Renal MRI: from Nephron to NMR signal

Abstract

Renal diseases pose a significant socio-economic burden on healthcare systems. The development of better diagnostics and prognostics is well-recognized as a key strategy to resolve these challenges. Central to these developments are MRI biomarkers due to their potential for monitoring of early pathophysiological changes, renal disease progression or treatment effects. The surge in renal MRI involves major cross-domain initiatives, large clinical studies, and educational programs. In parallel with these translational efforts, the need for greater (patho)physiological specificity remains, to enable engagement with clinical nephrologists and increase the associated health impact. The ISMRM 2022 Member Initiated Symposium (MIS) on renal MRI spotlighted this issue with the goal of inspiring more solutions from the ISMRM community. This work is a summary of the MIS presentations devoted to (1) educating imaging scientists and clinicians on renal (patho)physiology and demands from clinical nephrologists, (2) elucidating the connection of MRI parameters with renal physiology (3) presenting the current state of leading MR surrogates in assessing renal structure and functions as well as their next generation of innovation, and (4) describing the potential of these imaging markers for providing clinically meaningful renal characterization to guide or supplement clinical decision making. We hope to continue momentum of recent years and introduce new entrants to the development process, connecting (patho)physiology with (bio)physics, and conceiving new clinical applications. We envision this process to benefit from cross-disciplinary collaboration and analogous efforts in other body organs but also maximally leverage the unique opportunities of renal physiology.

1. Introduction

Kidney health is the unacknowledged key player in longevity and quality of life in countries of all development stages. Chronic kidney disease (CKD), claiming more than 1 million lives annually, is a global killer hidden in plain sight and an underappreciated challenge to healthcare systems (1). Kidney disease often progresses unnoticed until late into the disease course. Laboratory measures such as estimated glomerular filtration rate (eGFR) and albuminuria are key indicators of kidney dysfunction. Kidney biopsy remains the ‘gold standard’ for diagnostic (and prognostic) purposes but interpretation relies upon small tissue samples representing global kidney disease and exposes patients to procedural risks. Clinical evaluation of kidney disease widely employs ultrasound and computed tomography (CT) imaging to assess obstructive pathology, kidney size, blood flow, tumors, cysts, and stone burden. However, there is an unmet need for the development of non-invasive biomarkers to assess renal pathophysiology including interstitial fibrosis, hypoperfusion and hypoxia, inflammation, macrophage infiltration, and microvascular disease. The development of better diagnostics and prognostics which also limit procedural risk for patients is well-recognized as a key strategy to resolve these challenges.

Importantly, advancements in renal MR technology (Figure 1) and procedures allow assessment of microstructure (diffusion weighted imaging, T₁ mapping, MR elastography, diffusion tensor imaging, diffusion kurtosis imaging), hemodynamics (arterial spin labeling, intravoxel-incoherent motion, phase contrast MR, dynamic contrast-enhanced MR), oxygenation (blood oxygen-level dependent, quantitative susceptibility mapping), function (dynamic contrast-enhanced MRI), morphology, and inflammation. Collectively, MRI biomarkers have the potential to monitor early pathophysiological changes, renal disease progression or treatment effects, ideally in the entire kidney volume. One important example is total kidney volume (TKV), a prognostic enrichment biomarker approved by the US Food and Drug Administration (FDA) and European Medicines Agency (EMA), which informs disease altering therapy therein offering hope to individuals with polycystic kidney disease (PKD) (2,3). Ongoing kidney MRI advances expand non-invasive tools that propel CKD progression prediction, differentiate between acute tubular necrosis and rejection-induced injury in the kidney transplant, measure markers that may eventually replace current measures of kidney function (i.e. nephron number and single nephron GFR), and assess response to therapeutic interventions (i.e. hypoperfusion and hypoxia modification)(4). Kidney fibrosis, the final common outcome for all chronic diseases, remains a challenge to noninvasively quantify through MR elastography and other modalities (5). However, therapeutic targeting of multiple pathogenic processes is likely necessary to alter kidney fibrogenesis. Chronic sterile inflammation is increasingly recognized as a major contributor to progressive injury in several kidney diseases (6). For example, MR technology that non-invasively quantifies kidney macrophage density could reduce costs associated with kidney biopsy and histologic analyses and allow for repeated assessment over time. Thus, advances in MR imaging technology that quantify kidney macrophage abundance in vivo could greatly enhance clinical tools available to nephrologists.

Figure 1:

Diagram of renal functional markers from multiple approaches (nephrology, physiology, histopathology, quantitative MRI), with different but related connections to key features of renal health. Markers discussed in this review are shaded in green, and others not covered herein are shaded in blue. Continued efforts to flesh out these relationships, in both health and disease, will inform the development application of MRI as a surrogate or supplemental clinical marker. Abbreviations: ASL=arterial spin labelling; BOLD/T2*=blood oxygen level dependent; DCE=dynamic contrast-enhanced MRI; DWI=diffusion-weighted imaging; DKI=diffusion kurtosis imaging; DTI=diffusion tensor imaging; eGFR=estimated glomerular filtration rate; HEP=high energy phosphates; IVIM=intravoxel-incoherent motion model of diffusion; QSM=quantitative susceptibility mapping; MRE= MR elastography; MTR= magnetization transfer imaging; PC MR=phase-contrast MRI.

The emergence and standardization of renal MRI biomarkers involves major cross-domain initiatives (renalMRI.org, ISMRM Renal MRI study group, US NIDDK, UK Renal Imaging Network (UKRIN)), large-scale clinical studies and educational activities. These international collaborations have synthesized the current technical knowledge in a series of consensus technical review articles (7–12) and protocol descriptions (13), providing sample MRI protocols and guidelines for investigators who desire to start their own studies using quantitative MRI techniques. In parallel with these translational efforts, the need for greater (patho-)physiological specificity remains to enable engagement with clinical nephrologists and increase the associated health impact.

The ISMRM 2022 Member Initiated Symposium (MIS) “Renal MRI: from Nephron to k-space” spotlighted these developments with the goal of inspiring more solutions from the ISMRM and imaging sciences communities. This review article is a summary of the MIS presentations devoted to highlighting (1) the utilization of MRI by clinicians and imaging experts, in addressing current needs and possibly supplying new biomarkers to clinical use, (2) the needs for biological validation of renal MR markers to clarify the link to biology and physiology, (3) the current form of renal MRI technologies as well as next level innovations and (4) the potential added value of renal MR in patient management as a supplement or alternative to current clinical standards. As it summarizes the symposium presentations, this article, like the presentations themselves, is not a systematic review of kidney MRI literature to-date, nor is it a “how-to” technical implementation guide to the MRI techniques presented. For practical MRI protocol details, the readers are invited to consult the technical review references (7–12).

2. Renal physiology and pathophysiology – a primer for MR scientists

The major tasks the kidneys fulfill can be deduced from the classical symptoms of renal failure. These include retention of metabolic ‘waste’ products such as urea, retention of sodium and water with the ensuing volume expansion and arterial hypertension, potassium retention resulting in hyperkaliemia, and retention of protons resulting in disturbed acid-base-balance.

The nephron – the kidney’s functional unit for urine formation and for maintaining water and electrolyte balances – is comprised of the glomerulus, in which primary urine is produced by ultrafiltration of blood plasma, and the tubule, which reabsorbs the vast majority of filtered water and osmolytes, but also secretes some substances, to produce the final urine. Both processes rely on intricate interactions of the nephrons with the vasculature (14). In conjunction with the kidney’s unique vascular architecture, these interactions shape the renal-specific relationships among perfusion, energy metabolism and oxygenation (15,16). The particularities of renal hemodynamics and oxygenation are of paramount importance for the pathophysiology of kidney disorders (15,17–19).

Total renal blood flow (RBF) is greater than in virtually all other organs: the kidneys receive about 20% of the cardiac output under resting conditions (14). While the renal mechanisms of autoregulation largely prevent physiological changes in systemic arterial pressure from exerting effects on RBF and GFR, RBF and GFR are nevertheless quite variable: for instance, they increase following a protein-rich meal and decrease during heavy exercise (14,20). Renal vascular resistance is determined by active vasomotion and passive compression of intrarenal vessels. With regard to pathophysiology, vasoconstriction mediated by sympathetic nerves, epinephrine, and angiotensin II, plays a prominent role, e.g. during hypovolemic shock, as does lack of vasodilatory nitric oxide in diabetic kidney disease (DKD) (21–23). Passive compression of intrarenal vessels plays a pathophysiological role in what is called intrarenal “compartment syndrome”: as the renal capsule is relatively rigid, any increase in intrarenal pressure as occurs, e.g. due to obstruction of the urinary tract, tubular congestion, or occlusion of the renal vein, results in reduced perfusion and oxygenation (24–26). Both active vasomotion and passive compression of intrarenal vessels do not only alter blood flow but, at the same time, also the blood volume per tissue volume, i.e. the intrarenal blood volume fraction. This is of high relevance for physiological interpretation of MR-based assessments of renal oxygenation by blood oxygenation level-dependent (BOLD) techniques, as will be detailed in section 3b. The MRI relaxation times T₂* and T₂ provide surrogate markers of tissue oxygenation. However, T₂*, T₂ reflect the amount of deoxygenated hemoglobin (deoxyHb) per tissue volume, therefore the relationship of renal T₂*,T₂ with tissue oxygenation is also dependent on (changes in) the blood volume fraction (27–30).

GFR depends on the total filtration area of all glomeruli, their hydraulic conductance, and on the balance of four pressures. Typical examples for reductions of the filtration area are nephrectomy and the gradual loss of functional nephrons with age and in CKD. Following nephrectomy this is partially compensated for by hyperfiltration by the remaining kidney. Hyperfiltration of the (remaining) functional nephrons occurs also in the initial stages of CKD and in early DKD (15,23,31). The blood pressure in the glomerular capillaries drives glomerular filtration, whereas plasma oncotic pressure and fluid pressure of the primary urine in Bowman’s capsule hinder it. A fourth pressure comes into play in pathophysiological conditions only, namely, when sizable amounts of plasma proteins are filtered, which increases oncotic pressure in Bowman’s capsule. The balance of the four pressures – termed effective filtration pressure – determines present GFR (14). Each of the pressures may vary under (patho-)physiological conditions. Glomerular capillary pressure varies with both active vasomotion and passive vascular compression, whereby increased preglomerular vascular resistance decreases GFR while increased postglomerular resistance increases it (14). Plasma oncotic pressure increases, e.g. in dehydration, and decreases, e.g. in nephrotic syndrome. Increase in fluid pressure in Bowman’s capsule as occurs due to obstruction of the urinary tract or tubular congestion reduces GFR (14,18,32).

In healthy adults, GFR amounts to about 180 liters per day. This is about fourfold the amount of total body water of an average adult (14)! In order to maintain body water content, water balance must be achieved, i.e. renal water excretion must exactly match water intake minus extrarenal water loss, both of which are quite variable. On average about 99% of the filtered water must be reabsorbed by the tubules. The daily amount of sodium filtered in the glomeruli is the equivalent of about 1500 grams of table salt (14). Again, in order to maintain body sodium content, renal excretion must exactly match sodium intake minus extrarenal loss, both of which vary considerably (33). Considering that our daily average salt intake nowadays amounts to 10 grams, it becomes clear that at least 99% of the about 1500 grams of filtered sodium must be reabsorbed by the tubules. All tubular reabsorption ultimately relies on energy-depending processes: sodium reabsorption requires about 90% of the renal energy expenditure and, thus, of oxygen (O₂) consumption (15,16). In contradistinction to virtually all other organs, where O₂ consumption determines perfusion, it is the other way around in the kidney: the higher the GFR the more sodium must be reabsorbed by the tubules. As GFR, in general, varies with RBF, renal perfusion determines renal oxygenation (15,16).

In accordance with the huge total RBF, the whole kidneys’ O₂ extraction is much lower than in all other organs. Yet, blood perfusion within the kidneys is quite heterogeneous: while 100% of total RBF reaches the cortex, only 15% of blood that previously passes through the cortex, reaches the medulla. Medullary perfusion relies on capillaries and vasa recta only, as there are no arteries or arterioles supplying it. This unequal distribution is one reason behind the very low tissue partial pressure of O₂ (pO₂) in the medulla (15,16,28). The other reason is O₂ shunt diffusion from arteries to veins in the cortex and from descending to ascending vasa recta in the medulla (16,28).

These specifics of renal hemodynamics and oxygenation may explain why the kidneys are quite vulnerable to injury and disorders. Acute kidney injury (AKI) is triggered by a variety of causes, yet, irrespective of the etiology, there is a common early element in AKI pathophysiology: renal tissue hypoperfusion and hypoxia. Imbalance between renal O₂ delivery and demand also promotes progression from AKI to CKD. It furthermore plays a prominent pathophysiological role in DKD (15,17–19). Finally, the development of renal fibrosis – be it triggered by inflammatory responses to AKI and its progression to CKD, by DKD, autoimmune diseases, or hereditary diseases such as PKD – in general, leads to rarefication of intrarenal vessels and, thereby, to a vicious cycle of ever increasing renal O₂ deficits. In other words, the imbalance in the O₂ delivery/demand becomes self-perpetuating (5,19,34,35). As the fibrosis-induced rarefication of intrarenal vessels decreases the blood volume fraction and, thereby the amount of deoxyHb per tissue volume, this impact on T₂*, T₂ must be taken into account when interpreting renal BOLD-MRI.

2.a. Diagnosis: Current biomarkers and the role of MRI

Current options for effective prophylactic and therapeutic regimen for the majority of kidney disorders are disappointingly sparse. One generally recognized major obstacle is the notorious lack of clinically available diagnostic tools that would allow early recognition of patients undergoing AKI, and recognition of CKD with sufficient sensitivity and specificity. In accordance with current clinical practice guidelines, AKI and CKD are still diagnosed, and their staging determined, based largely upon measurements of the serum concentration of creatinine (SCrea), a surrogate marker for GFR. To overcome the well-known limitations of SCrea-based metrics including estimated GFR (eGFR), several alternative blood- or urine-based markers reflecting renal injury, inflammation, fibrosis, or repair have been proposed. Despite promise, the clinical performance of these markers has been modest, and none has advanced to provide point-of-care diagnosis. In general, these markers fail to reveal early events in AKI pathophysiology, such as tissue hypoxia (18,36,37). Recognizing these limitations, synergistic approaches that include MRI are warranted (38–42).

2.b. Histopathological diagnosis of kidney disease

Renal biopsy is the clinical “gold standard” for qualitative evaluation of kidney tissue. Despite its invasiveness, biopsy provides access to tissue for research purposes, such as advanced histopathology, immunohistochemistry, proteomics and metabolomics quantification. Research collaborations such as the Kidney Precision Medicine Project (KPMP) at the U.S. NIDDK conduct prospective kidney biopsies in acute and chronic disease for deep molecular (and imaging) phenotyping (43).

Kidney core needle biopsy is used since the 1950s, and is performed with automated devices under ultrasound guidance since the 1980s (43). Common indications include unexplained proteinuria, hematuria, rapid kidney function loss, or to monitor response to treatment. Kidney allograft biopsies are often performed as an initial post-implantation biopsy to assess the health of the allograft and thereafter “for cause” to assess dysfunction. A minority of transplant centers perform “protocol” biopsies at fixed intervals, particularly in patients at higher risk for rejection. A good quality biopsy specimen will generally contain 2 tissue cores of 1–2 cm in length, with mostly cortical (vs. medullary) tissue, and 10–15 visible glomeruli (43). To minimize the procedural risk, sample collection is usually limited to 2–3 but fewer than 5 needle passes despite adequacy of sample (43).

Stains are performed for cell nuclei (H&E), glomerular basement membranes (period acid-Schiff), collagen, fibrous tissue (hematoxylin phloxine saffron, Sirius red, Congo red, Masson’s trichrome), in addition to immunofluorescence studies used to identify glomerular diseases, according to each institution’s protocols. A renal pathologist scores the presence of fibrosis, inflammation, vascular and glomerular disorder qualitatively, on a scale of 0–3 according to increments of 10%, then 25% of tissue that show pathological changes (43). Kidney transplant biopsies are scored according to the Banff system (44), a complex scoring system with 15 scores classifying interstitial fibrosis, vascular, tubular and interstitial inflammation, as well as glomerular and vascular disease.

Less than 0.0001% of the renal tissue is sampled in a core needle biopsy. A report of 310 biopsies from 155 patients concluded that chronic allograft injury was misclassified in 25% of biopsy specimens (45). Histopathological lesions, especially interstitial fibrosis/tubular atrophy (IFTA), are patchy and heterogeneously distributed throughout the kidney allograft (45). Biopsy, particularly in allograft injury, is thus prone to profound sampling error and high inter/intra-observer variability (46,47). In terms of biopsy risks and complications, recent meta-analyses show that while risk of major complications (0.3% of patients with bleeding requiring intervention in native and transplant biopsy), or death is infinitesimal (1 in 1,667 patients with native and 5 in 38,593 with transplanted kidneys), there are non-negligible risks (perinephric hematoma in 11% of native and 1.63% of transplanted kidneys, macroscopic hematuria in 3.5% of native and 3.2% of transplant biopsy (48–50)).

Taken together, while traditional biopsy methods offer the advantage of direct evaluation of renal tissue, disadvantages are sampling bias, qualitative and subjective clinical interpretation, and a non-negligible rate of complications.

MRI facilitates non-invasive assessment of several structural and functional kidney features (Figure 1). As detailed below, several MRI tools provide physiologic specificity. A growing literature shows its potential to inform on the different stages of renal pathophysiology, improve prediction and interception of disease progression and evaluate treatment of renal disease.

3. Renal structural and functional MRI: what are we measuring, what is its added clinical value?

3.a. MRI-based assessment of kidney size

Since changes in kidney size (KS) are associated with several renal pathologies, an increasing body of literature assesses the potential of non-invasive imaging-derived KS as a clinical parameter for diagnosis, prognosis, and treatment evaluation (41,51,52). In patients suffering from PKD, KS correlates with disease progression and GFR (53). Consequently, KS has been approved as a prognostic marker for clinical trials of therapies for autosomal dominant PKD (2,3). Detecting KS reduction due to parenchymal atrophy, sclerosis, and fibrosis has been demonstrated to identify CKD and to determine its severity (51). Longitudinal MR-based KS monitoring is proposed as a key measure for several kidney disorders including hyperfiltration in early DKD, renal transplants, renal artery stenosis, and vesicoureteral reflux (41). If performed manually, segmentation of the kidneys is time consuming, and dependent on the experience and skill of the image analyst. Methods using deep learning convolution neural network (CNN) U-Net models are now becoming available for automated segmentation of KS to assess TKV in health and disease, as well as cortex and medulla segmentation. These models must be trained and tested on a given image contrast and slice orientation using accurate manual segmentations to achieve good accuracy. A number of studies have shown improvements for estimation of KS in PKD using T₁- and/or T₂-weighted imaging (54,55), including a study using MRI scans from the National Institute of Diabetes and Digestive and Kidney Disease (NIDDK) Consortium for Radiologic Imaging Studies of Polycystic Kidney Disease (CRISP) study (56). Recent studies have demonstrated the use of whole-body T₁-weighted dual echo gradient echo (GRE) sequences to estimate KS in in the UK Biobank (UKBB) (57) and German National cohort (NAKO) large scale population studies(58). Segmentation of TKV, and separation of the cortex and medulla is more challenging in Chronic Kidney Disease due to both the reduction in KS and reduction in cortical-medullary differentiation. Automated segmentation of T₂-weighted MR images using a CNN in healthy controls and CKD patients has been shown to be robust to accurately segment to calculate TKV (59). Most recently, a CNN has been shown to be able to automatically segment the renal cortex and medulla from MOLLI T₁ maps in transplanted CKD patients (60).

That KS changes are indicative of pathophysiologic developments has been demonstrated by serial MRI in a variety of models. These include models that emulate clinical conditions such as i) ureteral obstructions due to urolithiasis or during upper urinary tract endourologic procedures, ii) administration of X-ray contrast media (CM) for cardiac procedures, iii) obstructions of the renal vein during partial nephrectomy or due to renal cell carcinoma-derived thrombus formation, and iv) clamping of the suprarenal aorta or renal artery during surgery or the low arterial target pressure during cardiopulmonary bypass (25). MR-based KS assessment was also used in experimental diabetes, mutant models mimicking PKD, and renal allografts (61–63). A recent preclinical study demonstrated that monitoring KS allows for physiological interpretation of MRI-based oxygenation changes in acute pathophysiologically relevant scenarios (26).

3.b. MRI-based assessment of renal oxygenation

Pre-clinical studies

Given the pivotal and early prognostic pathophysiological role of renal tissue hypoxia, MRI-based assessment of renal oxygenation by blood oxygenation level-dependent (BOLD) techniques could become a vital assay for research into renal (patho-)physiology and for clinical application This approach is based on the magnetic susceptibility of blood, with deoxygenated hemoglobin (deoxyHb) being paramagnetic and oxygenated hemoglobin being diamagnetic. Increases in the concentration of deoxyHb induces reductions in the MRI relaxation times T₂* (=1/R₂*) and T₂ (=1/R₂). T₂*, T₂ provide surrogate markers of tissue oxygenation due to their dependence on the O₂ saturation of hemoglobin (StO₂) and its relationship to the pO₂ in blood and tissue (27,40). However, T₂*,T₂ reflect the amount of deoxyHb per tissue volume, therefore the relationship of renal T₂*,T₂ with tissue pO₂ is also dependent on the blood and tubular volume fractions (27–29,40). While an ever-increasing number of experimental and clinical studies utilize renal BOLD-MRI, it cannot yet be regarded as a quantitative biomarker. Such a designation requires standardization of MRI protocols for both preclinical and clinical studies as well as calibration against ‘gold standard’ quantitative methods (7,27,64). The ‘gold standard’ probes for the pathophysiologically relevant parameter tissue pO₂ are invasive, thus, calibration can only be done in preclinical studies (16,65). For this purpose, a multimodality setup (MR-PHYSIOL) was developed that combines invasive measurements including tissue pO₂ with parametric MRI, allowing simultaneous in vivo measurements in the same rat kidney. Dedicated acute (patho-)physiologically relevant test interventions performed by remote control proved instrumental to detail the link between renal tissue pO₂ and T₂*, T₂ including the role of the confounding factors vascular and tubular volume fraction in these scenarios (40,65,66).

Human studies on renal oxygenation necessitate non-invasive approaches, preferable, in a stand-alone MRI setting. Recognizing that events leading to acute renal hypoxia are often associated with changes in the blood and/or tubular volume fractions, and that these changes are mirrored by changes in kidney size, dynamic MRI to monitor KS in parallel with T₂*,T₂ was recently shown to enable physiological interpretation of acute T₂*,T₂ changes (26). In this preclinical study, serial in vivo mapping of T₂*,T₂ was performed during clinically realistic interventions which alter renal tissue oxygenation reversibly – including a brief occlusion of the suprarenal aorta (OA), the renal vein (OV), or both (OAV) – and with longer-lasting effects – including injection of an X-ray CM. As shown by Figure 2, OA resulted in a decrease in KS and a moderate decrease in T₂*,T₂, OV resulted in an increase in KS and a much more pronounced decrease in T₂*,T₂, and OAV left KS unchanged and resulted in an intermediate decrease in T₂*,T₂. Previous studies with ‘gold standard’ probes showed an equivalent decrease in tissue pO₂ upon the three occlusions (66). The reason for this discrepancy is that changes in T₂*,T₂ reflect changes in the amount of deoxyHb per tissue volume rather than directly mirroring StO₂. In addition to decreased StO₂, intrarenal blood volume is reduced upon OA, increased upon OV, and unchanged upon OAV. Thus, correct interpretation of T₂*,T₂ as surrogate markers for acute changes in renal tissue oxygenation must take into account changes in KS. If T₂*,T₂ decrease and KS remains unchanged, tissue oxygenation is reduced. If T₂*,T₂ decrease and KS also decreases, the pO₂ reduction is more severe than if KS is unchanged; if T₂*,T₂ decrease and KS increases, the pO₂ reduction is less severe.

Figure 2:

Time courses during kidney vascular occlusions and recovery. Time course of relative changes (mean±SEM) for kidney size (cross sectional area; left panels) and T₂ (blue) and T₂* (red) obtained for the renal cortex (right panels) before the occlusions (baseline), during the intervention (green area), and during recovery. Absolute baseline values (mean±SEM) are denoted; * P < 0.05; # P < 0.01; § P < 0.001. Data from Ref. 26 (Cantow et al., 2022).

Administration of an X-ray CM induced a sustained KS increase, with an initial increase in T₂*,T₂ followed by a moderate decrease (26). Measurements by ‘gold standard’ probes showed an immediate, massive and sustained drop in pO₂ (67). The explanation for this apparent discrepancy becomes clear from the observed KS increase, which reflects the ‘compartment syndrome’ that results in compression of the intrarenal vessels and thus in decreased deoxyHb. This result underscores how MRI-based measurement of renal oxygenation by T₂*,T₂ is crucially dependent on monitoring accompanying changes in KS. Taken together, this study demonstrated that monitoring KS allows physiological interpretation of acute renal oxygenation changes obtained by T₂*,T₂ (26).

Human studies

Given the pathophysiological role of renal tissue hypoxia for CKD progression, BOLD-MRI has been used in patients with CKD to assess its diagnostic potential. These studies have produced inconsistent results (68–72). Apart from different MRI acquisition and analysis protocols, major potential reasons for these inconsistencies include CKD-related reductions in the intrarenal blood volume fraction due to rarefication of vessels in fibrosis or reduced vascular lumen due to vascular remodeling, e.g. in kidney disease associated with hypertension. These alterations will greatly vary according to the disease etiology and stage. Moreover, in most CKD studies, the main clinical marker renal BOLD was benchmarked against, was notoriously insensitive SCrea/eGFR (73). However, in a prognostic study, Pruijm et al. measured BOLD-MRI in 112 patients with CKD, 47 with hypertension but without CKD and 24 healthy controls, and all participants underwent a physical examination, blood and urine sampling as well as an ultrasound scan (74). Pruijm et al. observed that CKD patients with increased R₂*=1/T₂* (low oxygenation) in the cortical layers and a flat corticomedullary R₂* slope had a faster yearly decline in eGFR and were ten times more likely to develop an adverse renal outcome. Therefore, R₂*-based assessment of renal tissue oxygenation might be used as a predictor of GFR decline in CKD.

Another area of application is in kidney transplantation, wherein discovery of non-invasive diagnostic measures to assess kidney allograft dysfunction is needed. Han et al. (75) tested whether acute rejection (AR) and acute tubular necrosis (ATN) could be differentiated by BOLD-MRI. They measured 110 patients, with 28 patients having a biopsy-proven AR (n = 21) and ATN (n = 7). AR kidney allografts showed significantly lower medullary R₂* values (higher oxygenation), while ATN allografts revealed higher R₂* in the cortex. Additional investigations applying BOLD-MRI in kidney transplants also identified a loss of corticomedullary R₂* differentiation in allografts with acute dysfunction (76). Finally, BOLD-MRI has been successfully used to assess response to novel therapeutics. In an early phase clinical trial in patients with atherosclerotic renovascular disease, intraarterial renal infusion of mesenchymal stem cells improved renal blood flow and fractional hypoxia (expressed as % R₂* >30s⁻¹) decreased in the stenotic kidney (77).

BOLD-MRI has very well-recognized limitations. Changes in T₂* are not only caused by differences in renal blood oxygenation but also by other physiological parameters such as renal blood volume fraction (28,78,79). Therefore, the question was raised whether the interpretation of BOLD-MRI as a surrogate for renal tissue oxygenation is universal (30). Second, the BOLD-signal is not only influenced by the paramagnetic deoxyHb, but also by other underlying tissue susceptibilities like diamagnetic proteins (80). Additionally, scanner related factors such as eddy currents or magnetic field inhomogeneity also affect T₂*. This makes it difficult to compare BOLD data across multiple sites (81).

Renal quantitative susceptibility mapping (QSM)

To overcome the constraints of BOLD-MRI, QSM facilitates direct probing and quantification of the underlying magnetic susceptibility of tissue while removing scanner related factors (82). QSM utilizes the phase information of the image data to determine magnetic susceptibility maps of the tissue. QSM provides a high structural contrast that is sensitive to changes in tissue microstructure and chemical composition (83). However, QSM requires complex, non-trivial post-processing, including solving ill-posed inverse problems (84–86). In essence, the information needed for QSM is encoded in any susceptibility sensitized BOLD protocol that uses a range of susceptibility weighting. This has the advantage that BOLD-MRI and QSM can be performed with a single measurement and without additional scan time.

There is a limited number of studies on abdominal QSM due to challenges including respiratory movement, substantial changes in susceptibility at tissue interfaces and the presence of subcutaneous and visceral fat. An early study used QSM to detect inflammation and fibrosis in ex-vivo mice kidneys (87). All mice kidneys with inflammation and fibrosis exhibited a more diamagnetic susceptibility in the cortex, outer and inner medulla when compared to healthy controls. This decrease in susceptibility accompanies the increase of diamagnetic lipids and proteins associated with inflammation and fibrosis (88).The literature shows very few examples on in-vivo renal QSM in the humans (80). One demonstrates for a single patient that fibrotic kidney displayed the same decrease in susceptibility as the above-mentioned mice study (Figure 3). Broader clinical studies are warranted to evaluate the potential role of QSM as a diagnostic marker for kidney pathologies.

Figure 3:

Magnitude images overlaid with QSM maps of the right kidney for a healthy volunteer (left image) and a patient with renal fibrosis (CKD V, eGFR < 15 ml/min/1.73 m², right image). Compared to the healthy renal tissue the fibrotic kidney shows a strong diamagnetic susceptibility (0.04 ± 0.07 ppm vs. −0.43 ± 0.02 ppm). Data from Ref. 80 (Bechler et al., 2021)

3.c. Renal blood flow and perfusion

Blood flow into the kidney can be assessed by measuring the bulk flow within the renal artery using phase contrast (PC)-MRI. Renal tissue perfusion can be measured with dynamic contrast enhanced (DCE)-MRI or arterial spin labelling (ASL)-MRI.

Phase contrast MRI (PC-MRI)

PC-MRI uses velocity-encoding (V_ENC) magnetic field gradients to generate phase shifts proportional to the velocity of protons in the direction of the gradient. Flow measurements are most accurate when the imaging plane is perpendicular (through-plane) to the vessel of interest (Figure 4A). The V_ENC must be higher than the peak velocity to avoid phase aliasing, typically 100–120cm/s to measure renal artery blood flow. PC-MRI data are typically collected in a breath hold with the use of retrospective or prospective cardiac gating to synchronize data acquisition with the cardiac cycle. For PC-MRI analysis, a region of interest (ROI) is placed in the renal artery. To obtain the arterial flux (in ml/min) at each cardiac phase, the velocity (in cm/s) can be multiplied by the area (in mm²) in each frame (Figure 4B). Global perfusion of the kidney (in ml/min/100 ml tissue) is obtained from the ratio of mean arterial flux by total kidney volume (ml) multiplied by 100.

Figure 4:

Phase contrast MRI is used to measure bulk flow within the renal artery. (A) The imaging plane must be perpendicular to the renal artery, planned using (i) orthogonal localisers or (ii) a vascular survey. (B) For each phase of the cardiac cycle, (i) magnitude and phase (proportional to velocity) images are acquired at each phase of the cardiac cycle, from which (ii) the arterial flux can be calculated.

Careful consideration is required for acquisition, quantification and reporting of PC-MRI (8,89). A good vascular survey is required, such as a time-of-flight (TOF), or inflow enhancement scan to position the imaging plane perpendicular to the renal artery and close to the aorta prior to any bifurcations (Fig 4A(ii)). The spatial resolution must be sufficient to include several voxels from within the vessel, whilst keeping the breath hold short and retaining sufficient signal-to-noise ratio (SNR). Parallel imaging can be used to reduce scan time or to increase the spatial resolution. Images affected by severe movement should be discarded, and if aliasing occurs data can be phase unwrapped. The between session coefficient of variation (CoV) of renal artery blood flow measured using PC-MRI ranges from 8 to 23% (89).

4D flow PC-MRI uses blood flow encoding in 3 directions. This removes the need for detailed planning and allows several vessels to be assessed simultaneously (90,91). Further work should focus on calibration of PC-MRI using flow phantoms, standardization and availability of analysis software including motion correction, and the comparison of semi-automatic and manual ROI selection.

Dynamic contrast enhanced MRI (DCE-MRI)

DCE-MRI uses an exogenous contrast agent, typically gadolinium-based, which is filtered in the glomeruli. A rapid series of T₁-weighted images is collected, before and during the passage of a contrast agent bolus. The signal-time-curve (STC) reflects the passage of the contrast through the kidney tissue (92). Analysis of DCE-MRI data comprises kidney segmentation, delineation of the arterial input function (AIF) from the aorta, calculation of T₁ maps to convert voxel-wise STCs to concentration time curves, before model-based quantification of renal perfusion (in ml/100ml/min) and glomerular filtration rate (in ml/min). Good SNR and high temporal resolution are required to define the AIF.

A study by de Boer et al.(93) measured DCE as part of a multiparametric repeatability study in healthy subjects and showed a correlation between visits. However, Bland-Altman limits of agreement were large (186 to −202 ml/100ml/min). Eikefjord et al.(94) performed a reproducibility study of healthy volunteers where a correlation was seen between scanning sessions, but the intraclass correlation coefficient (ICC) was low (0.34). DCE is widely used in animal research, but its clinical utility is limited by the risk of nephrogenic systemic fibrosis in patients with low GFR. Another limitation of renal DCE-MRI that hampers its clinical adoption is the complex post-processing pipeline, and the lack of consensus regarding the best pharmacokinetic model for fitting the data, even for the same contrast agent.

Arterial spin labeling (ASL)

ASL uses endogenous arterial blood water to measure renal perfusion, and is consequently safe for patients with renal impairment. Images with and without labeling of inflowing arterial blood are acquired, with the difference providing a perfusion-weighted image. Renal perfusion is quantified in units of ml/100g/min with appropriate modeling.

The two labeling schemes recommended for renal ASL (10) are pseudo-continuous ASL (PCASL) and the pulsed ASL (PASL) Flow-sensitive Alternating Inversion Recovery (FAIR) scheme. In FAIR ASL, a non-selective adiabatic inversion slab centered at the imaging plane is applied to label inflowing arterial blood while for the control acquisition a selective inversion slab inverts the magnetization in the imaging plane only (Figure 5A). The inversion time should be chosen to allow the labeled blood to arrive and exchange with the renal tissue. In PCASL, the blood is inverted as it flows through the labeling plane, which is positioned over the aorta (Figure 5B).

Figure 5:

Schematic to illustrate planning of labeling and imaging planes for (A) pseudo-continuous arterial spin labeling (ASL) and (B) flow sensitive alternating inversion recovery (FAIR) ASL.

ASL has intrinsically low SNR, therefore multiple repetitions are averaged. Short echo times increase SNR and reduce the impact of respiratory motion. Fast acquisitions allow optimal sampling of the highest ASL difference signal. Spin echo-echo planar imaging (SE-EPI) provides the highest temporal SNR and lowest variability in perfusion (95). Respiratory motion challenges ASL-MRI because it exceeds clinically acceptable breath-hold times. Free breathing with coaching, guided breathing or respiratory triggering are better suited. Background suppression (BGS) of static tissue is often used to improve ASL sensitivity, but this can affect the success of image realignment (96–98).

A simple perfusion model neglects tissue transit delays and MRI signal relaxation differences between blood and tissue, which results in underestimation of perfusion. Alternatively, data can be collected at multiple inversion times to determine tissue transit delays, however this approach is time consuming. A faster solution is to acquire a series of images following each inversion pulse (99) for iterative perfusion modelling (100). Renal cortical perfusion measured by ASL ranges from 200 to 350 ml/100g/min (101). The between-visit CoV of renal cortical perfusion is ~10% (102) but increases up to 20% when BGS is used (96). Reproducibility of FAIR and PCASL between two sessions showed CoV of 10% for FAIR and 34% for PCASL (97).

A review (103) compared DCE with ASL, with some studies reporting higher perfusion values from DCE than ASL, whilst others showed no difference. DCE has a larger CoV than ASL (27–32 vs. 14–18% (104), 17 vs. 10% (93)) with larger confidence limits for DCE compared with ASL (186 to −202 vs. 110 to −91 ml/100g/min(93)). Alhummiany et al. found no differences between PC-MRI, DCE and ASL in a cohort of type-2 diabetes patients (105). Bland-Altman and correlations between each of the measures showed PC-MRI and ASL to have the best agreement. The repeatability error is lower for PC-MRI than for ASL in healthy volunteers (79 ± 41 vs. 241 ± 85 ml/min/1.73m²).

Several reports highlighted that eGFR positively correlates with perfusion measured by ASL (95,106–109) and PC-MRI (95). Renal perfusion reduces with age, as shown using ASL (110,111), DCE (112), and PC-MRI (101). Renal perfusion is reduced in CKD (51,110,113,114) and at peak AKI (115). In these AKI patients, eGFR was normal after one year, but some participants still had lower than expected renal perfusion, potentially indicative of patients at risk of further AKI episodes or progression to CKD. The diagnostic potential of ASL to diagnose AKI was investigated (116). However, the correlation of SCrea with renal perfusion in healthy volunteers was lost in the patient group. PC-MRI blood flow was reduced in septic AKI compared with healthy volunteers (117). ASL perfusion and PC-MRI renal blood flow are reduced during AKI in COVID-19 (118).

3.d. Renal microstructure and microcirculation

Basic concepts of diffusion-weighted imaging: water motion and microstructure

MRI signal arising from water protons in vivo is sensitive to tissue composition and structure in a complex manner, both through longitudinal and transverse relaxation processes as well as through molecular motion and diffusion effects. These thermally or physiologically driven random movements of water molecules can be probed on a microscopic scale using diffusion-weighted imaging (DWI; Figure 6A), providing unique information on the functional architecture of tissues. In biological tissues, molecular water motion is restricted by many obstacles including cell membranes, interstitial matrix and directional structures such as the renal tubules and blood vessels in the kidney. As a result, the observed diffusion coefficient (a measure of mean-square displacement per unit time) in an imaging voxel containing a mix of all these different diffusion environments is called the apparent diffusion coefficient (ADC). As tissue microstructure is considered the major determinant of apparent diffusion, changes in cellular density, water content, cell membrane integrity, structural organization, and the presence of macromolecules can cause alterations in ADC and motivate the use of DWI for diagnosis and monitoring of kidney pathologies.

Figure 6:

Renal microstructure and microcirculation measured by diffusion weighted imaging (DWI). (A) A schematic of the Stejskal-Tanner pulsed gradient spin echo experiment, which forms the basis of the current DWI pulse sequences; (B) DW images acquired at different b-values (with different levels of diffusion-weighting); (C) An ideal mono-exponential ADC fit to the noiseless DWI data; and (D) the corresponding ADC map. (E) Signal fractions (f_slow and f_fast) and (pseudo-) diffusion coefficient (D_slow and D_fast) parameter maps derived from DWI using intravoxel incoherent motion (IVIM) modelling. (F) Mean diffusivity (MD), fractional anisotropy (FA) and direction-encoded FA maps obtained from diffusion tensor imaging (DTI).

ADC maps are derived from the voxel-wise fitting of the DWI signal acquired at different b-values (with different levels of diffusion weighting) to a mono-exponential model (Figure 6B, C and D). ADC represents the most widely used diffusion biomarker and its potential clinical value has been demonstrated in patients with CKD (119), kidney allograft dysfunctions (120,121), acute pyelonephritis (122,123), PKD (124), obstruction of the urinary tract (125), and renal artery stenosis (126). With the exception of PKD, all of the studies reported lower parenchymal ADCs in patients than in healthy controls. Moreover, ADC displayed a good correlation with the histopathologic degree of renal fibrosis (119,127), and SCrea/eGFR (128,129). Although ADC is a promising imaging marker, it should be noted that renal DWI signal is not truly mono-exponential and ADC is a purely empirical parameter representing a mix of all potential contributors to the diffusion NMR signal. This multifactorial dependence of ADC complicates the interpretation of the observed signal changes that can reflect impaired renal function and reduced perfusion and/or the degree of renal fibrosis, and/or other histologic features. This ambiguity motivates development and application of more advanced acquisition protocols and DWI signal modeling approaches.

Renal microcirculation and tubular flow: intravoxel incoherent motion (IVIM)

Urine formation by glomerular filtration and tubular water reabsorption involve considerable microscopic flow in blood vessels and tubules. Consequently, the contribution of the flow-driven pseudo-diffusion component to the measured DWI signal in the kidney is substantial. The effects of blood flow in the renal microvasculature and ‘true’ tissue diffusion can be separated and quantified using an intravoxel incoherent motion (IVIM) model (130), which improves the interpretation of the diffusion-attenuated MR signal versus a mono-exponential ADC model (131). IVIM imaging has been applied in CKD patients (132), renal artery stenosis (133), urinary tract infections (134), and kidney allograft dysfunctions (135). The original two-compartment IVIM model assumes that the incoherent blood flow in the microvasculature results in a fast-decaying signal component with pseudo-diffusion coefficient D_fast (or D_p, D*) and signal fraction f_fast (or f_p, f), whereas the water molecules motion in the renal tissue contributes to the slow water diffusion component described by a diffusion coefficient D_slow (or D_t, D). More recently, it has been recognized that the intra-tubular fluid compartment contributes to the pseudodiffusion term as well, and its contribution can be modeled by a third intermediate component characterized by D_inter (or D_i) and f_inter (or f_i) in the renal DWI signal (136,137) (Figure 6E).

While multi-compartment IVIM models are more suitable for the representation of diffusion properties of the kidney, the literature shows a rather large variability in reported IVIM parameter values. This divergence is mainly related to the diversity of acquisition and post-processing protocols, which ask for harmonization and standardization (9,138). In multi-platform studies, gradient non-linearity (in which the ratio of actual to nominal b-values varies with the square of the gradient amplitude, away from isocenter), shim and eddy currents introduce bias in fitted parameter values. Of these sources of bias, gradient non-linearity is the most important in off-isocenter organs such as the kidneys and can be assessed by an ice-water diffusion phantom and corrected by site or vendor-derived correction factors (139). Furthermore, renal DWI is sensitive to cardiac activity and blood pulsation (140,141) and to diuretic challenges (142).

Renal microstructure: diffusion tensor imaging (DTI)

Besides providing functional information on renal microcirculation and tubular flow, DWI can be used to probe diffusion directionality as an indicator of functional loss and structural damage in different renal pathologies. One of the most unique features of DWI is its sensitivity to water motion only along the direction of the magnetic field gradient. Consequently, it is possible to probe anisotropic microstructure by measuring preferential water diffusion directions. Diffusion tensor imaging (DTI) is particularly suitable for studying kidney microstructure including tubular structures and blood vessels that contribute to spatially restricted diffusion of water molecules. Overall, the radial orientation of tubular structures and vessels in the renal medulla results in more restricted water diffusion compared to cortex which contains more randomly oriented structures. Fractional anisotropy (FA; Figure 6F) and axial diffusivity (AD), which represent the degree of diffusion anisotropy and the rate of diffusion along the preferential diffusion direction, are higher in the medulla than the cortex. Conversely, mean diffusivity (MD) and radial diffusivity (RD), which describe the average mobility of water molecules and diffusivity perpendicular to the principal diffusion direction, show higher values in the cortex compared to medulla.

DTI supports the study of renal microstructure in healthy human subjects and patients with various kidney pathologies including CKD (143–145), DKD (146), glomerulonephritis (147), kidney transplant dysfunction (135,148,149) and PKD (150,151). Most of these studies reported lower medullary FA in diseased kidneys compared to the healthy ones, indicating a reduction in diffusion anisotropy due to pathological changes in the renal parenchyma, such as interstitial fibrosis and/or tubular atrophy. These findings suggest that FA may be a useful imaging marker of structural alterations regardless of the etiology of the underlying disease. Moreover, the results obtained in PKD patients and renal allograft recipients show that besides FA also the length, number and density of the DTI tracts can be used to evaluate renal microstructural damage quantitatively and visually.

Diffusion kurtosis imaging (DKI)

Conventional DWI assumes that diffusion-driven water displacement can be described by a Gaussian distribution. In reality, complex cellular microstructure greatly affects the random motion of water molecules and, consequently, their distribution deviates considerably from the normative pattern. This non-Gaussian diffusion behavior can be probed by means of diffusion kurtosis imaging (DKI), which extends conventional DWI by quantifying the degree to which the water diffusion probability distribution function varies from the normal distribution. DKI provides a dimensionless measure of the diffusion hindrance and restriction known as kurtosis, which can be considered as a reflective marker of tissue heterogeneity that vanishes for Gaussian diffusion. Qualitatively, higher kurtosis values suggest more impediments to normal diffusion and thus greater microstructural complexity. The potential diagnostic value DKI was investigated in patients with hyperuricemia (152), immunoglobulin-A-nephropathy (153) and CKD (154). Overall, higher renal parenchymal mean kurtosis (MK) were found in subjects affected by asymptomatic hyperuricemic injury and gouty arthritis than in controls (152). A similar trend was observed in CKD patients. A positive correlation between MK and glomerular injury and tubulointerstitial lesion, and a negative correlation between MK and eGFR were reported (154). These studies indicate that cortical MK values have higher sensitivity to renal pathology and function associated with CKD progression than ADC.

Advanced variants of renal DWI

Although DTI and IVIM can separately quantify microstructural anisotropy and microcirculation in the kidney, it is known that each of these representations provides only a partial description of the measured renal DWI signal. In particular, it has been shown that both tubular/vascular flow and tissue diffusion contribute to the medullary anisotropy (155). Therefore, there is a growing interest in developing more advanced DWI variants that more accurately describe the diffusion properties of the kidney. Among these, approaches such as a generalized intravoxel oriented flow (IVOF) (156,157) model and Renal Flow and Microstructure AnisotroPy (REFMAP) (141,155,158) use a hybrid IVIM/DTI methodology to simultaneously measure renal diffusion anisotropy and flow effects, accounting not only for diffusion anisotropy but also for the anisotropy of the flow compartment.

Recent developments resolved contributions from tubular and vascular compartments based on the differences in pseudo-diffusion rate using data-driven non-negative least squares (NNLS) spectral modeling (136,159,160). NNLS presents an alternative to the rigid model fitting approaches. It does not require a priori assumptions on the number of components of the signal decay nor starting parameters. This advantage can be particularly beneficial when evaluating renal pathologies in which the number of diffusion compartments or sources is not known in advance or might change during the disease progression or treatment. Although promising, these novel approaches should be validated in larger clinical studies and patients with various renal pathologies.

3.e. Renal tissue characterization: virtual MR biopsy

The identification of MRI markers with a strong correlation to tissue histopathology is essential for establishing MRI as a non-invasive, “virtual” biopsy approach for the assessment of renal structure and functions. This section focuses on renal MRI studies that showed a direct correlation with tissue fibrosis, inflammation, glomerular or vascular dysfunction, as derived from qualitative or quantitative assessment of renal biopsy samples.

For “virtual biopsy” centered on MRI, investigations focus on identifying MRI markers that distinguish between degree of severity of pathological conditions such as fibrosis and inflammation, as reflected by the pathology scoring of traditional biopsy samples. Others have sought to correlate quantitative MRI markers with quantitative pathology metrics from tissue staining, such as the percentage tissue fraction stained for markers of fibrosis, such as Sirius red or Masson’s trichrome (Figure 7). Thus, studies have established an association or negative correlation between either cortical ADC or corticomedullary difference in values (ΔADC) and interstitial fibrosis (IF) as measured by Banff scoring or Masson trichrome stain for native and transplanted kidneys (120,135,161,162). In human allografts, moderate negative correlation between cortical ADC and IF assessed by Masson trichrome stain, and corticomedullary difference in ADC values (ΔADC) correlated with eGFR and IF were reported (120,162,163). The inverse correlation of ADC with IF was significant (r=−0.77; p<0.001) in a study of 103 allografts with acute dysfunction undergoing indication biopsy, supporting the concept that DWI/ADC is useful when the study endpoint is IFTA, irrespective of other concomitant diagnoses (162). Corticomedullary ΔADC was found to be useful at separating high and low fibrosis as defined by the 40% tissue stained fraction for fibrosis, in both native (51) and transplanted kidneys (120).

Figure 7.

mpMRI in 55 year-old female patient with stable functional allograft (A-F; eGFR 78.4 ml/min/1.73m²) and a 38 year-old female patient with dysfunctional allograft (G-L; eGFR 19.9 ml/min/1.73m²) and fibrosis. ADC and D are decreased while T₁ and T₁ρ are increased in both cortex and medulla in the fibrotic allograft.

T₁ relaxation measurements depend on both the molecular environment of the water molecules and the pathological changes occurring in the tissue (Figure 7), and have correlated with tissue fibrosis (due to the association of collagen) and inflammation (cellular swelling and interstitial edema). In a recent study including animal and human allografts, both T₁ and ADC values were correlated with IF and interstitial inflammation (120). Corticomedullary differences ΔADC and ΔT₁ were predictive of IF by Masson trichrome and Sirius red stain, but not inflammation (120). The combination of ΔADC and ΔT₁ showed improved performance for fibrosis detection by imaging variables alone, suggesting that the two values measure slightly different processes (163). Human studies have confirmed significant associations for T₁ values and moderate/severe IF (AUC=0.76 for the diagnosis of interstitial fibrosis ci>1) and total inflammation (ti) (161). Corticomedullary ΔT₁ in combination with cortex ADC was found to distinguish stable allografts from allografts with chronic dysfunction and fibrosis with AUC=0.94 (164).

T_1ρ is an emerging imaging marker for noninvasive assessment of renal fibrosis (165) (Figure 7). T₁ρ is sensitive to the interactions between water molecules and macromolecules, including collagen. Only a few published studies have investigated T₁ρ in the kidney, with an initial study showing increase in T₁ρ relaxation time in patients with lupus nephritis compared to healthy controls, with very good/excellent measurement repeatability (166). A study in transplanted kidneys showed an AUC=0.77 of T₁ρ for differentiating between functional allografts and chronic dysfunction with fibrosis (165). T₁ρ may thus provide an indirect measurement of collagen deposition due to kidney fibrosis.

Magnetic resonance elastography (MRE) enables quantitative measurement of tissue stiffness by imaging the propagation of mechanical shear waves through tissue. MRE for the assessment of kidney fibrosis showed confounding results. Kennedy et al. reported that mean and median cortical MRE stiffness did not correlate with Banff scores, including ci+ct (167). Brown et al. showed that MRE renal stiffness decreased with CKD severity and fibrosis grade in a cohort of patients with DKD, with the decrease in stiffness mirroring decrease in renal perfusion, as measured concomitantly by ASL (168). Preliminary results indicate that MRE-derived stiffness is not solely or even dominantly dependent on renal interstitial fibrosis (169,170) and is confounded by blood flow, complex renal physiology, anisotropic renal anatomy and low spatial resolution parametric maps (171,172). Further constraints of MRE include the need for external hardware and specialized software, limiting widespread availability.

While most studies focus on the dyad of fibrosis and inflammation in the kidney, there has been interest in correlating IVIM-DWI parameters with Banff scores of vasculopathy, glomerulitis and peritubular capillary density (ptc) in transplants (164), and of ASL perfusion with ptc (162). MRI parameters have been combined with biopsy scores in models for the prediction of kidney transplant loss or re-listing at 2 years, emphasizing the complementary nature of quantitative MRI and traditional biopsy (164).

Last but not least, novel “virtual biopsy” applications in the kidneys have focused on direct quantification of nephron (glomeruli) number in-vivo in animals and in ex-vivo human kidneys, as the nephron constitutes the smallest functional unit in the kidney. Nephron number associates with CKD risk. Nephron quantification and single nephron GFR determination may allow for improved CKD progression risk assessment beyond GFR (173). Typically, nephron number in clinical studies is estimated by extrapolating the glomerular density measured on traditional patient biopsy samples, to the patient’s entire renal volume as measured by CT or MRI (174). Bennet et al. developed a method of measuring nephron number by cationic-ferritin-enhanced MRI in vivo in animals and in ex-vivo human kidneys (175). This is a major leap since there is no preferred sampling location for biopsy and as many as 200 core needle biopsies would be needed to estimate kidney nephron number within 20% of the ground truth nephron number measured by cationic-ferritin enhanced MRI, with 95% confidence (176). While cationic ferritin MRI quantification of nephron number remains an ex-vivo and pre-clinical application, radiolabed cationic ferritin makes in-vivo nephron number quantification with PET(-MRI) a possibility in humans (177).

As previously mentioned, kidney macrophage infiltration, in addition to reduction in nephron number, portends poor prognosis in several kidney diseases. Hence, quantification of macrophage density from histopathology may soon be integrated into clinical practice (6). For the past 25 years, MRI with ultra-small paramagnetic iron oxide (USPIO) contrast agents that are phagocytosed by macrophages has been proposed for the non-invasive assessment of macrophage distribution in animal models of transplant rejection (178) or renal disease (179,180). T₂* in the organ of interest is measured before, early (4h) and 20–24 hrs after administration of a USPIO contrast agent, to allow for macrophage uptake of the USPIO(181). T₂* is expected to decrease in areas of high USPIO and macrophage concentration. Translation of these pre-clinical findings to a study of renal transplant rejection in children found that USPIO-enhanced T₂* was higher in allografts with acute rejection than without rejection, while there were no significant differences in macrophage concentration on histopathology between the two groups (181). Elevated T₂* in the allografts with rejection was likely due to decreased perfusion and increased edema (181). This suggests that USPIO-enhanced T₂* mapping has similar limitations as BOLD MRI, possibly remediated by other quantitative methods such as QSM. Macrophage quantification by MRI with or without iron oxides or other macrophage-labeling contrast agents remains an open area of investigation. Novel approaches of imaging the deleterious role of phagocytes such as monocytes, neutrophils and dendritic cells in the progression of kidney disease include explorations into ¹⁹F MRI of the kidney, which represents an instrumental tool for the quantification of exogenous ¹⁹F substances in vivo. ¹⁹F MRI is commonly used in association with an intravenous administration of perfluorocarbons labelling of inflammatory cells - typically in nanoparticle form emulsions - to study the distribution of inflammatory cells that are migrating through the systemic circulation into inflamed organs in vivo (182).

3.f. Limitations of kidney MRI

Technical challenges and related confounding factors of MRI methods necessitate devoted efforts to advance the field. For example, some MRI methods like T₁ mapping do not distinguish inflammation and edema from fibrosis. In the assessment of kidney fibrosis, MRE and DWI are confounded by tissue perfusion. As mentioned above, MRE is currently limited in terms of spatial resolution to image the anisotropic renal anatomy and complex renal physiology (171,172). Furthermore, BOLD remains sensitive to the effects of perfusion and to changes in kidney size. Finally, as with any biomarkers, age, sex and race-related effects may influence renal MRI parameters (183). Other limitations should also be recognized. While the possibilities outlined above for the characterization of renal tissue in terms of structure and function by MRI are exciting to clinicians, MRI is an expensive imaging modality, available mostly in urban centers in middle- and high-income countries. Expertise in renal functional MRI is available only at tertiary medical care facilities. In addition to its high cost and low availability, pulse sequence implementations and protocol parameters for the MRI techniques presented vary among vendors and research centers. For most of the techniques presented, there is an unmet need for standardization and harmonization of protocols among sites and vendors.

4. Recent advancements and future directions in kidney MRI

Some of the limitations of renal MRI overviewed above can be partially overcome by new hardware and software developments in MRI. Imaging at 7T can substantially increase SNR compared to 3T and 1.5T, which allows faster acquisition times or higher spatial resolution. Recent developments in transmit-receive RF coils have made human abdominal imaging at 7T a possibility (184). Although anatomical imaging of the kidney at 7T remains challenging because of inhomogeneities of the radiofrequency field and SAR constraints, a few studies have shown promising results in functional imaging (184,185). Renal BOLD at 7T benefits from increased susceptibility effects: renal BOLD studies at 7T (186–188) found higher R₂* values compared to 3T (although cortico-medullary R₂* ratios were roughly the same), and more pronounced decrease in medullary R₂* following a water loading challenge (188). ASL at 7T benefits from longer T₁ relaxation times (186), which allowed longer total delay time (1200 ms) between the inversion pulse and acquisition time, without loss of perfusion SNR, in a feasibility study of single-breath-hold kidney ASL(189).

In addition to hardware developments that can increase SNR, software developments in acquisition and post-processing can decrease imaging time while maintaining the quality of the data and the accuracy of quantitative parameters. Accelerated imaging through under-sampling, combined with compressed-sensing reconstruction is a means of obtaining high-quality anatomical images with high spatial resolution and reduced acquisition times. Compressed sensing combined with use of radial trajectories allowed free-breathing, high-spatial resolution T₁-weighted imaging for DCE-MRI and T₁ mapping in abdominal applications (190,191). However, compressed sensing reconstruction settings designed to reduce artifacts can also over-smooth the signal temporally, which reduces the accuracy of functional parameter mapping in techniques such as DCE-MRI (192). Use of deep learning image reconstruction methods allows optimization of the reconstruction to maintain accuracy of DCE-MRI parameter estimation, while reducing artifacts (192).

5. Conclusions

This review provides a synopsis of the member-initiated symposium on renal MRI at ISMRM 2022. It summarizes state-of-the art technology, research directions, and applications of renal MRI. This survey reiterates that renal MRI has gone well beyond nascent stages of feasibility. The state of the field has evolved sufficiently, with accompanying evidence accumulation at the preclinical and clinical levels, to more deeply interrogate MRI markers of renal function. One source of deeper understanding comes in the form of more detailed models of each MRI technique (QSM, IVIM, etc), while another source of validation comes in comprehensive correlation with clinical standards of histopathology, physiology, and nephrology. Taken together, this body of work addresses the fundamental questions of the links between surrogate renal MRI markers, molecular profiles, biochemical markers and the signatures from physiological gold standard measurements, as well as what these relationships can tell us about the stages and evolution of kidney diseases over time, and responses to therapy and intervention. With these answers MRI adds value to patient management, as supplement or alternative to established standards such as serum eGFR and renal biopsy, and may ultimately provide markers to inform on the different stages of renal pathophysiology, improve prediction and interception of disease progression and evaluate treatment of renal disease.