Measurement of Absolute Epicardial Coronary Artery Flow and Flow Reserve With Breath-Hold Cine Phase-Contrast Magnetic Resonance Imaging
Background Noninvasive measurement of absolute coronary arterial flow and coronary flow reserve would be of considerable use in the diagnosis and management of patients with coronary artery disease. Phase-contrast magnetic resonance imaging (MRI) has been used to measure flow in a variety of vessels. The goal of the present study was to determine if MRI measurements of coronary artery flow in a single breath-hold can be used to determine flow reserve and the severity of pericardial stenosis.
Methods and Results In eight mongrel dogs, a closed chest model of partial left anterior descending coronary artery (LAD) occlusion was created. Coronary flows in the left circumflex artery (LCx) and LAD were measured at rest and during adenosine infusion using velocity-encoded, breath-hold MRI and perivascular ultrasound (US) flowmeters. MRI measurements of absolute coronary flow and coronary flow reserve were highly correlated with US (r=.96 and .94, respectively). Flow reserve measured in the constricted LAD was significantly lower than that in the unconstricted LCx by both US (P=.002) and MRI (P=.011).
Conclusions MRI measurements of coronary flow and flow reserve were in good agreement with US measurements. In addition, MRI measurements of coronary flow reserve successfully discriminated stenotic from normal vessels. These results indicate that MRI is a useful method for the noninvasive assessment of coronary flow and stenosis.
Epicardial stenoses do not significantly reduce absolute coronary arterial flow until a severe constriction occurs.1 Coronary flow reserve, defined as the maximum flow under conditions of maximal vasodilation divided by the resting flow, has been used as a means of assessing the capacity of a vessel to provide increased flow.2 Noninvasive measurements of myocardial perfusion, such as those obtained with nuclear medicine instruments, produce data from which the relative, rather than absolute, coronary flow reserve can be derived. Velocity-encoded magnetic resonance imaging (MRI) offers the potential for noninvasive measurement of absolute coronary artery flow and coronary flow reserve from the simultaneous determination of both mean flow velocities and vessel lumen cross-sectional areas in vivo. MRI measurement of blood flow in coronary arteries has been limited by the relatively long study times and by cardiac and respiratory motion. Edelman and coworkers3 proposed reducing the number of independent cardiac phases viewed by acquiring phase-encoding groups (PEG), a method they called k-space segmentation, to shorten the total scan time and allow single breath-hold acquisitions.
Rapid and accurate MRI measurements of pulsatile flow using the PEG approach combined with quantitative phase-contrast velocity mapping have been demonstrated in test objects.4 This method was used to measure coronary flow velocities in human volunteers during episodes of breath-holding5 6 ; however, the spatial resolution was not adequate for accurate flow measurements. High-performance gradient coils have recently become available, which allow the acquisition of breath-hold cine MRI velocity images with submillimeter pixel sizes.
In the present study, the following questions were addressed. (1) Can adequate spatial and temporal resolution be obtained to perform phasic breath-hold MRI flow measurements of coronary arterial flow in vivo? (2) How well do MRI measurements of absolute coronary arterial flow and coronary flow reserve agree with those obtained from perivascular ultrasound (US) flowmeters? (3) Are measurements of absolute coronary flow and flow reserve adequate to distinguish flow in normal vessels from those with a subcritical stenosis?
After approval from the Institutional Animal Care and Research Advisory Committee was granted, we obtained eight male mongrel dogs (mean weight, 36.3 kg) for study. Aspirin was administered for 1 week before surgery. A blood sample was taken on the day of the study to ensure a normal red blood cell count. The animals were tranquilized with acepromazine maleate (0.1 mg/kg SQ) and then anesthetized with pentobarbitol (28 mg/kg IV), intubated, and ventilated (12 breaths per minute; tidal volume, 20 to 25 mL/kg body wt). After a carotid cutdown, a catheter was placed in the jugular vein for drug infusion, and another was advanced into the ascending aorta for pressure monitoring. A thoracotomy was performed at the fourth or fifth intercostal space. The left anterior descending coronary artery (LAD) and the left circumflex coronary artery (LCx) were then isolated. Nonmagnetic perivascular US flow probes (Transonic Systems Inc) were placed around the LCx and LAD.7 8 US flow measurements were recorded. A constrictor, machined from Lexan, was placed on the LAD to simulate a subcritical stenosis distal to the flow probe.9 The constrictor size was determined for a particular vessel by choosing the maximum size with which the artery could maintain normal flow while producing a significant hyperemic response when the constrictor was removed. This size constrictor was deemed adequate to produce a hemodynamic condition equal to a subcritical stenosis. The chest was closed with the instrumentation in place and the animal was transported to the MRI system.
At the magnet, pressure and flow measurements were calibrated and recorded. Each animal was placed on its right side on the table of the MRI system and ventilated with a mixture of isoflurane (1% to 2%) and oxygen (2.5 L, 12 breaths per minute). US flow, aortic pressure, and ECG signals were recorded on a strip-chart. US flow measurements and aortic pressure were recorded continuously except during the MRI acquisition as these electronic instruments were found to produce unacceptable radiofrequency noise in the MR image. MRI baseline flow measurements were performed with the US flow data and aortic pressures recorded before the first adenosine bolus of 6 mg was administered intravenously. After a wait of 1 to 2 minutes, a second bolus of 12 mg was given intravenously. An intravenous adenosine infusion of 0.14 mg · kg−1 · min−1 was started, and pressure and US flow measurements were recorded until a consistent, high coronary flow rate was established in the LCx.
US and MRI flows were recorded during maximum coronary dilation. The adenosine infusion was stopped. Pressure and flows were monitored as they returned to baseline levels, and MRI flow measurements were repeated. The animal was then removed from the magnet and killed by a lethal dose of sodium pentobarbitol, and the heart was excised.
Magnetic Resonance Imaging
MRI signal acquisition was performed using a method of k-segmentation that allows the acquisition of several phase-encoding steps for each cine frame in a single heartbeat.3 In this method, data are acquired in PEG, and short repetition times are used to acquire multiple phase-encoding steps for each cardiac frame for each RR interval. A complete cardiac cine with four to six frames can be obtained during a single breath-hold. The scan time is reduced by a factor equal to the PEG size without reduction in spatial resolution or special modifications to the system hardware.
The specific type of k-space segmentation used in the present study involved symmetrical, centrally ordered PEG (SCOPEG).4 10 With this method, heartbeat-to-heartbeat misregistration of the data is avoided by acquiring the data in the central portion of k-space (low spatial frequencies) at the same time in the cardiac cycle for successive heartbeats. The effects of eddy currents, produced by rapid pulsing of the gradients, are minimized by designing the phase-encoding gradient pulses so that a positive pulse is always followed by a negative pulse, resulting in smoother transitions. To preserve uniform modulation of the gradient pulses, the positive and negative regions of k-space are sampled separately, resulting in a data set that is symmetrically ordered about the zeroth phase-encoding step. Thus, this segmentation scheme reduces the incidence of ghosting and eddy current effects while also minimizing the blurring due to motion between low spatial frequency phase-encode steps.
Phase-contrast cine MRI velocity measurement requires that at least two independent images are acquired.11 By manipulation of magnetic field gradient pulses, one image is “motion compensated” so that minimal dephasing of the echo signal due to first-order motion occurs. The second acquisition is designed so that the phase of the signal is “velocity encoded” for a predetermined range of velocities. The difference in the phases of the two complex MRI signals from moving spins is directly proportional to the velocity. To produce quantitative phase-contrast MRI velocity images, an image of the spatial distribution of the MRI signal’s phase, as distinct from the more commonly used signal amplitude, must be produced. The pixel-by-pixel difference between the motion-compensated and velocity-encoded phase images produces the velocity image. By interleaving the “motion-compensated” and “velocity-encoded” acquisitions, view-to-view misregistration artifacts are reduced.12
The present study was conducted with a 1.5-T Vista HPQ MRI system (Picker International) that was fitted with prototype high-performance gradient coils designed using methods described by Morich.13 These gradient coils consisted of a distributed winding with a peak gradient of 15 mT/m and a 500-μs minimum ramping time to 10 mT/m. A receive-only quadrature surface coil (20 cm×25 cm), designed for spine imaging, was placed on the left thorax of the animal over the incision. Multiphasic pilot images (TR=19.1 milliseconds, TE=11 milliseconds) were obtained using the breath-hold PEG method.
Starting with a sagittal view and continuing with oblique orientations, a series of high-resolution cine images with the LAD and LCx in the image plane were obtained. Breath-hold MRI pilot images were obtained with the ventilator stopped during the scan to determine the position of the LCx and LAD during a breath-hold. A perpendicular view of the vessel was acquired to confirm that the section was straight and adequate for the slice thickness of the anticipated flow measurement and that there was no excessive through-plane motion. Through-plane motion was evaluated by measuring the frame-to-frame displacement of landmarks on the myocardium near the arteries that were visible on the cine scout images.
MRI images could not be positioned at the same position as the US flow probes since a significant magnetic susceptibility gradient virtually eliminated signal in these regions. Placement of the velocity image acquisition was generally along the arterial segment, proximal to the US flow probe. In two animals, no sufficiently straight arterial segment proximal to US flow probe on the LAD artery was found. In these cases, the LAD flow velocity images were obtained distal to the US flow probe.
Flow was determined using interleaved velocity-encoded and motion-compensated MRI scans. The velocity encoding was calibrated for the range of flow velocities from −138.4 cm/s to +138.4 cm/s in the slice select direction. MRI image parameters were TR, 19.1 milliseconds; TE, 11 milliseconds; 40-degree nutation; and slice thickness, 7 mm. The following parameters varied with the specific image orientation and the animal’s heart rate: FOV=18 to 24 cm corresponding to a pixel size range of 0.7 to 0.94 mm2, with phase-encode aspect ratios ranging from 160:256 to 220:256, and PEG sizes of two or three. Each cine set of four to six images was acquired during a 30- to 40-second breath-hold. A total of 8 to 12 flow measurements were made in each of the animals using quantitative MRI velocity imaging while measuring flow by US just before and after MR image acquisition.
Flow measurements in the LAD were in the presence of a simulated subcritical stenosis. In the LCx, all measurements were made with no constrictor present, both before and during adenosine administration. In the course of a typical experiment, at least two MRI multiphasic velocity image sets under baseline and adenosine infusion conditions were acquired from each vessel. Seventy-five MRI coronary flow values were compared with the concurrent flow probe measurements, providing a database adequate for meaningful statistical analysis.
The coronary artery flow was estimated from the MRI data in the following manner. The magnitude images were displayed in a cine loop, and the position and shape of the artery were noted. To mitigate identification problems due to in-flow signal enhancement, the lumen cross-sectional area was determined from the magnitude image of the cine set in which the vessel appeared to be largest. Previous studies have demonstrated a mean variation in coronary vessel size of 3% from beat to beat and of 6% from systole to diastole,14 15 suggesting that beat-to-beat and phasic changes in coronary arterial diameter could not be measured at the MRI spatial resolutions used in the present study. Thus, the region of interest (ROI) size was kept constant and the frame-to-frame location of the vessel was determined using the magnitude image set. By placing the ROI at the identical position on the associated velocity image, the mean velocity in the vessel was measured. The flow during each phase of the cardiac cycle was calculated from the product of the mean velocity in the ROI and the ROI area. The MRI estimated coronary flow (in mL/min) was the mean of the calculated flow values for the entire set of images.
Coronary flow reserve in each artery was defined as the ratio between the average of the vasodilated coronary flow values and the mean resting coronary flow value. The predicted and US-measured flow reserves in each artery were calculated by taking the mean coronary flow measured during adenosine infusion and dividing it by the mean baseline coronary flow.
The correlation coefficients and the slope and intercept of the regression line were calculated by linear regression of the MRI-predicted and US-measured coronary flow and coronary flow reserve data.16 The MRI-predicted flow values were compared for agreement with the US flow data and analyzed using the method of Bland and Altman.17 The reproducibility of the MRI-predicted and US flow measurements was assessed using the data from the pairs of flow measurements performed under stable hemodynamic conditions. The standard deviation of the mean difference between two successive measurements multiplied by 2 yields the reproducibility coefficient.17
Relative coronary flow reserve was assessed by comparing the maximal coronary flow reserve in the LAD (with subcritical stenosis) with the maximal coronary flow reserve in the LCx (no stenosis) during the same measurement interval. An unpaired Student’s t test with equal variances12 was used to determine the probability of the null hypothesis that the coronary flow reserve in the two vessels would be measured as equal. A paired t test12 was used to determine the statistical significance of differences between vessel sizes, mean velocities, and mean flow values measured with MRI before and during adenosine infusion. Except where noted, all data are expressed as mean±SD. Statistical significance is defined as a value of P≤.05.
One animal began to breathe without the respirator, causing one LAD and one LCx image to be excluded due to excessive motion artifacts. In another animal, high coronary flow during adenosine infusion was maintained only briefly, and one LCx flow image was discarded from the study since the flow was rapidly decreasing during this scan.
Hemodynamic data are listed in Table 1⇓. There was an increase in the baseline coronary artery flow recorded at the beginning of MRI measurements compared with those recorded during surgery. This difference was presumed to be due to the vasodilatory effects of the isoflurane anesthesia used during the MRI scan.18
No problems were encountered with changing heart rate. There was a small, but not statistically significant, increase from baseline heart rate during adenosine infusion (Table 1⇑). The mean heart rate was 112.1 beats per minute before a breath-hold and 113.0 beats per minute after a breath-hold. The mean percent difference between heart rate before and after breath-holding was <1.0%.
LAD and LCx Coronary Flow
The MRI acquisition took approximately 90 minutes, including set-up and velocity imaging. During the course of each experiment, coronary arterial flow increased with both MRI and US flowmeter measurements (Fig 1⇓). The vessel mean diameters measured by MRI before and during adenosine infusion are listed in Table 2⇓. Both magnitude and velocity images depicted changes related to increases in coronary flow during adenosine infusion. The magnitude images were observed to have increased signal in the vessels. Although coronary flow velocities were observed to increase in the LAD velocity images (Fig 1⇓), a more noticeable effect was visualized in the velocity images of the LCx during adenosine infusion (Fig 2⇓).
The mean velocity between frames was 1.8±2.24 cm/s due to bulk motion in the slice select direction of the velocity image. The mean number of pixels per vessel was 12.5±7.2. The frequency distribution of the number of pixels per vessel was positively skewed. In the LCx, both vessel size and flow velocity increased significantly during adenosine infusion, whereas in the LAD, only flow velocity increased significantly. MRI-estimated flow values in both the LAD and LCx were similar to US-measured flows, during both baseline and adenosine infusion conditions (Table 2⇑).
Linear regression demonstrated the correlation of MRI and US flow measurements (Fig 4⇓) made at the same time. The mean difference between the coronary arterial flow values obtained using US and MRI was small, −1.2±12 mL/min. The limits of agreement, defined as ±2 SD from the mean difference, were +22.7 and −25.1 mL/min. If a normal distribution is assumed, 95% of the differences between the two measurement methods would fall within this range (Fig 5⇓).
Measured coronary arterial flow ranges were 15 to 178 mL/min for MRI and 17 to 180 mL/min for US. The reproducibility coefficients, calculated from successive measurements, were 16.8 mL/min for US and 26.4 mL/min for MRI.
Coronary Flow Reserve Measurements
The mean predicted flow reserve in the LAD arteries was 1.38±0.31 compared with 1.42±0.31 as measured by US. The mean predicted flow reserve in the LCx was 2.57±0.92, and the US-measured LCx flow reserve was 2.55±0.77. Regression of the MRI and US flow reserve values produced a correlation coefficient (r=.94, intercept=0.10, and slope=1.04; Fig 6⇓).
The mean difference between the two methods of coronary flow reserve measurement was 0.01. The limits of agreement between the US- and MRI-predicted flow reserve measurements were 0.64 to −0.61. Both measurement methods indicated a significant difference between the flow reserves of the LAD and LCx (US, P=.0017; MRI, P=.0107).
The ability of MRI coronary flow measurements to assess stenosis severity is important for diagnosing and managing patients with coronary artery disease. Absolute coronary flow reserve varies with aortic pressure and heart rate, independent of stenosis geometry, due to their differential pressure effects on resting and maximal coronary flow.2 Other ancillary factors affecting coronary arterial blood flow include metabolic demand and vasomotor tone. Relative maximum flow, defined as the maximum flow of the stenotic artery divided by the maximum flow of a nonstenotic artery, reflects integrated stenosis severity relatively independent of physiological conditions.9 Methods commonly used to measure coronary flow are not able to distinguish between an elevation in prevasodilation flow and a reduction in flow during vasodilation. Thus, a technique capable of measuring absolute flow rates for both cases should be especially beneficial for assessing coronary stenosis.19 The breath-hold MRI velocity imaging method reported in the present study is well suited to measure both absolute coronary flow and relative coronary flow reserve, thus reducing potential problems in the interpretation of flow test results.
The results of the present study demonstrate that despite the small size of the coronary vessels and the gross heart motion, it is possible to measure absolute flow in the coronary arteries using breath-hold MRI phase-contrast images. MRI flow measurements were in good agreement with perivascular US flowmeter measurements. Absolute coronary flow and flow reserve measurements obtained using MRI were sufficiently accurate to discriminate between normal arteries and arteries with simulated subcritical stenoses.
The data presented offer the first direct, quantitatively validated measurements of coronary flow reserve in individual coronary vessels using MRI phase-difference velocity imaging. Improved MRI velocity image spatial resolution has allowed measurement of coronary blood flow in a specific experimental setting. Tang et al20 demonstrated in phantoms that the error in measured volume flow rate will be <10% if the ratio between in-plane pixel dimension and vessel radius is <0.5. The spatial resolutions of the images obtained in the present study are at the margin of the above criterion. Diameter measurements, as reported in Table 2⇑, were determined from the number of pixels across a vessel. These pixel sizes, as stated in “Methods,” ranged from 0.7 to 0.94 mm and are known to have good accuracy. Thus, we report a vessel measured to be three 0.7-mm pixels in diameter as 2.1 mm. The large dogs in the present study had mean coronary artery cross-sectional areas of 9.0 mm2 for the LAD and 6.6 mm2 for the LCx. These measurements correspond to the range of the mean sizes of proximal coronary vessels in men 55 to 74 years old.21
Tang et al20 also noted that the error observed in conditions of laminar flow is much less than that seen in plug flow. Laser and high-frequency US Doppler velocimetric studies of normal coronary arteries in dogs22 and human patients23 have demonstrated that the velocity profile in the proximal coronary arteries, while not strictly laminar, has many features resembling laminar flow. Poststenotic flow, however, has been demonstrated to exhibit flow reversal and irregularities in the flow velocity profile.24 Previous MRI phantom studies in open and stenotic tubes have demonstrated a decrease in MRI signal25 and disruption of the phase/velocity relationship26 ; however, these effects have been observed only at Reynolds numbers much higher than those of the coronary arteries studied in the present study.
Negative flow velocities were measured in the MRI velocity images of the constricted LAD during the first phase of the cardiac cycle (early systole) in four of the eight animals in the present study. The negative excursions were not large in magnitude and were applied uniformly with the other measurements in determining the total coronary flows. The present study was not designed to evaluate the phasic variation of coronary flow over the cardiac cycle using MRI methods. However, phasic variation in the MRI flow measurement was routinely observed and was found to be similar to those recorded in flowmeter measurements.
Other approaches have been taken to perform measurements of absolute coronary flow and coronary flow reserve with MRI. Global coronary flow reserve in healthy subjects and in patients has been measured indirectly by phase-contrast MRI velocity imaging using a model that considers the antegrade and retrograde flows in the aorta during systole and diastole.27 This method has not been evaluated for absolute accuracy, and the global flow reserve is of limited usefulness.
Echo-planar MR images (EPI), which can be produced in ≤100 milliseconds, have spatial resolutions too poor for measuring arterial lumen cross-sectional areas. Recently, investigators have reported a breath-hold time-of-flight method using EPI to measure coronary flow velocities in healthy subjectsts.28 However, the pixel size reported was 1.5×3.0 mm, and the model used in this method relied on the assumptions that the vessel was straight and that flow in the coronary artery was laminar. Nevertheless, EPI methods are still under development and may be capable of producing accurate measurements of absolute coronary flow in the future. Other methods, especially radiographic angiography, intra-arterial Doppler US measurements, and nuclear medicine, have been used to assess coronary flow reserve. The catheterization methods offer the present gold standard but are invasive. Our coronary flow reserve results compare favorably with those obtained in a similar animal model using quantitative radiographic angiography correlated with perivascular Doppler US.29
Doppler wires are now being used to measure flow velocities in vessels <2.5 mm in diameter, and they offer the potential for evaluating coronary flow reserve in arterial branches. This capability is of interest for determining flow reserve in small vessels with diffuse and multiple stenoses and in helping to understand the details of the role of collateral flow in coronary artery disease. Improvements in MRI gradient coil performance and methods for enhancing signal-to-noise ratio, which are currently being proposed by MRI equipment vendors, will lead to improved temporal and spatial resolution of MRI phase-contrast velocity imaging, allowing evaluation of the distal coronary vasculature.
The present study had four methodological limitations. First, the conditions under which the MRI flow velocity measurements were performed were not the most favorable for demonstrating the ability of the method to measure coronary flow reserve. The baseline flow measured by MRI was relatively high due to the vasodilatory effects of the isoflurane anesthesia. It is also unlikely that the degree of adenosine-induced vasodilation was maximum in some cases.
Second, the coronary anatomy of the dog differs significantly from that of humans, in that the LAD and LCx supply the vast majority of blood to the canine heart. These vessels were isolated at surgery, and perivascular US flowmeters, which produced signal voids in the MRI images, were placed around the vessels. Thus, the coronary arteries were relatively easy to locate due to the presence of the US probes. Also, the MRI measurements were not performed at the most optimal locations on the arteries because the US flow probes were placed at these sites.
The third limitation of the present study was the length of time required to carefully position each arterial flow measurement—approximately 30 minutes. Holding still in the magnet for this length of time may be prohibitive for some patients. The length of the coronary artery in which the flow measurement is to take place is studied from at least two oblique orthogonal cine views during the pilot scan and set-up phase of the study, as described in “Methods.” The flow image slice is then positioned so that the coronary artery will appear in the center of the velocity image field of view. Careful positioning of the flow velocity image slice was found to be the greatest technical challenge to a high-quality breath-hold PEG study. It is extremely important to measure flow in a straight segment of the artery and to ensure that there is limited through-plane motion of the artery.
The coronary arteries in general exhibit complex curvature and are moving; therefore, one would have difficulty viewing them in other than multiple, parallel segments regardless of the method of flow measurement being used. Acquisition of phase-contrast velocity measurements with the vessel in the plane of the image has been studied by Kraft et al,30 who demonstrated that partial volume effects due to static tissue on either side of the vessel cause severe phase errors. They recommended that small vessels be imaged transversely, as was done in the present study.
The fourth limitation is the length of the breath-hold period. In healthy, sedated dogs, a breath-hold of 30 to 40 seconds presents no problem. In a previous study, patients with known myocardial infarction have been able to hold their breaths repeatedly for as long as 30 seconds.31 However, many patients do not have this capability, especially those with pulmonary complications. Two possible approaches to reducing the length of the breath-hold include decreasing TR using a very high-performance gradient system and reducing the number of frames obtained per cardiac cycle. However, reducing the number of frames significantly may create other problems because these vessels are much easier to perceive in a cine-loop presentation. The magnitude and velocity images can be projected side-by-side, and the cyclic motion of the vessels is better understood. The flow character, in terms of both the gray scale in the velocity image and the in-flow effects in the magnitude images, can also be more easily evaluated.
Investigators have previously reported problems with inadequate breath-holding or patient movement during a breath-hold in real-time cardiac imaging,32 and this was observed in the present study in one inadequately anesthetized animal. Patient motion causes ghosting in the phase-encode direction of MRI images and is a concern for the clinical application of MRI flow measurement methods in conscious patients undertaking voluntary breath-holds. This is another argument for improving the method to make the breath-holding durations as short as possible.
The results of the present study suggest that cine breath-hold MRI phase-contrast velocity images can be used for the accurate, noninvasive measurement of absolute flow and flow reserve in the major epicardial coronary arteries. Also, these data suggest that MRI can be used to assess the severity of a known stenosis.
This work was supported by Grant-in-Aid 92G-091 from the American Heart Association, Texas Affiliate, Inc, and by a National Institutes of Health Special Center of Research grant (Ischemic SCOR grant HL-17669). Adenosine used for this project was kindly supplied by MEDCO Research Inc, Research Triangle Park, NC.
- Received August 2, 1994.
- Revision received November 22, 1994.
- Accepted November 26, 1994.
- Copyright © 1995 by American Heart Association
Hoffman JIE. Maximal coronary flow and the concept of coronary vascular reserve. Circulation. 1984;70:153-159.
Edelman RR, Manning WJ, Burnstein D, Paulin S. Coronary arteries: breath-hold MR angiography. Radiology. 1991;81:641-643.
Li H-F, Clarke GD, Peshock RP, Lui H, NessAiver M. MRI k-space segmentation using phase encode groupings: the accuracy of quantitative measurements of pulsatile flow. Med Phys. 1995;22.
MacGibbon D, Drost CJ. Accuracy of Perivascular Volume Flowsensors. Ithaca, NY: Transonic Systems, Inc; 1992, technical note 11.
NessAiver M: Breath-hold cines using phase encode grouping. In: Pohost GM, ed. Cardiovascular Applications of Magnetic Resonance. Mount Kisco, NY: Futura Publishing; 1993:33-145.
Morich MA. MRI self-shielded gradient coils. US Patent 5,296,810, assigned March 22, 1994.
Richards KL, Dole WP, Hartley CJ, Cannon S, Bishop VS. Doppler coronary artery flow measurements in conscious man. In: Mates RE, Nerem RM, Stein PD, eds. Mechanics of the Coronary Circulation. New York, NY: The American Society of Mechanical Engineers; 1983:63-65.
Tomoike H, Ootsubo H, Sakai K, Kikuchi Y, Nakamura M. Continuous measurement of coronary artery diameter in situ. Am J Physiol. 1981;240:H73-H79.
Microsoft Excel Version 5 User’s Guide. Redmond, Wash: Microsoft Corporation; 1993:589-606.
Klocke FJ. Measurements of coronary flow reserve: defining pathophysiology versus making decisions about patient care. Circulation. 1987;73:562-571.
Tang C, Blatter DD, Parker DL. Accuracy of phase-contrast flow measurements in the presence of partial-volume effects. Magn Reson Med. 1993;3:377-385.
Leung W-H, Stadius ML, Alderman EL. Determinants of normal coronary artery dimensions in humans. Circulation. 1991;84:2294-2306.
Kajiya F. Evaluation of point blood flow velocity in canine coronary artery and vein by laser Doppler method. In: Mates RE, Nerem RM, Stein PD, eds. Mechanics of the Coronary Circulation. New York, NY: The American Society of Mechanical Engineers; 1983:55-58.
Kajiya F, Ogasawara Y, Tsujioka K, Kakai M, Goto M, Wada Y, Tadaoka S-I, Matsuoka S, Mito K, Fujiwara T. Evaluation of human coronary blood flow with an 80-channel 20-MHz pulsed Doppler velocimeter and zero-cross and Fourier transform methods during cardiac surgery. Circulation. 1986;74(suppl III):III-53-III-60.
Kajiya F, Tsujioka K, Ogasawara Y, Wada Y, Matsuoka S, Kanazawa S, Hiramatsu O, Tadaoka S-I, Goto M, Fujiwara T. Analysis of flow characteristics in poststenotic regions of the human coronary artery during bypass graft surgery. Circulation. 1987;76:1092-1100.
Bogren HG, Buonocore MH. Measurement of coronary artery flow reserve by magnetic resonance velocity mapping in the aorta. Lancet. 1993;341:899-900.
Kirkeeide RL, Gould KL, Parsel P. Assessment of coronary stenoses by myocardial perfusion imaging during pharmacologic coronary vasodilation, VII: validation of coronary flow reserve as a single integrated functional measure of stenosis severity reflecting all its geometric dimensions. J Am Coll Cardiol. 1986;7:103-113.
Hundley WG, Clarke GD, Landau C, Lange RA, Willard JE, Hillis LD, Peshock RM. Noninvasive determination of infarct artery patency by cine magnetic resonance angiography. Circulation. 1995;91:1347-1353.
Nissen SE, Gurley JC. Assessment of the functional significance of coronary stenoses: is digital angiography the answer? Circulation. 1990;81:1431-1435.