New magnetic resonance imaging methods in nephrology

Abstract

Established as a method to study anatomic changes, such as renal tumors or atherosclerotic vascular disease, magnetic resonance imaging (MRI) to interrogate renal function has only recently begun to come of age. In this review, we briefly introduce some of the most important MRI techniques for renal functional imaging, and then review current findings on their use for diagnosis and monitoring of major kidney diseases. Specific applications include renovascular disease, diabetic nephropathy, renal transplants, renal masses, acute kidney injury and pediatric anomalies. With this review, we hope to encourage more collaboration between nephrologists and radiologists to accelerate the development and application of modern MRI tools in nephrology clinics.

INTRODUCTION

Long established as a method to study structural changes in disease, such as renal tumors or atherosclerotic vascular disease, magnetic resonance imaging (MRI) to interrogate renal function has only recently begun to come of age. Functional renal MR approaches render added value for MRI over other conventional imaging modalities, with emerging applications in nephrology. As an example, low doses of the conventional contrast agent used in MRI − gadolinium chelates − can be used to measure glomerular filtration rate in individual kidneys ^{1, 2}. Techniques borrowed from functional brain mapping known as Blood Oxygen Level Dependent (BOLD) MRI have recently been applied to interrogate renal oxygenation and metabolic rate ^{3, 4}. Renal perfusion can be measured using MRI either by intravenous injection of an exogenous tracer or with endogenous blood labeling methods that do not require contrast injection.

In this article, we briefly review the principles of MR imaging, introduce some of the most important MRI techniques, and then review current findings on their use for diagnosis and monitoring of major kidney diseases. Specific applications include renovascular disease, diabetic nephropathy, renal transplants, renal masses, acute kidney injury and pediatric anomalies. Our purpose is to accelerate the application of modern MRI tools in the clinic and to strengthen the collaboration between MRI physicists, radiologists and nephrologists. We refer the reader to other reviews ^5–14 and the cited papers for more technical details.

MRI PRINCIPLES AND TECHNIQUES

Modern clinical MRI scanners are typically equipped with a large superconducting magnet, which provides a stable homogeneous magnetic field (1.5 – 3.0 Tesla), and multiple coils for different purposes, including signal transmission, reception and creating magnetic field gradients. For abdominal imaging, surface coils are available to be applied to the abdomen to improve signal reception. Due to the strong magnetic field, pacemakers are usually contraindicated with MRI scans, but most vascular stents are compatible.

Most MRI signals originate from hydrogen nuclei or protons, which behave like tiny magnets in the MRI system’s magnetic field and their behavior in response to changing magnetic fields forms the basis of MRI. MRI image contrast is achieved by manipulating the magnetic properties of hydrogen protons and thereby distinguishing among various tissue characteristics, including intrinsic MR properties like the relaxation times T₁ and T₂. Compared to tissues, water (e.g. cysts, cerebrospinal fluid) typically has longer T₁ and T₂ times (Figure 1).

Figure 1.

Conventional anatomic MRI of kidney. The cyst, with long T₁, is dark on T₁-weighted image (A) and, with long T₂, bright on T₂-weighted image (B).

Using MRI we can also visually distinguish tissues based on other inherent tissue properties such as diffusivity of water within those tissues, capillary perfusion, blood flow or velocity, and even oxygenation. Additionally, with the administration of exogenous contrast agents, other tissue characteristics can be explored. Exogenous agents like gadolinium (Gd) based chelates are useful for MR angiography and in MR renography because they shorten T₁ relaxation time, and for many years were regarded as one of the safest agents used in medicine. About a decade ago, the associations between high doses (typically double and triple the standard doses) of Gd contrast in patients with renal failure and nephrogenic systemic fibrosis (NSF) were first reported ¹⁵. Recommendations have been made on the use of Gd contrast in patients with renal impairment ^{16, 17}. In the last few years with adherence to guidelines, no report of new NSF cases has been published ¹⁸.

Below we review some of the MRI techniques most widely explored for applicability to functional renal imaging. Their characteristics are summarized in Table 1.

Table 1.

A summary of major MRI techniques and their capabilities

Techniques	Capability	Parameters
Dynamic contrast enhanced MRI (DCE MRI)	Tracer transit through vascular space and tubules	GFR, perfusion, vascular and tubular mean transit times (MTT)
Blood oxygen level dependent (BOLD)	Direct measure of deoxyhemoglobin, and reflects blood and tissue pO2	Spin-spin relaxation rate (R2* = 1/T2*), medulla-cortex R2* ratio (MCR =R2*Med/R2*Cx)
Arterial spin labeling (ASL)	Perfusion without injecting tracer	Perfusion
Diffusion weighted imaging (DWI)	Water diffusion in interstitial space; capillary flow	Apparent diffusion coefficient (ADC), anisotropy, perfusion fraction

MR Renography (MRR) and Dynamic Contrast-Enhanced (DCE) MRI

MR renography (MRR) is a term used to describe one application of dynamic contrast-enhanced (DCE) MRI, specifically, the use of Gd-based contrast agents for the noninvasive measurement of glomerular filtration rate (GFR). Most Gd chelates have favorable renal properties: freely filtered at the glomeruli without tubular secretion or resorption. This means that with the continuous acquisition of high resolution images through the kidney every few seconds, renal function can be visualized as the passage of contrast material from the aorta through the kidney and out of the collecting system. Calculation of the extraction fraction of the Gd contrast allows determination of GFR. Several technical limitations have slowed the widespread adoption of this method, although recent developments are promising. We review some of the challenges below.

Unlike CT or nuclear medicine, where tracer concentration is approximately linear to signal intensity, in DCE MRI, the relationship is more complex and it depends on the specific pulse sequence used. Methods converting MR signal intensity to concentrations of tracer have been reviewed elsewhere ^{19, 20}. Because of the sensitivity of MRI to Gd contrast agents, 3–4 ml of Gd contrast agent suffices for accurate renal functional measurements. In MRR, signal intensities can be recorded from the whole kidney to determine GFR. For more refined measurements, signal intensities can be recorded from the renal cortex, medulla and collecting system. With more detailed data, the transit of the bolus of contrast from the artery to the renal cortex can be used to estimate renal perfusion; the transit from cortex to medulla reflects glomerular filtration; and finally from the medulla into the collecting system reflects tubular function.

To extract these functional measurements from signal intensities in the kidney, the method typically assumes a tight bolus of contrast entering the renal artery. Gd contrast is administered intravenously, and therefore accurate measurement of renal function depends on knowing the shape of the bolus of contrast as it arrives in the abdominal aorta, following dispersion during transit through the pulmonary circulation, heart, and thoracic aorta. Direct measurement of the bolus as it arrives at the level of the renal arteries can be measured from images of the abdominal aorta and is termed as arterial input function (AIF). Using a mathematical operation called deconvolution we can eliminate the variable effects of bolus dispersion, as quantified by AIF, on the measurements of tracer in the kidney. Compartmental modeling is typically used to extract renal physiological parameters like GFR from enhancement curves. The details of the models and their comparison can be found in literature ^{2, 21–25}.

Several studies have shown good agreement between MRR measurements of GFR and reference methods. Hackstein et al ²⁶ found a correlation coefficient of 0.83 between MRR-GFR and GFR measured by iopromide plasma clearance. Using low Gd dose (4 ml) and multiple-compartment modeling technique, Lee et al ^{2, 21, 24} measured MRR-GFR with correlation coefficient of 0.82–0.84 with ^99mTc-DTPA plasma clearance.

One major issue with MRR is imprecise measurement of arterial input function (AIF). Cutajar et al ²⁷ found that AIF errors could severely lower the precision of the estimated GFR and renal blood flow. Zhang et al ²⁸ proposed a technique of using patient’s cardiac output (measured using phase-contrast MRI) and increased the correlation coefficient between two independent MRR-GFR measurements from 0.83 to 0.92. The cardiac-output approach ²⁸ is promising in correcting for AIF errors, but measurement of the patient’s cardiac output with phase-contrast MRI can be cumbersome ²⁹. The other parameters such as vascular and tubular mean transit times (MTT) measure the time it takes for unfiltered and filtered tracers to go through a kidney, respectively. While clinically able to distinguish different renal pathologies ³⁰, the accuracy of these parameters has not been validated.

Another hurdle to the clinical use of MRR is the absence of dedicated processing software on commercial workstations. For pediatric applications, several programs are freely downloadable on the web ^{31, 32}. Additionally, for analysis of adult data, recent software developments appear promising in their simplicity of use and potential widespread dissemination ^{33, 34}.

Blood oxygen level dependent (BOLD)

Renal BOLD MRI is a method similar to functional imaging in the brain ³⁵ that forms images of a particular tissue characteristic, the transverse relaxation time constant T₂*. T₂* is strongly affected by deoxyhemoglobin, whose paramagnetic effect shortens T₂*. This gives a potential window into the oxygen level throughout the kidney. Some investigators prefer the use of R₂* (=1/T₂*). Higher R₂* (or low T₂*) theoretically corresponds to higher deoxyhemoglobin, and in turn, lower tissue pO₂ level.

Several groups have demonstrated the sensitivity of BOLD imaging to different physiologic states in both human and animal kidneys. Hypoxia in the renal medulla of rat and pig models, measured with oxygen probes, has been shown to improve after administration of the loop diuretic furosemide ^{36, 37}. Corresponding increases in medullary T₂* (increase in oxygenation) with furosemide administration or with water diuresis have been reported in animal models ^{37, 38} and normal humans ³. Some studies have shown that the changes in T₂* after furosemide administration or water diuresis are attenuated in diabetic kidneys ^{39, 40} and in elderly subjects ⁴¹. Some other studies have shown that the basal T₂* values (without diuretic challenge) negatively correlate with the degree of diabetic nephropathy (eGFR level in humans ⁴², and days after induction of diabetes in a rat model ⁴³).

Given the central role of hypoxia in the progression of chronic kidney disease ^44–46, the prospect of assessing renal parenchymal oxygenation level by non-invasive techniques is exciting. However, published renal BOLD data appear to be somewhat contradictory, with some studies failing to show these aforementioned effects ^{47, 48}. The contradictions may be partly due to the technical challenges of BOLD MRI. One such challenge is the low signal-to-noise ratio (SNR) of conventional renal BOLD images. The kidneys move during respiration, and this excursion limits BOLD imaging to a single breath hold. A short imaging time results in the low SNR and consequently decreases in the precision of T₂* estimation. To overcome this problem, we have recently developed a new BOLD imaging technique during free breathing ⁴⁹ which allows longer imaging times, thus improving the T₂* precision significantly (Figure 2). Another factor that may have caused the contradiction in the publications is the multiple confounding factors other than tissue pO₂ that influence T₂* ^{50, 51}, such as blood perfusion, intrinsic transverse relaxation not due to deoxyhemoglobin, volume fraction of blood in a voxel and magnetic field strength (1.5T vs 3T). To quantify absolute tissue oxygenation based on BOLD T₂*, it is necessary to measure the above confounding factors in vivo and address their effects. In addition, the BOLD signal, which is primarily due to the microscopic susceptibility effect of deoxyhemoglobin, could sometimes be superimposed by signals from macroscopic susceptibility artifact. Magnetic field shimming to improve the magnetic field homogeneity ⁵² can help reduce the impact of such artifact.

Figure 2.

Kidney T₂* maps from BOLD imaging. The scale on the right side is T₂* value, with unit of milliseconds. Higher T₂* corresponds to lower deoxyhemoglobin concentration. A) Typical T₂* map with conventional BOLD scan (20-sec breath-hold, matrix size 256 × 256, FOV 32 × 32 cm). B) T₂* map with free-breathing prospectively navigated sequence (10-minute imaging time, matrix size 512 × 512, FOV 50 × 50 cm). The free-breathing images offer greater image quality.

Arterial spin labeling (ASL)

Originally used in the brain for perfusion and functional imaging, arterial spin labeling (ASL) MR imaging ^{53, 54} uses arterial blood as an endogenous contrast agent, and obviates any exogenous agent as in DCE MRI. In this approach, the MR system is used to regionally label inflowing arterial blood by altering its magnetic state. The labeled blood then transits to the tissue where its magnetization affects the measured signal intensity, and the degree of signal change reflects tissue perfusion.

Typically, ASL requires two types of images: label and control. The two images are acquired in exactly the same way, except that before acquiring the label image, magnetization of some arterial blood is altered or labeled. Subtraction of the control from the label images provides a difference image, and the difference is solely due to the labeled magnetization of arterial blood in the part of the body being imaged. Examples of difference image are shown in Figure 3 ^{55, 56}. To estimate renal perfusion, the labeling is typically done at abdominal aorta, and the labeled blood transits through renal vascular space, like an exogenous tracer. Unlike an exogenous tracer, the altered magnetization of labeled blood recovers toward its baseline pre-labeled state within a few seconds due to T₁ relaxation. This means that the transit time for the labeled arterial blood to move into the perfused organ should be short enough to maintain measurable difference signals.

Figure 3.

Difference images obtained from renal ASL scans. A) Acquired at 800 ms after arterial blood labeling, when the labeled blood is mostly in renal cortex; B) 1000 ms after the labeling, and some labeled blood reaches renal medulla. The images were acquired by a modified TrueFISP FAIR ASL sequence (8 averages, acquisition time ~24 sec, with breath hold).

Due to concerns about nephrogenic systemic fibrosis (NSF) ^57–59, non-contrast imaging techniques such as ASL are particularly attractive for patients with advanced renal diseases as well as allograft dysfunction, where it can be used for longitudinal perfusion evaluation (see ‘Renal Transplants’ section). One technical challenge with kidney ASL is the respiratory motion associated with long scan time (minutes) for achieving sufficient SNR. These errors can be mitigated, however, using a variety of approaches including: synchronized breathing, respiratory triggering, or retrospective sorting based on respiratory position ^60–62. For example, Gardener and Francis ⁶³ proposed a rapid multi-slice ASL technique for the kidneys and used retrospective image sorting to correct for the respiratory motion artifact.

Recent studies have tested ASL in native and transplanted kidneys with both normal and altered function ^{55, 60–62, 64–81}. Two main ASL methods have been used: FAIR ASL and pseudocontinuous ASL (pCASL), details of which can be found elsewhere ^{82, 83}. Renal perfusion measured using a FAIR ASL scheme, first demonstrated in the kidneys by Martirosian et al. ⁵⁵, has correlated with renal artery stenosis grades ⁶⁹, renal plasma flow ^{69, 76}, and microsphere perfusion measurements ⁶⁵. Artz et al. ⁶⁴ found that FAIR-ASL yielded reproducible cortical perfusion results in native and transplanted kidneys, although the reproducibility for medulla was moderate to poor. ASL has detected significantly lower perfusion in allografts vs. healthy kidneys ⁶⁰, native diseased vs. healthy kidneys ^{74, 78}, and in renal allografts experiencing an acute decrease in renal function (>20% increase in serum creatinine) compared to allografts with good function (serum creatinine < 2mg/dl) ⁷³. A background-suppressed, pseudocontinuous ASL (pCASL) method ⁶¹ has helped characterize renal masses ⁷⁵ and enabled distinction between histopathological diagnoses such as oncocytomas and renal cell carcinomas ⁷². Finally, ASL measurements have correlated with clinical outcome in patients with metastatic renal cell carcinoma undergoing antiangiogenic therapy ⁶⁷.

Overall, while early clinical ASL results appear promising for differentiating different disease states, validation and development of robust quantitative perfusion methods remain under investigation.

Diffusion weighted MRI (DWI)

Diffusion-weighted MR imaging probes micron-scale water motion in tissue ⁸⁴. During the imaging period, relative displacement of water molecules due to diffusion in the presence of magnetic field gradients results in a decay in the DWI signals. Such decay can be quantified by the “apparent diffusion coefficient” (ADC), which is computed as the decay constant of the exponential signal decay. ADC is influenced by microstructural barriers. In the case of anisotropic diffusion in ordered structures such as renal medulla, diffusion tensor imaging (DTI) ^85–87 can be applied to measure both the magnitude and the 3D direction of diffusion (Figure 4). DTI requires more data using multiple gradient directions. In another variant of DWI with MR parameters designed to be highly sensitized to perfusion, referred to as intravoxel incoherent motion (IVIM) imaging ⁸⁸, both tissue diffusion and perfusion-like characteristics can be extracted from DWI images.

Figure 4.

Kidney diffusion-weighted imaging using DTI methods. Following imaging processing, color-coded primary diffusion eigenvectors display radial pattern of medullary tubules.

DWI has recently found increasing application extracranially including characterization of kidney function ^89–91. Progress to date indicates promising sensitivity of DWI to renal function ^{70, 92}, but further experiments and analysis are needed to disentangle the relevant biophysical mechanisms and yield further diagnostic specificity.

APPLICATIONS FOR KIDNEY DISEASES

Renovascular Diseases

Renal MR angiography has long been established as an accurate, clinically-acceptable method for depicting renal artery stenosis ^93–95. Because of the high incidence of asymptomatic renal artery stenosis, several methods have been investigated as adjuncts to anatomic imaging to help determine the functional significance of the stenosis, to monitor therapy, and to develop predictive indices to identify patients likely to benefit from revascularization.

To assess the functional significance of the renovascular disease (RVD), renal blood flow (RBF) can be measured using ASL or DCE MRI, using as Gd contrast agents or ultra-small paramagnetic ironoxide (USPIO) particles that stay in the intravascular space. Using USPIO-enhanced imaging, Schoenberg et al found that with renal artery diameter narrowing less than 80%, intrarenal cortical perfusion did not change significantly (average 513 ml/100 g/min), while artery narrowing more than 80% caused a fall of more than 200 ml/100 g/min in cortical perfusion^{96, 97}. ASL-MRI uses spin-labeled arterial blood as the tracer, thereby avoiding the potential adverse effects of exogenous contrast agents. Its value in RVD remains to be determined.

The management of RVD, often asymmetric in nature, may benefit from the determination of single-kidney GFR using MRR. The detection of hemodynamically significant renal artery stenosis (RAS), often corresponding to >70% decrease in diameter, can be facilitated by the administration of angiotensin converting enzyme inhibitors (ACEI). ACEI attenuates GFR and thus the magnitude of signal enhancement in the presence of significant renal artery stenosis ⁹⁸, but this attenuation is difficult to determine in subjects with low basal signal enhancement. To overcome this barrier, a multi-compartmental modeling method for analyzing dual-injection MRI data has been developed to allow for GFR determinations in human kidneys with significant stenotic renal arteries and has shown that ACEI caused a significant decrease in GFR averaging ~26% in a group of kidneys with RAS≥50% ⁹⁹.

The effects of RVD on intrarenal hypoxia have been evaluated using BOLD MRI. Although it is relatively preserved in moderate RVD ¹⁰⁰, a decrease renal oxygenation becomes evident once severe stenosis develops ^{4, 101}. This is possibly because in severe vascular occlusion that threatens cortical perfusion, processes of compensating tissue oxygenation become overwhelmed. Histogram-based analysis over large cortical and medullary regions can be used to evaluate the distribution of tissue oxygenation in these regions and reduces sampling error ¹⁰². Moreover, a furosemide challenge enables the study of tubular oxygen-dependent transport ¹⁰³, which is blunted in damaged kidneys ¹⁰⁴, but enhanced in hyper-filtering kidneys ¹⁰⁵. Hence, BOLD MRI may be useful to assess the functional integrity of the renal medullary tubules ¹⁰⁶.

Other methods have also been explored in RVD. Magnetic resonance elastography (MRE) utilizes the translocation of mechanical shear waves to estimate elasticity, which may decrease in fibrotic kidneys. Interestingly, in swine with RVD overall kidney stiffness is unaltered, because a fall in RBF reduces renal turgor and masks decreased elasticity ¹⁰⁷. In contrast, medullary elasticity appears to be less dependent on hemodynamic variables, and may reflect kidney fibrosis ¹⁰⁸.

DWI detects changes in tissue property based on its sensitivity to restriction of free-water diffusion. In patients with RVD ¹⁰⁹, but not hypertension alone ¹¹⁰, apparent diffusion coefficient (ADC) declines and correlates with the extent of RVD, suggesting that significant kidney injury is required to become detectable by DWI.

The noninvasive and versatile nature of MRI positions this modality at the forefront for evaluation of RVD in humans. The rapid progress in this field will hopefully inspire the development of molecular and metabolic probes to assess mechanisms of injury and viability of the post-stenotic kidney. These developments would be for identifying the subset of patients with RVD who may benefit from revascularization of stenotic renal arteries ¹¹¹. The determination of tissue perfusion and oxygenation may facilitate the decision to perform the revascularization procedure and monitoring of the kidney responses after the procedure.

Diabetic Nephropathy (DN)

Recent advances suggest that progressive chronic kidney disease (CKD) eventually results in peritubular capillary injury, tubular hypoxia and atrophy, and interstitial fibrosis ¹¹², independent of the type of underlying primary kidney disease. Diabetic nephropathy (DN) is the most common form of CKD, and functional renal MRI is an attractive opportunity for noninvasive diagnosis and monitoring of potential therapeutic interventions. Most studies have focused on hypoxia using BOLD MRI ^{113, 114}, fibrosis using DWI, tubular damage using DTI ^{115, 116} or a combination of BOLD MRI and DWI ^{117, 118}. A recent preliminary report applied ASL to show reduced cortical blood flow in patients with CKD ⁸¹.

Studies using BOLD-MRI in rodent models suggest that, at least in the early stages, type 1 and type 2 diabetic nephropathy is associated with increased hypoxia ^118–120. DWI that measures water diffusion in the interstitial space, however, has failed to show any changes in early stage DN ¹¹⁸. Part of the reason may be related to the specific parameters of the diffusion MRI method ¹²¹.

Clinical results of BOLD MRI in DN have been strikingly variable. Consistent with rodent studies, one recent BOLD MRI study of 46 patients with type 2 diabetes using 3.0 T MRI found that the ratios of medullary-to-cortical R₂* (MCR) were higher in stages 1 and 2 CKD compared to controls ¹¹⁴, suggesting a greater than expected medullary tissue hypoxia relative to the cortex. Interestingly the MCR values were lower at later stages (3–5) of CKD compared to controls. The reason for this paradox is not apparent, but may provide interesting clues to understanding of the pathophysiology of DN. Another study of 20 diabetic patients (14 with stages 3–5 CKD) performed using 1.5 T MRI confirmed lower MCR values than healthy subjects ¹¹³. Yet another study of type-2 diabetic patients (CKD stages 1–4) using 3.0 T MRI found no change in MCR compared to controls reported in the literature ¹²². As noted above (BOLD section), the apparently contradictory results of BOLD imaging in these patients is likely due to technical challenges such as image artifacts and the oversimplification of interpretation of R₂* (or T₂*) values.

Diffusion-weighted imaging has also been applied to DN patients. A recent study of CKD patients with (n=43) or without (n=76) diabetes found a statistically significant correlation between ADC and estimated GFR values in both groups of patients ¹¹⁷. However, this correlation was only seen in the non-diabetic patients and not in the diabetic patients. A recent DTI study suggested changes in fractional anisotropy in the renal medulla with different levels of DN ¹¹⁶ probably related to glomerulosclerosis, interstitial fibrosis, and tubular damage ¹¹⁵.

Overall, these reports clearly demonstrate the complexities involved in translating results from animal models to humans. In addition to technical challenges inherent in the MR methods, these discrepancies may also be related to differences species, pathogenesis of diabetes, severity of kidney disease, comorbid conditions, use of medications such as renin-angiotensin system blockers, hydration and other preparations for imaging. Further refinement and validation of these techniques and their applications to larger, well-designed clinical studies ^{114, 117} may establish their utility in clinical research of DN and routine clinical use.

Renal Transplants

MRI has emerged as an attractive approach for evaluating the function of renal allografts due to its noninvasiveness and suitability for repeated application. The most promising results have involved near-term allograft complications, of which acute tubular necrosis (ATN) and acute allograft rejection (AR) are the most common. Szolar et al. ¹²³ observed that the first-pass cortical signal enhancement using DCE MRI was markedly reduced in allografts with AR compared to normally functioning kidneys, while allografts with ATN showed no difference compared to normal cases. Using a similar approach, Wentland et al. ¹²⁴ reported lower cortical and medullary blood flow in AR compared to normal kidneys and lower medullary blood flow in AR compared to ATN. Most recently, by applying a multi-compartmental tracer kinetic model to DCE MR images, Yamamoto et al. ³⁰ showed that mean transit times (MTT) could differentiate normal allografts from AR or ATN, where AR cases had higher ratio of vascular MTT over whole-kidney MTT while ATN cases had higher ratio of tubular MTT over whole-kidney MTT.

Several studies have investigated diagnosis of acute dysfunction using non-contrast techniques such as BOLD MRI. Sadowski et al. ¹²⁵ and Han et al. ¹²⁶ performed BOLD MRI in recent kidney transplant recipients and observed significantly lower medullary R₂*, or higher oxygenation, in cases of AR compared to ATN. As the authors suggested, this could be due to preferential blood shunting toward the medulla during acute rejection or reduced oxygen consumption rate due to subclinical medullary tubular injury or both. Together the results consistently show differentiation between ATN and AR based on perfusion and oxygenation parameters; however, robust differentiation of ATN from normally functioning allografts has not been demonstrated. Therefore, Chandarana et al. ⁶ have proposed a follow-up role for MRI only after acute dysfunction has been implicated by other tests such as elevated serum creatinine.

Studies of long-term allograft function have combined multiple functional MRI techniques, including BOLD, DWI and ASL MRI. In a cross-sectional BOLD MRI study of healthy volunteers and transplant recipients with chronic allograft nephropathy (CAN), Djamali et al. ¹²⁷ observed significantly reduced cortical and medullary R₂*, indicating increased oxygenation, in allografts affected by CAN. Long-term longitudinal studies, however, have produced mixed results. Vermathen et al. ¹²⁸ reported stable DWI parameters and an increase of cortical R₂* (decrease in oxygenation) in allografts between 7 and 32 months post-transplant. In a small pilot study involving matched donor and recipient pairs, Malvezzi et al. ¹²⁹ observed a reduction in cortical R₂* in both groups and a reduction in medullary R₂* (increase in oxygenation) in the transplanted kidney 1 month following transplant. In a recent study of 14 donor and recipient pairs, Niles et al. ¹³⁰ observed a reduction in both medullary R₂* (increase in oxygenation) and ASL-estimated cortical perfusion in transplanted kidneys 3 months post-transplant, and this reduction persisted for at least two years. The clinical significance of these changes was unclear, as all allografts were functioning well based on conventional clinical biomarkers. Further longitudinal studies with larger sample sizes will be necessary to determine whether long-term changes in MRI measures of allograft function are associated with clinical outcomes.

Renal Tumors

Renal masses are increasingly discovered incidentally, largely attributable to the increased use of medical imaging ^{131, 132}. Because tumors differ in biologic behavior, aggressiveness, and prognosis ^{133, 134}, their increased detection has led to a management dilemma. Accurate characterization of tumor aggressiveness can guide management. Although CT is the most widely used to diagnose renal lesions in clinical practice, advantages of MRI include superior soft tissue contrast, avoidance of ionizing radiation and iodinated contrast media, and most importantly the availability of different techniques such as DCE, DWI and BOLD MRI to probe different aspects of tumor such as vascularity, microstructure and oxygenation (6).

Widely used clinically, DCE MRI is one of the most robust techniques for evaluating the aggressiveness of renal tumors ^{135, 136}. Studies have shown that by imaging at three time points following contrast administration, the low level and homogeneous enhancement of papillary renal cell carcinoma (RCC) can help distinguish it from clear cell RCC ^{137, 138}. With a higher temporal resolution of ~30 sec per acquisition, a distinct pattern of enhancement was identified for angiomyolipomas: an early enhancement peak followed by lower level enhancement ¹³⁹. Using a 2-compartmental model to analyze the high temporal resolution data, Notohamiprodio et al. ¹⁴⁰ estimated perfusion and permeability of renal tumors. These parameters could help differentiate tumor subtypes and identify tumor features such as necrosis and vessel invasion.

Cellular renal lesions, such as renal cell cancer (RCC), restrict water diffusion in interstitial space, which explains the associated lower ADC values in the lesion compared to normal tissue ^141–143. Kim et al. found significantly lower ADC values in malignant lesions compared to benign lesions (1.75 ± 0.57 vs. 2.50 ± 0.53 × 10⁻³ mm²/sec) ¹⁴⁴. Sandrasegaran et al. ¹⁴⁵ found similar results. Taouli et al. reported lower ADC values in Bosniak category 3 and 4 lesions compared to category 1 simple cysts, although a statistically significant difference between Bosniak 2F and 3–4 lesions was not detected ¹⁴². Low ADC values have been shown in the papillary subtype of RCC compared to non papillary subtypes ¹⁴². Wang et al. ¹⁴⁶ found that clear cell RCCs showed a significantly higher mean ADC (1.85 × 10⁻³ mm²/sec) than papillary (1.09 × 10⁻³ mm²/sec) and chromophobe (1.31 × 10⁻³ mm²/sec) RCCs. ADC has also been reported to be significantly lower in high nuclear grade (III and IV) than low nuclear grade (I&II) clear cell RCCs ¹⁴⁷. With advanced DWI methods, Chandarana et al. showed that DWI has potential in assessing renal tumor cellularity as well as vascularity, and can help discriminate RCC subtypes ^{148, 149}.

Acute Kidney Injury (AKI)

AKI, previously termed as acute renal failure, refers to a rapid and reversible decline of GFR within days or weeks and has recently been defined and classified more specifically by the RIFLE (risk, injury, failure, loss, end stage) criteria ^{150, 151}. Causes of AKI include renal ischemia and renal parenchymal diseases such as contrast-induced nephropathy (CIN) and acute tubular necrosis (ATN). AKI predisposes the patients to chronic kidney diseases ^{152, 153}.

Although DCE MRI with low Gd dose is capable of measuring single-kidney GFR with higher accuracy than serum creatinine, it is typically not used for assessing AKI as lowered GFR in severe AKI patients could potentially increase the risk of NSF. He et al. ¹⁵⁴ developed an innovative non-contrast MRI technique for estimating GFR based on ASL. Although not yet validated against any gold standard, this approach showed a promising 28% increase in the estimated GFR after protein loading, as would be expected. In a rat model of ischemic AKI, Zimmer et al ¹⁵⁵ observed that although their ASL-estimated cortical perfusion was ~30% lower than that from DCE MRI, both were able to differentiate between healthy and AKI cases. Prowle et al. ¹⁵⁶ used phase-contrast MRI, a non-contrast MRI method to measure blood flow rate through the renal artery, and found that RBF in ischemic AKI patients was significantly lower than that in normal volunteers (335–1137 ml/min vs. 791–1750 mL/min).

Tissue oxygenation is another physiologic parameter of interest for AKI ^{157, 158}. BOLD MRI enables non-invasive mapping of renal tissue oxygenation for human subjects (section “BOLD MRI”). Assessment of acute tubular necrosis by BOLD is also discussed in section “Renal Transplants”. Contrast induced nephropathy (CIN) is another form of AKI. Using BOLD and a rat model of CIN, Li et al. ¹⁵⁹ found that BOLD measurements could detect the effects of viscosity and dose of iodinated contrast on subsequent CIN.

Pediatric Kidney Imaging

Congenital anomalies of the kidney and urinary tract are frequent in children. Ultrasonography remains the primary imaging modality to image these disorders. An extension of MRR, MR urography (Figure 5) is increasingly used in practice as a complementary tool since it can combine exquisite anatomic depiction and functional evaluation in a single examination without radiation exposure. Heavily T₂ weighted images allow a complete visualization of the urinary tract in a few seconds (with 2D acquisitions) to few minutes (with 3D acquisitions and respiratory synchronization). With 3D isotropic acquisitions, multiplanar and volumetric reconstructions that are easily understandable for urologists have made intravenous urography obsolete. Additionally, the renal parenchyma can be studied in detail: corticomedullary differentiation, thickness, cortical scarring, and cysts.

Figure 5.

Primitive right megaureter on a bifid ureter in a 6-month old boy. (A) T₂-weighted image with fat saturation. (B) Coronal view of volume-rendered T₂-weighted images. (C) Oblique view of volume-rendered T₂-weighted images. (D) Maximum intensity projection of T₁-weighted images at excretory phase. (E) Renography before contrast arrival. (F) Renography at arterial phase. (G) Renography at tubular phase. (H) Renography at excretory phase. Symmetric enhancement and excretion of contrast bilaterally suggests that the marked dilatation of the right collecting system and ureter is not a functional obstruction.

In pediatric urology, one of the most challenging issues is to identify whether dilated systems have true obstruction and therefore require surgery. True chronic obstruction is defined in practice by a decrease in the split (differential) renal function (SRF) on serial functional imaging, such as renal scintigraphy, MRR or MR urography ^160–162. Using gadolinium contrast agents and a tracer compartmental model of analysis, the relative filtration of the parenchyma can be calculated for each side with classical scintigraphic-derived estimates such as the integral method and/or the Rutland-Patlak method ^{32, 163, 164}. These results have to take into account the volume of renal parenchyma on both sides. As with scintigraphy, many estimates have been developed to assess the drainage such as the shape of the renograms or transit times ^{165, 166}. These parameters turned out to be of poor value, and the presence of chronic obstruction remains based on an evolving decrease in SRF.

Conclusion

With high diagnostic reliability, anatomic MRI of the kidneys and their blood vessels has achieved widespread clinical adoption. On the other hand, the functional MRI techniques for the kidneys require more work for their clinical application. Of the functional methods available, the low-dose gadolinium-enhanced imaging (MR renography or MR urography) are closest to clinical adoption. In pediatrics, these methods have been proven to be useful for determining functional obstruction in the setting of hydronephrosis or ureterectasis. In both pediatrics and adults, numerous studies have validated the high agreement between the glomerular filtration of gadolinium chelates as a marker of GFR and other more established techniques. Among the other functional methods, BOLD imaging promises the most valuable insights into renal disease with the opportunity to measure hypoxia noninvasively. Addressing technical challenges for BOLD is the focus of many laboratories, and it is likely that in addition to improving acquisition techniques for reducing image artifacts, the development of appropriate physiologic models to interpret the BOLD data will be critical. The jury is still out on methods such as ASL and DWI, both of which have either technical challenges or image interpretation issues. Given the prevalence and growing rate of renal diseases together with MRI’s advantages of providing both high resolution anatomic imaging without exposure to ionizing radiation, functional renal MRI is worthy of intensive research effort which will be most successful where nephrologists and urologists collaborate with radiologists and MR scientists.