Normal Three-Dimensional Pulmonary Artery Flow Determined by Phase Contrast Magnetic Resonance Imaging

Transcription

1 Annals of Biomedical Engineering, Vol. 26, pp , /98 $ Printed in the USA. All rights reserved. Copyright 1998 Biomedical Engineering Society Normal Three-Dimensional Pulmonary Artery Flow Determined by Phase Contrast Magnetic Resonance Imaging VICTORIA L. MORGAN,* ROBERT J. ROSELLI, and CHRISTINE H. LORENZ *Department of Radiology and Radiological Sciences and Department of Biomedical Engineering, Vanderbilt University, Nashville, TN and Cardiovascular Division, Barnes Jewish Hospital at Washington University Medical Center, St. Louis, MO (Received 2 April 1997; accepted 22 December 1997) Abstract In this study, an application was developed to measure three-dimensional blood flow in the main, right, and left pulmonary arteries of seven healthy volunteers using phase contrast magnetic resonance imaging MRI. Presently, no other noninvasive technique is capable of providing this information. Flow, mean velocity, kinetic energy, and crosssectional area were measured at multiple phases of the cardiac cycle and were consistent with previously reported values measured with one-dimensional velocity encoded MRI and Doppler echocardiography. Additionally, axial, circumferential, and radial shear stresses near the wall of the vessel at multiple phases of the cardiac cycle were estimated using the in-plane velocities. All three shear stresses were relatively constant along the vessel wall and throughout the cardiac cycle ( 7 dyn/cm 2 ). This three-dimensional characterization of normal pulmonary blood flow provides a base line to which effects of altered pulmonary artery flow patterns in disease can be compared. Morgan, V. L., T. P. Graham, Jr., and C. H. Lorenz. Circulation Suppl. 94:I 417 abstract, Biomedical Engineering Society. S Keywords Shear stress, Blood velocity, Kinetic energy. INTRODUCTION Although pulmonary artery flow can be measured invasively with cardiac catheterization and noninvasively using echocardiography, neither of these techniques gives a full three-dimensional description of the flow characteristics. The directional patterns of the flow, in addition to its magnitude, can be important in understanding phenomena such as thrombus formation, atherosclerosis, and changes in endothelial cell structure or function. For example, in patients who have undergone the Fontan procedure for univentricular heart or tricuspid atresia, changes in blood flow paths cause alterations in three-dimensional 3D flow patterns in the pulmonary arteries. 2,10,13,25 These alterations may result in energy losses. To better understand the implications of these Address correspondence to Christine H. Lorenz, Ph.D., Center for Cardiovascular Magnetic Resonance, Cardiovascular Division, Barnes Jewish Hospital at Washington University Medical Center, St. Louis, MO Electronic mail: chl@ccmr.wustl.edu 557 differences, the physiologic range of flow characteristics in the pulmonary arteries must be first evaluated. Thus, the purpose of this paper was to characterize the threedimensional blood flow in the main, right, and left pulmonary artery in healthy volunteers as a first step towards understanding the effects of altered pulmonary flow patterns in disease. MATERIALS AND METHODS Image Acquisition In this study phase contrast magnetic resonance imaging MRI was used to determine three-dimensional velocities in the main pulmonary artery and left and right pulmonary artery of seven normal volunteers 4 M male, 3F female, yrs. Table 1 describes the study population. Phase contrast MRI utilizes the phase of the MR signal rather than the magnitude of the signal used in conventional MRI. The phase shift of the MR signal is induced by moving tissue and this shift can be used to create images where pixel intensity is proportional to velocity in each pixel in a single selected direction x, y, or z. 15,16,19 This technique can be expanded to threedimensional velocity encoding by measuring onedimensional velocity along three perpendicular axes in the same imaging slice. Walker et al. 26 and Kim et al. 8 have applied this technique to the measurement of threedimensional blood flow in the normal left ventricle. The images were acquired using the body coil of a 1.5 T Siemens Magnetom SP 4000 scanner Siemens Medical Systems, Inc., Iselin, NJ. The imaging protocol began with three sagittal images through the chest to locate the heart. Next, 7 mm thick T1-weighted spin-echo transverse images were acquired from the top of the aortic arch to the apex of the heart. The main pulmonary artery and its bifurcation into the right and left pulmonary arteries were identified in a transverse image. An additional scout image was acquired along the axis of the

2 558 MORGAN, ROSELLI, and LORENZ TABLE 1. Study population. Note: yrs years, bsa body surface area, and avg HR average heart rate. Subject Gender Age (yrs) Height (in) Weight (lbs) bsa (m 2 avg HR 1 M M M F F F M main pulmonary artery in the transverse image to create a sagittal image showing the length of the main pulmonary artery Fig. 1. From this image, a velocity encoding imaging slice was positioned perpendicular to the axis of the main pulmonary artery Fig. 2. To locate the position for the velocity encoded imaging slices in the left and right pulmonary arteries, the vessels were identified in the transverse image and the imaging slice was positioned perpendicular to the vessel immediately distal to the bifurcation of the main pulmonary artery. Test scans run on normal volunteers prior to this experiment showed aliasing in images acquired using a maximum z velocity of 150 cm/s and maximum x and y velocities of 50 cm/s. Therefore, the maximum velocities chosen for this study were 200 cm/s in the z direction and 75 cm/s in the x and y directions. No aliasing was seen using these encoding velocities in any subjects. Other pulse sequence parameters for velocity encoded cine imaging included the following: FOV 300 mm, slice thickness 10 mm, matrix size , TE 10 ms, FIGURE 2. Scout image created from imaging slice shown in Fig. 1. The lines represent imaging slices perpendicular to the axis of the main pulmonary artery. The center slice position was used for velocity encoding. TR 66 ms, number of averaged acquisitions 1, and number of phase encoding lines per segment 3. These pulse sequences resulted in approximately 12 images of the cross section of each vessel at successive time points through the cardiac cycle cardiac phases. The time required to acquire data to create a single image is one TR interval per heartbeat. Therefore, the maximum number of images acquired during the cardiac cycle is equal to the subject s R R interval divided by the TR. The cardiac phase is then identified by the percent of the R R interval at the end of this TR interval. The spatial resolution was 1.2 mm 2. Axial Velocity Measurements FIGURE 1. Axial T1-weighted image showing the main pulmonary artery and its bifurcation into the left and right pulmonary artery of a normal volunteer. The line represents a scout imaging slice positioned along the axis of the main pulmonary artery. A three-dimensional velocity vector was calculated for each pixel in the vessel cross section at all the imaged time points through the cardiac cycle using methods developed previously. 14 To calculate flow at each phase of the cardiac cycle, the average axial velocity in the cross section cm/s was multiplied by the cross-sectional area (cm 2 ). These flows were plotted versus time. The total flow over one cardiac cycle in cc/min was defined as the area under the flow curve multiplied by the heart

3 Normal 3D Pulmonary Artery Flow 559 In-plane Velocity Measurements Wall Shear Stress Calculations FIGURE 3. A schematic of the vessel showing a single pixel and its axial velocity v z, circumferential velocity v, radial velocity v r, the axial shear stress rz, the circumferential shear stress r, and the radial shear stress rr. rate in min 1. Similarly, the percent of reverse flow was calculated for each vessel by calculating flow using only the negative velocities and dividing this flow by the total flow and multiplying by 100%. The highest average velocity over the cross section through the cardiac cycle was identified as the peak velocity. Therefore, assuming a parabolic flow profile, this average peak velocity is approximately half of the peak velocity in a single pixel. The total flow, peak flow, peak velocity, mean velocity, diameter, and cross-sectional area over the cardiac cycle were determined for each subject for each vessel. The distensibility of the vessel was calculated as the difference between the maximum cross-sectional area of the vessel and the minimum cross-sectional area of the vessel divided by the maximum cross-sectional area of the vessel multiplied by 100%. 18 The percent right pulmonary artery RPA flow was calculated as the flow in the RPA divided by the sum of the left pulmonary artery LPA and RPA flow multiplied by 100%. The volume of the blood in the cross section was estimated as the average cross-sectional area multiplied by 10 mm thickness of voxel. The mass of the blood was calculated as the volume of the blood in the cross section multiplied by the density of blood ( 1 g/cc). An estimate of the kinetic energy of the flow in each vessel at each phase of the cardiac cycle was determined by multiplying 1/2 by the mass of the blood by the average axial velocity squared. 27 The total flow, peak flow, peak velocity, mean velocity, and cross-sectional area were compared between vessels using a one-way analysis of variance ANOVA with Bonferroni subtest where applicable. The flow, velocity, and kinetic energy over the cardiac cycle were compared between vessels using the multifactoral ANOVA test. Where statistical significance was determined, the data were compared using the one-way ANOVA with Bonferroni subtest. Wall shear stresses are created by blood velocity gradients at the wall of the vessel and are measured in units of force per unit area. High shear stresses have been associated with changes in endothelial cell function 4,20 and structure 5,20,29 as well as dilation of the vessel by increasing the release of vasodilators and inhibiting the release of vasoconstrictors. 11,12,17,20,21,24 Low shear stresses are associated with intimal thickening and development of atherosclerosis. 22,28 Three types of shear stresses at the wall were examined in this study. These are the axial shear stress, rz, which opposes flow in the axial direction; the circumferential shear stress, r, which opposes circumferential flow in the cross section; FIGURE 4. Axial velocity across the cross section in the main pulmonary artery of a normal volunteer at peak systole a, end systole b, and diastole c measured using threedimensional phase contrast MRI.

4 560 MORGAN, ROSELLI, and LORENZ FIGURE 5. Axial velocity across the cross section in the right pulmonary artery of a normal volunteer at peak systole a, end systole b, and diastole c measured using threedimensional phase contrast MRI. FIGURE 6. Axial velocity across the cross section in the left pulmonary artery of a normal volunteer at peak systole a, end systole b, and diastole c measured using threedimensional phase contrast MRI. and radial shear stress, rr, which opposes flow in the radial direction. These shear stresses were measured at each pixel at the wall of the vessel using the following equations: 6 rz v z r v r z, r r r v r 1 r v r, rr 2 v r r 2 3 v, where is the coefficient of viscosity of the blood 0.03 g/cm s, v z is the axial velocity, v r is the radial velocity, v is the circumferential velocity, r is the distance between the pixel and the center of the vessel, and is the angle between the horizontal plane through the center of the cross section and the radius of the pixel as shown in Fig. 3. Using matrices of rotation, 3 the values for v r and v can be computed from v x and v y. To calculate the ratios of the form v/ r near the wall of the vessel, the pixel closest to the wall and its neighboring pixel closest to the center of the vessel were identified. The ratio was estimated as the difference in their velocities divided by the difference in their radii. Using the pixel closest to the wall and its neighboring pixel in a counterclockwise rotation, the expressions in the form v/ were estimated in a similar manner. The v r / z term in Eq. 1 was assumed to be negligible with respect to v z / r, especially near the wall. It could not be calculated with only one imaging slice acquired

5 Normal 3D Pulmonary Artery Flow 561 RESULTS Axial Velocity Measurements The axial velocity at each pixel in the cross section is graphed as a mesh at systole, end systole, and diastole in the main pulmonary artery MPA, RPA, and the LPA of one subject in Figs. 4, 5, and 6. In all three vessels the velocity is highest during systole and lowest in diastole as expected. In the MPA, the axial profile was nearly flat with a smaller magnitude in the inferior region of the vessel. Similarly, in the RPA, the axial profile was also FIGURE 7. Flow in main pulmonary artery a, right pulmonary artery b, and left pulmonary artery c of seven normal volunteers measured using 3D velocity encoded MRI only six subjects MPAs were scanned due to time constraints. per vessel. By assuming blood to be an incompressible fluid, the expression ( v) in Eq. 3 was assumed to be zero. Both the absolute value of the shear stress and the shear stress without the absolute value of each pixel at the wall was averaged over the entire wall length and this mean was plotted versus the percent R R interval. The absolute value of the shear stress was used to show the magnitude of shear acting on the vessel without regard to its direction. Similarly, the shear stress without absolute value measurements were used to determine relative dispersion standard deviation/mean of the shear stress along the vessel wall. This is a measurement of how much the shear stress changed along the vessel wall length. The plots of percent R R versus the absolute values of the axial, circumferential, and radial shear stress were compared between vessels using the multifactoral ANOVA test. Where statistical significance was determined, the one-way ANOVA with Bonferroni subtest was used. FIGURE 8. The average kinetic energy in seven normal volunteers throughout the cardiac cycle in the main pulmonary artery a, right pulmonary artery b, and left pulmonary artery c measured using velocity encoded MRI. The error bars represent plus or minus one standard deviation.

6 562 MORGAN, ROSELLI, and LORENZ TABLE 2. Axial flow and velocity parameters in the main, right, and left pulmonary arteries of normal volunteers determined with phase contrast cine MRI. Total flow over one cardiac cycle (L/min) Regurgitant flow (%) Peak flow (L/min) Peak velocity (cm/s) Mean velocity (cm/s) Diameter (cm) Crosssectional area (cm 2 Distensibility (%) MPA * R ** NS NS ** RPA LPA **Statistically different than RPA and LPA (p 0.05). * R Statistically different than RPA (p 0.05). NS Not statistically different from RPA or LPA. smaller along the inferior wall, while in the LPA the axial velocity was smaller in the right inferior region of the vessel. A plot of the flow versus percent R R interval for each vessel is given in Fig. 7. The systolic peak in flow at 20% 30% of the R R interval was evident in most vessels evaluated. The 20% R R phase was statistically the highest in flow (p 0.05) over all vessels. Reverse flow is indicated by negative values. Figure 8 shows the kinetic energy of each vessel through the cardiac cycle. Its peak correlates with the peak flow at the 20% R R phase and, like velocity, is not statistically different between vessels. Table 2 gives the results of the axial velocity and flow parameters for each vessel. The parameters reaching statistical significance are indicated. Table 2 indicates that the total and peak flow in the MPA is approximately twice that of the RPA and LPA. Also, the total flow in the RPA is statistically lower than that of the MPA (p 0.05) as anticipated. The total flow in the LPA was not statistically different from the total flow in the MPA or the RPA due to the large variations between subjects. The regurgitant flow in the MPA and RPA are similar and are much greater and higher in standard deviation than those in the LPA. Similar to the total flow, the peak flow through the LPA and RPA is approximately half of that of the MPA. In-plane Velocity Measurements Shear Stress Calculations Figure 9 shows typical in-plane velocity vectors in a cross section of each vessel. Ordered, coherent flow with little or no swirling was observed in the in-plane velocity vectors in all three vessels. Very minimal disordered, random flow was seen in any vessel at any time. The mean absolute value of axial shear stress at the wall over the cardiac cycle for each vessel is plotted in Fig. 10. The statistical analysis showed that the axial shear stress was significantly higher (p 0.05) in the MPA than the LPA 7.56 vs 5.27 dyn/cm 2, but that axial shear in the RPA was not different from either vessel. Axial shear stress was not statistically different throughout the cardiac cycle. The mean absolute value of circumferential shear stress at the wall over the cardiac cycle for each vessel is plotted in Fig. 11. There was no significant difference in the circumferential shear stress between the vessels or between the time points in the cardiac cycle. The mean absolute value of the radial shear stress at the wall over the cardiac cycle is shown in Fig. 12. In the MPA, the standard deviation of the radial shear stress between subjects increases greatly in diastole. There was no statistical difference in radial shear stress between vessels and the radial shear stress was constant throughout the cardiac cycle. The relative dispersion, a measure of variation in a parameter, of the axial shear stress along the wall of the vessel was less than 10% in all subjects in the MPA, less that 15% in all subjects in the RPA, and less than 6% in all subjects in the LPA. The relative dispersion of the circumferential shear stress along the wall of the vessel was less than 11% in all subjects in the MPA, less than 20% in all subjects in the RPA, and less than 30% in all subjects in the LPA. The relative dispersion of the radial shear stress along the wall of the vessel was less than 20% in six of the seven subjects in all vessels but was 74% in the seventh subject. DISCUSSION Comparison with Other Studies The axial velocity profiles in the MPA showed smaller velocities along the inferior wall of the vessel. This is similar to findings by Sloth et al. 23 who reported lower axial velocities in the right inferior region of the MPA during systole using MRI. Our findings also showed lower velocities in the inferior region of the RPA and the right inferior regions of the LPA. The flow versus percent R R graphs show the systolic peak in all three vessels 20% of the R R interval and the end of systole at approximately 40% of the R R

7 Normal 3D Pulmonary Artery Flow 563 FIGURE 10. Average axial shear stress averaged over the vessel wall through the cardiac cycle in the main pulmonary artery a, right pulmonary artery b, and left pulmonary artery c of seven normal volunteers. The error bars represent plus or minus one standard deviation. FIGURE 9. Typical in-plane velocity vectors obtained at 30% of the cardiac cycle in systole in a cross section of the main pulmonary artery a, right pulmonary artery b, and left pulmonary artery c of normal volunteers measured using 3D phase contrast MRI. The length of the vector represents the magnitude of the velocity and the direction of the vector represents the direction of the velocity. Each point represents a 0.6 mm 2 pixel. interval. The total flow through the MPA L/ min fell within ranges reported by Paz et al L/min, Sloth et al L/min, and Kondo et al ml stroke volume using velocity encoded MRI. The total flow in this study through the RPA and LPA L/min RPA, L/min LPA also fell within ranges reported by Paz et al L/ min RPA, L/min LPA. However, the standard deviation of the measurements in this study were greater than those calculated by previous researchers. This higher standard deviation may be responsible for the discrepancy between previously published values of relative right and left pulmonary blood flow approximately 55% of flow to the right lung and our value

8 564 MORGAN, ROSELLI, and LORENZ FIGURE 11. Average circumferential shear stress averaged over the vessel wall through the cardiac cycle in the main pulmonary artery a, right pulmonary artery b, and left pulmonary artery c of seven normal volunteers. The error bars represent plus or minus one standard deviation. FIGURE 12. Average radial shear stress averaged over the vessel wall through the cardiac cycle in the main pulmonary artery a, right pulmonary artery b, and left pulmonary artery c of seven normal volunteers. The error bars represent plus or minus one standard deviation. approximately 40% 47% L/min right pulmonary artery flow divided by L/min MPA flow or L/min divided by approximately 5.2 L/min mean LPA RPA flow. Some reasons for this may be the longer imaging time required for the threedimensional velocity encoding in this study which may result in more variation in heart rate throughout the exam and the lower number of subjects in this study compared to Paz et al subjects, Sloth et al subjects, and Kondo et al subjects. In addition, Sloth et al. 23 used respiratory gating as well as cardiac triggering in their study. The percent reverse flow found by Bogren et al. 1 was 2% and Sloth et al. 23 was 3.6% 2.0% in the MPA using MRI. These values are lower than the percent reverse flow calculated in this study in the MPA and the RPA and may be due to residual eddy current induced velocity offsets shifting the zero flow base line. There was a L/min difference between the flow in the MPA and the sum of the flow in the RPA and the LPA. These errors may be caused in part by not selecting an imaging plane perpendicular to the axis of the vessel. In this case, some of the axial flow may be measured as in-plane flow, thus reducing axial flow and increasing in-plane flow. Further, in imaging the main pulmonary artery, an additional scout was acquired to ensure that the velocity encoding image was perpendicular to the vessel in both the transverse and sagittal planes. In imaging the left and right pulmonary arteries, the velocity encoding slice was positioned perpendicular to the vessel in the transverse image only. It is possible that the vessel may have been curved through the plane of the transverse image without detection. This could cause more error in the measurement of the left and right pulmonary artery flow than in the main pulmonary artery

9 Normal 3D Pulmonary Artery Flow 565 flow. Also, error in determining vessel cross-sectional area and different eddy current behavior in the different imaging planes may have created some error in these measurements. The diameters of the vessels measured in this study MPA diam cm, RPA diam cm, LPA diam cm correlate well with values reported by Paz et al. 18 MPA diam cm systole, cm diastole, RPA diam cm systole, cm diastole, LPA diam cm systole, cm diastole. The average peak distensibility of all three vessels MPA 35% 13%, RPA 41% 14%, LPA 36% 11% is somewhat higher than that reported by Paz et al. 18 MPA 25.6% 10.7%, RPA 21.4% 10.7%, LPA 24.5% 7.8% and Bogren et al. using MRI 1 MPA 23%, LPA 28%. The highest discrepancy in these measurements was found in the RPA which had the smallest cross-sectional area of the three vessels studied. This difference may be due in part to the manual identification of the vessel cross section in the images. To minimize these errors, the images were magnified to allow for more precise manual definition of the vessel wall, and the regions of interest were identified on the magnitude images instead of the phase images. Shear Stress Measurements The in-plane velocity vector plots in Fig. 9 indicate ordered, coherent flow with minimal or no swirling in all vessels. In addition, shear stress is relatively constant throughout the cardiac cycle and axial, circumferential, and radial shear stress all averaged approximately 7 dyn/cm 2 in magnitude. The slightly higher axial shear stresses found in the MPA may be due to its higher flow and its proximity to the ventricular pumping chamber. There was no statistical difference between vessels in circumferential or radial shear stress which correlates with the findings of little or no circular swirling in the vessel cross section. The relative dispersion measurements show that the shear stresses did not vary greatly along the vessel wall. These results imply that shear stress at the wall is normally a constant both longitudinally and circumferentially along the wall. These findings provide a basis from which to determine what types of changes in shear stress may affect endothelial function. Both changes in magnitude and variation in shear stress over the cardiac cycle may play a role. Potential errors in the shear stresses, which are calculated at the wall of the vessel, may be due to errors in cross-section identification. However, Hofman et al. found that in a vessel with a diameter of at least three pixels, accurate flows can be measured using MRI. 7 In general, the heterogeneity in the data presented could be attributed to either subject heterogeneity or to systematic errors. Subject heterogeneity includes differences in the hemodynamic state of the subject during scanning and variations in pulmonary artery branching geometry among subjects. It is beyond the scope of this study to quantify these differences and their effects. To quantify the systematic errors, the technique has been validated in vitro in a constant flow phantom resulting in a standard error of 8.62 cm/s within the ranges cm/s. In vivo, previous validation was performed by comparison of one-dimensional main pulmonary flow to right ventricular stroke volume. 14 These validation results suggest that subject heterogeneity may account for between one-third and one-half of the observed heterogeneity in the data. CONCLUSIONS In conclusion, this study provides a characterization of normal main, right, and left pulmonary artery flow in three dimensions using phase contrast MRI for comparison to values measured in various disease states. The advantage of three-dimensional phase contrast MRI over one-dimensional phase contrast MRI and Doppler echocardiography is that it provides the in-plane velocities which allow calculation of shear stresses at the wall of the vessel. There is currently no other method capable of yielding these data in vivo. These shear stresses may be altered in various diseases affecting the pulmonary arteries such as pulmonary hypertension or congenital heart disease where abnormal anatomy results in unusual flow patterns within the vessels. Therefore, a threedimensional characterization can provide additional significant information for evaluation of blood flow under these conditions over one-dimensional blood velocity measurements and provides a noninvasive research tool for in vivo characterization of vascular disease. ACKNOWLEDGMENTS This work was partially supported by Department of Health and Human Services Public Health Service Grant No. HL and the Patricia Roberts Harris Fellowship V.L.M.. REFERENCES 1 Bogren, H. G., R. H. Klipstein, R. H. Mohaiddin, D. N. Firmin, S. R. Underwood, R. S. O. Rees, and D. B. Longmore. Pulmonary artery distensibility and blood flow patterns: A magnetic resonance study of normal subjects and of patients with pulmonary arterial hypertension. Am. Heart J. 118: , Chuen-Neng, L., H. V. Schaff, G. K. Danielson, F. J. Puga, and D. J. Driscoll. Comparison of atriopulmonary versus atrioventricular connections for modified Fontan/Kreutzer repair of tricuspid valve atresia. J. Thorac. Cardiovasc. Surg. 92: , 1986.

10 566 MORGAN, ROSELLI, and LORENZ 3 Craig, J. J. Introduction to Robotics: Mechanics and Controls, 2nd ed. Reading, MA: Addison Wesley, Davies, P. F., C. F. Dewey, Jr., S. R. Bussolari, E. J. Gordon, and M. A. Gimbrone, Jr. Influence of hemodynamic forces on vascular endothelial function. J. Clin. Invest. 73: , Fry, D. L. Acute vascular endothelial changes associated with increased blood velocity gradients. Circ. Res. 22: , Fung, Y. C. Biodynamics Circulation. New York, NY: Springer, Hofman, M. B. M., F. C. Visser, A. C. van Rossum, Q. M. Vink, M. Sprenger, and N. Westerhof. In vivo validation of magnetic resonance blood volume flow measurements with limited spatial resolution in small vessels. Magn. Reson. Med. 33: , Kim, W. Y., P. G. Walker, E. M. Pedersen, J. K. Poulsen, S. Oyre, K. Houlind, and A. P. Yoganathan. Left ventricle blood flow patterns in normal subjects: A quantitative analysis by three-dimensional magnetic resonance velocity mapping. J. Am. Coll. Cardiol. 26: , Kondo, C., G. R. Caputo, R. Semelka, E. Foster, A. Shimakawa, and C. B. Higgins. Right and left ventricular stroke volume measurements with velocity-encoded cine MR imaging: In vitro and in vivo validation. Am. J. Roentgenol. 157:9 16, Low, H. T., Y. T. Chew, and C. N. Lee. Flow studies on atriopulmonary and cavopulmonary connections of the Fontan operations for congenital heart defects. J. Biomed. Eng. 15: , Masuda, H., K. Kawamura, T. Sugiyama, and A. Kamiya. Effects of endothelial denudation in flow-induced arterial dilatation. Front. Med. Bio. Eng. 5:57 62, Melkumyants, A. M., S. A. Balashov, and S. P. Kartamyshev. Anticonstrictor effect of endothelium sensitivity to shear stress. Pflugers Arch. Eur. J. Physiol. 427: , Morgan, V. L., T. P. Graham, Jr., and C. H. Lorenz. Alterations in pulmonary artery flow patterns in Fontan patients determined with three dimensional phase contrast magnetic resonance imaging. Circulation Suppl. 94:I 417 abstract, Morgan, V. L., R. R. Price, and C. H. Lorenz. Application of linear optimization techniques to MRI phase contrast blood flow measurements. Magn. Reson. Imaging 14: , Nayler, G. L., D. N. Firmin, and D. B. Longmore. Blood flow imaging by cine magnetic resonance. J. Comput. Assist. Tomogr. 10: , O Donnell, M. NMR blood flow imaging using multiecho, phase contrast sequences. Med. Phys. 12:59 64, Ohno, M., G. H. Gibbons, V. J. Dzau, and J. P. Cooke. Shear stress elevates endothelial cgmp. Role of a potassium channel and G protein coupling. Circulation 88: , Paz, R., R. H. Mohiaddin, and D. B. Longmore. Magnetic resonance assessment of pulmonary arterial trunk anatomy, flow, pulsatility and distensibility. Euro. Heart J. 14: , Pelc, N. J., R. J. Herfkens, A. Shimakawa, and D. R. Enzmaann. Phase contrast cine magnetic resonance imaging. Magn. Reson. Q. 7: , Rossitti, S., J. Frangos, P. R. Girard, and J. Bevan. Regulation of vasculature tone. Can. J. Physiol. Pharmacol. 73: , Rubanyi, G. M., J. C. Romero, and P. M. Vanhoutte. Flowinduced release of endothelium-derived relaxing factor. Am. J. Physiol. 250:H1145 H1149, Sabbah, H. N., F. Khaja, J. F. Brymer, E. T. Hawkins, and P. D. Stein. Blood velocity in the right coronay artery: Relation to the distribution of atherosclerotic lesions. Am. J. Cardiol. 53: , Sloth, E., K. C. Houlind, S. Oyre, W. Y. Kim, E. M. Pedersen, H. S. Jorgensen, and J. M. Hasenkam. Threedimensional visualization of velocity profiles in the human main pulmonary artery with magnetic resonance phasevelocity mapping. Am. Heart J. 128: , Snow, H. M., S. J. McAuliffe, J. A. Moors, and R. Brownlie. The relationship between blood flow in the iliac artery of the anaesthetized dog: The role of endothelium-derived relaxing factor and shear stress. Exp. Phys. 79: , Stein, D. G., H. Laks, D. C. Drinkwater, L. C. Permut, H. W. Louie, J. M. Pearl, B. L. George, and R. G. Williams. Results of the total cavopulmonary connection in the treatment of patients with a functional single ventricle. J. Thorac. Cardiovasc. Surg. 102: , Walker, P. G., G. B. Cranney, R. Y. Grimes, J. Delatore, J. Rectenwald, G. M. Pohost, and A. P. Yaganathan. Three dimensional reconstruction of the flow in a human left heart by using magnetic resonance phase velocity encoding. Ann. Biomed. Eng. 24: , White, F. M. Fluid Mechanics, 2nd ed. New York, NY: McGraw-Hill, Zarins, C. K., D. P. Giddens, B. K. Bharadvaj, V. S. Sottiurai, R. F. Mabor, and S. Glagov. Carotid bifurcation atherosclerosis quantitative correlation of plaque localization with flow velocity profiles and wall shear stress. Circ. Res. 53: , Ziegler, T., R. W. Alexander, and R. M. Nerem. An endothelial cell-smooth muscle cell model for use in the investigation of flow effects on vascular biology. Ann. Biomed. Eng. 23: , 1995.