With the dual-bolus approach, the arterial input function is preserved with the first, low-concentration, bolus, and myocardial enhancement that allows comparison of analysis techniques is concurrently generated. High-quality time–signal intensity curves on which to fit the estimate of absolute blood flow also are generated with the dual-bolus approach. To simplify the calculations, alternative approaches to quantification other than the use of a Fermi function can be used (22).
Practical application: Application of the described dual-bolus MR imaging method in human subjects is currently in development. The dual-bolus approach to perfusion MR imaging offers the potential to acquire images with high signal-to-noise ratios for qualitative interpretation and to perform fully quantitative analyses.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Abbreviations: CER = contrast enhancement ratio, MBF = myocardial blood flow, LV = left ventricle
Author contributions: Guarantors of integrity of entire study, T.F.C., R.S.B., A.E.A.; study concepts, T.F.C., D.R., A.H.A., R.S.B., A.E.A.; study design, T.F.C., A.H.A., R.S.B., A.E.A.; literature research, T.F.C., A.E.A.; experimental studies, T.F.C., D.R., A.H.A., S.L., J.L.T., A.E.A.; data acquisition, T.F.C., D.R., A.H.A., S.L., J.L.T., A.E.A.; data analysis/interpretation, T.F.C., D.R., A.H.A., R.S.B., A.E.A.; statistical analysis, T.F.C., A.E.A.; manuscript preparation, T.F.C., A.H.A., A.E.A.; manuscript definition of intellectual content, T.F.C., D.R., A.H.A., S.L., R.S.B., A.E.A.; manuscript editing, T.F.C., D.R., A.H.A., J.L.T., R.S.B., A.E.A.; manuscript revision/review and final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  N. Beohar, A. K. Erdogan, D. C. Lee, H. N. Sabbah, M. J. Kern, J. Teerlink, R. O. Bonow, and M. Gheorghiade Acute heart failure syndromes and coronary perfusion. J. Am. Coll. Cardiol., July 1, 2008; 52(1): 13 - 16. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. M. Gonzalez, S. Shiva, P. S. Vincent, L. A. Ringwood, L.-Y. Hsu, Y. Y. Hon, A. H. Aletras, R. O. Cannon III, M. T. Gladwin, and A. E. Arai Nitrite Anion Provides Potent Cytoprotective and Antiapoptotic Effects as Adjunctive Therapy to Reperfusion for Acute Myocardial Infarction Circulation, June 10, 2008; 117(23): 2986 - 2994. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. E. Petersen, M. Jerosch-Herold, L. E. Hudsmith, M. D. Robson, J. M. Francis, H. A. Doll, J. B. Selvanayagam, S. Neubauer, and H. Watkins Evidence for Microvascular Dysfunction in Hypertrophic Cardiomyopathy: New Insights From Multiparametric Magnetic Resonance Imaging Circulation, May 8, 2007; 115(18): 2418 - 2425. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. A. Barmeyer, A. Stork, K. Muellerleile, C. Tiburtius, A. K. Schofer, T. A. Heitzer, T. Hofmann, G. Adam, T. Meinertz, and G. K. Lund Contrast-enhanced Cardiac MR Imaging in the Detection of Reduced Coronary Flow Velocity Reserve Radiology, May 1, 2007; 243(2): 377 - 385. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S Watkins, K G Oldroyd, and S Frohwein Magnetic resonance myocardial perfusion imaging: a new era in the detection of reversible myocardial ischaemia Heart, January 1, 2007; 93(1): 7 - 10. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
), normal myocardium (
), and occluded region (
). After initial contrast material bolus, signal intensity in LV cavity increases proportionally with contrast agent concentration. However, owing to nonlinearity of the MR imaging examination, the LV cavity signal intensity does not accurately reflect the distribution of the high-concentration contrast material bolus. After administration of the high concentration of contrast material, the signal intensity is only approximately four times higher than that observed after administration of the low concentration, despite the 40-fold increase in concentration. The signal intensity of the normal myocardium changes little after the low-concentration bolus (between 5 and 18 seconds), but it changes substantially following the high-concentration bolus. The occluded zone shows less enhancement following the high-concentration bolus.
), the signal intensity of the LV cavity is underestimated by a factor of approximately 10. Right: Resulting myocardial enhancement after the high-concentration bolus (
) is a mathematical procedure that can be used to describe how the input in a system changes into the output, as described by the system’s transfer function. In the case of perfusion, the transfer function characterizes the myocardial circulation that transforms the delivery of contrast material into the LV cavity (input) into the observed myocardial enhancement (output). Middle: The deconvolution process extracts the myocardial transfer function by iteratively convolving the scaled signal intensity of the LV cavity resulting after the low-concentration bolus by using a three-parameter Fermi transfer function until the myocardial enhancement can be predicted.
