MR Imaging of Liver Fibrosis, Fat and Iron

Introduction

Chronic liver disease (CLD) is an increasing concern for both the US and world healthcare systems, with approximately 2 million people dying each year worldwide from CLD³. Importantly, this problem is not going away or lessening. While prevention and treatment of viral hepatitis has significantly improved over the past few decades, the growing global problem of obesity and type 2 diabetes has led to continued increase in CLD, particularly due to non-alcoholic fatty liver disease (NAFLD)^3,5.

With more and more CLD patients, it is imperative to develop methods to screen and monitor for disease progression efficiently and effectively. Although histological analysis through liver biopsy remains the gold standard⁷, practical limitations such as availability, cost, and risk of complications limit its utility as a widespread or long-term solution. Therefore, noninvasive imaging, both underlying etiology (e.g., hepatic steatosis and iron deposition) and the end result (hepatic fibrosis), has and will continue to be an important part of CLD management. In the past, imaging assessment of CLD was typically limited to somewhat rudimentary qualitative assessment, such as evaluating liver morphology as a sign of fibrosis or decreased hepatic attenuation on computed tomography (CT) as a sign of hepatic steatosis. However, advances in magnetic resonance (MR) imaging have allowed for accurate noninvasive assessment of CLD. In this review, we will discuss current MR techniques for evaluation of hepatic fibrosis, steatosis, and iron content as well as future avenues for advancement currently being studied.

MR Imaging of Liver Fibrosis

Current Techniques

Routine or morphologic based techniques

Advanced fibrosis or cirrhosis has long been identified on cross-sectional imaging by macro-alterations to the liver morphology and architecture. For instance, surface nodularity, caudate and left lateral lobe hypertrophy with right and left medial lobe atrophy, widening of the fissures, and increased parenchymal reticulation on pre-and post-contrast MR images are all signs of advanced fibrosis^8,9. However, these features are typically only present on advanced fibrosis cases when it may be too late for effective treatment and to prevent progression of CLD⁹. Secondary effects of hepatic fibrosis such as sequela of portal hypertension (e.g. ascites, splenomegaly, and upper abdominal varices) are mostly seen in advanced cases and are also more non-specific, as they can be seen in non-fibrotic processes such as nodular regenerative hyperplasia (NRH)¹⁰. Therefore, while these morphologic signs can be used to indicate advanced fibrosis, they are not suitable for diagnosis or monitoring of liver fibrosis.

MR Elastography

MR elastography (MRE) is the primary MR technique currently utilized to evaluate hepatic fibrosis. The increased extracellular matrix and collagen deposition that occurs with hepatic fibrosis leads to increased liver stiffness (LS)¹¹. This increased stiffness can be measured using MRE, allowing for indirect but accurate fibrosis assessment. Examples of MRE exams in different stages of fibrosis are shown in Figures 1–3.

Figure 1.

MR elastogram in a patient with suspected primary sclerosing cholangitis. Magnitude (A), color wave (B), color elastogram with confidence map (C), and gray-scale elastogram (D) are shown. Liver stiffness was 1.6 kPa, in the normal range.

Figure 3.

MR elastogram in a patient with alcohol related liver disease. Magnitude (A), color wave (B), color elastogram with confidence map (C), and gray-scale elastogram (D) are shown. Liver stiffness was 10.3 kPa, in the stage 4 fibrosis or cirrhosis range.

Currently, two-dimensional (2D) MRE is the predominant technique for clinical evaluation. MRE can be performed as a standalone exam or as part of a more comprehensive MR evaluation. While MRE at 1.5T overall has better technical performance due to decreased artifacts¹², both 1.5T and 3T systems offer accurate and reproducible liver stiffness measurement (LSM).

MRE has been shown to be highly accurate, and is widely regarded as the most accurate noninvasive method for diagnosing hepatic fibrosis⁴. Studies have consistently shown high diagnostic accuracies greater than 85%, both in detecting fibrosis as well as discriminating severe fibrosis (F3–4) from mild fibrosis^13–15. In addition, numerous studies have shown the results of MRE are highly reproducible with high interobserver agreement, with studies showing intraclass correlation coefficients typically >0.85, and >0.9 for repeatability and reproducibility^16–19. One advantage of MRE is that it is not significantly affected by hepatic steatosis, an important issue with the high prevalence of hepatic steatosis²⁰. MRE also benefits from a rather large volume of liver sampled compared to other, non-MR based, noninvasive methods. This larger sample size leads to more accurate and consistent results, particularly in patients with heterogenous fibrosis such as in primary sclerosing cholangitis²¹.

MRE does have a few limitations. First, any patient contraindication to MRI, such as an MR incompatible device, will preclude the use of MRE. The primary sequence obtained in MR is a gradient-recalled echo (GRE) sequence. GRE is susceptible to short T2* relaxation times with elevated hepatic iron content, especially in 3T systems, which can make a GRE MRE nondiagnostic. A spin-echo echo planar imaging (SE-EPI) based MRE, which is much less affected by high hepatic iron content may be useful and studies have shown an overall greater technical success rate for SE-EPI MRE due to fewer technical failures in elevated hepatic iron patients¹². Another limitation of MRE are several confounding factors that can increase hepatic stiffness, unrelated to hepatic fibrosis. Examples include hepatic inflammation, hepatic congestion, biliary obstruction, portal hypertension, and infiltrative diseases such as amyloidosis^22,23. These confounding factors should be kept in mind and the radiologist consider clinical context when evaluating an MRE.

Diffusion weighted imaging

Diffusion weighted imaging (DWI) is a well-established MRI technique, which measures the Brownian motion of water molecules in a voxel of tissue. A fibrotic liver restricts the intravoxel motion of water, likely due to the increased connective tissue/collagen. This leads to a lower apparent diffusion coefficient (ADC), which can be measured and used to assess for and stage fibrosis. A few studies have shown accuracy in detecting significant fibrosis with DWI^24,25, though the technique does not perform as well as MRE²⁶. A main advantage of DWI is its wide availability without additional hardware needed. However, it does have its drawbacks. ADC cutoff values can vary between studies and scanners, leading to difficulty establishing value ranges. DWI can also be affected by noise, motion artifacts, and hepatic iron/fat content^27,28.

Recent Advancements and Future MR Techniques

Three-dimensional MRE

The current predominantly used MRE technique is a 2D technique, specifically 2D GRE MRE which measures shear waves moving in a two-dimensional plane. However, as shear waves propagate in three dimensions in tissue, a technique measuring shear wave properties in all three dimensions could be more accurate. This 3D-MRE is best accomplished using an SE-EPI technique, as the large number of slices needed necessitates a fast speed of acquisition. Early studies have shown similar accuracy between 3D SE-EPI MRE and 2D GRE MRE²⁹.

3D MRE derived parameters

As discussed in the MR elastography section, there are several confounding factors in the liver that can increase liver stiffness without fibrosis. Using advanced techniques of 3D MRE, it is possible to evaluate other tissue mechanical properties with a goal of producing a more accurate assessment of fibrosis. The technical details are out of scope for this review, though the technique utilizes multiple frequencies and complex inversion algorithms. Among the parameters obtained include liver viscosity, dispersion of shear wave velocity, damping ratio, and volumetric strain³⁰. Of these, shear modulus has been shown to correlate with fibrosis; while others correlate more with steatosis (dispersion of shear wave velocity) or necroinflammation (damping ratio and loss modulus), for example^30,31. More research is needed to bring these factors into clinical use, though their potential to improve MRE is promising.

Hepatocyte function using hepatobiliary contrast agents

Hepatobiliary contrast agents (HBA) are taken up by functioning hepatocytes, with a peak hepatobiliary phase enhancement of about 20 minutes with gadoxetate disodium and about 60–180 minutes with gadobenate dimeglumine. As the liver becomes more fibrotic, there is a decrease in functional hepatocytes and therefore decreased HBA uptake³². The hepatobiliary phase enhancement can be compared to precontrast images or internal comparisons such as the spleen or muscle to detect and stage fibrosis. While the research into staging fibrosis by HBA uptake is somewhat sparse, several studies have shown the ability to accurately detect fibrosis and further research is needed^32–34.

T1 Mapping

T1 mapping is a technique that takes advantage of the prolonged T1 relaxation times in tissue with higher water content. The increased extracellular space associated with fibrosis creates this increased tissue water content that can be measured and used to correlate with stages of fibrosis³⁵. As hepatic iron will shorten T1 relaxation time, this technique is often performed with a compensatory algorithm to account for the iron, referred to as corrected T1 (cT1). Early studies have shown T1/cT1 to be predominantly inferior to MRE in fibrosis evaluation, though research is ongoing³⁶.

Intravoxel incoherent motion (IVIM)

Earlier, we discussed DWI assessment of hepatic fibrosis. A closely related technique also being studied for evaluation of fibrosis is intravoxel incoherent motion (IVIM). Both DWI and IVIM use gradients to evaluate the motion of water molecules in tissue. However, DWI uses stronger gradients and assesses the diffusibility of water molecules, primarily affected by cell membranes and macromolecules. On the other hand, IVIM uses smaller gradients to assess microperfusion in the capillary network³⁷. Several studies have shown high accuracy in IVIM detecting hepatic fibrosis, though cohort sizes are relatively small and further research is needed^37,38.

Other Techniques

A few additional MR techniques in evaluating hepatic fibrosis are described in Table 2.

Table 2:

Additional MR techniques for the evaluation of hepatic fibrosis.

Technique	Description
Spin-Lattice Relaxation in the Rotating Frame (T1ρ)	Measures the tissue transverse magnetization relaxation while under a constant, spin-lock radiofrequency field65. Sensitive to the macromolecular components of tissue, such as collagen in a fibrotic liver. Few studies have shown promise in the technique at detecting and staging liver fibrosis, though there have been conflicting results and further research is needed66,67.
T2 Mapping	Fibrosis prolongs T2 relaxation time, likely due to increased extracellular water. Several studies have failed to show high accuracy, though research is ongoing35,68.
Susceptibility-Weighted Imaging (SWI)	Gradient-echo based sequence which demonstrates susceptibility changes from substances such as iron, calcification, or blood products. A few studies have shown that SWI can be effective at evaluating for hepatic fibrosis, presumably due to changes in the extracellular matrix and/or increased iron within the fibrotic liver69,70
MR Perfusion Techniques	Assesses different parameters related to hepatic parenchymal perfusion and associated enhancement characteristics. Multiple studies have shown good results, though data is relatively limited71–73.

MR Imaging of Hepatic Steatosis and Iron

Background

Hepatic steatosis, the pathognomonic and earliest histological feature of NAFLD, is characterized by the intracellular accumulation of triglycerides within hepatocytes. In some patients, the resultant inflammatory response can drive progression to NASH and, ultimately, cirrhosis and its associated complications, such as hepatocellular carcinoma and liver failure. Furthermore, there is mounting evidence that there is significant interplay between the development of hepatic steatosis and pathologic processes in other organ systems. One recent study documented a strong association between hepatic steatosis on noncontrast CT and coronary artery disease³⁹, which accounts for a greater fraction of mortality in patients with NAFLD than liver disease, while several recent analyses have suggested that steatosis may be a causal factor of both the metabolic syndrome and type II diabetes^40,41.

Excessive iron deposition within the body typically stems from conditions that result in disordered intestinal absorption, such as hereditary hemochromatosis, ineffective erythropoiesis seen with inherited anemias, or repeated transfusions in patients with profound anemia such as that seen in bone marrow suppression. These mechanisms result in pathologic accumulation of iron in the liver, damaging hepatocytes and predisposing patients to progressive hepatic fibrosis and the development of cirrhosis⁴², and a surplus of iron may accelerate hepatic injury in patients with NASH⁴³. In the heart, toxic iron accumulation can result in cardiomyopathies and fatal arrythmias, while in the pancreas, iron overload can cause exocrine dysfunction and lead to type I diabetes; iron accumulation in the anterior pituitary gland disrupts gonadotropin synthesis, yielding gonadal dysfunction and decreased serum levels of estrogen and testosterone⁴².

Accordingly, the accurate quantification of hepatic fat and iron has long been important in the detection, prevention and treatment of chronic liver disease. Traditional methods of diagnosing steatosis and iron overload, such as non-targeted liver biopsy, are subject to sampling error and can be complicated by pain, bleeding and high cost^44–46. As a result, there has been increasing interest in recent years in the development of imaging biomarkers for the non-invasive identification and quantification of hepatic steatosis and iron overload.

Conventional MRI Techniques

In-phase and opposed-phase imaging

For decades, conventional MRI imaging sequences, including in-phase and opposed-phase (IOP) imaging and fat suppression, have utilized the different resonance frequencies of fat and water to provide pseudo-quantitative estimates of hepatic fat content.

IOP imaging utilizes multiple echo times to acquire images in which the signals from water and fat are either in-phase (IP = |W + F|) or out-of-phase (OP = |W – F|), where the vertical bars indicated the absolute value, ie: magnitude.

However, a number of factors confound measurement of signal fat-fraction^47,48, which may not accurately reflect hepatic fat content. This approach is also limited by a dynamic range of 0–50% hepatic fat fraction; in practice, hepatic fat fractions in excess of 50% are rare but do occur⁴⁹.

MR Spectroscopy

Magnetic resonance spectroscopy (MRS) has also been used to quantify hepatic fat. With MRS, the resonant frequencies of water and fat protons are demonstrated on a spectral tracing and the signal intensities of the protons at their respective frequencies are quantified, yielding a signal fat fraction⁴⁸. MRS is limited by the need for specialized expertise and time-intensive postprocessing. Furthermore, the MRS signal is usually obtained with a small voxel, meaning that sampling error is a potential problem in patients with heterogeneous hepatic fat distribution⁴⁹.

T2 and T2* Imaging

The accumulation of iron in tissue impacts MRI signal by shortening the relaxation times T2*, T1, and T2. A qualitative estimate of hepatic iron overload can be obtained with either T2- or T2*-weighted images. In cases of iron overload, the liver demonstrates decreased signal intensity relative to a normal liver on T2-weighted images, which are usually acquired with fast spin echo (FSE) sequences. IOP gradient echo (GRE) sequences have also been used to qualitatively estimate excessive iron accumulation within hepatocytes by taking advantage of T2* weighting. In IOP imaging, the second echo, given its longer echo time, is more T2* weighted than the first; hepatic parenchymal signal loss between the first echo (opposed-phase images) and the second echo (in-phase images) suggests short T2* decay and is indicative of iron deposition⁴².

Quantitative assessment of liver iron content has been performed using signal intensity ratio (SIR) techniques using either T2- or T2*-weighted imaging, MR susceptometry, and quantitative relaxometry. A comprehensive review of these methods is beyond the scope of this text but can be found elsewhere in the literature⁴².

Recent Imaging Techniques

Quantitative chemical shift-encoded MRI

Quantitative chemical shift-encoded (CSE) MRI is an imaging technique that facilitates the separation of MR signal into fat and water components by acquiring images at three or more echo times, although typically 6 are used. CSE MRI corrects for an assortment of technical and biological factors—such as T1 relaxation, T2* decay, noise bias, eddy currents, and spectral complexity of fat^47,48,50,51—that can otherwise confound accurate and reproducible measurements of signal fat-fraction within the liver. In doing so, the resultant signal fat-fraction becomes equivalent to the proton density fat-fraction, which is a fundamental property of tissue, defined as the ratio of mobile protons attributable to fat to the total number of mobile protons^47,52.

The two major approaches to perform CSE MRI can be characterized as either magnitude-based or complex-based: the magnitude-based technique^48,53,54, which is generally easier to implement, consists only of magnitude images and thus is insensitive to phase errors. The main drawbacks of magnitude-based CSE MRI are lower SNR performance and a dynamic range limited to PDFF between 0–50%. Complex-based technique^54–56 incorporates both phase and magnitude information, enabling estimation of PDFF with higher SNR performance and a full dynamic range between 0–100%.

CSE MRI data are frequently obtained with a short breath hold and a 3D acquisition of the liver, from which a parametric map can be produced at the level of the scanner. The same acquisition used to produce an estimate of hepatic fat content in the form of PDFF can also be used to generate a confounder-corrected map of R2* (1/T2*) that is corrected for the presence of fat^42,57,58. These maps provide a visual representation of both hepatic fat fraction, expressed as a percentage, and hepatic iron content in the form of R2*, which is directly proportional to liver iron content (LIC). Regions of interest (ROIs) can be drawn within the hepatic parenchyma on these maps to quantitate hepatic fat and iron content. Examples of PDFF and R2* images are depicted in Figures 4–6.

Figure 4.

1.5T MRI in a patient with elevated liver enzymes. In-phase (A) and opposed-phase (B) images demonstrate only a small drop in signal intensity (SI) on opposed-phase images. Proton density fat fraction (C) images demonstrate a fat fraction (FF) of approximately 3.2%, within normal limits. R2* images (D) demonstrate a R2* of approximately 33 s⁻¹, also within normal limits.

Figure 6.

1.5T MRI in a patient for elevated liver enzymes. In-phase (A) and opposed-phase (B) images demonstrate slightly increased signal intensity (SI) on opposed-phase images. T2-weighted images (C) demonstrate diffusely decreased signal throughout the liver. R2* images (D) demonstrate a R2* of approximately 132 s⁻¹, moderately increased.

There is no current consensus with respect to the best technique by which to analyze PDFF and R2* parametric maps, as wide variability exists with regard to the size and number of ROIs used to measure hepatic fat and iron content and the locations in which they are placed^59–61. These differences may impede accurate comparison of PDFF and R2* values across institutions. One recent investigation⁶⁰ found that intra- and interobserver variability was minimized by maximizing the volume of liver sampled through a large number of ROIs (in either the medial, lateral, posterior, and anterior segments of the liver or in each of the nine Couinaud segments). In the future, automated whole-liver segmentation may eliminate the need for manual ROI placement⁴⁹.

Confounder-corrected CSE MRI techniques have demonstrated excellent correlation with PDFF as calculated by MRS⁵², have been validated in both phantoms and animals^62,63, and, with a dynamic range of 0–100%, obviate challenges associated with heterogeneous hepatic fat deposition. Recently, CSE MRI techniques have been utilized to devise limited imaging protocols that permit quantification of liver fat and iron in as little as three breath-holds and five total minutes of table time, at a cost similar to that of most laboratory tests⁶⁴, highlighting the promise of this imaging method as a means of accurate, noninvasive diagnosis and monitoring of both hepatic steatosis and iron overload.

Figure 2.

MR elastogram in a patient with non-alcoholic steatohepatitis. Magnitude (A), color wave (B), color elastogram with confidence map (C), and gray-scale elastogram (D) are shown. Liver stiffness was 3.8 kPa, in the stage 2–3 fibrosis range.

Figure 5.

1.5T MRI in a patient with suspected hepatic steatosis. In-phase (A) and opposed-phase (B) images demonstrate a significant drop in signal intensity (SI) on opposed-phase images. Proton density fat fraction (C) images demonstrate a fat fraction (FF) of approximately 41% consistent with severe steatosis. R2* images (D) demonstrate a R2* of approximately 41 s⁻¹, within normal limits.

Table 1:

MR elastography liver stiffness measurement (LSM) correlation with fibrosis stage

Mean LSM	Fibrosis Stage
<2.5 kPa	Normal
2.5–3.0 kPa	Normal or Inflammation
3.0–3.5 kPa	Stage 1–2 Fibrosis
3.5–4.0 kPa	Stage 2–3 Fibrosis
4.0–5.0 kPa	Stage 3–4 Fibrosis
>5.0 kPa	Stage 4 Fibrosis or Cirrhosis

Synopsis.

Chronic liver disease is a large and ever growing problem in both the US and world healthcare systems. While histological analysis through liver biopsy is the gold standard for hepatic parenchymal evaluation, this is not feasible in such a large population of patients or as a way of monitoring change over time. This review discusses MRI based techniques for assessing hepatic fibrosis, hepatic steatosis, and hepatic iron content, with discussions of both current techniques and future advancements.

Key Points.

Chronic liver disease is a large and increasing problem in the world, and non-invasive imaging modalities are needed for assessment and long-term monitoring.

MR elastography is the predominant non-invasive MR method for detecting and staging hepatic fibrosis and shows high accuracy and reproducibility.

Multiple additional MR techniques are being researched for assessment of hepatic fibrosis.

Recent advances in hepatic fat and iron measurement using quantitative chemical shift-encoded MRI have made measuring hepatic iron and fat straightforward and accessible.

MR Elastography Technique.

• Available on all major vendor platforms and as an addition to existing MR system or part of a new MR installation.

• MRE is available for both 1.5T and 3T clinical scanners

• Uses an active driver system to produce and transmit acoustic pressure waves (60Hz frequency) into the liver.

• A passive driver is strapped snugly to the patient’s lower chest over the liver. Passive driver is connected to active driver with a plastic tube that transmits the acoustic waves.

• The propagating shear waves are visualized using a special phase-contrast MR sequence. As shear waves have a higher velocity in more stiff tissues, they are visualized as waves with longer wavelength⁴.

• An inversion algorithm automatically converts wave information (velocity, amplitude, and wavelength) into stiffness maps.

• Typically, 4 slices are obtained and for each slice, several images are obtained, both raw data and post-processed images.

• Magnitude images demonstrate anatomy and confirm appropriate location of the prescribed slices.

• Phase images can be used to confirm presence of propagating shear waves in the liver. Gray-scale and color elastograms are used for liver stiffness measurement and visual representation of degrees of fibrosis respectively.

MRE Liver Stiffness Measurement.

• There are both automated and manual methods for liver stiffness measurement (LSM) using MRE¹, though currently a manual process is the widely used method clinically.

• The process for manual LSM varies depending on the manufacturer and sequences obtained. In the end, regions of interest (ROIs) are drawn on the gray-scale elastogram to obtain the LSM, with careful correlation with the magnitude image to ensure the ROI is avoiding the liver edge and major vessels and correlation with the wave images to ensure only good quality waves are being sampled.

• If a confidence overlay is available, the ROI should be within the available (non-hashed out) area.

• ROIs should be as large as possible to ensure the largest area of liver is sampled⁶.

• The process is repeated for all four slices, with a weighted arithmetic mean calculated. This mean LSM can then be used to correlate with the histological stages of fibrosis (Table 1).

Hepatic Signal Fat-Fraction Calculation.

$$\eta =\left(\text{IP}-\text{OP}\right)/\left(2\text{IP}\right)=\text{ F }/\left(\text{W }+\text{ F}\right)$$

• Where η is the signal fat fraction, IP is the signal from the fat protons (F) plus the signal from the water protons (W), and OP is the signal from the water protons minus the signal from the fat protons.

• Given that the IP and OP signals are magnitude-only signals, there is a 50% ambiguity of the fat-fraction estimate for fat concentrations greater than 50%.

Pseudo-quantitative Estimation of Fat Fraction.

• Requires acquisition of two different sets of images, with identical parameters save for one set employs fat-saturation.

• The difference in signal intensity between the sequences is assumed secondary to fat, and an approximate fat fraction can be calculated by:

  $$\eta =\left(\text{SNFS}-\text{SFS}\right)/\left(\text{SNFS}\right)$$

• η is the signal fat fraction, SNFS is the liver signal intensity without fat suppression, and SFS is the signal intensity with fat suppression.

• Potential issues with this technique include the unintentional suppression of water signal and the incomplete or inhomogeneous suppression of fat, problems that would render the signal fat-fraction invalid².

Abbreviations:

CLD

Chronic liver disease

MRE

Magnetic resonance elastography

LS

Liver stiffness

GRE

Gradient-recalled echo

SE-EPI

Spin-echo echo planar imaging

IOP

In-phase and opposed-phase

CSE

Chemical shift-encoded

PDFF

Proton density fat fraction