Functional MR imaging for the native and transplanted kidney can be performed without injecting a contrast agent and may be helpful in the longitudinal assessment of kidney function in patients with renal disease. Before these methods will be clinically useful, applicable techniques must be developed. Our group has implemented a FAIR-ASL technique to measure perfusion in both native and transplanted kidneys over a broad range of function. Specifically we chose a sagittal orientation in many of the transplanted kidneys to avoid major feeding vessels, utilized a respiratory triggered/coached breathing acquisition, and performed automatic image registration using NMI. We were able to measure the cortical and medullary perfusion in all but 1 of 35 kidneys with this technique and found correlations between cortical perfusion and eGFR. Groupwise comparisons indicate cortical perfusion differences between native and transplant subjects with eGFR > 60 ml/min/1.73m2 and medullary perfusion differences for both eGFR > 60 and eGFR < 60 ml/min/1.73m2. We also successfully implemented a free breathing protocol in a subset of transplanted kidneys to assess robustness of the technique in patients who may not be able to breathe at a consistent rate or follow respiratory coaching (i.e. very ill patients).
While perfusion values in this study are similar to other renal ASL work [6
], cortical perfusion values for healthy native subjects in this study are slightly higher than values reported by Fenchel et al. using a similar pulse sequence [6
]. This may partly be due to variation in acquisition and processing. As for acquisition differences, the present study allowed two additional seconds (for a total of 5 s) for magnetization recovery between inversions. This is more representative of the compartment modeling, which assumes complete recovery and should provide a larger difference between tag and control, increasing the perfusion value. As for processing differences that would lead to higher perfusion values, a more inclusive threshold for cortical perfusion of 1000 ml/min/100g (only < 2% of pixels were rejected) was used compared to 600 ml/min/100g used in the Fenchel et al. study.
Prospective respiratory coaching and triggering was applied to minimize motion artifacts, but it proved necessary to use post-acquisition image registration in most cases and this was critical in certain native and transplant subjects. Registering the MC
images eliminated blurring, allowing better tissue segmentation and perfusion measurement. Without registration, the perfusion-weighted difference image would be blurred throughout the kidney as well at the kidney border, where blurring could lead to artificially low cortical perfusion values. Retrospective respiratory sorting was an alternative, as it has been shown to reduce motion artifacts in renal ASL imaging compared to a simple coached acquisition (without respiratory triggering) [14
]. With a FAIR acquisition, the inversion slab must encompass the imaging slice so prospective respiratory triggering was used in addition to coaching for this study.
The oblique-sagittal orientation was found to be more practical for transplant subjects whose feeding vessel trajectories precluded the possibility of a coronal acquisition. For example, in subjects with native kidneys, an oblique slice could be positioned both coronal to the kidneys and posterior to the renal arteries allowing coverage of both kidneys with one FOV. However, in many transplanted kidneys, a coronal slice orientation placed the feeding vasculature in-plane. In these cases, a coronal acquisition would have led to artificially low perfusion measurements by slice-selectively inverting the inflowing blood spins which are presumed to be at equilibrium magnetization. An oblique-sagittal slice omitted major feeding vessels in these transplant subjects and allowed for sufficient coverage of the transplanted kidney (). Slice orientation would not be restricted for other ASL techniques that label the blood upstream, such as proximal inversion with a control for off-resonance effects (PICORE) or pulsed-continuous methods. Unfortunately, labeling via these methods would lengthen the transit time of the tagged blood and increase the variability in arrival time from subject to subject. Since diseased kidneys are likely to demonstrate more variable flow, adjusting the delay time is also a source of potential error that is avoided by use of the FAIR technique.
The respiratory triggered and coached perfusion measurements in this investigation were comparable to a study by Lanzman et al. [9
] which used a free-breathing FAIR-ASL approach in transplanted kidneys. The free-breathing, sagittal perfusion measurements in this study were also similar to respiratory triggered and coached measurements, although slightly lower. Thus free-breathing may be a reasonable option for patients who are unable to follow the regimented breathing pattern. The free-breathing perfusion measurements may be lower because the slice selective inversion thickness was larger than the standard 20 mm to increase the tolerance for through-plane motion. Consequently, more blood spins outside of the imaging slice were inverted when they flowed into the imaging plane.
As expected there was a positive correlation between cortical perfusion measurements and eGFR in both transplant and native kidneys supporting the regulation of glomerular filtration rate by renal blood flow. However, perfusion values were significantly reduced in transplanted kidneys compared to native kidneys for subjects with eGFR > 60 ml/min/1.73m2
(). This may result from differential regulation of blood flow in transplanted kidneys or the vasoconstrictive effects of calcineurin inhibitors that are commonly used in kidney transplantation to prevent rejection [21
] and is an area of future study.
Several limitations apply to the current study. No gold-standard was available to compare the perfusion results, however other studies using FAIR-ASL in the kidneys have shown good correlation with standard references [6
]. The results in this work were obtained from a limited number of native and transplant subjects and must be extended to a larger population before drawing firm conclusions regarding practicality and clinical significance. This study was performed on a 1.5T system which has limited SNR compared to a 3T system but should have fewer banding artifacts. Banding was visible in four of the kidneys, although only a small portion of the kidney body was affected, and unaffected measurements were obtained for the majority of the kidney. Higher order shimming or adjusting the center frequency of the radiofrequency pulse would likely shift these artifacts outside of the kidney [27
]. This FAIR ASL sequence did not apply bi-polar gradients before readout to suppress signal from tagged spins in the arterioles. This may have introduced a bias towards higher perfusion measurements as well.
Perfusion values presented in this study were determined using a one compartment model which requires many assumptions, such as rapid water exchange between the intravascular and extravascular space. A two compartment model [28
] would reduce those assumptions and allow more accurate perfusion quantification, however it requires measurements at multiple delay times which is not possible to achieve in a clinically feasible scan time. Further assumptions used in this study for perfusion measurement include a constant tissue T1
and inversion efficiency, α. In reality, the α can vary depending on rf-amplifier performance, stability and pulse design, while the T1
is a patient-specific parameter that may vary regionally and with kidney disease and setting. The cortical T1
assumed in this study may have introduced a bias towards higher perfusion for lower functioning kidneys because cortical T1
can increase with renal insufficiency [29
]. Incorporating a patient-specific measured T1
and pulse-specific α into the model should increase the quantitative accuracy of these measurements with the trade-off of slightly increased scanning time.
In conclusion, FAIR-ASL was able to measure renal perfusion in subjects with native and transplanted kidneys, potentially providing a clinically viable technique for monitoring kidney function. Results from this study demonstrate that medullary perfusion was systematically lower in transplant vs native kidneys, and that cortical perfusion was lower for transplanted kidneys as well for kidneys with higher function, eGFR > 60 ml/min/1.73m2. Cortical perfusion correlated with eGFR in both native and transplanted kidneys. Future work will compare FAIR-ASL with microsphere perfusion in a swine model, assess reproducibility in human subjects, and measure the ability of this technique to characterize renal disease and provide earlier detection of functional change.