- Email: [email protected]
MRI-based quantification of renal perfusion in mice: Improving sensitivity and stability in FAIR ASL
ZEMEDI-10698; No. of Pages 6
ARTICLE IN PRESS
TECHNISCHE MITTEILUNG
MRI-based quantification of renal perfusion in mice: Improving sensitivity and stability in FAIR ASL Fabian Tobias Gutjahr 1,∗ , Stephan Michael Günster 1 , Thomas Kampf 1,4 , Patrick Winter 1 , Volker Herold 1 , Wolfgang Rudolf Bauer 2 , Peter Michael Jakob 1,3 1
Universität Würzburg, Lehrstuhl für Experimentelle Physik 5, Am Hubland, 97074 Würzburg, Germany Universität Würzburg, Medizinische Klinik und Poliklinik I, Oberdürrbacher Straße 6, 97080 Würzburg, Germany 3 Research Center Magnetic-Resonance-Bavaria, Am Hubland, 97074 Würzburg, Germany 4 Institut für Diagnostische und Interventionelle Neuroradiologie, Universitätsklinikum Würzburg, Josef-Schneider-Straße 11, 97080 Würzburg, Germany 2
Received 26 August 2016; accepted 6 February 2017
Abstract
Zusammenfassung
Purpose: The importance of the orientation of the selective inversion slice in relation to the anatomy in flow-sensitive alternating inversion recovery arterial spin labeling (FAIR ASL) kidney perfusion measurements is demonstrated by comparing the standard FAIR scheme to a scheme with an improved slice selective control experiment. Methods: A FAIR ASL method is used. The selective inversion preparation slice is set perpendicular to the measurement slice to decrease the unintended labeling of arterial spins in the control experiment. A T1∗ -based quantification method compensates for the effects of the imperfect inversion on the edge of the selective inversion slice. The quantified perfusion values are compared to the standard experiment with parallel orientation of imaging and selective inversion slice. Results: Perfusion maps acquired with the perpendicular inversion slice orientation show higher sensitivity compared to the parallel orientation. The T1∗ -based quantification method removes artifacts arising from imperfect inversion slice profiles. The stability is improved. Conclusion: Adjusting the labeling technique to the anatomy is of high importance. Improved sensitivity and reproducibility could be demonstrated. The proposed method provides a solution to the problem of FAIR ASL measurements of renal perfusion in coronal view.
Ziele: Die Bedeutung der Lage der schicht-selektiven Inversion des Kontrollexperiments in FAIR-ASL Nierenperfusionsmessungen soll demonstriert und ein verbessertes Schema vorgeschlagen werden. Methoden: Es wird eine FAIR-ASL Methode angewandt. Die Inversionsschicht des Kontrollexperiments wird senkrecht auf die Bildgebungsschicht gesetzt und dadurch die ungewollte Markierung von einfließenden Spins reduziert. Eine T1∗ -basierte Quantifizierung wird verwendet, um die Effekte von variabler Inversionsgüte innerhalb der Bildgebungsschicht zu kompensieren. Die quantifizierten Werte werden mit denen des Standardexperiments mit parallel orientierter Inversions- und Bildgebungsschicht verglichen. Ergebnisse: Die mit der neuen Methode gemessenen Perfusionskarten zeigen eine höhere Sensitivität, als die mit der Standard-FAIR-ASL Methode aufgenommenen. Die T1∗ -basierte Quantifizierung eliminiert Artefakte, die durch die reduzierte Inversionsgüte am Rand der selektiven Inversionsschicht auftreten. Die Stabilität der Messung wird erhöht. Schlussfolgerung: Das Anpassen der Experimente an die untersuchte Anatomie ist von großer Bedeutung. Es konnte eine gesteigerte Sensitivität und Reproduzierbarkeit der Messung demonstriert werden. Die vorgeschlagene Methode ist eine mögliche Lösung für das Problem der
∗ Corresponding
author at: Universität Würzburg, Lehrstuhl für Experimentelle Physik 5, Am Hubland, 97074 Würzburg, Germany. E-mail: [email protected] (F.T. Gutjahr).
Z. Med. Phys. xxx (2017) xxx–xxx http://dx.doi.org/10.1016/j.zemedi.2017.02.001 www.elsevier.com/locate/zemedi
ARTICLE IN PRESS 2
F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
FAIR-ASL Nierenperfusionsmessung in koronalen Schnitten. Keywords: Arterial spin labeling, FAIR ASL, renal, kidney, perfusion, coronal view
Introduction Perfusion measurements in kidneys prove to be a valuable tool in preclinical research to assess their function [1], as the filtration is closely linked to the perfusion. There have been several publications on the subject in the recent years [2–6]. Among other methods such as pseudo continuous arterial spin labeling (pCASL) [5,6] a flow sensitive alternating inversion recovery (FAIR) [7] arterial spin labeling (ASL) method was used [2–4]. ASL is well suited for applications in the kidney, as contrast agent based methods can be contraindicated in subjects with renal insufficiency [8]. In FAIR ASL the magnetization is usually prepared using a global inversion which is referred to as the labeling experiment and a slice selective inversion which is referred to as the control experiment. In the standard labeling scheme the selective inversion slice is in the same plane as the imaging slice with a thickness of 1.5–5 times the thickness of the imaging slice. This warrants a high inversion quality throughout the imaging slice. For renal imaging in the coronal view this proves to be problematic, as the selective inversion slice can unintentionally pass through the heart, lungs and vasculature containing a large amount of blood, which may flow into the kidney and mitigate the inflow effect of non-inverted spins on T1 . This reduces the quality of the control experiment and thereby the difference between the global and slice selective experiment, which will ultimately lead to underestimation of perfusion [2,9]. The coronal view is desirable because the long axis of both kidneys can be imaged in a single experiment. In Fig. 1(a) a maximum intensity projection from three adjacent slices with a slice thickness of 0.75 mm is shown. The resulting thickness of 2.25 mm is still below the typical thickness of the selective inversion slice. Nevertheless large parts of the aorta are visible and would therefore be included in the selective inversion slice, deteriorating the quality of the control experiment. The problem is also visualized in Fig. 1(b) where a sagittal FLASH image through a kidney and the heart with the position of the selective inversion slice is shown. Large vessels and parts of the lungs which contain a lot of blood are in the marked area. This problem has been noted in literature before [2,5,9]. In [2] it is mentioned that the inclusion of feeding arteries in the selective inversion sliced leads to perfusion signal loss. In [9] the inversion and imaging slice was positioned so it would not include the aorta in human subjects. However, this poses limits to the imaging slice positioning or slice thickness and can be impossible in small rodents. In [5] a possible solution using pseudo-continuous ASL (pCASL) was demonstrated.
Schlüsselwörter: Arterielles Spin Labeling, FAIR ASL, Niere, Perfusion, Koronaler Blick
However, as FAIR ASL is well established in preclinical research, a solution to the problem for FAIR ASL in coronal view is desirable. This allows the use of existing FAIR ASL protocols. In this work we propose a new solution within the framework of FAIR ASL. By adapting the orientation of the selective inversion slice (see blue area in Fig. 1(a) and (b) right) to minimize the problem of undesired manipulation of inflowing spins by the slice selective inversion the labeling efficiency is improved in comparison to the standard protocol.
Methods Animal experiments All experimental procedures were in accordance with institutional guidelines and were approved by local authorities. Mice were placed in prone position. Anesthetization was
Figure 1. (a) A maximum intensity projection from three slices (thickness 0.75 mm, no gap) that cover a thickness of 2.25 mm. The position and size of the perpendicular inversion slice is marked by the blue bar. Everything visible in this image would be labeled in the slice selective inversion of the standard FAIR experiment. (b) Exemplary image of possible positions of the imaging slices with a 5◦ tilt (thick red and green lines) and their corresponding selective inversion slices (red and green shading) for the parallel (left) and perpendicular orientation (right). The area marked with the line pattern is only labeled in one of the two slightly titled slices. It is clearly visible that the volume labeled in the perpendicular selective inversion orientation is less than that of the parallel orientation. Also the difference in the amount of labeled inflowing blood, due to a small tilt in the imaging slice, is reduced in the perpendicular orientation. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
ARTICLE IN PRESS F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
induced using 4% isoflurane diluted in oxygen and maintained at 1.5–2%. Vital parameters were acquired using ECG leads attached to the forepaws for cardiac monitoring. A pressure sensor was placed underneath the mouse for breath monitoring. ECG and breath data were recorded for retrospective gating. The mice were kept at physiological temperatures using the gradient cooling system and a breathing gas temperature control. Three healthy female mice (C57BL/6, Charles River Laboratories, Sulzfeld Germany) were examined. All animals were sacrificed after the measurement.
MRI All measurements were performed on a horizontal bore 7 T small animal scanner (BioSpec, Bruker, Ettlingen Germany) using a custom built 35 mm quadrature birdcage coil. The retrospectively triggered T1 -based ASL method using a model based reconstruction described in [10] was used. The standard FAIR approach, in which the selective inversion slice is in the same plane as the imaging slice, was adapted to enable an inversion slice with arbitrary positioning. The sequence parameters for all measurements were as follows: 16 inversions each for the slice selective and the global inversion experiment followed by 2500 FLASH-readouts with repetition and echo time TR /TE = 4.0/1.9 ms, inversion pulse: adiabatic hyperbolic secant with a duration of 2.8 ms and a bandwidth of 6.3 kHz, readout pulses: sinc with a duration of 0.4 ms, nominal flip angle of 5◦ and a bandwidth of 15.5 kHz, readout bandwidth 59 kHz, matrix size 96 × 96, field of view (FOV) 2.6 × 2.6 cm2 , 1.5 mm imaging slice thickness and a waiting time of 8 s between inversions, which results in a native resolution of (0.271 mm)2 and a measurement time of less than 10 min. A coronal view was chosen to show both kidneys in their long axis. For each animal a perfusion map was acquired with the standard labeling scheme with the selective inversion slice in parallel to the imaging slice and another map with the same FOV but a perpendicular selective inversion slice. The imaging slice remained the same. The inversion slice thickness in the parallel measurement was 4.5 mm. For the perpendicular measurement the inversion slice was rotated to cover both kidneys. Three animals were used to investigate the stability of the perfusion measurement. In the experiments with perpendicular orientation the inversion slice thickness was fixed to 12 mm (see Fig. 1). Thus the selective inversion slice extended the kidneys by about 1 mm in both directions. To assess the stability the imaging slice was repositioned by another experimenter (see Fig. 1b). This resulted in a slight tilt of 5◦ for animal one, 3◦ for animal two and 1.4◦ and a 0.2 mm shift for animal three. Both the parallel and perpendicular inversion measurements were repeated. The inversion slices for the selective measurement were shifted accordingly to remain either parallel
3
or perpendicular to the measurement slice. Altogether four perfusion maps were acquired for each animal. Post processing As described in [10] the data used for reconstruction are selected retrospectively and a model based interpolation method is used to cope with the undersampling, inherent in retrospective methods. In the maps shown in this work the data acquired during breathing were discarded while data from all positions in the heart cycle were included. The model based interpolation allows images to be reconstructed on any inversion time TI. In this work 500 images were reconstructed for each the control and the labeling experiment on a grid with a spacing of five TR, i.e. TIn = 6 ms+5 · TR · n with n = 1 . . .500. The reconstructed images are fitted using a non-linear least square fit in MATLAB (Matlab Release 2015b, the MathWorks, Inc., Natick, USA) according to the model:
TI S(TI) = M∞ − (M∞ − M0 ) exp − ∗ T1
(1)
This yields the apparent decay constant T1∗ , the equilibrium magnetization M0 and the steady state magnetization M∞ under the influence of the FLASH readout pulses. These values are usually used to obtain T1 -values T1 =
M0 ∗ T M∞ 1
(2)
from which perfusion can be calculated [11,12]. P=
λ T1blood
·
T1glob −1 , T1sel
(3)
where λ is the blood-tissue partition coefficient and T1blood the T1 -value of blood. However, this requires a high inversion quality across the region of interest, in this case the kidneys. While this is usually achieved in the global inversion experiment, the slice-selective inversion requires careful adjustment of the inversion slice thickness. In the case of coplanar inversion- and imaging-slice, the inversion slice is made as thick as necessary to ensure a sufficient inversion quality inside the imaging slice, but as narrow as possible to keep the amount of inverted blood outside the imaging slice at a minimum. In the experiment with the inversion slice perpendicular to the imaging slice, the inversion slice thickness is chosen to only cover the kidneys. This ensures that only a minimal amount of the blood pool is influenced by the slice-selective inversion but it also causes an incomplete inversion of the magnetization close to the edge of the inversion slice. In this area the correction given in (2) fails in the slice selective experiment as M0 is underestimated at the edges of the selective inversion
ARTICLE IN PRESS 4
F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
derived, which allows to calculate perfusion directly from T1∗ -values: P=
λ 2 · T1blood
M∞ 1+ M0
T1∗ glob T1∗ sel
−1
(4)
This circumvents the problem of variable inversion quality in the slice-selective experiment. For comparisons between the different slice orientations, median perfusion values are calculated within regions of interest (ROI). The ROIs covering the cortex of the kidneys were selected manually in the anatomical images. Figure 2. T1 - and T1∗ -map [s] after slice selective inversion. It can be seen that the T1 -correction fails at the edges of the inversion slices. The T1∗ -map is stable over a larger area.
slice due to incomplete inversion. This leads to underestimation of T1 -values which again results in an overestimation of the perfusion values in this region using (3). This problem is visualized in Fig. 2. The corrected T1 -map is shown together with the corresponding T1∗ -map. The T1 values at the edges of the kidney are strongly underestimated rendering perfusion quantification based on T1 impossible with this inversion slice thickness. The corresponding T1∗ map is stable over the kidney. In [13] a new equation was
Results Four perfusion maps were acquired for each animal. The coefficient λ/T1blood was assumed to be 5.8 ml/(g · s). These maps are shown in Fig. 3. The higher effective SNR and higher perfusion values acquired with the perpendicular labeling are clearly visible. The median and the median absolute deviation (MAD) are shown in Table 1. Additionally, the relative deviation (xi /x¯i · 100) between values acquired for the slightly different slice positions (x1 and x2 ) is given in percent. The median and MAD were chosen as measures due to their robustness and the fractality of perfusion [14] which could govern the standard deviation.
Figure 3. Perfusion maps in mL · (g · min)−1 . For every animal the maps generated after parallel and perpendicular slice selective inversion are shown for two slightly tilted slice positions. It is visible that for the perpendicular inversion slice set-up the acquired perfusion values are generally higher and the inter slice variation between two similar slices is lower.
ARTICLE IN PRESS F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
5
Table 1 Median ± MAD (median absolute deviation) for perfusion values in the cortex of the left and right kidney [mL · (g · min)−1 ] for the two slices and two selective inversion methods and the deviation between two different slices relative to their corresponding mean. Animal
Slice
1
⊥
1 2 Deviation [%] 1 2 Deviation [%]
2
3
Left
Right
Left
Right
Left
Right
5.7 ± 1.2 2.9 ± 0.9 65.1 4.8 ± 1.3 4.1 ± 1.3 16.0
4.8 ± 1.5 2.9 ± 0.8 50.1 3.7 ± 1.1 3.7 ± 1.2 0.3
3.4 ± 1.1 3.0 ± 1.1 11.1 5.2 ± 1.6 5.2 ± 1.5 0.3
3.3 ± 1.0 2.8 ± 1.0 17.2 5.6 ± 1.4 5.3 ± 1.6 5.3
5.5 ± 1.4 4.1 ± 1.3 29.1 5.7 ± 1.6 5.4 ± 1.7 5.7
5.3 ± 1.6 3.5 ± 1.4 39.1 5.9 ± 1.6 5.5 ± 1.7 7.6
The mean deviation between the slightly different imaging slice orientations is 35% in the parallel measurement and 6% in the perpendicular measurement. Additionally, mean perfusion values are 32% higher for the perpendicular inversion than the parallel inversion slice measurement. The higher sensitivity, which is expressed by the higher perfusion values, as well as the improved stability, shown by the reduced deviation between slightly changed slice positions, are in good accordance to the expectations of an improved control experiment.
Discussion The aim of this work was to provide a straightforward and effective solution to the issue of quantitative renal perfusion measurements in coronal view using FAIR ASL as noted in [2,5,9]. It was shown that the quality and stability of perfusion maps can be improved without increasing the complexity of the measurement or the post processing. This was achieved by decoupling the imaging slice orientation from the selective inversion slice orientation in the control experiment. By changing the orientation of the selective inversion experiment but not it’s position relative to the imaging slice the method remains insensitive to magnetization transfer effects (MTC) as the inversion pulses for both the labeling and the control experiment are on-resonant within the region of interest. The higher perfusion values in the perpendicular orientation of the selective inversion slice support the original thesis that the parallel selective inversion slice labels too much of the inflowing blood. Perfusion values were acquired for both the standard parallel orientation of the inversion slice and a perpendicular orientation. The mean value over all kidneys was 3.9±1.1 mL · (g · min)−1 for the standard and 5.0 ± 0.8 mL · (g · min)−1 for the perpendicular orientation. The values for the parallel orientation of the selective inversion slice correspond to the values reported previously 3.97 ± 0.36 mL · (g · min)−1 [2] and 4.4 mL · (g · min)−1 [4]. The values for the perpendicular orientation are higher than the previously reported values acquired using FAIR
ASL, but not as high as the values reported in [5] 5.66 ± 0.48 mL · (g · min)−1 and within the range of the values reported in [6] (4.66 ± 1.11 to 6.79 ± 1.49) mL · (g · min)−1 that were acquired using pseudo-continuous ASL [15] (pCASL) with varying read out methods. In [2–4] perfusion is calculated from the difference between images acquired in the selective and non selective measurement. While a different approach for quantification was used here, the experimental results are relevant to these techniques as well, as they profit from the increased stability of the control experiment and a larger difference between the control and labeling experiment. The T1∗ -based approach was chosen as it is able to reconstruct perfusion values even with varying inversion efficiency within the imaging slice in the control experiment. This is desirable since it allows a thinner selective inversion slice as perfusion values can still be obtained at the edge of the selective inversion slice. For preclinical studies the best inversion slice thickness should be assessed beforehand, taking into account the maximum size of the kidneys of the chosen animal model. It is advisable to reduce the slice thickness as much as possible in order to minimize the influence of unintentionally labeled inflowing blood. The thickness of 12 mm is a good compromise between minimal blood labeling in the selective inversion experiment and full coverage of the kidneys in coronal view for the chosen animals. The deviation between two very similar slices was used to show the improved stability of the perfusion values acquired using the perpendicular slice selective orientation. This effect can be attributed mainly to the fact that a slight tilt of the selective inversion slice affects a much larger portion of the mouse in the parallel than in the perpendicular setup (see Fig. 1b). Given the aim of this work acquisition parameters were not optimized for short protocol length. In [10] it was shown that reliable perfusion values could be obtained using as low as eight inversions despite triggering to the cardiac cycle. This would lead to a reduction in protocol length by 50%. As no triggering to the heart cycle is necessary in renal perfusion measurements, even further reduction is possible without changing acquisition parameters. Optimizing these
ARTICLE IN PRESS 6
F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
could accelerate the acquisition time for each inversion leading to additional shortening of the protocol length. In [5] another possible solution for the problem of renal perfusion in coronal view was proposed and several advantages for pCASL in comparison to a single TI-implementation of FAIR ASL were shown and the problematic control experiment in the standard FAIR ASL approach in renal perfusion measurements in coronal view was noted. This was mainly attributed to the coupling of selective inversion slice orientation to the imaging slice. By removing this constraint, reliable quantification and an increased sensitivity in coronal view with FAIR ASL could be demonstrated in our work.
Conclusion In this paper a straightforward and effective way to improve the quality of FAIR ASL kidney perfusion measurements has been demonstrated. By decreasing the amount of unwanted labeling of the inflowing blood pool in the perpendicular orientation of the inversion slice, the difference between the control and labeling experiment is increased, leading to higher sensitivity. Also the method becomes more stable against variations in the slice positioning, which is of great importance for comparability in longitudinal or inter animal measurements.
Acknowledgements The authors would like to thank Sabine Voll for animal handling. This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB 688 B5, Z2) and the Bundesministerium für Bildung und Forschung (BMBF01 EO1004).
References [1] Bennett HF, Li D. MR imaging of renal function. Magn Reson Imaging Clin N Am 1997;5(1):107–26. [2] Rajendran R, Lew SK, Yong CX, Tan J, Wang DJJ, Chuang KH. Quantitative mouse renal perfusion using arterial spin labeling. NMR Biomed 2013;26(10):1225–32.
[3] Winter JD, St Lawrence KS, Cheng HLM. Quantification of renal perfusion: comparison of arterial spin labeling and dynamic contrastenhanced MRI. J Magn Reson Imaging 2011;34(3):608–15. [4] Hueper K, Gutberlet M, Rong S, Hartung D, Mengel M, Lu X, Haller H, Wacker F, Meier M, Gueler F. Acute kidney injury: arterial spin labeling to monitor renal perfusion impairment in micecomparison with histopathologic results and renal function. Radiology 2014;270(1):117–24. [5] Duhamel G, Prevost V, Girard OM, Callot V, Cozzone PJ. Highresolution mouse kidney perfusion imaging by pseudo-continuous arterial spin labeling at 11.75T. Magn Reson Med 2014;71(3):1186–96. [6] Prevost VH, Girard OM, Callot V, Cozzone PJ, Duhamel G. Fast imaging strategies for mouse kidney perfusion measurement with pseudocontinuous arterial spin labeling (PCASL) at ultra high magnetic field (11.75 tesla). J Magn Reson Imaging 2015;42(4):999–1008. [7] Kim SG. Quantification of relative cerebral blood flow change by flowsensitive alternating inversion recovery (FAIR) technique: application to functional mapping. Magn Reson Med 1995;34(3):293–301. [8] Wang JJ, 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(6):1783–91. [9] Cutajar M, Thomas DL, Banks T, Clark Ca, Golay X, Gordon I. Repeatability of renal arterial spin labelling MRI in healthy subjects. Magn Reson Mater Phys Biol Med 2012;25(2):145–53. [10] Gutjahr FT, Kampf T, Winter P, Meyer CB, Williams T, Jakob PM, Bauer WR, Ziener CH, Helluy X. Quantification of perfusion in murine myocardium: a retrospectively triggered t1-based ASL method using model-based reconstruction. Magn Reson Med 2015;74(6):1705–15. [11] Deichmann R, Haase A. Quantification of T1 values by snapshot-flash NMR imaging. J Magn Reson 1992;96:608–12. [12] 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(6):1240–5. [13] Kampf T, Helluy X, Gutjahr FT, Winter P, Meyer CB, Jakob PM, Bauer WR, Ziener CH. Myocardial perfusion quantification using the t1-based fair-ASL method: the influence of heart anatomy, cardiopulmonary blood flow and look-locker readout. Magn Reson Med 2014;71(5):1784–97. [14] Bauer WR, Hiller KH, Galuppo P, Neubauer S, Köpke J, Haase a, Waller C, Ertl G. Fast high-resolution magnetic resonance imaging demonstrates fractality of myocardial perfusion in microscopic dimensions. Circ Res 2001;88(3):340–6. [15] Dai W, Garcia D, de Bazelaire C, Alsop DC. Continuous flow driven inversion for arterial spin labeling using pulsed radiofrequency and gradient fields. Magn Reson Med: Off J Soc Magn Reson Med/Soc Magn Reson Med 2008;60(6):1488–97.
Available online at www.sciencedirect.com
ScienceDirect
ARTICLE IN PRESS
TECHNISCHE MITTEILUNG
MRI-based quantification of renal perfusion in mice: Improving sensitivity and stability in FAIR ASL Fabian Tobias Gutjahr 1,∗ , Stephan Michael Günster 1 , Thomas Kampf 1,4 , Patrick Winter 1 , Volker Herold 1 , Wolfgang Rudolf Bauer 2 , Peter Michael Jakob 1,3 1
Universität Würzburg, Lehrstuhl für Experimentelle Physik 5, Am Hubland, 97074 Würzburg, Germany Universität Würzburg, Medizinische Klinik und Poliklinik I, Oberdürrbacher Straße 6, 97080 Würzburg, Germany 3 Research Center Magnetic-Resonance-Bavaria, Am Hubland, 97074 Würzburg, Germany 4 Institut für Diagnostische und Interventionelle Neuroradiologie, Universitätsklinikum Würzburg, Josef-Schneider-Straße 11, 97080 Würzburg, Germany 2
Received 26 August 2016; accepted 6 February 2017
Abstract
Zusammenfassung
Purpose: The importance of the orientation of the selective inversion slice in relation to the anatomy in flow-sensitive alternating inversion recovery arterial spin labeling (FAIR ASL) kidney perfusion measurements is demonstrated by comparing the standard FAIR scheme to a scheme with an improved slice selective control experiment. Methods: A FAIR ASL method is used. The selective inversion preparation slice is set perpendicular to the measurement slice to decrease the unintended labeling of arterial spins in the control experiment. A T1∗ -based quantification method compensates for the effects of the imperfect inversion on the edge of the selective inversion slice. The quantified perfusion values are compared to the standard experiment with parallel orientation of imaging and selective inversion slice. Results: Perfusion maps acquired with the perpendicular inversion slice orientation show higher sensitivity compared to the parallel orientation. The T1∗ -based quantification method removes artifacts arising from imperfect inversion slice profiles. The stability is improved. Conclusion: Adjusting the labeling technique to the anatomy is of high importance. Improved sensitivity and reproducibility could be demonstrated. The proposed method provides a solution to the problem of FAIR ASL measurements of renal perfusion in coronal view.
Ziele: Die Bedeutung der Lage der schicht-selektiven Inversion des Kontrollexperiments in FAIR-ASL Nierenperfusionsmessungen soll demonstriert und ein verbessertes Schema vorgeschlagen werden. Methoden: Es wird eine FAIR-ASL Methode angewandt. Die Inversionsschicht des Kontrollexperiments wird senkrecht auf die Bildgebungsschicht gesetzt und dadurch die ungewollte Markierung von einfließenden Spins reduziert. Eine T1∗ -basierte Quantifizierung wird verwendet, um die Effekte von variabler Inversionsgüte innerhalb der Bildgebungsschicht zu kompensieren. Die quantifizierten Werte werden mit denen des Standardexperiments mit parallel orientierter Inversions- und Bildgebungsschicht verglichen. Ergebnisse: Die mit der neuen Methode gemessenen Perfusionskarten zeigen eine höhere Sensitivität, als die mit der Standard-FAIR-ASL Methode aufgenommenen. Die T1∗ -basierte Quantifizierung eliminiert Artefakte, die durch die reduzierte Inversionsgüte am Rand der selektiven Inversionsschicht auftreten. Die Stabilität der Messung wird erhöht. Schlussfolgerung: Das Anpassen der Experimente an die untersuchte Anatomie ist von großer Bedeutung. Es konnte eine gesteigerte Sensitivität und Reproduzierbarkeit der Messung demonstriert werden. Die vorgeschlagene Methode ist eine mögliche Lösung für das Problem der
∗ Corresponding
author at: Universität Würzburg, Lehrstuhl für Experimentelle Physik 5, Am Hubland, 97074 Würzburg, Germany. E-mail: [email protected] (F.T. Gutjahr).
Z. Med. Phys. xxx (2017) xxx–xxx http://dx.doi.org/10.1016/j.zemedi.2017.02.001 www.elsevier.com/locate/zemedi
ARTICLE IN PRESS 2
F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
FAIR-ASL Nierenperfusionsmessung in koronalen Schnitten. Keywords: Arterial spin labeling, FAIR ASL, renal, kidney, perfusion, coronal view
Introduction Perfusion measurements in kidneys prove to be a valuable tool in preclinical research to assess their function [1], as the filtration is closely linked to the perfusion. There have been several publications on the subject in the recent years [2–6]. Among other methods such as pseudo continuous arterial spin labeling (pCASL) [5,6] a flow sensitive alternating inversion recovery (FAIR) [7] arterial spin labeling (ASL) method was used [2–4]. ASL is well suited for applications in the kidney, as contrast agent based methods can be contraindicated in subjects with renal insufficiency [8]. In FAIR ASL the magnetization is usually prepared using a global inversion which is referred to as the labeling experiment and a slice selective inversion which is referred to as the control experiment. In the standard labeling scheme the selective inversion slice is in the same plane as the imaging slice with a thickness of 1.5–5 times the thickness of the imaging slice. This warrants a high inversion quality throughout the imaging slice. For renal imaging in the coronal view this proves to be problematic, as the selective inversion slice can unintentionally pass through the heart, lungs and vasculature containing a large amount of blood, which may flow into the kidney and mitigate the inflow effect of non-inverted spins on T1 . This reduces the quality of the control experiment and thereby the difference between the global and slice selective experiment, which will ultimately lead to underestimation of perfusion [2,9]. The coronal view is desirable because the long axis of both kidneys can be imaged in a single experiment. In Fig. 1(a) a maximum intensity projection from three adjacent slices with a slice thickness of 0.75 mm is shown. The resulting thickness of 2.25 mm is still below the typical thickness of the selective inversion slice. Nevertheless large parts of the aorta are visible and would therefore be included in the selective inversion slice, deteriorating the quality of the control experiment. The problem is also visualized in Fig. 1(b) where a sagittal FLASH image through a kidney and the heart with the position of the selective inversion slice is shown. Large vessels and parts of the lungs which contain a lot of blood are in the marked area. This problem has been noted in literature before [2,5,9]. In [2] it is mentioned that the inclusion of feeding arteries in the selective inversion sliced leads to perfusion signal loss. In [9] the inversion and imaging slice was positioned so it would not include the aorta in human subjects. However, this poses limits to the imaging slice positioning or slice thickness and can be impossible in small rodents. In [5] a possible solution using pseudo-continuous ASL (pCASL) was demonstrated.
Schlüsselwörter: Arterielles Spin Labeling, FAIR ASL, Niere, Perfusion, Koronaler Blick
However, as FAIR ASL is well established in preclinical research, a solution to the problem for FAIR ASL in coronal view is desirable. This allows the use of existing FAIR ASL protocols. In this work we propose a new solution within the framework of FAIR ASL. By adapting the orientation of the selective inversion slice (see blue area in Fig. 1(a) and (b) right) to minimize the problem of undesired manipulation of inflowing spins by the slice selective inversion the labeling efficiency is improved in comparison to the standard protocol.
Methods Animal experiments All experimental procedures were in accordance with institutional guidelines and were approved by local authorities. Mice were placed in prone position. Anesthetization was
Figure 1. (a) A maximum intensity projection from three slices (thickness 0.75 mm, no gap) that cover a thickness of 2.25 mm. The position and size of the perpendicular inversion slice is marked by the blue bar. Everything visible in this image would be labeled in the slice selective inversion of the standard FAIR experiment. (b) Exemplary image of possible positions of the imaging slices with a 5◦ tilt (thick red and green lines) and their corresponding selective inversion slices (red and green shading) for the parallel (left) and perpendicular orientation (right). The area marked with the line pattern is only labeled in one of the two slightly titled slices. It is clearly visible that the volume labeled in the perpendicular selective inversion orientation is less than that of the parallel orientation. Also the difference in the amount of labeled inflowing blood, due to a small tilt in the imaging slice, is reduced in the perpendicular orientation. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
ARTICLE IN PRESS F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
induced using 4% isoflurane diluted in oxygen and maintained at 1.5–2%. Vital parameters were acquired using ECG leads attached to the forepaws for cardiac monitoring. A pressure sensor was placed underneath the mouse for breath monitoring. ECG and breath data were recorded for retrospective gating. The mice were kept at physiological temperatures using the gradient cooling system and a breathing gas temperature control. Three healthy female mice (C57BL/6, Charles River Laboratories, Sulzfeld Germany) were examined. All animals were sacrificed after the measurement.
MRI All measurements were performed on a horizontal bore 7 T small animal scanner (BioSpec, Bruker, Ettlingen Germany) using a custom built 35 mm quadrature birdcage coil. The retrospectively triggered T1 -based ASL method using a model based reconstruction described in [10] was used. The standard FAIR approach, in which the selective inversion slice is in the same plane as the imaging slice, was adapted to enable an inversion slice with arbitrary positioning. The sequence parameters for all measurements were as follows: 16 inversions each for the slice selective and the global inversion experiment followed by 2500 FLASH-readouts with repetition and echo time TR /TE = 4.0/1.9 ms, inversion pulse: adiabatic hyperbolic secant with a duration of 2.8 ms and a bandwidth of 6.3 kHz, readout pulses: sinc with a duration of 0.4 ms, nominal flip angle of 5◦ and a bandwidth of 15.5 kHz, readout bandwidth 59 kHz, matrix size 96 × 96, field of view (FOV) 2.6 × 2.6 cm2 , 1.5 mm imaging slice thickness and a waiting time of 8 s between inversions, which results in a native resolution of (0.271 mm)2 and a measurement time of less than 10 min. A coronal view was chosen to show both kidneys in their long axis. For each animal a perfusion map was acquired with the standard labeling scheme with the selective inversion slice in parallel to the imaging slice and another map with the same FOV but a perpendicular selective inversion slice. The imaging slice remained the same. The inversion slice thickness in the parallel measurement was 4.5 mm. For the perpendicular measurement the inversion slice was rotated to cover both kidneys. Three animals were used to investigate the stability of the perfusion measurement. In the experiments with perpendicular orientation the inversion slice thickness was fixed to 12 mm (see Fig. 1). Thus the selective inversion slice extended the kidneys by about 1 mm in both directions. To assess the stability the imaging slice was repositioned by another experimenter (see Fig. 1b). This resulted in a slight tilt of 5◦ for animal one, 3◦ for animal two and 1.4◦ and a 0.2 mm shift for animal three. Both the parallel and perpendicular inversion measurements were repeated. The inversion slices for the selective measurement were shifted accordingly to remain either parallel
3
or perpendicular to the measurement slice. Altogether four perfusion maps were acquired for each animal. Post processing As described in [10] the data used for reconstruction are selected retrospectively and a model based interpolation method is used to cope with the undersampling, inherent in retrospective methods. In the maps shown in this work the data acquired during breathing were discarded while data from all positions in the heart cycle were included. The model based interpolation allows images to be reconstructed on any inversion time TI. In this work 500 images were reconstructed for each the control and the labeling experiment on a grid with a spacing of five TR, i.e. TIn = 6 ms+5 · TR · n with n = 1 . . .500. The reconstructed images are fitted using a non-linear least square fit in MATLAB (Matlab Release 2015b, the MathWorks, Inc., Natick, USA) according to the model:
TI S(TI) = M∞ − (M∞ − M0 ) exp − ∗ T1
(1)
This yields the apparent decay constant T1∗ , the equilibrium magnetization M0 and the steady state magnetization M∞ under the influence of the FLASH readout pulses. These values are usually used to obtain T1 -values T1 =
M0 ∗ T M∞ 1
(2)
from which perfusion can be calculated [11,12]. P=
λ T1blood
·
T1glob −1 , T1sel
(3)
where λ is the blood-tissue partition coefficient and T1blood the T1 -value of blood. However, this requires a high inversion quality across the region of interest, in this case the kidneys. While this is usually achieved in the global inversion experiment, the slice-selective inversion requires careful adjustment of the inversion slice thickness. In the case of coplanar inversion- and imaging-slice, the inversion slice is made as thick as necessary to ensure a sufficient inversion quality inside the imaging slice, but as narrow as possible to keep the amount of inverted blood outside the imaging slice at a minimum. In the experiment with the inversion slice perpendicular to the imaging slice, the inversion slice thickness is chosen to only cover the kidneys. This ensures that only a minimal amount of the blood pool is influenced by the slice-selective inversion but it also causes an incomplete inversion of the magnetization close to the edge of the inversion slice. In this area the correction given in (2) fails in the slice selective experiment as M0 is underestimated at the edges of the selective inversion
ARTICLE IN PRESS 4
F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
derived, which allows to calculate perfusion directly from T1∗ -values: P=
λ 2 · T1blood
M∞ 1+ M0
T1∗ glob T1∗ sel
−1
(4)
This circumvents the problem of variable inversion quality in the slice-selective experiment. For comparisons between the different slice orientations, median perfusion values are calculated within regions of interest (ROI). The ROIs covering the cortex of the kidneys were selected manually in the anatomical images. Figure 2. T1 - and T1∗ -map [s] after slice selective inversion. It can be seen that the T1 -correction fails at the edges of the inversion slices. The T1∗ -map is stable over a larger area.
slice due to incomplete inversion. This leads to underestimation of T1 -values which again results in an overestimation of the perfusion values in this region using (3). This problem is visualized in Fig. 2. The corrected T1 -map is shown together with the corresponding T1∗ -map. The T1 values at the edges of the kidney are strongly underestimated rendering perfusion quantification based on T1 impossible with this inversion slice thickness. The corresponding T1∗ map is stable over the kidney. In [13] a new equation was
Results Four perfusion maps were acquired for each animal. The coefficient λ/T1blood was assumed to be 5.8 ml/(g · s). These maps are shown in Fig. 3. The higher effective SNR and higher perfusion values acquired with the perpendicular labeling are clearly visible. The median and the median absolute deviation (MAD) are shown in Table 1. Additionally, the relative deviation (xi /x¯i · 100) between values acquired for the slightly different slice positions (x1 and x2 ) is given in percent. The median and MAD were chosen as measures due to their robustness and the fractality of perfusion [14] which could govern the standard deviation.
Figure 3. Perfusion maps in mL · (g · min)−1 . For every animal the maps generated after parallel and perpendicular slice selective inversion are shown for two slightly tilted slice positions. It is visible that for the perpendicular inversion slice set-up the acquired perfusion values are generally higher and the inter slice variation between two similar slices is lower.
ARTICLE IN PRESS F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
5
Table 1 Median ± MAD (median absolute deviation) for perfusion values in the cortex of the left and right kidney [mL · (g · min)−1 ] for the two slices and two selective inversion methods and the deviation between two different slices relative to their corresponding mean. Animal
Slice
1
⊥
1 2 Deviation [%] 1 2 Deviation [%]
2
3
Left
Right
Left
Right
Left
Right
5.7 ± 1.2 2.9 ± 0.9 65.1 4.8 ± 1.3 4.1 ± 1.3 16.0
4.8 ± 1.5 2.9 ± 0.8 50.1 3.7 ± 1.1 3.7 ± 1.2 0.3
3.4 ± 1.1 3.0 ± 1.1 11.1 5.2 ± 1.6 5.2 ± 1.5 0.3
3.3 ± 1.0 2.8 ± 1.0 17.2 5.6 ± 1.4 5.3 ± 1.6 5.3
5.5 ± 1.4 4.1 ± 1.3 29.1 5.7 ± 1.6 5.4 ± 1.7 5.7
5.3 ± 1.6 3.5 ± 1.4 39.1 5.9 ± 1.6 5.5 ± 1.7 7.6
The mean deviation between the slightly different imaging slice orientations is 35% in the parallel measurement and 6% in the perpendicular measurement. Additionally, mean perfusion values are 32% higher for the perpendicular inversion than the parallel inversion slice measurement. The higher sensitivity, which is expressed by the higher perfusion values, as well as the improved stability, shown by the reduced deviation between slightly changed slice positions, are in good accordance to the expectations of an improved control experiment.
Discussion The aim of this work was to provide a straightforward and effective solution to the issue of quantitative renal perfusion measurements in coronal view using FAIR ASL as noted in [2,5,9]. It was shown that the quality and stability of perfusion maps can be improved without increasing the complexity of the measurement or the post processing. This was achieved by decoupling the imaging slice orientation from the selective inversion slice orientation in the control experiment. By changing the orientation of the selective inversion experiment but not it’s position relative to the imaging slice the method remains insensitive to magnetization transfer effects (MTC) as the inversion pulses for both the labeling and the control experiment are on-resonant within the region of interest. The higher perfusion values in the perpendicular orientation of the selective inversion slice support the original thesis that the parallel selective inversion slice labels too much of the inflowing blood. Perfusion values were acquired for both the standard parallel orientation of the inversion slice and a perpendicular orientation. The mean value over all kidneys was 3.9±1.1 mL · (g · min)−1 for the standard and 5.0 ± 0.8 mL · (g · min)−1 for the perpendicular orientation. The values for the parallel orientation of the selective inversion slice correspond to the values reported previously 3.97 ± 0.36 mL · (g · min)−1 [2] and 4.4 mL · (g · min)−1 [4]. The values for the perpendicular orientation are higher than the previously reported values acquired using FAIR
ASL, but not as high as the values reported in [5] 5.66 ± 0.48 mL · (g · min)−1 and within the range of the values reported in [6] (4.66 ± 1.11 to 6.79 ± 1.49) mL · (g · min)−1 that were acquired using pseudo-continuous ASL [15] (pCASL) with varying read out methods. In [2–4] perfusion is calculated from the difference between images acquired in the selective and non selective measurement. While a different approach for quantification was used here, the experimental results are relevant to these techniques as well, as they profit from the increased stability of the control experiment and a larger difference between the control and labeling experiment. The T1∗ -based approach was chosen as it is able to reconstruct perfusion values even with varying inversion efficiency within the imaging slice in the control experiment. This is desirable since it allows a thinner selective inversion slice as perfusion values can still be obtained at the edge of the selective inversion slice. For preclinical studies the best inversion slice thickness should be assessed beforehand, taking into account the maximum size of the kidneys of the chosen animal model. It is advisable to reduce the slice thickness as much as possible in order to minimize the influence of unintentionally labeled inflowing blood. The thickness of 12 mm is a good compromise between minimal blood labeling in the selective inversion experiment and full coverage of the kidneys in coronal view for the chosen animals. The deviation between two very similar slices was used to show the improved stability of the perfusion values acquired using the perpendicular slice selective orientation. This effect can be attributed mainly to the fact that a slight tilt of the selective inversion slice affects a much larger portion of the mouse in the parallel than in the perpendicular setup (see Fig. 1b). Given the aim of this work acquisition parameters were not optimized for short protocol length. In [10] it was shown that reliable perfusion values could be obtained using as low as eight inversions despite triggering to the cardiac cycle. This would lead to a reduction in protocol length by 50%. As no triggering to the heart cycle is necessary in renal perfusion measurements, even further reduction is possible without changing acquisition parameters. Optimizing these
ARTICLE IN PRESS 6
F.T. Gutjahr et al. / Z. Med. Phys. xxx (2017) xxx–xxx
could accelerate the acquisition time for each inversion leading to additional shortening of the protocol length. In [5] another possible solution for the problem of renal perfusion in coronal view was proposed and several advantages for pCASL in comparison to a single TI-implementation of FAIR ASL were shown and the problematic control experiment in the standard FAIR ASL approach in renal perfusion measurements in coronal view was noted. This was mainly attributed to the coupling of selective inversion slice orientation to the imaging slice. By removing this constraint, reliable quantification and an increased sensitivity in coronal view with FAIR ASL could be demonstrated in our work.
Conclusion In this paper a straightforward and effective way to improve the quality of FAIR ASL kidney perfusion measurements has been demonstrated. By decreasing the amount of unwanted labeling of the inflowing blood pool in the perpendicular orientation of the inversion slice, the difference between the control and labeling experiment is increased, leading to higher sensitivity. Also the method becomes more stable against variations in the slice positioning, which is of great importance for comparability in longitudinal or inter animal measurements.
Acknowledgements The authors would like to thank Sabine Voll for animal handling. This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB 688 B5, Z2) and the Bundesministerium für Bildung und Forschung (BMBF01 EO1004).
References [1] Bennett HF, Li D. MR imaging of renal function. Magn Reson Imaging Clin N Am 1997;5(1):107–26. [2] Rajendran R, Lew SK, Yong CX, Tan J, Wang DJJ, Chuang KH. Quantitative mouse renal perfusion using arterial spin labeling. NMR Biomed 2013;26(10):1225–32.
[3] Winter JD, St Lawrence KS, Cheng HLM. Quantification of renal perfusion: comparison of arterial spin labeling and dynamic contrastenhanced MRI. J Magn Reson Imaging 2011;34(3):608–15. [4] Hueper K, Gutberlet M, Rong S, Hartung D, Mengel M, Lu X, Haller H, Wacker F, Meier M, Gueler F. Acute kidney injury: arterial spin labeling to monitor renal perfusion impairment in micecomparison with histopathologic results and renal function. Radiology 2014;270(1):117–24. [5] Duhamel G, Prevost V, Girard OM, Callot V, Cozzone PJ. Highresolution mouse kidney perfusion imaging by pseudo-continuous arterial spin labeling at 11.75T. Magn Reson Med 2014;71(3):1186–96. [6] Prevost VH, Girard OM, Callot V, Cozzone PJ, Duhamel G. Fast imaging strategies for mouse kidney perfusion measurement with pseudocontinuous arterial spin labeling (PCASL) at ultra high magnetic field (11.75 tesla). J Magn Reson Imaging 2015;42(4):999–1008. [7] Kim SG. Quantification of relative cerebral blood flow change by flowsensitive alternating inversion recovery (FAIR) technique: application to functional mapping. Magn Reson Med 1995;34(3):293–301. [8] Wang JJ, 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(6):1783–91. [9] Cutajar M, Thomas DL, Banks T, Clark Ca, Golay X, Gordon I. Repeatability of renal arterial spin labelling MRI in healthy subjects. Magn Reson Mater Phys Biol Med 2012;25(2):145–53. [10] Gutjahr FT, Kampf T, Winter P, Meyer CB, Williams T, Jakob PM, Bauer WR, Ziener CH, Helluy X. Quantification of perfusion in murine myocardium: a retrospectively triggered t1-based ASL method using model-based reconstruction. Magn Reson Med 2015;74(6):1705–15. [11] Deichmann R, Haase A. Quantification of T1 values by snapshot-flash NMR imaging. J Magn Reson 1992;96:608–12. [12] 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(6):1240–5. [13] Kampf T, Helluy X, Gutjahr FT, Winter P, Meyer CB, Jakob PM, Bauer WR, Ziener CH. Myocardial perfusion quantification using the t1-based fair-ASL method: the influence of heart anatomy, cardiopulmonary blood flow and look-locker readout. Magn Reson Med 2014;71(5):1784–97. [14] Bauer WR, Hiller KH, Galuppo P, Neubauer S, Köpke J, Haase a, Waller C, Ertl G. Fast high-resolution magnetic resonance imaging demonstrates fractality of myocardial perfusion in microscopic dimensions. Circ Res 2001;88(3):340–6. [15] Dai W, Garcia D, de Bazelaire C, Alsop DC. Continuous flow driven inversion for arterial spin labeling using pulsed radiofrequency and gradient fields. Magn Reson Med: Off J Soc Magn Reson Med/Soc Magn Reson Med 2008;60(6):1488–97.
Available online at www.sciencedirect.com
ScienceDirect