Magnetic Resonance Versus Radionuclide Pharmacological Stress Perfusion Imaging for Flow-Limiting Stenoses of Varying Severity
Background— Although magnetic resonance first-pass imaging (MRFP) has potential advantages in pharmacological stress perfusion imaging, direct comparisons of current MRFP and established radionuclide techniques are not available.
Methods and Results— Graded regional differences in coronary flow were produced during global coronary vasodilation in chronically instrumented dogs by partially occluding the left circumflex artery. Regional differences in full-thickness flow quantified using microspheres were compared with regional differences obtained with MRFP and radionuclide SPECT imaging (99mTc-sestamibi and 201Tl). Relative regional flows (RRFs) derived from the initial areas under MRFP signal intensity-time curves were linearly related to reference microsphere RRFs over the full range of vasodilation (y=0.93x+4.3; r2=0.77). Relationships between 99mTc-sestamibi and 201Tl RRFs and microsphere RRFs were curvilinear, plateauing as flows increased. The high spatial resolution of the MRI enabled transmural flow to be evaluated in 3 to 5 layers across the myocardial wall. Reductions in subendocardial flow were visually apparent in MRFP images for ≥50% reductions in full-thickness flow. Endocardial-to-epicardial gradients in MRFP flow increased progressively with stenosis severity, whereas transmural flow patterns in remote normally perfused myocardium remained normal. Flow reductions of ≥50% not identified by radionuclide imaging were apparent in MRFP full-thickness and transmural analyses.
Conclusions— High-resolution MRFP can identify regional reductions in full-thickness myocardial blood flow during global coronary vasodilation over a wider range than current SPECT imaging. Transmural flow gradients can also be identified; their magnitude increases progressively as flow limitations become more severe and endocardial flow is compromised increasingly.
Received June 30, 2003; de novo received January 29, 2004; revision received March 17, 2004; accepted March 22, 2004.
Radionuclide stress perfusion imaging now plays a major role in the noninvasive diagnosis of coronary artery disease. As the capabilities of magnetic resonance scanners have improved, the possibility has arisen that magnetic resonance first-pass perfusion (MRFP) imaging during pharmacological coronary vasodilation can provide similar information. Our laboratory has recently demonstrated that MRFP imaging can identify 2-fold regional differences in flow over the full range of coronary vasodilation.1 Because of the known plateau in radionuclide uptake at flow rates exceeding twice resting levels,2,3 this finding suggested that MRI may be able to identify limitations in regional flow reserve over a wider range of coronary stenoses than radionuclide imaging. The improved spatial resolution of MRIs also raises the possibility of identifying transmural differences in perfusion produced by stenotic lesions.
The present study was undertaken to compare MRFP and 99mTc-sestamibi and 201Tl SPECT imaging in the quantification of regional differences in full-thickness vasodilated blood flow in viable myocardium. Additionally, we tested the hypothesis that regional endocardial-to-epicardial differences in MRFP flow patterns can be helpful in identifying flow-limiting stenoses. Studies were conducted in a canine model in which high-resolution MR images could be obtained while regional differences in vasodilated flow were varied over a wide range and quantified independently using systemically administered microspheres.
Studies were conducted in 18 chronically instrumented dogs of both sexes (R&R Research, Howard City, Mich, and Covance Research Products, Denver, Pa) using procedures and protocols in accord with the Position of the American Heart Association on Research Animal Use.
Surgery and Anesthesia
Random and purpose-bred hounds weighing 17 to 27 kg were instrumented as reported previously.4 The proximal portion of the left circumflex artery (LC) was instrumented with an external hydraulic occluder and external cuff-type Doppler flowmeter. Left and right atrial catheters were placed for administration of fluorescent microspheres and MR contrast agent, respectively. Animals were allowed to recover for ≥7 days before study. On days of study, animals were brought to the MR laboratory or SPECT scanner, anesthetized with methohexital sodium (11 mg/kg IV) or sodium thiopental (10 to 20 mg/kg IV), and ventilated with a commercial anesthesia machine and oxygen-isoflurane (1.5% to 2.5%) gas mixture. End tidal Pco2 was maintained at 30 to 32 mm Hg. Systemic o2 saturation always exceeded 95%. Heart rates varied between 80 and 110 bpm.
Magnetic Resonance Imaging
Fifty-three MRFP studies were performed using a 1.5T scanner (Sonata, Siemens Medical Systems) with animals in the right lateral decubitus position. To maximize spatial resolution and signal-to-noise ratios, we confined observations to a single 8-mm short-axis slice positioned in the mid-portion of the posterior papillary muscle and imaged using a single-shot steady-state free-precession (SSFP) inversion recovery sequence.1 Typical parameters were repetition time 500 to 600 ms, inversion time 270 to 280 ms, field of view 150×300 mm, matrix 77×256, and voxel size 1.9×1.2×8 mm (in-plane resolution, 2.3 mm2). Global coronary vasodilation was produced using carbocromen (5 to 10 mg/kg IV, n=17) or adenosine (140 to 840 μg/kg per min IV, n=36). The dose of adenosine was that producing the maximum LC Doppler velocity in a preliminary study. LC stenoses sufficient to produce regional flow limitations were produced by partially inflating the LC hydraulic occluder while monitoring the Doppler velocity signal. MRFP images were obtained with ventilation interrupted at end expiration. Using an automatic injection device (Medrad, Spectris), Gd-HP-DO3A (gadoteridol, n=28) or Gd-DTPA (n=25) 0.054±0.015 mmol/L per kg was injected into the right atrial catheter at a rate of 1 mL/s and followed by a 12-mL saline flush also at 1 mL/s. Images were obtained during each diastole during the first passage of contrast through the central circulation. MRFP image acquisition and regional flow analysis are depicted in Figure 1. A total of 3×106 fluorescent microspheres5 were injected into the left atrium immediately after each MR measurement.
Radionuclide perfusion tracers were administered either during or shortly after MR studies. In 18 studies, 99mTc-sestamibi (25 mCi) was injected intravenously immediately after MRFP imaging under the same conditions of global coronary vasodilation and LC stenosis with the animal still in the MR scanner. Additional 99mTc-sestamibi studies (n=8) and 201Tl studies (3 mCi, n=12) were performed during global coronary vasodilation and similar degrees of flow reduction outside of the MR scanner, with additional concomitant microsphere flow measurements (including reference arterial blood sampling, which was ordinarily not possible in the scanner). Flows in normally perfused myocardium averaged 3.1±0.5 mL/min per g. 201Tl SPECT images were acquired beginning 5 to 10 minutes after tracer injection. 99mTc-sestamibi SPECT images were acquired in the first 2 hours after tracer injection and included any degree of tracer redistribution that occurred between the time of injection and the time of sampling.6 Images were obtained using a single crystal gamma camera (Siemens Orbiter) and high-resolution collimator interfaced with an ICON acquisition and processing system (64 views over 180 degrees, 40 seconds per view, 6.5-mm pixel size and slice thickness). When animals were studied on more than one occasion, at least 72 hours (>10 half lives of 99mTc-sestamibi) were allowed between studies.
Postmortem Microsphere and Radionuclide Analyses
After completion of studies, animals were euthanized with an overdose of pentobarbital, and the heart was retrieved for additional analysis. Two contiguous 1-cm mid-left ventricular slices encompassing the posterior papillary muscle were positioned so that the papillary muscle and posterior attachments of the right ventricle matched the contour of MRI images. Each slice was then divided into 12 30-degree sectors. In 2 animals, hearts were fixed in formalin and individual sectors were additionally divided into 4 transmural layers. Concentrations of fluorescent microspheres in individual samples were quantified fluorometrically5 and expressed on a per-gram basis. Data from corresponding samples in the 2 slices were averaged for comparisons with MRI and SPECT data. In 9 animals in which a 99mTc-sestamibi perfusion scan was obtained just before euthanasia, myocardial samples underwent well counting for 99mTc (Auto-Gamma Counting System, Packard Instrument Co) before microsphere analysis.
Full-Thickness Flow Analysis (MRI and SPECT)
Left ventricle endocardial and epicardial borders were drawn manually, and MRI signal intensity-time curves were generated for each 30-degree full-thickness sector. Baseline values measured before contrast administration were used to correct for any coil-induced differences in signal, which rarely exceeded 15%. The initial areas under each curve were measured from the onset of contrast appearance to the peak of the curve showing the most rapid upstroke.1 These values were taken to represent relative regional flows. They were then normalized to the highest sector value and compared with relative regional concentrations of fluorescent microspheres normalized to their highest sector value.
For SPECT images, 3 contiguous short-axis midventricular images were similarly divided into 12 30-degree sectors. The average signal intensity of the 3 images for each sector was taken to represent relative regional flow. These values were then normalized to the highest sector value and compared with relative regional concentrations of fluorescent microspheres normalized to their highest sector value. In animals in which a 99mTc-sestamibi perfusion scan was obtained just before euthanasia, SPECT intensities were also compared with values of 99mTc-sestamibi measured using postmortem well counting.
Transmural Flow Analysis (MRI)
Transmural flow patterns across the myocardial wall in the inferior myocardium in the area of the posterior papillary muscle (distal LC bed) were compared with those in normally perfused anteroseptal myocardium. As outlined in Figure 2, MRFP signal intensity-time curves could be generated for 3 to 5 (and occasionally 6) transmural layers in the inferior myocardium and papillary muscle and 3 to 4 transmural layers in anteroseptal myocardium. The initial areas in the signal intensity-time curves across each wall were taken to represent relative flows. Areas were normalized to the peak value within the wall and subjected to linear regression. The endocardial-to-epicardial slope of the regression was taken to represent the transmural flow gradient across the wall.
Figure 3 illustrates MRFP images, regional MRFP signal-intensity curves, 99mTc-sestamibi SPECT images, and regional microsphere concentrations in animals with normal and regionally reduced vasodilated flows. In the normal situation, regional microsphere concentrations show minimal variation, as do regional MRFP and 99mTc-sestamibi SPECT signals, and MRFP signal-intensity curves have similar initial areas. When circumflex microsphere flow is reduced by ≥50%, a perfusion defect is apparent in MRFP images and signal-intensity curves but not the SPECT image. When microsphere flow is reduced by ≥85%, MRFP images, MRFP signal-intensity curves, and SPECT images all show marked abnormalities. A transmural gradient in flow reduction is visually apparent in MRFP images for both degrees of flow limitation.
Figure 4A shows the relationship between MRFP and microsphere relative regional flows (RRF [30-degree sectors]) in all MRFP studies. MRFP RRFs are linearly related to the reference microsphere RRFs over the full range of vasodilation (y=0.93x+4.3; n=608, r2=0.77, SEE=13). Figures 4B and 4C show similar comparisons between in vivo 99mTc-sestamibi and 201Tl RRFs and microsphere RRFs. Both relationships are curvilinear, plateauing as flows increase (y=23 Ln x−11; n=310, r2=0.74; y=28 Ln x−35; n=142, r2= 0.70). Figure 4D compares in vivo SPECT and ex vivo well counting values of 99mTc-sestamibi with microsphere RRFs. The relationships agree closely.
Figure 5 illustrates similar agreement between MRFP and microsphere flows in individual transmural layers to that in full-thickness analyses. Figure 6 shows representative analyses of MRFP endocardial-to-epicardial flow gradients in normally perfused and stenotic myocardium. Figure 7 shows the relationship between inferior and anteroseptal transmural MRFP slopes and microsphere RRFs in all studies. Transmural gradients were apparent visually in regions of MRFP images in which full-thickness flow was reduced by ≥50%. MRFP slopes in these regions increased progressively with severity of stenosis but were unchanged in remote normally perfused myocardium.
Although cardiac applications of MRI have increased substantially in recent years, the strengths and limitations of MRI in identifying regional limitations in coronary flow reserve in viable myocardium supplied by stenotic coronary arteries remain incompletely defined. Comparisons of MRFP with radionuclide stress perfusion imaging are especially pertinent because of the established clinical utility of the radionuclide modality. The present study examined entire short-axis slices as would be done in routine clinical examinations. The full-thickness MRI findings (Figure 4A) extend our earlier report that relative initial areas under MRFP curves in selected regions of interest are linearly related to relative regional flows determined with microspheres over the full range of coronary vasodilation.1 In contrast to MRFP findings, Figures 4B and 4C document the expected falloffs in 99mTc-sestamibi and 201Tl deposition at higher flow rates.2,3 In vivo SPECT and ex vivo well-counting data for 99mTc correspond closely (Figure 4D). Taken in context with Figure 4A, these data suggest that MRFP can be advantageous in the detection of stenoses producing moderate limitations in flow reserve.
The identification of regional flow limitations was additionally facilitated by the ability of MRFP to identify transmural differences in perfusion when high spatial resolution is achieved. When regional full-thickness flow was reduced by ≥50%, a transmural gradient in perfusion was consistently apparent in unprocessed images (Figures 3 and 6⇑, middle and lower panels). Transmural analysis was aided by the development of an MRFP M-mode display to replace freehand drawing of regions of interest. The display allowed analysis of transmural layers in 2-voxel steps from endocardium to epicardium. Additionally, the ability to visualize regional signal intensity changes over time in a single frame (rather than a 30- to 40-frame cine loop) was helpful in recognizing motion-related artifacts and signal contamination originating outside the myocardium (eg, right ventricle and left ventricle cavity or pericardial structures). As illustrated in Figures 6 and 7⇑, transmural MRFP slopes increased progressively with increasing stenosis severity, whereas transmural patterns in remote normally perfused myocardium remained unchanged.
The observed transmural MRFP slopes in normally perfused and stenotic beds are consonant with factors known to limit perfusion progressively from epicardium to endocardium. The pressure drop in arteries upstream of the microcirculation is intrinsically greater in arteries supplying the subendocardium than the subepicardium, presumably because of the longer intramural path of the vessels supplying the inner myocardial layers.7 Compressive effects of ventricular contraction narrow intramural arteries and other vessels, displacing a significant fraction of their contained blood and greatly increasing systolic inflow resistance.8 A finite period in early diastole is required for reexpansion and refilling of these vessels, thereby also delaying early diastolic microcirculatory perfusion. Finally, inner wall flow can be further limited if intraventricular diastolic pressure is elevated.9
Although these effects are normally counterbalanced by autoregulatory adjustments in transmural microvascular tone, they are unopposed when the coronary bed is maximally vasodilated. The resulting transmural gradients in flow are usually modest in normally perfused areas, as reflected in the present study by minimal transmural flow gradients in normally perfused myocardium (Figure 7). In stenotic areas, transmural flow limitations are magnified by the reductions in regional epicardial artery pressure that result from flow-related pressure gradients across the stenoses. These gradients increase exponentially with severity of stenosis,10,11 causing correspondingly larger reductions in distal coronary artery pressure. The resulting differences between left ventricular and distal coronary artery pressures progressively accentuate the transmural effects outlined above. Thus, even when absolute flows remain above resting values in all myocardial layers, regional gradients in transmural flow increase progressively with severity of stenosis. A regionally increased gradient can be helpful in interpreting the significance of a borderline reduction in full-thickness flow.
Radionuclide approaches for distinguishing viable myocardium with limited flow reserve from areas of myocardial infarction usually involve comparisons of perfusion images obtained at rest with those during vasodilation. Although a similar approach could be taken by performing MRFP at rest as well as during vasodilation, it is now possible to visualize areas of infarction directly by collecting contrast-enhanced inversion recovery images 10 to 15 minutes after MRFP.12 The high spatial resolution of MR facilitates the identification of small or nontransmural infarctions13 and can thereby more precisely define areas with limited vasodilator flow reserve, which represent viable myocardium. The ability to distinguish between viable and infarcted myocardium can also be useful in interpreting regional deficits in MRFP under resting conditions, eg, in the setting of acute ischemia or previous infarction.
Several limitations of this canine study need to be kept in mind when considering human clinical populations. We used an inversion recovery SSFP sequence to maximize signal-to-noise ratios as well as spatial resolution. However, inversion recovery sequences can image only a single myocardial slice during each cardiac cycle and require constant R-R intervals. Sequences using saturation recovery14 are insensitive to arrhythmia and capable of multislice imaging but involve significant tradeoffs in signal-to-noise and spatial resolution. The degree of resolution in the present study also benefited from a smaller field of view than is usually feasible in humans and did not have to contend with diaphragmatic instability during breath holds. These various limitations will need to be minimized or eliminated if the present findings are to be extended to the clinical setting.
Efforts to improve options in human first-pass imaging continue. Schreiber et al14 recently reported an in-plane resolution of 2.4×2.1 mm in single-slice imaging using a saturation recovery SSFP sequence. Using saturation recovery and a segmented echo-planar gradient echo readout, Kwong et al15 have achieved 3.0- to 3.3-mm in-plane resolution while imaging 7 to 9 locations every 2 heart beats. Using a similar sequence and readout, Nagel et al16 have reported a spatial resolution of 1.7 to 2.2×1.9 to 2.4 mm while imaging 5 8-mm slices each heart beat.
In summary, MRFP can identify regional reductions in full-thickness myocardial blood flow during pharmacological coronary vasodilation over a wider range than current SPECT imaging. The detection of flow-limiting stenoses is additionally facilitated by the ability of MRFP to quantify regional transmural gradients in flow that are accentuated as the severity of flow limitation increases.
Supported by the Feinberg Cardiovascular Research Institute and NIH RO1-HL63268 (to R.M. Judd).
Klocke FJ, Simonetti OP, Judd RM, et al. Limits of detection of regional differences in vasodilated flow in viable myocardium by first-pass magnetic resonance perfusion imaging. Circulation. 2001; 104: 2412–2416.
Leppo JA, Meerdink DJ. Comparison of the myocardial uptake of a technetium-labeled isonitrile analogue and thallium. Circ Res. 1989; 65: 632–639.
Glover DK, Ruiz M, Edwards NC, et al. Comparison between 201Tl and 99mTc sestamibi uptake during adenosine-induced vasodilation as a function of coronary stenosis severity. Circulation. 1995; 91: 813–820.
Glenny RW, Bernard S, Brinkley M. Validation of fluorescent-labeled microspheres for measurement of regional organ perfusion. J Appl Physiol. 1993; 74: 2585–2597.
Li Q-S, Solot G, Frank TL, et al. Myocardial redistribution of technetium-99m-methoxyisobutyl isonitrile (sestamibi). J Nucl Med. 1990; 31: 1069–1076.
Chilian WM. Microvascular pressures and resistances in the left ventricular subepicardium and subendocardium. Circ Res. 1991; 69: 561–570.
Yada T, Hiramatsu O, Kimura A, et al. In vivo observation of subendocardial microvessels of the beating porcine heart using a needle-probe videomicroscope with a CCD camera. Circ Res. 1993; 72: 939–946.
Gould KL. Pressure-flow characteristics of coronary stenoses in unsedated dogs at rest and during coronary vasodilation. Circ Res. 1978; 43: 242–253.
Klocke FJ. Measurements of coronary flow reserve: defining pathophysiology vs. making decisions about patient care. Circulation. 1987; 76: 1183–1189.
Kim RJ, Fieno DS, Parrish TB, et al. Relationship of MRI delayed contrast enhancement to irreversible injury, infarct age, and contractile function. Circulation. 1999; 100: 1992–2002.
Kwong RY, Schussheim AE, Rekhraj S, et al. Detecting acute coronary syndrome in the emergency department with cardiac magnetic resonance imaging. Circulation. 2003; 107: 531–537.
Nagel E, Klein C, Paetsch I, et al. Magnetic resonance perfusion measurements for the noninvasive detection of coronary disease. Circulation. 2003; 108: 432–437.