Author + information
- Received January 10, 2001
- Revision received November 29, 2001
- Accepted December 14, 2001
- Published online March 6, 2002.
- Tareq Ibrahim, MD*,
- Stephan G. Nekolla, PhD*,
- Karin Schreiber, MD*,
- Kenichi Odaka, MD*,†,‡,
- Stefan Volz, MD†,
- Julinda Mehilli, MD‡,
- Martin Güthlin, MD†,
- Wolfram Delius, MD† and
- Markus Schwaiger, MD, FACC*,* ()
- ↵*Reprint requests and correspondence:
Dr. Markus Schwaiger, Nuklearmedizinische Klinik der Technischen Universität München, Ismaningerstr. 22, D 81675 Munich, Germany.
Objectives The study compared flow reserve indices by magnetic resonance imaging (MRI) with quantitative measures of coronary angiography and positron emission tomography (PET).
Background The noninvasive evaluation of myocardial flow by MRI has recently been introduced. However, a comparison to quantitative flow measurement as assessed by PET has not been reported in patients with coronary artery disease (CAD).
Methods Two groups of healthy volunteers and 25 patients with angiographically documented CAD were examined by MRI and PET at rest and during adenosine stress. Dynamic MRI was performed using a multi-slice ultra-fast hybrid sequence and a rapid gadolinium-diethylenetriaminepenta-acetic acid bolus injection (0.05 mmol/l). Upslope and peak-intensity indices were regionally determined from first-pass signal intensity curves and compared to N-13 ammonia PET flow reserve measurements.
Results In healthy volunteers, the upslope analysis showed a stress/rest index of 2.1 ± 0.6, which was higher than peak intensity (1.5 ± 0.3), but lower than flow reserve by PET (3.9 ± 1.1). Localization of coronary artery stenoses (>75%, MRI <1.2), based on the upslope index, yielded sensitivity, specificity and diagnostic accuracy of 69%, 89% and 79%, respectively. Upslope index correlated with PET flow reserve (r = 0.70). A reduced coronary flow reserve (PET <2.0, MRI <1.3) was detected by the upslope index with sensitivity, specificity and diagnostic accuracy of 86%, 84% and 85%, respectively.
Conclusions Magnetic resonance imaging first-pass perfusion measurements underestimate flow reserve values, but may represent a promising semi-quantitative technique for detection and severity assessment of regional CAD.
The noninvasive detection and functional assessment of regional coronary artery disease (CAD) continue to be of great clinical importance in cardiology. Experimental and clinical studies have shown that the functional severity of a coronary stenosis can be determined by measuring flow reserve—that is, the ratio of maximum myocardial blood flow to baseline conditions (1). Noninvasive assessment of myocardial flow by nuclear imaging techniques is a widely accepted clinical tool with documented high diagnostic and prognostic value (2). Whereas single-photon emission computed tomography provides only relative myocardial blood flow estimates, positron emission tomography (PET) has been shown to allow regional quantification of absolute coronary flow reserve (CFR) in the human heart (3,4), which has been used for the sensitive detection of early abnormalities in coronary vascular reactivity associated with CAD (5).
Contrast-enhanced magnetic resonance imaging (MRI) represents a promising alternative to established nuclear methods for evaluating regional blood flow. Using a first-pass technique, the wash-in and wash-out of contrast agents can be qualitatively and quantitatively assessed by dynamic data acquisition. This concept has been validated in animal studies in comparison to microsphere flow measurements (6,7). Based on this approach, MRI has shown the ability to detect regional flow abnormalities in patients with CAD (8,9). In addition, first quantitative methods have been introduced, which were developed to measure absolute myocardial blood flow and flow reserve (10,11). The purpose of this study was to validate further the MRI estimates of regional myocardial blood flow in patients with CAD compared to PET-flow measurements.
Two different groups (group I and group II) of healthy volunteers with a low likelihood for CAD based on the history and clinical examination were either examined by MRI or PET in order to provide reference data. Twenty-five clinically stable patients with angiographically documented CAD were examined by MRI and PET within the same day. The study protocol was reviewed and approved by the local ethics committee. Written informed consent was obtained prior to inclusion in the study.
Selective coronary angiography was performed in multiple views according to the Judkins technique. Quantitative analysis of coronary angiograms was performed (CMS 3.0, Medis Medical Imaging Systems, Leiden, The Netherlands). The luminal diameter of the stenosed artery in the projection showing maximal severity was measured at end-diastole. The degree of stenosis was expressed as percentage reduction of internal luminal diameter in relation to diameters at the proximal and distal boundaries of stenosis (12).
Magnetic resonance imaging
All subjects were examined in supine position with a 1.5-tesla tomograph (ACS-NT Philips Medical Systems, Best, The Netherlands). A fast hybrid, gated-imaging sequence consisting of three short-axis slices was used (3 slices/heart beat, single-saturation prepulse, repetition/echo time 12.5/4 ms, flip angle 30°, 11 echoes per excitation, prepulse delay 200 ms, 1.95·1.95·10 mm3) (13). Use of a single prepulse offers the advantage of a significantly reduced acquisition time; however, the second and third slice will show a reduced signal intensity (SI).
A rapid bolus (3 ml/s) of 0.05 mmol/l gadolinium-diethylenetriaminepenta-acetic acid (Gd-DTPA) (Magnevist, Schering-AG, Berlin, Germany) was injected into a peripheral vein (Spectris MR-Injector, Medrad Inc., Indianola, Pennsylvania). In a subset of healthy volunteers, an additional bolus of 0.005 mmol/l Gd-DTPA was injected. After 30 min, adenosine (Adrekar, Sanofi-Winthrop, Notre Dame de Bondeville, France) at 140 μg/kg/min was intravenously infused for 6 min, and the MRI protocol was repeated.
Using a program developed at our institution (MunichHeart/MRI) epi- and endocardial contours were manually traced and corrected for motion due to breathing or diaphragmatic shift. Each slice was subdivided into 18 equiangular regions of interest (ROI). The SI curves were automatically generated from the mid-myocardial section and further analyzed. The first pass through the left ventricular (LV) chamber and myocardium was modeled to a generic function in order to reduce signal fluctuations (see Appendix). After calculation of its first derivative, the maximal upslope in myocardial tissue was determined. Myocardial peak SI was calculated as the difference of the maximal SI to baseline. Parametric polar maps were constructed from three acquired slices. Flow ratios, calculated by dividing stress by rest values, were defined as indices for CFR. To assess estimates of the arterial input function, an ROI was additionally drawn in the central LV. The SI curves of the blood-pool region were quantitated by the ratio of amplitudes of first pass and first recirculation.
Positron emission tomography
Data were acquired using either ECAT 951 or ECAT EXACT scanner (Siemens/CTI, Knoxville, Tennessee). Voxel size was 2.34·2.34·3.38 mm3. After a transmission scan of 15 min, N-13 ammonia (740 MBq) was injected (30-s bolus). After 50 min, adenosine infusion (140 μg/kg per min) was started over 6 min. Three minutes after the initiation of infusion, a second dose of N-13 ammonia (740 MBq) was administered. Each acquisition consisted of 21 frames (12 × 10 s, 6 × 30 s, 3 × 300 s). Using an analysis program developed at our institution (MunichHeart/NM) (14), the absolute flow was calculated based on a validated three-compartment tracer kinetic model (15,16).
Polar maps were subdivided into six segments. Based on the individual coronary supply, the segments were combined into three regions representing the right coronary artery (RCA), left circumflex (LCX) and left anterior descending artery (LAD). The MRI and PET polar maps were compared in corresponding segments.
Hemodynamic changes were evaluated by the two-tailed paired Student ttest. The unpaired ttest was used for comparison between MRI and PET in volunteers. Linear regression analysis was used for comparison of MRI with coronary angiography and PET and for determination of intra- and interobserver variability. A receiver operating characteristic (ROC) analysis was performed to define the optimal sensitivity and specificity pairs for localization of coronary artery stenosis and reduced PET CFR by MRI flow indices (MedCalc-Software, Maria Kerke Belgium). All values were expressed as mean ± SD. A p value <0.05 was regarded as statistically significant.
Magnetic resonance imaging
Twenty healthy volunteers (group I: 13 men; 31 ± 7 years) underwent MRI. During adenosine infusion the rate-pressure roduct significantly increased compared to baseline conditions (7,333 ± 1,221 vs. 10,998 ± 2,092; p < 0.001). In a subset of 14 volunteers an additional 0.005 mmol/l bolus of Gd-DTPA was injected prior to the defined amount of 0.05 mmol/l. The SI ratios of the arterial input function (first-pass peak divided by recirculation peak), obtained from an ROI in the LV blood pool, averaged 3.5 ± 0.8 for the low-dose injection and 1.6 ± 0.2 for the diagnostic bolus.
Table 1shows the global MRI flow indices in the volunteers. The intra- and interobserver variability rates were very low and yielded a correlation coefficient of r = 0.98 (slope = 0.97) and r = 0.97 (slope = 0.95) for upslope index as well as r = 0.95 (slope = 0.98) and r = 0.93 (slope = 0.97) for peak intensity index. Because of the use of a single saturation prepulse, as described above, slightly higher upslope and peak intensity indices were found in the apical slice, which was acquired first (mean upslope index apical: 2.1; mid: 1.8; basal: 1.8; mean peak intensity index: apical: 1.6; mid: 1.4; basal: 1.4). To account for this slice dependency, scaling factors (apical: 1.0; mid: 1.15; basal: 1.15) were introduced to realize homogeneous, slice-independent parameters.
Positron emission tomography
Fourteen different healthy volunteers (group II: 12 men; 36 ± 13 years) as those, which were examined by MRI underwent PET. The rate-pressure product significantly increased under adenosine (9,048 ± 2,133 vs. 12,646 ± 1,863; p < 0.001). The mean absolute flow at rest averaged 0.8 ± 0.2 ml/g/min. During adenosine infusion, the flow increased to 2.9 ± 0.8 ml/g/min, yielding a CFR of 3.9 ± 1.2 (Table 1). The regional analysis in different segments showed a homogeneous flow distribution. Both the intra- and interobserver variability rates were very low and yielded a correlation coefficient of r = 0.97 (slope = 0.98) and r = 0.92 (slope = 0.98). The MRI flow ratios in normal volunteers were significantly smaller than the CFR as assessed by PET (p < 0.0001) (Table 1).
Patients with CAD
Twenty-five patients (19 men; 63 ± 13 years) with angiographically documented CAD and normal LV function (ejection fraction 64 ± 13.5%) were examined by MRI and PET. Hypercholesterolemia (n = 17) was the most common risk factor. Eleven patients had arterial hypertension, a positive family history for coronary events, or smoked. Obesity occurred in eight patients and diabetes mellitus in seven patients.
Eleven patients had single-vessel disease and 14 had multivessel disease. The distribution of stenoses is shown in Table 1. Three patients had a history of myocardial infarction (MI), 10 patients had balloon angioplasty, and 4 patients underwent coronary artery bypass graft surgery at least six months before entering the study.
The rate-pressure product during stress was comparable between MRI and PET studies (11,209 ± 2,192 vs. 11,707 ± 2,328; p = 0.2). Eleven patients developed typical angina and four had a significant ST-segment depression (>0.2 mV), which recovered rapidly after cessation of adenosine.
Comparison of MRI estimates with severity of stenosis
Figure 1shows an example of a patient with single-vessel disease. The MRI indices of flow reserve showed a progressive reduction with increasing severity of coronary artery stenosis as assessed by quantitative angiography with a negative correlation yielding a coefficient of r = −0.57 for the upslope index and r = −0.55 for the peak-intensity index. Compared to healthy volunteers, patients with CAD showed reduced MRI indices, even in regions without angiographically detectable lesions (Table 1, Fig. 2a). The sensitivity, specificity and diagnostic accuracy rates for the detection of regional stenosis (>75%) were 69%, 89% and 79% for the upslope index (cutoff: 1.20) and 69%, 74% and 72% for the peak-intensity index (cutoff: 1.10) (Fig. 3a).
Comparison of MRI estimates with PET
The MRI indices in CAD patients were significantly lower than the CFR in PET, which is consistent with findings in the healthy volunteers (p < 0.001) (Table 1). Both indices decreased progressively with decreasing CFR in PET (Fig. 2b). Poor correlation was found for the peak-intensity index in comparison to PET flow measurements (r = 0.23; slope = 0.06) (Fig. 4a). The upslope index showed a significant correlation to CFR, yielding a correlation coefficient of r = 0.70 (slope = 0.25). The MRI upslope index and the CFR by PET were comparably correlated within different vascular territories (LAD: r = 0.62, slope = 0.23; RCA: r = 0.68, slope = 0.28; LCX: r = 0.72, slope = 0.23) (Fig. 4b).
The sensitivity, specificity and diagnostic accuracy levels for detection of regional CFR <2.0 were 86%, 84% and 85%, respectively, for the upslope index (cutoff: 1.3) and 68%, 73% and 71%, respectively, for the peak-intensity index (cutoff: 1.1) (Fig. 3b).
The results of this study indicate that contrast-enhanced MRI provides estimation of regional myocardial flow reserve, which can be used to describe functionally regional CAD in agreement with the results of quantitative angiography as well as flow reserve measured by PET. However, the flow reserve is underestimated due to a low extraction of extravascular contrast agents such as Gd-DTPA. Further technical development allowing simultaneous imaging of the entire heart in conjunction with contrast agents that have better physiological characteristics will make MRI an attractive noninvasive tool to quantitatively assess perfusion and function in patients with suspected or proven CAD.
Contrast-enhanced MRI: assessment of myocardial perfusion
The only currently approved MRI contrast agents for clinical use are paramagnetic compounds such as Gd-DTPA, which rapidly distribute from the intravascular to the extracellular space. Retention of Gd-DTPA in the extracellular space depends on the extraction fraction and coronary flow (17,18). In normally perfused myocardium, the extraction fraction varies from 40% to 60% but decreases to only 20% with increasing flow (17). The decrease of extraction fraction with increasing flow is a well-known phenomenon for most flow tracers. Clinically used radiotracers such as thallium-201 and Tc-99m sestamibi display similar behavior, yielding a nonlinear relationship between tracer tissue retention and myocardial perfusion. Therefore, using tissue retention as a marker of myocardial blood flow leads to an underestimation of perfusion at high flow rates.
First-pass contrast-enhanced MRI: comparison to PET flow measurements
The data on MRI peak-intensity index observed in volunteers, which averaged only about 1.5, indicates a significant underestimation of myocardial blood flow using the SI as a marker of myocardial perfusion compared to PET-measured coronary flow. In contrast to Gd-DTPA, the initial tissue extraction of N-13 ammonia decreases only slightly, even at high flow rates (19). Experimental studies have shown that kinetic analysis of N-13 ammonia yields flow values in very close agreement with microspheres-measured blood flow (16). In contrast to peak-intensity index, upslope index was less attenuated with increasing blood flow as compared to PET. Because the early part of the SI curve following Gd-DTPA injection primarily reflects delivery of contrast agent and, to a lesser extent, the diffusion from the vascular to the extracellular space, this parameter appears to be most appropriate for the assessment of myocardial blood flow consistent with previous reports. The results of PET and MRI upslope in patients with CAD, however, indicate that, despite significant underestimation (slope = 0.25) of the true CFR, a close relationship exists between both measurements, especially in the low and medium range of myocardial flow reserve, whereas a comparison of peak intensity with PET reveals that it is not suitable for the assessment of blood flow. Nuclear techniques using single-photon tracers, such as Tc-99m sestamibi, display a similar pattern without affecting their diagnostic value (20).
Contrast-enhanced MRI: myocardial perfusion modeling
Other investigators assessed myocardial perfusion based on the calculation of mean transit time using a gamma-variate fit (7). However, using extracellular contrast agents, the estimation of the mean transit time may be difficult as the washout phase is increasingly influenced by diffusion. This may especially play a role in regions with stenotic coronary arteries. Thus, this measurement of transit time may be more suitable for strictly intravascular contrast agents.
The application of kinetic models for calculation of myocardial blood flow using extracellular agents has been reported (10,11). An accurate delineation of the arterial input function is required for this approach. The results of using different concentrations of Gd-DTPA in the present study indicate that the arterial input function is significantly underestimated at concentrations of 0.05 mmol/l using saturation-recovery sequences. The SI is primarily a function of the longitudinal relaxation time (T1) of tissue and the local concentration of contrast agent. For the accurate analysis of the arterial input function, a linear relationship between blood Gd-DTPA concentration and MRI signal is required. However, for shorter T1 below an imaging-sequence–dependent limit, an attenuated signal is observed. Conversely, Gd-DTPA bolus concentrations below 0.05 mmol/l are associated with increasing noise in the tissue SI curves. Thus, to balance between linearity and signal quality, a concentration of 0.05 mmol/l was used to maintain linearity of myocardial tissue signal and acceptable signal-to-noise ratio. Improved imaging characteristics by the introduction of faster gradient systems may overcome this physical limitation and enhance the linear range of the imaging signal, allowing the application of kinetic modeling.
Contrast-enhanced MRI: assessment of CAD
In patients with CAD, there was a considerable scatter between MRI flow reserve indices and the results of quantitative coronary angiography. This finding is not surprising; limitations of anatomic information such as stenosis severity in predicting the functional significance of angiographically detected stenosis are well appreciated. Many studies using coronary flow measurements have shown considerable variability between anatomic and functional measurements (1,3,4). In our population, estimates of CFR both for MRI and PET showed a gradual decline with increasing severity of CAD, which was more obvious for PET measurements owing to the wider flow range. In our patients, flow reserve measurements in myocardial regions without angiographically detectable lesions displayed reduced MRI and PET flow reserve compared to control subjects, indicating an impaired vascular reactivity, which may be related to endothelial dysfunction, as has been described in other PET studies (3,21).
Recent studies in patients with CAD suggest that selection for revascularization based on functional parameters may be superior to the use of anatomic criteria (22). Confirming the results of these previous studies, our data indicate a closer agreement between PET and MRI flow measurements as compared to the relationship with angiography. Despite the observed scatter, the diagnostic accuracy of 85%, using a threshold of 1.2 for the upslope index, indicates the diagnostic potential of this technique not only to localize disease, but also to assess the functional severity of a given stenosis.
Furthermore, these data suggest that, despite the underestimation of myocardial blood flow by MRI, advanced coronary artery stenosis can be detected with acceptable diagnostic accuracy, which compares favorably with results obtained by alternative noninvasive techniques. Parametric display of quantitative, regional upslope indices, as shown in Figure 1, may allow not only the detection of perfusion abnormalities but also the assessment of extent and severity of a given stenosis. Absolute measurements of flow ratios offer distinct advantages over relative evaluation, as demonstrated by the application of PET (5). Further development of the MRI equipment is necessary to provide imaging of the entire heart with high spatial and temporal resolution, and thus accurate assessment of perfusion defect size. Because the coronary vascular tree can be defined with the same imaging modality, correlation of perfusion defects and coronary anatomy may be possible, which represents an attractive diagnostic aspect in patients with suspected CAD.
The diagnostic thresholds were determined based on an ROC analysis yielding 1.2 for the upslope index and 1.1 for the peak-intensity index, which represents a 1.5 SD for the upslope index based on the results in volunteers. Other investigators used higher cutoff values for the detection of ischemic regions, which were defined using normal databases (23). As these thresholds are lower than those used in PET, further studies are necessary to refine diagnostic criteria prospectively in larger patient cohorts to optimize the diagnostic performance of the test.
The selected patient population had a very low incidence of prior MI and impaired LV function, which may not represent the entire spectrum of clinical disease representation in which both the localization and the characterization of coronary stenoses are important. However, the observed diagnostic accuracy suggests the feasibility of this approach, which needs to be confirmed in other patient groups.
Data in volunteers were obtained in two separate cohorts, which limits the direct comparison of individual data. However, both groups were closely matched in terms of age and low likelihood profile for CAD as well as hemodynamics. Because of ethical considerations, it was believed unnecessary to expose the MRI group of volunteers to radiation in the presence of PET data already acquired in normal volunteers. The differences in flow index ratios between both groups cannot be explained by the small differences in hemodynamics, but rather reflect the above-discussed limitations of Gd-DTPA as flow agent.
This study shows that stress/rest flow ratios derived from first-pass myocardial time-intensity curves based on contrast-enhanced MRI are stable and reproducible parameters. Although the upslope index showed the highest values among the tested MRI parameters, the CFR was underestimated as compared to PET, reflecting the low extraction fraction of Gd-DTPA by the myocardium. A close relationship was observed between MRI upslope index and PET estimates of flow reserve, yielding acceptable diagnostic performance for localization of CAD. Further technical developments are necessary to exploit fully the potential of MRI to provide functional and anatomic characterization of CAD.
We acknowledge the statistical support of Birgit Flatau and Raymonde Busch. We thank the technical staff: Martina Gildehaus, Gesa Klotzbach, Brigitte Dzewas and Coletta Kruschke in the MRI and PET suites. We also appreciate the excellent secretarial assistance of Gabriele Sonoda.
Several model functions for the first pass of contrast agent were proposed in the literature (e.g., bivariates) in order to extract quantitative parameters. Essentially, those techniques reduce the signal fluctuations of individual data points. Similarly, we model a rather generic function, namely a 5th order polynomial:to the SI curve using a nonlinear, least-squares-fit routine (24). Subsequently, we use the analytic derivative:to assess the parameter maximal upslope.
☆ This study was supported in part by the Deutsche Forschungsgemeinschaft (grant SCHW 235/4-1), Kennedyallee 40, 53170 Bonn, Germany.
- coronary artery disease
- coronary flow reserve
- gadolinium-diethylenetriaminepenta-acetic acid
- left anterior descending
- left circumflex
- left ventricle/ventricular
- myocardial infarction
- magnetic resonance imaging
- positron emission tomography
- right coronary artery
- receiver operating characteristic
- regions of interest
- signal intensity
- Received January 10, 2001.
- Revision received November 29, 2001.
- Accepted December 14, 2001.
- American College of Cardiology Foundation
- Gould K.L.,
- Kirkeeide R.L.,
- Buchi M.
- Beanlands R.S.,
- Muzik O.,
- Melon P.,
- et al.
- Dayanikli F.,
- Grambow D.,
- Muzik O.,
- et al.
- Manning W.J.,
- Atkinson D.J.,
- Grossman W.,
- et al.
- Lauerma K.,
- Virtanen K.S.,
- Sipila L.,
- et al.
- Cullen J.H.,
- Horsfield M.A.,
- Reek C.R.,
- et al.
- Kosa I.,
- Blasini R.,
- Schneider-Eicke J.,
- et al.
- Fischer S.,
- Groen J.,
- Henson R.
- Muzik O.,
- Beanlands R.S.,
- Hutchins G.D.,
- et al.
- Schelbert H.R.,
- Phelps M.E.,
- Huang S.C.,
- et al.
- Melon P.G.,
- Beanlands R.S.,
- DeGrado T.R.,
- et al.
- Uren N.G.,
- Marraccini P.,
- Gistri R.
- Ferrari M.,
- Schnell B.,
- Werner G.S.,
- et al.
- Al-Saadi N.,
- Nagel E.,
- Gross M.,
- et al.