Introduction

Nonalcoholic fatty liver disease (NAFLD) has become a major cause of chronic liver disease worldwide. Its prevalence is currently estimated to be 25% in the general population [1, 2], 90% in those with obesity, and 60% in those with type 2 diabetes mellitus [3,4,5]. The prevalence of NAFLD is expected to increase because of the current epidemics of obesity and diabetes [6]. Additionally, liver fibrosis has been reported to be strongly associated with the long-term outcomes in patients with NAFLD [7, 8].

Liver biopsy is the recommended gold standard method in the diagnosis of nonalcoholic steatohepatitis (NASH) and staging of liver fibrosis in patients with NAFLD [9]. However, because of high costs, possible risks, and requirement for healthcare resources, an invasive liver biopsy is a poorly suited diagnostic test for such a prevalent condition [10]. Therefore, alternatives to liver biopsy, including biochemical tests and assessments of liver stiffness measurement (LSM), are being developed [11]. LSM is a promising surrogate biomarker of the stage of liver fibrosis, and several elastography techniques are currently available for the same, including magnetic resonance elastography (MRE) [12, 13]. In this review, we discuss the recent studies on the use of magnetic resonance imaging (MRI), including MRE, in the assessment of liver fibrosis and other pathological findings, such as steatosis, inflammation, and ballooning in NAFLD.

Magnetic resonance elastography for the assessment of liver fibrosis in NAFLD

MRE was developed at the Mayo Clinic in 1995 [14], introduced into clinical practice in 2007, and approved by the FDA in 2010. It is an MRI-based technique for quantitative imaging of tissue stiffness and is currently the most accurate noninvasive imaging method available for diagnosing liver fibrosis [15,16,17,18]. MRE is currently available on MR scanners of three major manufacturers (General Electric, Milwaukee, WI, USA; Philips Medical Systems, Best, Netherlands; Siemens Healthineers, Erlangen, Germany) with 1.5-T and 3-T field strengths.

If there is a dedicated device and software for the generation and analysis of liver elastic wave propagation, quantitative stiffness images (elastograms) of the liver can be rapidly obtained during breath-holding and can, therefore, be readily included in the conventional liver MRI protocols [19]. The volume of the liver that is measurable using MRE is typically ≥ 250 mL and up to a third of the liver volume [20, 21]. A more advanced version of three-dimensional (3D) MRE, which is commercially not available at present, can evaluate the entire liver volume and was used in a recent prospective study [22]. Therefore, MRE can be used to assess the entire liver with a high success rate [23]. Furthermore, unlike ultrasonography (US)-based techniques, the success of MRE is operator-independent [18] and is minimally affected by obesity [15]. MRE is also highly repeatable; the inter-observer and intra-observer reproducibility among the scanners is high [24,25,26,27]. The failure rate of MRE was estimated to be approximately 1% in an unscreened population [28] and 5% in patients with various liver diseases [29], with substantial iron deposition in the liver being the most common cause of failure. The technical failure rate associated with iron deposition is higher with 3 T MR systems than with 1.5 T MR systems because of the stronger susceptibility effect of iron. Patients who are claustrophobic and have MR-incompatible implants cannot tolerate MR exams. Additionally, motion artifacts such as cardiac impulses are also a cause of failure because MRE is a motion-sensitive technique. MRE should be conducted after ≥ 4 h of fasting because LSM may increase due to postprandial portal blood flow [30].

An increasing number of studies have demonstrated that MRE is an accurate method for diagnosing and staging hepatic fibrosis in NAFLD (Table 1). Among the studies that used MRE, the AUROC for the diagnosis of stage > 1, > 2, > 3, and 4 was 0.772–0.869, 0.856–0.919, 0.870–0.981, and 0.882–0.993, respectively [31,32,33,34,35,36,37,38]. In the most recent meta-analysis by Xiao et al., which included 5 studies and 628 patients with NAFLD, the AUROCs for the diagnosis of stages 2, 3, and 4 fibrosis using MRE were 0.88, 0.93, and 0.92, respectively [39]. In a systematic review, the authors concluded that MRE might have the highest diagnostic accuracy for the staging of liver fibrosis (Fig. 1) [38]. The optimal MRE thresholds were 2.61, 2.97, 3.62, and 4.69 kPa, respectively, for the detection of any (stage ≥ 1), significant (stage ≥ 2), and advanced (stage ≥ 3) fibrosis and cirrhosis (stage ≥ 4) in patients with NAFLD [38].

Table 1 Magnetic resonance elastography in patients with NAFLD
Full size table
Fig. 1
figure 1

Pathological fibrosis and MRE images with optimal threshold values from representative patients with stage 0, 1, 2, 3, and 4 fibrosis, respectively

Full size image

MRE has several advantages over US-based elastography in the evaluation of liver fibrosis. Xiao et al. [39] conducted a systematic review and meta-analysis, which included 64 articles and 13,046 patients with NAFLD, to compare the diagnostic performances of noninvasive indices (aspartate aminotransferase-to-platelet ratio index (APRI], fibrosis-4 index (FIB-4], BARD score, NAFLD fibrosis score (NFS], vibration-controlled transient elastography (VCTE) [M and XL probe], shear wave elastography (SWE), and MRE in the prediction of significant fibrosis, advanced fibrosis, and cirrhosis; they found that MRE offered the best diagnostic performance for the staging of liver fibrosis. Other studies have also demonstrated that MRE is superior to VCTE and noninvasive indices in the diagnosis of liver fibrosis in patients with NAFLD [22, 34,35,36]. Since MRE has the highest accuracy in the diagnosis of liver fibrosis, it is increasingly regarded as a promising surrogate for the monitoring of disease progression and assessment of therapeutic endpoints [40]. The most recent prospective cohort study by Ajmera et al. [41] investigated the clinical utility of MRE in predicting the progression of fibrosis in patients with NAFLD with paired biopsies and paired MRE measurements. The authors reported that a 15% increase in MRE was associated with histologic progression of fibrosis. More recently, Honda et al. suggested in their review article, which summarized the meta-analysis of MRE, that MRE had good diagnostic accuracy over US-based elastography techniques in the assessment of liver fibrosis [42].

Proton density fat fraction (PDFF) measurement to assess steatosis in NAFLD

PDFF is the ratio of MRI-visible protons bound to fat to all protons in the liver (bound to fat and water); it is an MRI-based method for quantitatively assessing hepatic steatosis and is available from several manufacturers of MRI scanners. Chemical shift imaging is applied to separate the liver signal into water and fat components by acquiring the gradient echoes at appropriately spaced echo times. In some variants of this approach, only the magnitude data are retained while the phase data are discarded. These variants accurately quantify hepatic PDFF from 0 to 50%, which fortuitously captures the biological range of human hepatic steatosis, which rarely exceeds 50% [43]. Scan protocol parameters, such as flip angle and echo times, the fat signal model, and T2*-corrections are sensitive variables influencing the PDFF measurement [44]. Grimm et al. indicated that multi-point Dixon sequences, but not two-point sequences, should be used for PDFF measurements [44]. The precision and reproducibility of MRI-PDFF measurement have been explored. Negrete et al. [45] demonstrated high inter-examiner agreement in participants with obesity for each hepatic segment (intraclass correlation (ICC) ≥ 0.992, standard deviation (SD) ≤ 0.66%, range ≤ 1.24%), lobe (ICC ≥ 0.998, SD ≤ 0.34%, range ≤ 0.64%), and whole liver (ICC = 0.999, SD ≤ 0.24%, range ≤ 0.45%). Similar intra- and inter-examiner precisions were demonstrated in overweight and obese participants by Tyagi et al. [46]. Bannas et al. [47] further demonstrated significantly smaller variance with excellent intra- and inter-observer agreement and repeatability with MRI-PDFF compared with histologic grading of steatosis (P < 0.001). Vu et al. [48] suggested that MRI-PDFF quantification methods should sample each liver segment in both lobes and include a total surface area ≥ 5 cm2 to provide accurate estimate of the mean liver PDFF.

MRI-PDFF correlates with the histologically determined grade of steatosis in patients with NAFLD. Studies that used MRI-PDFF measurement reported that AUROCs for the diagnosis of grade > 1, grade > 2, and stage 3 were 0.960–0.990, 0.825–0.90, and 0.79–0.92, respectively. The corresponding MRI-PDFF cutoffs for mild steatosis (grade > 1) ranged from 3.5 to 8.9%, with a sensitivity of 89%–97% and specificity of 88%–100% (Table 2) [34, 36, 49,50,51]. Imajo et al. [34] directly compared and demonstrated that MRI–PDFF measurement had higher accuracy than VCTE-based controlled attenuation parameter (CAP) in diagnosing steatosis in patients with NAFLD. However, they only assessed VCTE using the M probe. More recently, using a well-characterized, prospective cohort of American adults with biopsy-proven NAFLD, Park et al. [36] compared the accuracy of VCTE-based CAP measurement using both M and XL probes and compared it with that of MRI-PDFF measurement in diagnosing steatosis in patients with NAFLD; they demonstrated that MRI-PDFF measurement was superior to CAP measurement using M and XL probes. More recently, Runge et al. [52] demonstrated that MR spectroscopy-derived PDFF measurement was superior to CAP measurement in detecting and grading liver steatosis in human NAFLD.

Table 2 Proton density fat fraction in patients with NAFLD
Full size table

Multiparametric MRI (mpMRI)

The utility of MRE has been demonstrated for identifying patients with NASH from those with simple steatosis [52] and from those with advanced fibrosis in chronic liver disease [53]. However, it has not exhibited sufficient utility for longitudinal monitoring of fibrosis [54, 55]. Furthermore, there is conflicting evidence regarding the effects of body mass index (BMI) on the MRE failure rates and its diagnostic performance [29, 56, 57].

Multiparametric MRI (mpMRI) measurements of hepatic steatosis (PDFF) and iron-corrected T1 (cT1) are emerging as promising quantitative imaging biomarkers for NASH. MpMRI is now standardized across all major MRI platforms and is available in several countries. MRI-PDFF has excellent correlation with histologically graded steatosis across the clinical range seen in NAFLD [58] and high diagnostic accuracy in stratifying all grades of liver steatosis [59,60,61], and cT1 has been demonstrated to correlate with ballooning [62]. More importantly, it has been demonstrated to predict the clinical outcomes in patients with chronic liver disease [63]. Additionally, the AUROCs in mpMRI were 0.69, 0.74, and 0.80, respectively, while differentiating between NASH and NAFL and while considering cT1 as an index test in the differentiation between NAS < 5 and ≥ 5 for the same index test, and in the diagnosis of NASH using liver inflammation and fibrosis (LIF) score [62, 63]. An optimal cutoff for LIF has recently been identified with a sensitivity of 91% and specificity of 52%. For cT1, as well, an optimal cutoff (875 ms) has been suggested to distinguish between low- and high-risk (NASH or fibrosis > 1) patients with sensitivity and specificity of 97% and 50%, respectively. Both metrics have demonstrated excellent technical validity with high repeatability and reproducibility across MRI manufacturers and field strengths [64]. The technical validation and precision afforded by mpMRI techniques, as well as their sensitivity to subtle changes in hepatic fat and fibro-inflammation, have resulted in their increasing inclusion as endpoints in NASH clinical trials, as well as inclusion in the FDA and EMA Biomarker Qualification Programs.

The diagnosis of NASH is currently based on the histological presence of steatosis, lobular inflammation, and ballooning. The presence of fibrosis, in addition to these pathological findings, suggests more advanced disease. It is important to note, however, that the only biomarkers demonstrated to predict the outcomes in these patients are histological fibrosis and MRI cT1 [65]. Of these, cT1 is sensitive to steatosis, ballooning, and inflammation, as well as fibrosis, and, therefore, cannot be a pure biomarker of fibrosis. An actual measurement image and details on the above report on mpMRI in the assessment of the pathogenesis of NAFLD are presented in Fig. 2 and Table 3.

Fig. 2
figure 2

Examples of multiparametric MRI (mpMRI) in patients with NAFLD. Representative images from patients in each severity category (NAS = 3 + Fibrosis = 1, NAS = 5 + Fibrosis = 2, NAS = 7 + Fibrosis = 3), produced by analysis of the raw data using mpMRI

Full size image
Table 3 Non-elastographic techniques in patients with NAFLD
Full size table

MRI technologies for the assessment of steatohepatitis in NAFLD

MRE was evaluated for diagnosing NASH in six studies; the AUROCs ranged from 0.70 to 0.79 with sensitivity and specificity of 72% and 87%, respectively, for NAS ≥ 5 in studies that did not include fibrosis in the NASH definition, and similar results were observed in a subset of patients without fibrosis [22, 34, 36, 53, 54]. Among MR non-elastographic techniques (Table 3), the 31PMRS-derived ratio between nucleotide triphosphates (α-peak) and triphosphates (αNTP/TP), which reflects cellular energy failure [66], and the concentration of specific metabolites (e.g., alanine, lactate, and triglycerides), which are assessed using 1H-MRS [67], were used to diagnose the severity of NAFLD. It demonstrated AUROCs of 0.71 for αNTP/TP and 1.00 for alanine, with the latter being evaluated in a small sample of 26 patients with NAS ≥ 5.

Other MRI approaches include quantitative susceptibility imaging, intravoxel incoherent motion (IVIM) diffusion-weighted MRI, and morphological evaluations, such as liver volume and preperitoneal fat area, all evaluated in a single study, with AUROCs of 0.61, 0.68, and 0.74 for different IVIM parameters and 0.91 for susceptibility; the last one was tested in a small sample of 32 patients [68,69,70,71]. Furthermore, a score based on MRI optical analysis estimators produced an AUROC of 0.83 with sensitivity and specificity of 87% and 60%, respectively [72]. In the case of contrast media-based approaches, gadoxetic acid enhancement in the hepatobiliary phase exhibited sensitivity and specificity of 97% and 63% in a retrospective study in 81 patients, whereas superparamagnetic iron oxide (SPIO) and ultrasmall SPIO (USPIO)-enhanced MRI-derived ΔR2* yielded sensitivity and specificity of up to 91% and 73%, respectively, for USPIO in a study in 25 patients with NAS ≥ 5 [73,74,75]. Nowadays, USPIO is not commercially available all over the world, and SPIO is available only in Japan. The direct consequences of the test on the health, qualitative analysis of resource consumption, operator-dependence, and state-of-the-art level of the techniques are summarized in Table 3.

Multifrequency magnetic resonance elastography

Diagnostic threshold levels for staging fibrosis vary as a result of varying technical setups and vibration frequencies in MRE. Previous studies in volunteers and a limited number of patients have demonstrated that the complex shear modulus of liver, G*, depends on the dynamic test range by demonstrating a clear frequency dispersion [76, 77]. The real part, G’, of the complex modulus is determined by the restoration of the mechanical energy owing to the elastic properties of the material, whereas the imaginary part, G’’, of the complex modulus is associated with viscous properties due to the inherent mechanical friction of the tissue [78]. Both parameters are independent and important measures of the mechanical constitution of materials and tissues. Viscoelasticity models the dispersion of G’ and G”, which results in generalized material parameters sensitive to the mechanical connectivity and integrity of tissue on a microstructural level. To date, elasticity-based staging of hepatic fibrosis has been performed using vibration data that are accumulated at a single frequency and, therefore, our current knowledge about the frequency-dependent viscoelastic properties of the different stages of liver fibrosis is limited. Multifrequency MRE (MMRE) has recently been demonstrated to be sensitive in the early detection of subtle alterations in the viscoelasticity of the brain due to physiological aging and multiple sclerosis [79, 80]. Therefore, MMRE is a promising technique for staging liver fibrosis and distinguishing normal liver from that in the early stages of fibrosis. Several investigations have demonstrated that liver stiffness can have a static component that is primarily determined by extracellular matrix composites and structure (e.g., hepatic fibrosis) and a dynamic component that is affected by intrahepatic hemodynamic changes (e.g., inflammation, ballooning, congestion, and portal hypertension) [77,78,79,80]. The ability to distinguish how these components contribute to tissue stiffness and how the contributions change over the course of different diseases will have important diagnostic and prognostic implications and will direct translational research in NAFLD. However, the value of mechanical properties, other than shear stiffness, in distinguishing the different pathophysiologic states of the liver is yet to be established in NAFLD.

MRE in the assessment of portal hypertension in chronic liver disease including NAFLD

Liver fibrosis is the most important risk factor of portal hypertension in chronic liver diseases, including NAFLD. At the Baveno IV Consensus Workshop in Europe, clinically significant portal hypertension (CSPH) was defined using LSM obtained using VCTE [81]. MRE can noninvasively estimate LSM as well as VCTE, and LSM has been reported to correlate significantly with wedge hepatic venous pressure (WHVP), hepatic venous pressure gradient (HVPG), and portal hypertension in patients with chronic liver diseases, including NAFLD [82]. In a Korean study of 126 patients with chronic liver diseases, including NAFLD, an AUROC of 0.859 was observed to have good diagnostic performance in the diagnosis of esophageal varices (EVs) when the cutoff value of LSM of MRE was set at 4.63 kPa [83]. In our previous report, in 276 patients with chronic liver diseases, including NAFLD, a cutoff value of 4.2 kPa for LSM of MRE in the diagnosis of EV resulted in an AUROC of 0.850 and a cutoff value of 4.8 kPa for LSM of MRE for the diagnosis of EV for treatment indication (high-risk EV), and the AUROC had good diagnostic performance of 0.840 for EV. Additionally, comparing the criteria using VCTE (Baveno IV criteria) proposed at the Baveno IV Consensus workshop with the modified Baveno IV criteria using MRE, we found that the diagnosis of the presence of EV and the diagnosis of high-risk varices was better in MRE based criteria than VCTE-based criteria. We reported that the modified Baveno IV criteria have a high diagnostic ability [84]. These results suggest that LSM of MRE is useful in predicting portal hypertension in chronic liver diseases, including NAFLD.

It is known that portal hypertension is associated with splenomegaly; however, recently it has been reported that spleen stiffness measurement (SSM) is elevated in these patients. Although the details of the pathogenesis remain unclear, it has been suggested that spleen congestion and spleen fibrosis may be involved in the pathogenesis of EV; a study of VCTE reported that SSM was more useful in the diagnosis of EV than LSM [85]. However, measurement of SSM using VCTE has several drawbacks. One is that VCTE is difficult to measure in the absence of splenomegaly. Another disadvantage is that the measurement limit of VCTE is 75 kPa; therefore, if SSM exceeds 75 kPa, it is difficult to measure it using VCTE [86]. Although VCTE for measuring SSM is currently being developed, it is difficult to use in Japan. However, MRE can measure LSM and evaluate SSM if the vibration reaches the spleen simultaneously (Fig. 3). It has been reported that SSM correlates more strongly with HVPG than with LSM, and that it can aid in diagnosing CSPH with high accuracy [87]. Shin et al. [88] suggested that SSM should be above the cutoff of 7.6 kPa and reported higher diagnostic ability of EVs than measurement of spleen length and volume. The authors also evaluated the usefulness of SSM in diagnosing EV in 511 patients with chronic liver diseases, including NAFLD (Fig. 3); they reported an AUROC of 0.92 at a cutoff of 9.4 kPa and 0.91 at a cutoff of 10.3 kPa in the diagnosis of high-risk EV (unpublished data, American Association for the Study of Liver Disease (AASLD) 2018.). Table 4 summarizes previous reports on portal hypertension and MRE in patients with chronic liver diseases, including NAFLD. SSM may be more useful in the diagnosis of portal hypertension than LSM. The use of criteria based on LSM (modified Baveno IV criteria in combination with platelets) or SSM for MRE, such as the diagnosis of CSPH exclusion using VCTE and platelets as advocated at the Baveno VI Consensus Workshop, can prevent unnecessary endoscopies in chronic liver diseases, including NAFLD. This would provide significant benefits, such as reduced medical costs and avoidance of complications.

Fig. 3
figure 3

Measurements of liver stiffness and spleen stiffness for the assessment of portal hypertension (left: wave image, right: elastogram)

Full size image
Table 4 The assessment of portal hypertension using MRI in patients with chronic liver disease including NAFLD
Full size table

Benefits and limitations of MRI in the assessment of NAFLD

A comparison between US elastography and MRI including MRE is presented in Table 5. One of the benefits of MRI is that it allows a much larger sampling compared with US techniques and liver biopsy, which may reduce the sampling variability secondary to the heterogeneity of fibrosis. Additionally, it has been proven that MRE generally provides more reliable measurements and fewer failures in patients with obesity or ascites. In a recent retrospective review of a large series of 1377 cases of MRE from the Mayo Clinic, the reported failure rate was less than 6%, with no effects of BMI on the failure rate [89]. MRI may also be a better candidate than US elastography in assessing the response to new therapies for NASH.

Table 5 Comparison between US elastography and MR elastography
Full size table

The limitations of MRE include the possibility of failure in patients with iron overload (using gradient echo sequence), cost and availability, and possible contraindications in patients with devices such as metallic splinters, vascular clips, and cochlear implants (Table 5). Additionally, LSM obtained using MRE may be influenced by extrahepatic cholestasis and acute liver injury [90, 91]. However, all major vendors now propose MRE capabilities, and new sequences such as echoplanar imaging have been demonstrated to decrease the failure rate in the presence of hepatic iron deposition. Although the US or MRI technique has advantages and limitations, VCTE and MRE are believed to be the methods of choice. According to the clinical practice guidelines published by the European Association for the Study of the Liver (EASL), VCTE is an acceptable noninvasive procedure for the identification of patients at low risk of advanced fibrosis or cirrhosis [92]. Additionally, according to the practice guidelines published by AASLD, VCTE and MRE are clinically useful tools in the identification of advanced fibrosis in patients with NAFLD [11]. However, other US elastographies are not recommended in the current guidelines for NAFLD. One of the reasons is that there are no follow-up data using other US elastographies in patients with NAFLD. Additionally, VCTE also has the advantage that it can be used to evaluate liver fibrosis and liver steatosis using a controlled attenuation parameter (CAP), which can measure the degree of ultrasound attenuation [93]. Recently, the FibroScan-AST (FAST) score combined with measurement of liver stiffness and CAP measured using VCTE and aspartate aminotransferase (AST) was proposed [94]. This score can identify patients with NASH (NAFLD activity score ≥ 4 and fibrosis stage ≥ 2) and has been validated in large global cohorts, even in Japan [95]. Furthermore, MRI-PDFF measurement is an MRI-based method for quantitatively assessing hepatic steatosis and is available as an option in MRI scanners from several manufacturers [49, 96].

Conclusions

MRI, including MRE, provides higher diagnostic performance in noninvasive detection of not only liver fibrosis, but also steatosis, inflammation, and ballooning in patients with NAFLD compared with other noninvasive methods. However, MRI techniques are relatively recent and have not been widely validated in NAFLD. Additionally, there are few reports on the usefulness of other technologies, including mpMRI and MMRE, in diagnosing NAFLD. Therefore, there is no consensus regarding the use of these elastography techniques in clinical practice in place of liver biopsy. Nevertheless, MRI appears to be best suited for the evaluation of pathological findings of the liver in patients with NAFLD. Several clinical algorithms for the diagnosis and monitoring of patients with NAFLD using MRE have been proposed [53, 97, 98]. Further research will validate these observations in patients with NAFLD.