Quantitation of renal perfusion using arterial spin labeling with FAIR-UFLARE

Magnetic Resonance Imaging 18 (2000) 641– 647

Quantitation of renal perfusion using arterial spin labeling with FAIR-UFLARE N. Karger*, J. Biederer, S. Lu¨sse, J. Grimm, J.-C. Steffens, M. Heller, C.-C. Glu¨er Klinikum an der CAU zu Kiel, Klinik fu¨r Diagnostische Radiologie, Michaelisstr. 9, 24105 Kiel, Germany Received 3 January 2000; accepted 4 May 2000

Abstract Quantitative perfusion imaging of human kidneys was performed using arterial spin labeling MRI with a fast spin echo readout-sequence. Perfusion maps of centrally located single slices were obtained in axial and coronal orientations. In ten healthy volunteers, the mean value of perfusion was 213 ⫾ 55 mL/(100g min) with a range from 140 to 319 mL/(100g min). These results are in accordance with literature data, considering the fact that FAIR only measures the perfusion component normal to the imaging plane. Intra-individual reproducibility errors of ⫾ 11% were smaller than the natural interindividual variability of renal perfusion (SD ⫽ ⫾ 25%). Perfusion in the cortex was approximately 3– 4 times higher compared to the medulla. Considering the relatively high resolution of 2 ⫻ 2 ⫻ 10 mm3, the ability to quantify perfusion, and the lack of ionizing radiation and contrast media, this technique should prove useful in diagnosing renal pathologies that are associated with reductions in tissue perfusion. © 2000 Elsevier Science Inc. All rights reserved. Keywords: Perfusion; FAIR; Kidney; Arterial spin labeling

1. Introduction Quantitative changes in the tissue perfusion rate can provide relevant information about the function of organs in many pathologic conditions. In the kidneys, not only acute infarction due to vessel stenosis, but also some cases of renal failure, particularly in transplanted kidneys, may be detected early by reduced perfusion [1,2]. A completely noninvasive, repeatable and easily applicable method to detect and quantify renal tissue perfusion alterations, combined with high resolution imaging of the kidneys would thus be a desirable diagnostic tool. In this respect, a promising technique has recently become feasible: perfusion imaging with arterial blood spin labeled magnetic resonance imaging (ASL-MRI) uses the water spins of blood as an endogenous tracer. This method does not need a contrast agent and is therefore completely noninvasive. Also, as results of preliminary studies show, quantification of perfusion is possible with reasonable spatial resolution and precision: ASL-MRI has already been applied to assess tissue perfusion in the brain [3], lung [4],

skeletal muscle [5], heart [6] and placenta [7]. A few studies were focused on perfusion in the human or animal kidney [8 –12]. The first results support the capability of ASL-MRI to quantify renal tissue perfusion, but establishment of reasonable normal values still remains difficult due to large variations of the measured values. These may be due to interindividual variability or technical problems. As a preparation for developing a reference database of the healthy population, and to establish the performance characteristics of ASL-MRI of the kidneys, this study on ten healthy volunteers was designed to optimize and standardize the examination protocol. Specifically, we investigated the intra- and interindividual variability and studied whether absolute levels of perfusion are in agreement with literature data. Lastly we tested whether expected differences in perfusion rates between cortex and medulla could be measured using this technique.

2. Methods 2.1. Theory

* Corresponding author. Tel.: ⫹⫹49-431-597-3088; fax: ⫹⫹49-431597-3127. E-mail address: [email protected] (N. Karger).

For spin labeling we chose the FAIR (Flow-sensitive Alternating Inversion Recovery)-method combined with a

0730-725X/00/$ – see front matter © 2000 Elsevier Science Inc. All rights reserved. PII: S 0 7 3 0 - 7 2 5 X ( 0 0 ) 0 0 1 5 5 - 7

642

N. Karger et al. / Magnetic Resonance Imaging 18 (2000) 641– 647

UFLARE (Ultra Fast Low Angle Rare)-readout sequence. The FAIR-method [13,14] is based on the acquisition of two images: The first image is preceded by a nonselective 180°pulse (global inversion) and taken after a certain inversion time Tinv, during which the spin system returns to equilibrium with the spin-lattice relaxation time T1. The second image is preceded by only selective inversion of the imaging slice, while all other parameters remain unchanged. In the presence of blood flow, noninverted spins enter the imaging slice during Tinv in the case of slice-selective inversion, while there are no noninverted spins that could enter the imaging slice after global inversion. In the subsequent image acquisition, these noninverted spins result in an increased signal after slice-selective inversion compared to global inversion. Subtraction of the image preceded by nonselective inversion from the image taken after selective inversion gives a signal intensity mapping reflecting local differences in spin inflow due to tissue perfusion. The time-evolution of the magnetization after nonselective (Mns (t)) and sliceselective (Mss (t)) inversion is given by: M ns共t兲 ⫽ M 0共1 ⫺ 2 exp共 ⫺ t/T 1兲兲

(1)

M ss共t兲 ⫽ M 0共1 ⫺ 2 exp共 ⫺ t/T ss 1 兲兲

(2)

T1 is the spin-lattice relaxation time and T1ss is the apparent relaxation time after slice selective inversion of the spins, given by [15]: 1 1 f ⫽ ⫹ T ␭ T ss 1 1

(3)

with ␭ the tissue/blood water distribution coefficient (assumed to be 0.8 mL/g10), and the assumption, that the capillary-tissue exchange of blood water is instantaneous. f is the perfusion rate in ml/(g s). The difference image intensity at the inversion time Tinv, ⌬M(Tinv), is then given by: ⌬M共T inv兲 ⫽ M ss ⫺ M ns

冉

⫽ ⫺ 2M 0exp

⫺ T inv T1

冊冉

1 ⫺ exp

冉

⫺ T invf ␭

冊冊 (4)

Since f/␭ is small (f ⬇ 0.05 mL/(g s)), the last term can be expanded, and the equation now reads:

冉

冊

f ⫺ T inv ⌬M共T inv兲 ⫽ ⫺ 2T invM 0 exp ␭ T1

(5)

The perfusion rate f is then given by f⫽

冉 冊

⫺ ⌬M ␭ T inv exp 2T invM 0 T1

(6)

Perfusion maps can thus be calculated pixel-by-pixel from a knowledge of the magnetizations Mns and Mss, the

undisturbed intensity M0, and the spin lattice relaxation time T1. 2.2. Measurements The FAIR-spin tagging sequence was combined with a UFLARE-readout pulse sequence, and implemented on a clinical 1.5 T MRI unit (Siemens Magnetom Vision, Siemens Medizintechnik, Erlangen, Germany). UFLARE is a single-shot RARE sequence, with the refocusing pulses being smaller than 180° (150° in this study), to reduce rf power deposition. Alternate centric phase encoding was employed to maximize the signal-tonoise ratio and to minimize the T2 contrast [16]. Other imaging parameters were: TR 6500 ms, TE 4.5 ms, excitation pulse 90°. The commercially available body array coil was used for all measurements. The imaging slice thickness was set to 10 mm, with an inversion slice thickness of 30 mm, to avoid artifacts from not fully inverted spins at the margins of the readout slice, which would lead to artificially high perfusion values. A matrix of 128 ⫻ 128 was used, with a field of view of 250 –350 mm. For each slice, 12 or 24 image pairs were accumulated to obtain the two averaged FAIR images, with an inversion time Tinv of 1600 ms and a repetition time between images of 6500 ms. The repetition time was chosen to be almost five times the longest spin-lattice relaxation time in kidney (⬇ 1.4 s), to ensure a complete relaxation of the spins between measurements. Tinv was chosen to give the maximum contrast between the two images. T1 -images were recorded with the same pulse sequence, with global inversion of the spins, using ten different Tinv -values between 100 ms and 7000 ms, and accumulating 2 images per Tinv. The total measurement time for one slice, including about 2 min for the T1 -measurement, was 5 or 8 min, depending on the number of accumulations. Ten healthy, adult volunteers were enrolled in the study (3 female, 7 male, age 27–33 years, average age 30.4 years). Subjects with a clinical history of renal malfunction were excluded. Informed consent after detailed explanation of the procedure was obtained from all participants. The imaging protocol included acquisition of two single slice images in perpendicular orientations, positioned in the central part of the kidneys. Axial slices were positioned to transsect both kidneys at the hili and therefore differed from the exact transverse orientation in some subjects. The coronal slices also usually differed from the vertical orientation, as they were tilted to match the longitudinal axis of the kidneys. Fig. 1 gives an example of the slice positioning. During image acquisition the volunteers were asked to breathe evenly, as the measurement time would not allow imaging in breathhold-technique. 2.3. Image processing For calculation of the relaxation maps, the following procedure was used: the inversion time resulting in the

N. Karger et al. / Magnetic Resonance Imaging 18 (2000) 641– 647

643

Fig. 2. Illustration of the procedure to obtain perfusion maps of the kidneys (volunteer 7); images recorded after global (upper left) and selective inversion of the spins (upper right); difference image (center left); T1 -image (center right); quantitative perfusion image (bottom) Fig. 1. Positioning of the measurement-slice (solid rectangle) and sliceselective inversion region (dashed rectangle) on the localizer images for the coronal (top) and axial (bottom) orientations

minimum intensity of the magnitude image was identified for each voxel. All intensities from inversion times below this value were set negative; the minimum intensity was not used in the fitting routine. The relaxation maps were then calculated by fitting the single voxel values to a monoexponential, three-parameter fit curve using a program written in IDL (Research Systems, Boulder, Co). Perfusion images were calculated according to eq. [6]. To better delineate the kidneys from the surrounding tissue in the perfusion images, voxels with a T1 -value of less than 700 ms were set to zero. This did not influence the perfusion evaluation of the kidney, since all areas of the kidney showed T1 -values greater than ca. 800 ms. However, this procedure effectively excludes artifactually high perfusion values of the surrounding tissue, which arise from the strong weighing factor exp(Tinv/T1) for short T1. The resulting

images represent perfusion maps of the kidneys, together with signal from other tissues, including blood vessels. To illustrate the process of obtaining perfusion maps, the various steps of the procedure are shown in Fig. 2. ROIs for the whole kidney were selected by interactively drawing a boundary on the T1 -images, and transferring the ROIs to the perfusion image. Pixels with perfusion values of over 450 mL/(100g min) were thought to arise from blood vessels and were not included in the analysis. This value is to a certain amount arbitrary. However, it is necessary to exclude blood vessels, since their inclusion would offset the perfusion results for kidney tissue. When perfusion measurements of the whole kidney become feasible, procedures can be developed to more accurately exclude larger blood vessels from the perfusion analysis. 3. Results Sets of transverse and coronal perfusion images of 4 volunteers are presented in Fig. 3. The images show signal

644

N. Karger et al. / Magnetic Resonance Imaging 18 (2000) 641– 647

Fig. 3. Coronal and transverse perfusion images of volunteers 1– 4; pixels were set to zero, when they met one or both of the following criteria: T1 relaxation time smaller than 700 ms; the T1 -fitting routine did not converge

tween medulla and cortex; secondly, partial volume effects due to the large slice thickness of 10 mm reduce the perfusion result for the cortex, while contributions of cortical segment walls, and blurring of the images resulting from the UFLARE sequence lead to an apparent increase of the perfusion rate in the medulla. Still, substantial contrast between cortex and segment walls on the one hand and medulla on the other can be observed, with typical values of 200 to 270 mL/(100 g min) for the cortex and 50 to 90 mL/(100g min) for medulla sections, excluding segment walls. The ROI’s were defined interactively, avoiding regions close to the boundary with different tissue types. A more robust value can be calculated for whole kidneys (portion imaged on a given slice). The mean values of the perfusion rates are given in Table 1 (perfusion obviously resulting from macroscopic vessels was not included in the calculations, see above). The mean value ⫾ standard deviation for the perfusion rate of whole kidneys was 213 ⫾ 55 mL/(100 g min). The difference in single slice left and right kidney perfusion was 10% on average (mean absolute value of percentage difference), exceeding 20% in only three of 20 cases. Two kidneys of two different volunteers were overlayed by obvious artifacts and thus were not included in the evaluation. As a test for the reproducibility of this method, 3 volunteers were measured again after a period of 4 weeks. Volunteer 6 underwent a third measurement after another two weeks. The results are included in Table 1. The difference in perfusion for a single kidney between two measurements ranged from 0 to 30%, with an average reproducibility error of 11%. Besides random measurement error, there are two other factors that contribute to this variation: a different positioning of the slices in the 2nd measurement (see below), and possibly naturally varying perfusion rates in the volunteers.

4. Discussion resulting from various organs and vessels. In most images the kidneys are outlined clearly. Apart from the kidneys, some other structures as displayed by their perfusion rates can be seen: the spleen as well as some vertebral vessels can be clearly identified in most coronal images. The liver was filtered out due to its short spin-lattice relaxation time T1 and can not be seen. Bright spots in the images result from macroscopic blood vessels traversing the slice. For a better visibility in the displayed images, the scale of the perfusion values is truncated at a value of 500 mL/(100g min). In the perfusion images, the higher perfusion of the cortex compared to medulla can generally be clearly distinguished. A quantitative differentiation between perfusion in the cortex and medulla is somewhat difficult to accomplish. On the one hand, the perfusion rate changes gradually be-

Using ASL-MRI we have been able to obtain quantitative perfusion maps of the kidney. Image contrast was sufficient to delineate regions with different perfusion rates, especially cortex versus medulla. Dworkin and Brenner [17] report an average value of 400 mL/(100 g min) for the total renal blood flow obtained by clearance measurements, with large interindividual variations, resulting in a standard deviation of approximately ⫾ 25%. The results obtained in the present study are somewhat lower than their mean value. For a direct comparison, however, it has to be kept in mind that the FAIR-method is only sensitive to the perfusion component perpendicular to the imaging slice, so that the values obtained here have to be lower than measurements of the total blood flow into the organ. Moreover, the selective saturation slice always includes some of the main vessels providing blood inflow into

N. Karger et al. / Magnetic Resonance Imaging 18 (2000) 641– 647

645

Table 1 Kidney perfusion rates in healthy volunteers obtained by FAIR-UFLARE volunteer 1 1 2 2 3 3 3, 2nd meas. 4 4 5 5 6 6, 2nd meas. 6, 3rd meas. 6 7 8 9 9 10 10 10, 2nd meas. average ⫾ SD

age [years]

sex

31

f

33

m

32

m

29

m

27

f

31

m

30 31 32

m m m

28

f

slice orientation trans cor trans cor trans cor cor trans cor trans cor trans trans trans cor cor trans trans cor trans cor cor

30 ⫾ 2

left kidney

right kidney

ratio: left/right

175 225 281 144 155 182 180 318 212 — 175 196 210 157 192 146 322 238 — 160 209 160 202 ⫾ 51

178 235 296 135 164 165 194 319 208 283 211 188 194 159 208 155 300 301 351 258 196 230 224 ⫾ 59

0.98 0.96 0.95 1.07 0.95 1.10 0.93 1.00 1.02 — 0.83 1.04 1.08 0.99 0.92 0.94 1.07 0.79 — 0.62 1.07 0.70 0.95 ⫾ 0.13

Perfusion rates f [ml/(100g min)] for whole kidneys (for the slices evaluated) and ratio of left/right kidney perfusion in healthy volunteers; two kidneys were overlayed by artifacts and could not be measured; in some volunteers, only one slice could be quantified, due to errors in the measurements of the relaxation maps (cor ⫽ coronal, trans ⫽ transverse orientation)

the kidneys, and therefore the tissue perfusion would also include some presaturated blood resulting in lowered measured perfusion values. Further studies would have to be done to evaluate the influence of different inversion times giving more or less time to blood inflow into the imaging slice. The thickness of the selective inversion slice of 30 mm implies that at least in the coronal images almost the whole kidney tissue was presaturated. The specific hilifugal vessel architecture of the kidney would not allow significant transfer of non-saturated blood spins from other parts of the organ into the imaging slice. Therefore most of the signal gain would reflect the contribution of blood inflow from the aorta and parts of the renal arteries, which were not included into the selective inversion slice. Under physiologic conditions the selected inversion time of 1.6 s should give enough time for significant blood inflow into the kidney. Roberts et al. [10] reported perfusion values obtained on seven volunteers by MRI, using steady-state inversion of arterial water. They give values for perfusion in cortex (278 ⫾ 55 mL/(100 g min)) and medulla [55 ⫾ 25 mL/ (100 g min)], obtained from transverse slices only. These values are comparable with the results obtained in this study, although our results for perfusion in medulla are higher, possibly due to partial inclusion of segment walls. Unfortunately, Roberts et al. do not provide perfusion ratios of left versus right kidneys, which would help to compare the precision of the two techniques.

The results of our study demonstrate that quantitative imaging of human kidney tissue perfusion can be realized with spin labeling of blood water protons. Compared to other modalities (e.g., radionuclide scintigraphy) the FAIRUFLARE protocol provides a higher spatial resolution (2 ⫻ 2 ⫻ 10 mm3) and a reasonable sensitivity to assess the spatial distribution of perfusion rates within the kidney. Renal cortex and medulla could be clearly separated by differences in their specific perfusion rates. Still, before the method can be recommended for routine applications, several limitations have to be overcome. First of all, the method measures only the perfusion component perpendicular to the imaging slice. Therefore, different results may be obtained for the same region of the kidney, depending on the orientation of the imaging slice relative to the main perfusion direction. However, as discussed above, with an inversion-slice thickness of 30 mm, blood exchange within the kidney itself should have no significant contribution to the signal gain, but with thinner slices this effect may get significant. Further development will have to concentrate on evaluation of relevant correction factors. Another important point with significant influence on the measured values is the positioning of the imaging slices. As the FAIR-technique measures the signal of blood flowing from outside of the inversion slice into the imaging slice, inclusion of the large supplying arteries into the inversion field may result in lower signal, depending on the length of the included part of the vessel, the flow within the vessel

646

N. Karger et al. / Magnetic Resonance Imaging 18 (2000) 641– 647

and the time delay between inversion and readout. Observed differences of the perfusion values between transverse and coronal slices in the same part of a kidney may reflect this positioning effect. Some reduction of this effect could probably be achieved by decreasing the inversion and imaging slice thicknesses. This, however, would also result in a reduced signal to noise ratio. The same positioning effects may also have contributed to the interindividual variability of the perfusion values, as well as to the differences in perfusion observed in the reproducibility measurements. Care was taken to position the slices similarly in all volunteers; however, their different anatomy precludes the full compensation of this effect. The presented FAIR-UFLARE protocol is a single slice technique, and it may be challenged that the selected slice is representative for the perfusion state of the entire organ. However, a reasonable contrast-to-noise ratio can be obtained within 5 min and it seems thus feasible to measure several slices consecutively in order to cover the whole kidneys. With a slice thickness of 10 mm, 3 slices would suffice to cover the entire kidney. This would also allow to calculate an average value of perfusion for the whole kidney, thus minimizing the effect of slice positioning on the measured perfusion rates. Finally, the measurements with the FAIR-technique are sensitive to motion artifacts. Depending on the orientation of the slices, any movement may result in transfer of noninverted water contents into the readout slice and thus simulating an inflow of water by tissue perfusion. As a result, motion artifacts generally result in an overestimation of the perfusion values. Measurement times are too long for breathhold imaging, and therefore motion artifacts have to be overcome by a resonable thickness of the inversion and readout slices, thus limiting the spatial resolution. The cooperation of the patient is needed to avoid deep breathing or gross movement. Even peristaltic motion of the intestinum may overlay the perfusion maps of the kidneys. As a completely noninvasive technique, ASL-MRI is promising for addressing a variety of clinical questions, and may be employed for control of disease progression or therapeutic effects. One possible indication would be renal hypertension due to renal artery stenosis. Hemodynamically relevant narrowing of the main vessels can be compensated by the kidney itself due to its capability of autoregulation (reducing the own peripheral resistance and elevating the systemic arterial pressure via the renin-angiotensin pathway leading to severe arterial hypertension). Detection of normal or reduced renal tissue perfusion could possibly provide relevant information about a kidney’s capability to overcome a given stenosis by its autoregulative mechanisms and thus give the indication for invasive procedures to treat renal artery stenosis [9]. In transplanted kidneys, graft failure is mainly caused by four conditions, namely acute rejection episodes, acute tubular necrosis, chronic rejection, and drug-induced nephro-

toxicity - all of them demanding different, early and intensive treatment [8]. Renal biopsy remains the gold standard to detect allograft rejection but subjects the patient to the risks of invasive diagnostics such as bleeding and infection. A direct relationship between renal perfusion and different tissue pathologies in transplanted kidneys has been shown with dynamic contrast-enhanced MRI perfusion imaging. Szolar et al. [1] found a smaller increase in cortical signal intensity after contrast medium delivery in patients with acute allograft rejection when compared to normal allografts. On the other hand, patients with acute tubular necrosis showed a slightly delayed and diminished cortical enhancement and a lesser medullary enhancement pattern. Still, contrast-enhanced MRI only provides a relative indication of blood flow, but no exact quantification, as is possible with ASL-MRI [18]. Moreover, the contrast medium itself may represent a problem, particularly in patients with impaired renal function [19]. Other invasive methods providing quantitative information about renal perfusion include urinary clearance of paraaminohippuric acid, dye dilution methods and radionuclide scintigraphy. Doppler sonography with flowmetry or color-encoded duplex sonography with pulse-wave doppler are useful noninvasive adjuncts to detect vessel stenosis or hemodynamic signs of alterations in the peripheral vessel resistance (pulsatility or resistance-indices). However, they do not allow exact quantification of tissue perfusion rates [20]. In view of the limitations of alternative methods, the ASL-MRI protocol used in this study appears very promising. It would be a completely noninvasive, repeatable and easily applicable method to detect and quantify renal tissue perfusion alterations with reasonably high spatial resolution. The reproducibility error is smaller than the populationbased standard deviation of kidney perfusion. The accuracy of determining the left/right ratio of kidney perfusion should be sufficient to detect significant one-sided stenoses. A combination of ASL-MRI of the kidneys with high resolution MR-imaging of the parenchyma and direct visualization of the renal vessels by MR-angiography would represent a single study providing comprehensive information on the kidney status, that otherwise could only be achieved with a combination of invasive and noninvasive methods. Before the method could be used for routine applications, development work would have to be focused on reducing the sensitivity to artifacts and on reducing measurement times. A large body of reference data would have to be collected as a basis to evaluate the degree of renal impairment in patients. Modification of the proposed technique may allow for a more complete evaluation of the organ, overcoming current limits in reproducibility and providing a more representative figure of renal perfusion.

N. Karger et al. / Magnetic Resonance Imaging 18 (2000) 641– 647

Acknowledgments The authors like to thank Dr. Bernd Ku¨hn, Siemens, Erlangen, for the implementation of the pulse sequence.

References [1] Szolar DH, Preidler K, Ebner F, Kammerhuber F, Horn S, Ratschek M, Ranner G, Petritsch P, Horina JH. Functional magnetic resonance imaging of human renal allografts during the post-transplant period: preliminary observations. Magn Reson Imaging 1997;15:727–35. [2] Neimatallah MA, Dong Q, Schoenberg SO, Cho KJ, Prince MR. Magnetic resonance imaging in renal transplantation. J Magn Reson Imaging 1999;10:357– 68. [3] Wong EC, Buxton RB, Frank LR. A theoretical and experimental comparison of continuous and pulsed arterial spin labeling techniques for quantitative perfusion imaging. Magn Reson Med 1998;40:348 – 55. [4] Mai VM, Hagspiel KD, Christopher JM, Do HM, Altes T, KnightScott J, Stith AL, Maier T, Berr SS. Perfusion imaging of the human lung using flow-sensitive alternating inversion recovery with an extra radiofrequency pulse (FAIRER). Magn Reson Imaging 1999;17:355– 61. [5] Frank LR, Wong EC, Haseler LJ, Buxton RB. Dynamic imaging of perfusion in human skeletal muscle during exercise with arterial spin labeling. Magn Reson Med 1999;42:258 – 67. [6] Belle V, Kahler E, Waller C, Rommel E, Voll S, Hiller KH, Bauer WR, Haase A. In vivo quantitative mapping of cardiac perfusion in rats using a noninvasive MR spin-labeling method. J Magn Reson Imaging 1998;8:1240 –5. [7] Gowland PA, Francis ST, Duncan KR, Freeman AJ, Issa B, Moore RJ, Bowtell RW, Baker PN, Johnson IR, Worthington BS. In vivo perfusion measurements in the human placenta using echo planar imaging at 0.5 T. Magn Reson Med 1998;40:467–73.

647

[8] Wang J-J, Hendrich KS, Jackson EK, Ildstad ST, Williams DS, Ho C. Perfusion quantitation in transplanted rat kidney by MRI with arterial spin labeling. Kidney Int 1998;53:1783–91. [9] Prasad PV, Priatna A. Functional imaging of the kidneys with fast MRI techniques. Eur J Radiol 1999;29:133– 48. [10] Roberts DA, Detre JA, Bolinger L, Insko EK, Lenkinski RE, Pentecost MJ, Leigh JS. Renal perfusion in humans: MR imaging with spin tagging of arterial water. Radiology 1995;196:281– 6. [11] Chen Q, Siewert B, Bly BM, Warach S, Edelman RR. STARHASTE: perfusion imaging without magnetic susceptibility artifact. Magn Reson Med 1997;38:404 – 8. [12] Berr SS, Hagspiel KD, Mai VM, Keilholz-George S, Knight-Scott J, Christopher JM, Spinosa DJ, Angle JF, Matsumoto AH. Perfusion of the kidney using extraslice spin tagging (eST) magnetic resonance imaging. J Magn Reson Imaging 1999;10:886 –91. [13] Kim S-G. Quantification of relative cerebral blood flow change by flow-sensitive alternating inversion recovery (FAIR) technique: application to functional mapping. Magn Reson Med 1995;34:293–301. [14] Kim SG, Tsekos NV. Perfusion imaging by a flow-sensitive alternating inversion recovery (FAIR) technique: application to functional brain imaging [published erratum appears in Magn Reson Med 1997 May;37(5):675]. Magn Reson Med 1997; 37: 425–35. [15] Detre JA, Leigh JS, Williams DS, Koretsky AP. Perfusion imaging. Magn Reson Med 1992;23:37– 45. [16] Niendorf T. On the application of susceptibility-weighted ultra-fast low-angle RARE experiments in functional MR imaging. Magn Reson Med 1999;41:1189 –98. [17] Dworkin LD, Brenner BM. The renal circulations. In: Brenner BM, Rector FC, editors. The kidney. Philadelphia: Saunders, 1981. [18] Schoenberg SO, Essig M, Bock M, Hawighorst H, Sharafuddin M, Knopp MV. Comprehensive MR evaluation of renovascular disease in five breath holds. J Magn Reson Imaging 1999;10:347–56. [19] Shellock FG, Kanal E. Safety of magnetic resonance imaging contrast agents. J Magn Reson Imaging 1999;10:477– 84. [20] Restrepo-Schafer IK, Schwerk WB, Muller TF, Prinz H, Gorg C, Arnold R. [Intrarenal doppler flow analysis in patients with kidney transplantation and stable transplant function]. Ultraschall Med 1999; 20:87–92.