Magnetic resonance imaging T₁- and T₂-mapping to assess renal structure and function: a systematic review and statement paper

Abstract

This systematic review, initiated by the European Cooperation in Science and Technology Action Magnetic Resonance Imaging Biomarkers for Chronic Kidney Disease (PARENCHIMA), focuses on potential clinical applications of magnetic resonance imaging in renal non-tumour disease using magnetic resonance relaxometry (MRR), specifically, the measurement of the independent quantitative magnetic resonance relaxation times T₁ and T₂ at 1.5 and 3Tesla (T), respectively. Healthy subjects show a distinguishable cortico-medullary differentiation (CMD) in T₁ and a slight CMD in T₂. Increased cortical T₁ values, that is, reduced T₁ CMD, were reported in acute allograft rejection (AAR) and diminished T₁ CMD in chronic allograft rejection. However, ambiguous findings were reported and AAR could not be sufficiently differentiated from acute tubular necrosis and cyclosporine nephrotoxicity. Despite this, one recent quantitative study showed in renal transplants a direct correlation between fibrosis and T₁ CMD. Additionally, various renal diseases, including renal transplants, showed a moderate to strong correlation between T₁ CMD and renal function. Recent T₂ studies observed increased values in renal transplants compared with healthy subjects and in early-stage autosomal dominant polycystic kidney disease (ADPKD), which could improve diagnosis and progression assessment compared with total kidney volume alone in early-stage ADPKD. Renal MRR is suggested to be sensitive to renal perfusion, ischaemia/oxygenation, oedema, fibrosis, hydration and comorbidities, which reduce specificity. Due to the lack of standardization in patient preparation, acquisition protocols and adequate patient selection, no widely accepted reference values are currently available. Therefore this review encourages efforts to optimize and standardize (multi-parametric) protocols to increase specificity and to tap the full potential of renal MRR in future research.

INTRODUCTION

Kidneys are morphologically complex organs. Renal pathologies induce (micro-) structural and functional changes that may be captured with magnetic resonance imaging (MRI) owing to its exceptional soft tissue contrast. Despite the frequent and successful use of magnetic resonance relaxometry (MRR) in other organs (e.g. cardiac MRI) to assess oedema, amyloid deposition and fibrosis, the application of renal MRR is still scarce.

Renal MRR holds the promise to non-invasively quantify tissue inflammation and alterations, such as interstitial or cellular oedema and/or fibrosis, as well as renal function. This review article evaluates and summarizes data on renal T₁ and T₂ mapping using clinical 1.5 and 3Tesla (T) systems and provides recommendations for upcoming research efforts to promote MRR in clinical practice.

MATERIALS AND METHODS

The European Cooperation in Science and Technology (COST) Action Magnetic Resonance Imaging Biomarkers for Chronic Kidney Disease (PARENCHIMA) (www.renalmri.org) initiated this systematic review by an extended PubMed search regarding renal mapping (see Supplementary data) on 25 October 2017 to identify human in vivo T₁ and T₂ measurements at 1.5 and 3T. Titles and abstracts of 357 publications were processed to identify matches aligning with the aim of this article. Furthermore, relevant references within the acquired papers and selected studies by the authors were added. Our analysis reaches back to the year 1983 and includes studies with field strengths below 1.5T. Some handpicked qualitative studies and preclinical studies were also included to present readers with relevant trends in the measurement of renal T₁ and T₂ values. Studies regarding renal neoplasms and/or dynamic contrast-enhanced MRI were excluded. For details on data collection, see Supplementary data.

BASIC PRINCIPLES OF MAGNETIC RELAXATION MECHANISMS

MRI is a non-invasive technique to map the human body using the interaction of three magnetic fields: (i) a strong static field (B₀ or main magnet) to magnetize the whole sample and to allow the signal to be measured; (ii) gradient coils producing three (G_{x, y, z}) linear, orthogonal gradients to allow the signal to be registered in space; and (iii) a dynamic radio frequency (RF) field (B₁ or excitation field) to change steady-state magnetization produced by B₀ and to enable the readout of the measured signal (using an appropriately frequency-tuned coil or antenna) [1].

When subjects are placed inside the MRI scanner, nuclear spins align with B₀ (Figure 1a and b). The application of an RF pulse (B₁; usually in the range of milliseconds and millitesla) changes this macroscopic magnetization and proton spins are perturbed (i.e. tipped away from B₀). RF pulses are named after their effect on the net magnetization vector, i.e. an RF pulse tilting the net magnetization vector by 90° from the z direction (B₀) into the x/y plane is called a 90° pulse and a 180° pulse inverts the magnetization vector (i.e. z to −z; Figure 1a and b). The return of the tipped net magnetic vector to the steady-state equilibrium along the B₀ axis and the decay of net transverse magnetization, respectively, are two independent processes that can be measured [2]; namely spin–lattice (T₁) and spin–spin (T₂) relaxation time. T₁ and T₂ relaxation times are characteristic for the tissue composition (i.e. local microstructural magnetochemical environment) and provide the main sources of tissue contrast in morphological MRI.

In order to actually quantify T₁ or T₂ relaxation times (i.e. T₁ and T₂ mapping) different clinical MRI protocols are available with specific advantages and disadvantages; the chosen method is often determined by the available MRI hardware, sequence and acquisition time. However, over the years, a plethora of measurement sequences and acronyms have been published, and the interested reader should refer to an in-depth textbook written particularly for medical doctors [3].

T₁ relaxation time

The gold standard for T₁ measurement, the inversion recovery (IR) technique, first inverts the magnetization in the z direction using a 180° pulse, which is followed by a waiting time, TI (inversion time), and a successive 90° pulse to initiate data readout with further 180° pulses. This IR preparation module has to be repeated several times by incrementing TI to acquire three to eight data points using a long TR (repetition time; i.e. five to seven times T₁), to ensure full relaxation before each inversion pulse, which leads to long overall IR-T₁ measurement times (Figure 1a and c).

The desire for faster T₁ measurement compatible with individual breath-holds has given rise to several efficient methods, the most common being variable flip-angle (VFA) and modified Look-Locker imaging (MOLLI).

In VFA, two or more spoiled gradient recalled-echo acquisitions with differing excitation pulse flip-angles give rise to signals modulated by T₁ [4]; while substantially faster than IR-T₁, care must be taken before considering VFA to provide quantitative, rather than relative, T₁ measures [5]. VFA measurements are susceptible to B₁ inhomogeneity and thus require additional B₁ mapping. Also, the accuracy of the resulting T₁ depends on the relation of the chosen flip-angles with respect to the observed T₁ range.

The MOLLI sequence and its variants, based on the technique developed in 1970 by Look and Locker [6], sacrifice the requirement of pre-excitation equilibrium to save time and report a modified, shorter, apparent T₁ (often denoted T₁*) derived from repeated efficient sampling of a single excitation pulse. This type of sequence is sufficiently fast, so it is well suited for cardiac imaging, but the comparability between T₁ and T₁* is limited [7, 8].

T₂ relaxation time

The most common method to measure T₂ relaxation time is a multi-echo (fast) spin-echo sequence, which first applies a 90° pulse to tilt the magnetization into the x/y plane and then applies several 180° pulses in the x/y plane to recover (echo) magnetization and hence enables T₂ estimates from the signal envelope (Figure 1b and d). This approach is achieved within one TR, which is much faster than T₁ (IR) measurements, and allows full kidney coverage within a few breath-holds.

However, T₂ measurements are sensitive to imperfect slice selection pulse profiles, diffusion, flow and field inhomogeneities [9]. A T₂ preparation module decreases the influence of imperfect slice selection profiles, diffusion and flow. Carr–Purcell–Meiboom–Gill (CPMG) and similar preparations can help to compensate for field inhomogeneities. Therefore T₂ preparations yield more accurate (but slightly higher) T₂ values as compared with a multi-echo spin-echo approach. T₂ preparations are widely used in cardiac imaging to visualize oedema after myocardial infarction [10], and can be performed during free breathing, although image registration prior to T₂ calculation is required. Commonly at least three source images with different echo times are recommended for accurate T₂ estimation using two- or three-parameter exponential fittings [10–13].

RENAL T₁ MAPPING

Reference values and physiological modulations

In the early 1980s, renal MRI detected relatively increased T₁ values in the medulla compared with the cortex in healthy subjects. This corticomedullary differentiation (CMD) is presumably caused by the higher free water content, i.e. higher mobility of water molecules, in the medullary tubules and collecting ducts [14, 15]. Additionally Hricak et al. [14] reported that hydration and the water balance management of the kidneys are important influencing factors, because T₁ CMD decreases during dehydration (relative cortical T₁ increase) and increases after rehydration, i.e. forced diuresis [14], but the impact in healthy subjects or patients was never reassessed at 1.5 T and 3T. Another inevitable variation is caused by the increase of B₀ from 1.5 and 3T, as T₁ generally increases. Further variation of renal T₁ values was reported due to different MRI acquisition schemes and breathing strategies [16, 17], even though high interexamination repeatability for single acquisition schemes was proven [18–20]. Therefore no widely accepted reference values are published and the given limitations have to be considered when comparing different studies (Table 1).

T₁ modulation by the inhalation of oxygen and carbogen

T₁ and T₂* relaxation times are modulated by oxygen level changes in the blood and/or tissue, although caused by different mechanisms [21]. T₂*, i.e. blood oxygen level dependent (BOLD) MRI, associated changes are reviewed by Pruijm et al. [30].

To our knowledge, modulations of renal T₁ values during the inhalation of pure oxygen (O₂) and carbogen (5% carbon dioxide mixed with 95% O₂) were only observed in healthy volunteers. In 2002, Jones et al. [21] reported a significant decrease in cortical T₁ values during O₂ inhalation at 1.5T. These findings were confirmed in 2007 and 2009 with an even more pronounced reduction in cortical T₁ values following the inhalation of O₂ and carbogen [23, 24]. In these studies, the lack of a renal hydration protocol [except in O’Connor et al. [24]], the free breathing acquisition, the VFA method and dyspnoea during the carbogen inhalation (leading to increased breathing motion), as well as the temporal and spatial acquisition constraints, can be considered as important limitations [23, 24].

The first 3T study was carried out by Ding et al. [27] when healthy subjects were evaluated during exposure to normoxia and O₂. Thereafter a multiparametric renal MRI study evaluated five healthy volunteers who underwent a hyperoxia challenge (∼80% O₂); again cortical T₁ values decreased, but unlike previous publications, no statistical significance was observed [18].

These studies show that cortical T₁ is sensitive to oxygenation level changes. However, the contribution of vasoconstriction and vasodilatation as well as perfusion changes during O₂ and carbogen inhalation, as well as the evaluation of renal oxygen delivery (ischaemia), were never directly assessed, which could have caused the reported ambiguous findings [18, 27]. Final conclusions regarding medullary T₁ modulations are currently not possible. Last but not least, it has not been clarified yet as to what extent alterations in T₁ reflect tissue and/or blood oxygenation. These questions remain a target for future evaluations.

Clinical studies

Renal transplants—early qualitative and semi-quantitative MRI studies

Imaging of renal transplants in the iliac fossae is less confounded by breathing motion, which enabled renal MRI evaluations in the 1980s [31]. Early qualitative and/or semi-quantitative renal MRI studies revealed a reduced T₁ CMD in acute allograft rejection (AAR), and even diminished T₁ CMD in chronic allograft rejection (CAR) [15, 31–34]. However, acute tubular necrosis (ATN) could not be sufficiently differentiated from AAR [32, 34–36], and even diminished T₁ CMD was reversible in some cases of ATN and AAR [36]. Thus scrutiny of the reduced T₁ CMD linked both oedema and fibrosis to prolonged T₁ values, which partially explains the low specificity of these renal transplant evaluations [37].

Another interesting finding on renal transplant observation was the clearly preserved T₁ CMD during an acute decline in renal function under cyclosporine therapy, which was linked to cyclosporine nephrotoxicity (CN) [32, 34]. However, three successive studies presented ambiguous outcomes [33, 37, 38]. Thereafter, no further research efforts were made, so no final conclusion can be made.

All these envisioned early MRI studies on renal transplants applied field strengths <1.5T, which today are not frequently in clinical use. However, in contrast to recent MRI evaluations, all of these studies applied histological validation. A low specificity was observed due to different acquisition settings (e.g. vendors and protocols), low reproducibility of the two-point method to calculate T₁ [31] and lack of a standardized patient preparation (e.g. hydration protocol) [14, 15]. In addition, loss of T₁ CMD was reversible after clinical improvement in some cases of ATN and ARR, which could have decreased the specificity further [36]. Therefore recommendations could not advocate qualitative and/or semi-quantitative MRI evaluations over ultrasound and scintigraphy [34].

Renal transplants—quantitative MRI studies

T₁ measurements on renal transplants at 1.5T were presented by Huang et al. [19] in 2011, when renal transplants and native kidneys with unknown underlying renal disease confirmed the trend of higher cortical and medullary T₁ values in renal transplants. They also achieved a high short-term in vivo repeatability (∼±10%). In addition, strong correlations were observed between estimated glomerualr filtration rate (eGFR) and cortical T₁ in both groups (native cortex: r = −0.83, P = 0.0001; transplant cortex: r = −0.80, P = 0.0017), but medullary T₁ values only significantly correlated with eGFR in the transplant group (r = −0.94, P < 0.0001) [19].

The second quantitative T₁ assessment of renal transplant was presented by Friedli et al. [29]. A total of 29 patients underwent a multiparametric MRI approach at 3T, including a validation against histological samples. With regard to T₁, only T₁ CMD showed a moderate correlation with renal interstitial fibrosis (R² = 0.29, P < 0.001) and eGFR (R² = 0.22, P < 0.05). No correlation was established between T₁ values and cellular inflammation [29]).

In 2018, renal T₁ was evaluated in 49 renal transplant patients, 52 patients after lung transplantation (LuTx; native kidneys) and 14 healthy volunteers [26]. Their aim was to assess acute kidney injury (AKI) after LuTx (reported incidence ∼60%), and after a 3- and 6-month follow-up. T₁ CMD was significantly decreased and mean cortical and medullary T₁ were significantly higher in renal transplants compared with healthy volunteers and the LuTx group (P < 0.001). However, T₁ CMD was also reduced in the LuTx group compared with volunteers (P < 0.05), which was linked to the incidence of AKI after LuTx. All patients and healthy volunteers were further grouped according to Kidney Disease Outcomes Quality Initiative (KDOQI) stages. Remarkable were the significantly lower cortical T₁ values in subjects with eGFR ≥60 mL/min/1.73 m² as compared with <60 mL/min/1.73 m² and that cortical T₁ negatively correlated (r = −0.642, P < 0.001) and T₁ CMD positively correlated (r = 0.542, P < 0.001) with eGFR for all participants. In contrast, medullary T₁ showed only a weak correlation with eGFR (r = −0.341, P < 0.001). During the 3- and 6-month follow-up, cortical T₁ and T₁ CMD exhibited a significant correlation with eGFR (P < 0.001 and < 0.01, respectively) in the LuTx and renal transplantation groups [26].

In summary, we identified only three quantitative T₁ studies on renal allografts at 1.5 and 3T. In contrast to early qualitative and semi-quantitative MRI studies, only one quantitative study applied a histological validation, in which it was shown that state-of-the-art T₁ measurements, i.e. T₁ CMD, could be used to assess renal interstitial fibrosis in allografts [29]. Another important finding was that T₁ values were sensitive to presumable AKI alterations in the context of post-LuTx [26]. However, the specificity of renal MRR regarding AAR, CAR, ATN or drug-induced toxicity was not further assessed or improved. Furthermore, these studies show that T₁ mapping has the potential to estimate renal function.

Non-invasive assessment of renal function

The first quantitative T₁ measurements on patients at 1.5T were published in 2007 [22]. A small and unbalanced group was primarily enrolled for the evaluation of a renal artery stenosis: one patient with CKD and hypertension and nine patients with hypertension alone. A loose hydration protocol was applied before the MRI acquisition, and afterwards all patients underwent a ^99mTc-diethylene triamine pentaacetic acid renography to measure the single-kidney GFR (SKGFR). A significant correlation was depicted only between cortical T₁ values and the SKGFR (r = −0.5, P = 0.03) [22].

In 2015 the association between cortical T₁, renal perfusion (from arterial spin labeling (ASL); see also Odudu et al. [39]) and eGFR in patients with chronic heart failure (HF) and control subjects with different levels of renal impairment was evaluated [25]. Renal perfusion was similar in chronic HF patients with and without renal impairment, but cortical T₁ showed a significant correlation with eGFR (r = −0.41, P = 0.015), which reflects the potential to assess CKD. Chronic HF patients had significantly higher cortical T₁ compared with all control subjects, and chronic HF patients with renal impairment had significantly higher cortical T₁ compared with chronic HF patients without renal impairment [25].

After the ASL reproducibility study of Gillis et al. in 2014 [20], a follow-up study evaluated renal perfusion and cortical T₁ in healthy volunteers and CKD patients with different a etiologies at 3T. Significantly higher cortical T₁ values were found in CKD patients and a correlation between cortical T₁ and eGFR was observed (r = −0.58, P < 0.001) [28].

One year later a multiparametric renal MRI study assessed T₁ in healthy subjects and CKD patients with various renal diseases after a short fasting period (>2 h) at 3T [18]. Compared with volunteers, CKD showed significantly higher cortical T₁, and T₁ CMD was reduced (P < 0.01). They achieved an interscan coefficient of variation of <2.9% and high intraclass correlation for the cortex and medulla (0.848 and 0.997, respectively, using spin-echo echo-planar imaging) [18].

As previously envisioned also, three renal transplant studies assessed the correlation of T₁ values and the renal function at 1.5 and 3T (see above) [19, 26, 29].

In summary, the envisioned studies show that the degree of renal impairment correlates moderately to strongly with cortical T₁ and T₁ CMD in CKD with various renal diseases [18, 22, 28], renal transplants [19, 26, 29], and chronic HF patients [25]. These findings are also in line with some qualitative assessments in the 1990s [40, 41], but not with all [42], due to the fact that renal T₁ values are modulated by many confounders, such as the degree of fibrosis [29], comorbidities (e.g. liver cirrhosis) [43, 44], the acquisition protocol (e.g. breathing motion) and fastening and hydration level [14], which all together seem to be responsible for the accomplished correlations in the envisioned quantitative studies at 1.5 and 3T. To our knowledge, only one study correlated renal T₁ values with measured GFR [22]. It should be noted that adequate patient preparation (e.g. hydration protocol, medication intake), patient selection in the context of comorbidities and acquisition protocols (e.g. triggered breath-hold) together with reference measurement of the renal function can improve T₁ renal function correlations, which advocates for further research in this field.

RENAL T₂ MAPPING

Reference values and physiological modulations

In healthy subjects, medullary T₂ is consistently longer than cortical T₂. As previously envisioned, Hricak et al. [14] evaluated the effect of fasting and hydration and showed that T₂ CMD decreased during hydration (i.e. forced diuresis), but these findings were never re-evaluated. Additional variation can also be found due to the increase in B₀ from 1.5 and 3T, which is accompanied by a general decrease in T₂, and by the fact that different MRI acquisitions and breathing strategies report unequal values. But for healthy subjects a high day-to-day repeatability was shown by a multi-echo spin-echo method with a mean variability of <4% for both cortex and medulla at 1.5T [45].

Closely linked to T₂ is T₂′, which is thought to reflect tissue oxygenation [45, 46]. For measurement of T₂′, both T₂ and T₂* are required. T₂*, i.e. renal BOLD MRI, is discussed by Pruijm et al. in this issue [30].

These variations have to be considered when comparing different studies (Table 2).

Clinical studies

In the 1980s renal transplants were evaluated regarding T₂, and MRI was shown to be useful to identify fluid collections in necrotic transplant, perinephric lymphoceles and haematoma [31].

To our knowledge, the first quantitative clinical, i.e. renal transplant, study on T₂ values at 1.5Tesla (T) was reported in 2011. One of two T₂ acquisition protocols identified a significant increase in cortical T₂ in 15 renal transplants compared with 6 healthy subjects. However, no significant difference was observed with regards to the allograft function [46].

In 2017, whole kidney T₂ values in animals with juvenile cystic kidneys and nine autosomal dominant polycystic kidney disease (ADPKD) patients were reported. A strong significant increase in T₂ values was seen in early-stage ADPKD patients compared with healthy volunteers. Based solely on T₂ values, early-stage ADPKD patients with a kidney volume <300 mL could be distinguished from healthy volunteers, which was not possible based on total kidney volume (TKV) [47].

In summary, human in vivo measurements of renal T₂ are relatively scarce. Therefore no final conclusion can be made regarding renal function estimation or renal transplant assessments. Nevertheless, interesting findings were obtained, which clearly advocate for future research. Early-stage ADPKD patients could benefit from the T₂ evaluations and the potentially improved assessment of early-disease progression compared with TKV [47]. This might be of special interest in the evaluation of novel therapeutic agents such as tolvaptan. The assessment of AKI in the context of ischaemia reperfusion injury, e.g. induced kidney damage during renal allograft surgery, also seems to be a potential application for T₂, as in vivo measurements were shown to be feasible [46]. Animal studies have shown that T₂ is sensitive to ischaemia–reperfusion injury [48, 49]. During initial ischaemia, T₂ decreases, probably due to deoxygenation, followed by an increase during reperfusion [50]. In the longer term, an elevation of T₂ that is more pronounced in the medulla compared with the cortex has been found [51, 52], which was attributed to consecutive inflammation and oedema (T₂ increase) [50–52]. Human studies are necessary to determine whether the T₂ changes following AKI can predict the recovery of renal function.

DISCUSSION

In recent decades, quantitative renal T₁ and T₂ mapping have been shown not only to be feasible, but also to provide non-invasive valuable information regarding renal structure and function in healthy, AKI, CKD, renal transplant and ADPKD patients at 1.5 and 3T (Tables 1 and 2).

Renal T₁ has been shown to be modulated by hydration and, in particular, cortical T₁ was sensitive to oxygenation. T₁ CMD is a potential candidate biomarker to assess AAR, CAR, ATN, CN, fibrosis and renal function. Renal T₂ was measured in only a few studies but showed the potential to evaluate renal transplants and to improve the diagnosis and progression of early-stage ADPKD.

However, the variation in T₁ and T₂ values is large, mainly due to the great diversity of the MRR methods applied, but also due to physiological (e.g. water balance management during fasting and forced diuresis) and pathological alterations (e.g. fibrosis) of the renal parenchyma. In virtually all renal diseases, renal function and microstructure are altered together, and this review on T₁ and T₂ unveiled the high sensitivity towards each of these processes as well as the complicated interpretation of the acquired data due to the low specificity.

In conclusion, currently available data suggest that the full potential of renal T₁ and T₂ mapping has not yet been tapped and adequate patient selection, with regard to comorbidities, alongside technical and physiological standardization, will significantly increase the specificity of renal MRR. On route towards renal T₁ or T₂ mapping as a biomarker it will be necessary to validate renal MRR against widely accepted reference measurements (e.g. nuclear medicine evaluations) as well as against histological findings, when possible. Last but not least, the integration of different quantitative renal MRI data into a multiparametric approach will likely enable us to gain the best insight into renal pathophysiology. The COST Action PARENCHIMA (www.renalmri.org) is working on standardization of multiparametric renal MRI techniques to tackle these challenges.

SUPPLEMENTARY DATA

Supplementary data are available at ndt online.

ACKNOWLEDGEMENTS

This article is based upon work from the COST Action Magnetic Resonance Imaging Biomarkers for Chronic Kidney Disease (PARENCHIMA), funded by COST (European Cooperation in Science and Technology). www.cost.eu. For additional information, please visit PARENCHIMA project website: www.renalmri.org.

FUNDING

M.W. was supported by the Austrian Science Fund (FWF; project P28867). A.d.B. was supported by an Alexandre Suerman scholarship for MD/PhD students at the University Medical Center Utrecht, Utrecht, The Netherlands. K.S. was supported by the Biomarker Enterprise to Attack Diabetic Kidney Disease project funded by the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement 115974. This joint undertaking received support from the European Union’s Horizon 2020 research and innovation programme and European Federation of Pharmaceutical Industries and Associations. P.B. was supported by Deutsche Forschungsgemeinschaft (BO 3755/6-1, SFB/TRR57, SFB/TRR219) and by the German Ministry of Education and Research (BMBF Consortium STOP-FSGS number 01GM1518A). N.P.J. was supported by the Liaison Committee for Education, Research and Innovation in Central Norway (grant number 90065000).

CONFLICT OF INTEREST STATEMENT

None declared.

REFERENCES