Magnetic resonance imaging of hepatic fibrosis: Emerging clinical applications


  • Jayant A. Talwalkar,

    Corresponding author
    1. Advanced Liver Diseases Study Group, Miles and Shirley Fitterman Center for Digestive Diseases, Mayo Clinic College of Medicine, Rochester, MN
    • Mayo Clinic College of Medicine, 200 First Street SW, Rochester, MN 55905
    Search for more papers by this author
    • fax: 507-284-0538.

  • Meng Yin,

    1. Department of Radiology, Mayo Clinic College of Medicine, Rochester, MN
    Search for more papers by this author
  • Jeff L. Fidler,

    1. Advanced Liver Diseases Study Group, Miles and Shirley Fitterman Center for Digestive Diseases, Mayo Clinic College of Medicine, Rochester, MN
    2. Department of Radiology, Mayo Clinic College of Medicine, Rochester, MN
    Search for more papers by this author
  • Schuyler O. Sanderson,

    1. Division of Anatomic Pathology, Department of Pathology and Laboratory Medicine, Mayo Clinic College of Medicine, Rochester, MN
    Search for more papers by this author
  • Patrick S. Kamath,

    1. Advanced Liver Diseases Study Group, Miles and Shirley Fitterman Center for Digestive Diseases, Mayo Clinic College of Medicine, Rochester, MN
    Search for more papers by this author
  • Richard L. Ehman

    1. Department of Radiology, Mayo Clinic College of Medicine, Rochester, MN
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.


Chronic liver disease and cirrhosis remains a major public health problem worldwide. While the majority of complications from chronic liver disease result from progressive hepatic fibrosis, the available diagnostic tests used in clinical practice are not sensitive or specific enough to detect occult liver injury at early or intermediate stages. While liver biopsy can stage the extent of fibrosis at diagnosis, its utility as a tool for longitudinal monitoring will be limited at the population level. To date, a number of methods including serum marker panels and ultrasound-based transient elastrography have been proposed for the non-invasive identification of hepatic fibrosis. Novel techniques including magnetic resonance (MR) spectroscopy, diffusion weighted MR, and MR elastography have also emerged for detecting fibrosis. In contrast to other non-invasive methods, MR imaging holds the promise of providing functional and biological information about hepatic pathophysiology as it relates to the natural history and future treatment of hepatic fibrosis. (HEPATOLOGY 2007.)

Chronic liver disease and cirrhosis remains a major public health problem worldwide. In the United States during 2004, disease-related complications were associated with nearly 40,000 deaths1 and greater than 1.4 billion dollars being spent on medical services.2 Furthermore, these trends are expected to increase based on an aging population, the growing epidemic of obesity, and the continued emergence of clinical manifestations among individuals with longstanding chronic hepatitis C infection.3, 4 In terms of patients as well as populations, there remains a great need to develop and identify methods of risk stratification and prognosis for individuals with chronic liver disease. To date, the noninvasive diagnostic tests available from clinical practice are not sensitive or specific enough to function as screening tests for detecting occult yet significant liver injury. In turn, this often results in the application of invasive procedures such as liver biopsy and hepatic venous pressure gradient (HVPG) measurement as diagnostic tests. Given the inability to target the appropriate use of these procedures for asymptomatic patients at high risk for progressive disease, their widespread application and clinical usefulness will remain hampered by issues related to procedure-related complications, patient acceptance, and cost. Nevertheless, there remains an opportunity to improve the detection rate of individuals with progressive fibrosis who might benefit from early intervention.

In this regard, the development of several noninvasive methods to detect the major consequences of chronic hepatic injury has been ongoing. Although most recent investigations have focused on serum indices (indirect and direct) for detecting hepatic fibrosis, the common thread underlying these studies is that assessments are done in isolation of other manifestations related to chronic liver injury such as inflammation and portal blood flow.5, 6

Current Methods for Detecting Hepatic Fibrosis

Liver Biopsy.

To date, liver biopsy has been the gold standard for detecting hepatic fibrosis. However, the focus in clinical practice today has shifted from diagnosis to risk stratification and monitoring. For many situations, the application of conventional and disease-specific laboratory and imaging tests allow for the recognition of entities such as chronic viral hepatitis and nonalcoholic fatty liver disease. Distinguishing nonalcoholic steatohepatitis from nonalcoholic fatty liver disease or determining extent of fibrosis in hepatitis C for consideration of treatment are more common indications for liver biopsy.

Most classification systems recognize 5 stages of fibrosis, graded as F0 (no fibrosis), F1 (portal fibrosis), F2 (periportal fibrosis), F3 (bridging fibrosis), and F4 (cirrhosis). Clinically significant fibrosis is generally defined by F2 involvement or greater. However, a number of studies have demonstrated excessive rates of sampling error (25%-40%) resulting in poor reproducibility regardless of underlying liver disease origin.7 In addition, the extent of variation from observer interpretation by expert histopathologists may be as high as 20%.8

However, there is mounting evidence that liver biopsy has a number of limitations for its use in these roles as well. These include (1) the effect of reduced biopsy size (<25 mm) and complete portal tract number (<11) on understaging fibrosis: (2) interobserver variation in histological interpretation; and (3) the qualitative nature of assessing fibrosis in 2 dimensions with descriptive staining techniques. Existing histological scoring systems such as the Ishak9 and Metavir10 classifications provide a semiquantitative assessment of fibrosis extent, which, in theory, has the potential to allow for serial assessments to document progression. However, there has been little documentation of strong reproducibility for existing methods despite their widespread use in clinical practice and controlled trials. Finally, these limitations make reliable and valid serial assessments to document progressive fibrosis nearly impossible to perform.11 Although computerized image analysis protocols may offer the potential for quantitative analysis of liver fibrosis,12 this method requires further study to provide standardized approaches to ensure reliability.13 Ultimately, the method of percutaneous liver biopsy is an invasive procedure with poor acceptance by patients. The associated morbidity from this technique is estimated at 3% with a mortality rate of 0.03%.14 In turn, the strict requirement for using liver biopsy will not allow for the ability to assess populations at risk.

Despite these concerns, the assessment of prognostic histological features including degree of steatosis and inflammation can only be provided by histology from liver biopsy.11, 12 Most current imaging strategies looking at fibrosis, however, have been unable to decipher these histological correlates separately.15

Serum Markers.

A number of serum markers representing the process of hepatic fibrosis have been studied. However, most of these studies are limited by retrospective study designs, low rates of liver biopsy reproducibility, and the inclusion of patients with a narrow spectrum of disease severity. In turn, a number of indices contain parameters that only correspond to an indirect assessment rather than to direct quantification of hepatic fibrosis.15, 16 Nevertheless, the detection of advanced fibrosis with more recent serum marker panels has been considered good to excellent among individual studies. Combined indirect and direct serum marker panels can avoid the need for liver biopsy in up to 50% of cases.5, 15 The detection of intermediate fibrosis has been most difficult for most assays.15 Pooling estimates from selected studies, however, is noted for between-study heterogeneity and only modest diagnostic power in providing a diagnosis.16

Based on its potential for widespread clinical availability and economic cost, there would be incentive to enhance the diagnostic performance of serum marker tests. Emerging data have shown that a sequential algorithm-type approach to improve the posttest likelihood of detecting advanced fibrosis may be more effective than application of a single test alone.17

Conventional Imaging Techniques.

Advances in cross-sectional imaging that focus on detecting hepatic morphologic alterations and features of portal hypertension can be used to identify cases of established cirrhosis. However, the ability to detect early and intermediate stages of fibrosis with these techniques remains limited.18 Conventional ultrasound with standard Doppler assessment of the hepatic vasculature is insufficient for detecting early and intermediate stages of fibrosis.19, 20 Emerging data on hepatic vein blood flow pattern assessment by Doppler21 or contrast-enhanced22 methods suggest that notable improvements in detecting advanced fibrosis or cirrhosis are possible. Helical and multidetector row computed tomography provide improved resolution of early morphological changes with cirrhosis (in the absence of portal hypertension) but remain unhelpful with respect to fibrosis detection.23 Specific features on magnetic resonance imaging (MRI) including hepatic vein narrowing,24 caudate to right lobe ratio,25 and expanded gallbladder fossa,26 reliably identify cirrhosis but remain absent in earlier stages of fibrosis.


ADC, apparent diffusion coefficient; ATP, adenosine triphosphate; HVPG, hepatic venous pressure gradient; MR, magnetic resonance; MRI, magnetic resonance imaging; PDE, phosphodiesters; PME, phosphomonoesters; RF, radio frequency; SPIO, superparamagnetic iron oxide; T, tesla.

Novel MR Applications for In Vivo Detection of Hepatic Fibrosis

Over the past decade, a number of technological advances have been made in developing clinical applications for MRI of the liver. Recent improvements have focused on exploiting the physiological and biomechanical properties of human liver tissue to improve the detection of focal and diffuse pathological conditions.

Contrast-Enhanced Magnetic Resonance Imaging.

The introduction of breath-hold sequences with gadolinium chelates have facilitated the development of dynamic contrast-enhanced imaging of the liver in clinical practice.27, 28 In turn, a number of studies have described changes in hepatic parenchyma texture after contrast-enhanced imaging by MRI (Table 1).27–35 Early patchy enhancement may correlate with increased numbers of macrophages and necroinflammation on liver histology. In contrast, a delayed, heterogeneous enhancement pattern is associated with bright-appearing reticulations suggestive of hepatic fibrosis.28, 29 Reduced distribution of gadolinium in hepatic extracellular spaces may explain this observed finding.28

Table 1. Summary of Contrast-Based MRI Investigations for Detecting Hepatic Fibrosis in Human Subjects
AuthorYearMethodMR HardwareMR TechniqueNMajor EtiologiesRef StdOutcomeResults
  1. Abbreviations: MR, magnetic resonance; Ref Std, reference standard; Gd, gadolinium; SPIO, superparamagnetic iron oxide; T, Tesla; SSEP, single-shot echo planar; N/A, not available; GRE, gradient echo; FSE, fast spin echo, SNR, signal to noise ratio; SGE, spoiled gradient echo; HASTE, half-Fourier acquisition single-shot turbo spin echo; ROI, region of interest; CTP, Child-Turcotte-Pugh

Elizondo et al.1990SPIO0.3 T Body coilSpin echo ROI 250 pixels 60 min post infusion7 patients 8 normalN/ABiopsyCirrhosisReduced SNR in normal versus compensated cirrhosis versus decompensated cirrhosis
Yamashita et al.1996SPIO1.5 T Body coilSpin echo11 patients 2 normaln/aClinical and BiopsySerum tests, fibrosis stageNo significant correlations between SNR and fibrosis stage/ serum tests
Hundt et al.2000SPIO1.5 T Circular polarized body coilGRE, FSE ROI 0.6 cm2 15 min47 patients 30 normalAlcohol ViralBiopsyCirrhosisReduced SNR in normal versus cirrhosis Reduced SNR across CTP class
Semelka et al.2001Gd1.5T Body coilSGE, HASTE 45s, 90s, and 10 min29 patientsVariousBiopsyFibrosis stageGd enhancement patterns correlated with inflammation and fibrosis
Tanimoto et al.2002SPIO1.5 T Torso-array coilGRE, FSE ROI 0.6 cm2 40,109 min12 patients 12 normalViralClinical and BiopsyCirrhosisReduced SNR in normal vs. cirrhosis
Lucidarme et al.2003SPIO1.5 T Body or torso coilFSE, GRE 60 min ROI 256 or 512 pixels68 patients 9 normalViral AlcoholBiopsyFibrosis stageModerate correlation (kappa = 0.51) between reticular pattern and fibrosis
Aguirre et al.2006Gd + SPIO1.5 T Anterior, posterior phase array coilsSpin echo, SGE 30 min after SPIO 180 s after Gd101 patientsViral AlcoholClinical and BiopsyFibrosis stageDouble-enhanced image scores best to detect advanced versus mild fibrosis

In contrast, the administration of superparamagnetic iron oxide (SPIO) contrast agents creates hypointensity within liver parenchyma based on its accumulation in reticuloendothelial cells.30–32, 34, 35 Recently, the sequential administration of gadolinium followed by SPIO contrast media was done to determine whether improved detection of liver fibrosis architecture could be observed (Fig. 1). Although SPIO-based images were highly accurate (85%) for detecting fibrosis compared with histopathological features, the combination of gadolinium-based and SPIO contrast agents yielded even greater accuracy (93%) overall.32 Further refinements in image acquisition, additional validation of semiquantitative fibrosis criteria including nodularity and reticulation stage, and prospective application to verify the diagnostic performance of this technique is anticipated.33

Figure 1.

Transverse MR images of liver in 39-year-old woman with METAVIR score of F4 on liver biopsy. Double-enhanced spoiled gradient-echo image shows hyperintense reticulations (arrows) in addition to hypointense nodules. Findings are thought to represent regenerating nodules surrounded by fibrotic septal tissue. (Adapted from Aguirre DA, Behling CA, Alpert E, Hassanein TI, Sirlin CB. Liver fibrosis: noninvasive diagnosis with double contrast material-enhanced MR imaging. Radiology 2006;239:425-437.)

Diffusion-Weighted Magnetic Resonance Imaging.

Diffusion-weighted magnetic resonance imaging (DWI) is a technique that assesses the degree of molecular diffusion in tissues with primary application for the early detection of cerebral ischemia. Recent advances have reduced image distortion and increased signal-to-noise ratio to allow for abdominal imaging.36 Diffusion is the term used to describe the random thermal motion of molecules (also known as Brownian motion) in solid matter. In hepatic fibrosis, it is thought that collagen is not proton rich and will have tightly bound nuclei overall. In turn, a reduction in diffusion of water content would be associated with increased fibrosis, which restricts Brownian motion within hepatic parenchyma.36, 37

Recent advances in MRI technology have facilitated the performance of diffusion-weighted MRI of the liver. Previous experience with physiological motion artifacts and poor image resolution have been addressed by the use of faster pulse sequences, cardiac and respiratory triggering, and use of a single breath-hold.38 The hardware required for diffusion-weighted MRI includes a magnet, a specialized radio frequency (RF) coil for transmitting/receiving signals, and a computerized system for data collection and processing. Most investigations assessing liver fibrosis have used horizontal bore magnets with a field strength of 1.5 Tesla (T). The RF coil may envelope the patient or lie on the anterior surface to facilitate liver imaging. Electronic hardware within the coil amplifies and processes the RF signals for transmission. Once typical contraindications for MRI are excluded, an individual patient lies supine on the MRI table with RF coils positioned appropriately. After entry into the magnet, a standard MRI examination can be performed that includes views to locate the appropriate anatomical location and volume (defined by number of voxels in mm3) for measurement of diffusion. Subsequently, there are specific motion-probing gradients applied before and after rapid MR pulse sequences that generate data for acquisition.39 Reported times for breath-holding are usually between 10 and 20 seconds per measurement. Typical examination time for both standard and diffusion-weighted MRI is approximately 45 to 60 minutes.

The primary variable measured with diffusion-weighted MRI is the apparent diffusion coefficient (ADC), which is thought to represent the hydration and metabolic status of imaged tissue. The ADC, is calculated by the formula ADC = −(lnS2 − lnS1)/(b2 − b1), where ln is the natural log with S1 and S2 as the signal intensities in the region of interest placed on sections corresponding to 2 different b values (b1 and b2). The term b value represents a defined time interval where proton dephasing occurs to facilitate measurement of diffusion and is measured in seconds/mm3. In general, lower b values will correspond to higher mean ADC values, whereas higher b values may result in underestimation of ADC based on signal degradation from poor image quality.39, 40 A number of ADC values, therefore, can be generated for a series of b values. Quantitative analysis of ADC values requires identification of a region of interest, which is typically positioned in the right hepatic parenchyma to avoid major vascular structures. Triplicate values with reporting of the mean ADC value is often performed using computerized software.37, 39, 40

Several reports have described lower ADC value in patients with varying degrees of cirrhosis as compared with healthy individuals undergoing diffusion-weighted MRI (Table 2). Recent studies among patients with varying degrees of chronic liver disease from hepatitis C, however, are mixed with respect to demonstrating a specific relationship between ADC values and fibrosis stage.40, 41 Sampling error from liver biopsy may be responsible, in part, for this result.41 Using a single-shot spin-echo planar sequence to improve reproducibility of measurement, the mean ADC in patients with cirrhosis was lower compared with healthy individuals. In this experience, a 93% sensitivity and 100% specificity for detecting cirrhosis was associated with a cutoff ADC value of 1.31 × 10-3 mm2/s.39 Notably, the change in ADC with greater degrees of fibrosis may not be related to reductions in water diffusion. Recent data suggest that decreased ADC values in rodents with hepatic fibrosis were observed in vivo but not when DWI was performed in euthanized rats or fixated livers. This suggests that reduced hepatic perfusion may explain the decrease seen for in vivo ADC values with increasing degrees of fibrosis, which has been theorized when using lower b values.38

Table 2. Summary of Diffusion-Weighted MRI Investigations for Detecting Hepatic Fibrosis in Human Subjects
AuthorYearMethodMRI HardwareMRI TechniquenMajor CausesRef StdOutcomeResults
  1. Abbreviations: MRI, magnetic resonance imaging; Ref Std, reference standard; Sens, sensitivity; Spec, specificity; DWI, diffusion-weighted imaging; MRE, magnetic resonance elastography; T, Tesla; GRE, gradient echo; SSEP, single-shot echo planar; MPG, motion probing gradient; ADC, apparent diffusion coefficient; n/a, not available; CTP, Child-Turcotte-Pugh; HA, hyaluronic acid.

Girometti et al.2007DWI1.5 T Phased array surface coil and spine array coilSpoiled GRE SSEP b values: 0, 150, 250, 400 s/mm228 cirrhosis 29 normalViral AlcoholBiopsyCirrhosisMean ADC lower in cirrhosis (1.11 × 10−3 versus 1.54 × 10−3) Sens 93%/ Spec 100%
Koinuma et al.2005DWI1.5 T Circular polarization body array coilSSEP MPG pulses b values:.01, 128 s/mm2119 patients 34 normalViralBiopsy or clinicalFibrosis stageADC decreased with increasing fibrosis
Aube et al.2004DWI1.5 T Body coilSpin echoplanar b values: 200, 400, 600, 800 s/mm213 cirrhosis 14 normalAlcoholBiopsyCirrhosisMean ADC lower in cirrhosis (2.05 × 10−3 versus 2.9 × 10−3) ADC correlated with CTP score, serum HA
Boulanger et al.2003DWI1.5 T Body coilSpin echoplanar b values: 50, 100, 150, 200, 250 s/mm218 patients 13 normalHepatitis CBiopsyFibrosis stageMean ADC lower in cirrhosis (2.3 × 10-3 versus 1.8 × 10−3) No correlation between ADC, inflammation, or fibrosis

Several aspects of diffusion-weighted MRI liver imaging require attention before extending its application in human subjects. Studies have used small numbers of patients and various hardware and sequencing profiles, rendering comparisons between experiences difficult. Significant hepatic iron accumulation appears to disrupt the magnetic field, rendering problems with diffusion assessment for patients with genetic or secondary hemochromatosis.37 Hepatic steatosis also may increase the number of free protons and impact ADC values unless fat suppression techniques are employed to minimize this phenomenon.39 Reproducibility has not been widely established given the limited experience thus far. Technical factors such as accounting for cardiac motion despite breath-holding as well as employing surface rather than body coils to reduce signal-to-noise ratio are being addressed to allow further study.37, 40, 41 High-field (3T) imaging also may enhance signal-to-noise ratio, reduce image misregistration, and subsequently improve hepatic ADC detection.42 Selection of the appropriate b value to capture diffusion without effects of perfusion remains an important area for development.

Magnetic Resonance Spectroscopy.

The technique of in vivo magnetic resonance (MR) spectroscopy has been available for over 2 decades to investigate the metabolic processes of organ tissue.43 Based on its anatomical location and increased metabolic demands, the liver is considered an ideal organ for MR spectroscopy investigation.44, 45 Various isotopes are used for MR spectroscopy including hydrogen (1H), carbon (13C), nitrogen (15N), and fluorine (19F). These nuclei behave as magnetic dipoles that align parallel or perpendicular to an applied magnetic field. With removal of the field, the nuclei return to their baseline state and generate a weak RF signal. These signals are detected by MRI and transformed into a frequency spectrum to provide biochemical information Although 1H-based MR spectroscopy allows for quantification of lipid profiles and 13C- based techniques focus on carbohydrate/amino acid metabolism, 31P-based MR spectroscopy provides insights on processes including cell turnover and energy state based on substantial concentrations within hepatocytes.43–45

Within the spectrum of 31P compounds, there are 6 discrete signals analyzed by MR spectroscopy in human subjects, including (1) phosphomonoesters (PME), (2) inorganic phosphate, (3) phosphodiesters (PDE), (4) adenosine triphosphate (ATP), alpha-ATP, and beta-ATP. The chemical precursors phosphocholine, phosphoethanolamnine, adenosine monophosphate, and glycolytic intermediates such as glucose-6-phosphate contribute to the PME peak. Glycerophosphorylcholine, glycerophosphorylethanolamine, and mobile phospholipids from the endoplasmic reticulum are the main components of the PDE peak. Both PME and PDE appear to provide information on cellular degradation.46, 47

The methodology used for performing MR spectroscopy has varied over time and between studies. Hardware required for MR spectroscopy includes a magnet, RF coil for transmitting/receiving signals, and computerized system for data collection and processing. All in vivo MR spectroscopy investigations to date for assessing liver fibrosis used horizontal bore magnets with a field strength of 1.5 Tesla (T). The RF coil may envelope the patient or lie on the anterior surface to facilitate liver imaging. Electronic hardware within the coil amplifies and processes the RF signals for transmission. Once typical contraindications for MRI are excluded, an individual patient lies supine on the MRI table with RF coils positioned appropriately. After a standard MRI examination, a number of MR pulse sequences are applied to generate spectroscopic data within the appropriate anatomical location and volume (defined by voxels) of interest. As with diffusion-weighted imaging, a typical examination will take 45 to 60 minutes.43–45

The spectral analysis of data requires filtering to reduce noise before conversion to produce a frequency spectrum. Metabolite concentrations can be expressed in absolute or relative terms. In general, the peak area of a metabolite signal is directly related to its concentration. Because the absolute quantitation of metabolites is difficult to achieve in vivo, many studies have used metabolite ratios for assessing spectral profiles.43, 46 Peak areas in a spectrum are referenced to standards for correlation with MR signal intensities. An internal standard such as adenosine triphosphate (ATP) can be used, given its natural occurrence in tissue.43

Whereas a number of in vivo studies have examined the diagnostic performance of MR spectroscopy for characterizing hepatic lesions,44 there is growing interest in the role of MR spectroscopy for detecting hepatic fibrosis (Table 3). Among patients with established cirrhosis, an increased hepatic PME signal measured by MR spectroscopy has been reported.48–51 The relationship between PDE and cirrhosis, however, is not well understood.45 For patients with varying degrees of involvement with chronic hepatitis, an increasing PME/PDE ratio was found to correlate with necroinflammatory and fibrosis scores on liver histology compared with healthy volunteers.52–55 Although differences between means for each group were statistically significant, there was some overlap between the patient groups.45

Table 3. Summary of MRI Spectroscopy Investigations for Detecting Hepatic Fibrosis in Human Subjects
AuthorYearMethodMRI HardwareMRI TechniqueNMajor CausesRef StdOutcomeResults
  1. Abbreviations; MRI, magnetic resonance imaging; Ref Std, reference standard; Sens, sensitivity; Spec, specificity; 1H, labeled hydrogen; 31P, labeled phosphorus; ISIS, image selected in vivo spectroscopy; TR, repetition time; NA, number of acquisitions; PBC, primary biliary cirrhosis; CSI, chemical shift imaging; PME, phosphomonoester; Pi, inorganic phosphate; Glx, glutamate; Glu, glucose; GPE, glycerophosphorylethanolamine; GPC, glycerylphosphorylcholine; PE, phosphoethanolamine; PC, phosphocholine; PDE, phosphodiester; ATP, adenosine triphosphate; P, inorganic phosphorus; n/a, not available; CTP, Child-Turcotte-Pugh; HA, hyaluronic acid; MRE, magnetic resonance elastography.

Schlemmer et al.200531P MRS1.5 T 14 cm 1H/31P surface coilISIS TR 1100 ms 500 slice width Voxel 45 ml40 patients 13 normalAlcoholBiopsyFibrosis stageGPE/GPC, PE/PC, and PME/PDE ratios increase from normal to cirrhosis
Lim et al.200331P MRS1.5 T 14 cm 1H/31P surface coilISIS TR 800, 10,000 ms Voxel 7 cm348 patients 15 normalHepatitis CBiopsyFibrosis stagePME/PDE ratio increased with fibrosis stage Sens 80%, Spec 80%
Cho et al.200131P MRS1.5 T No data on body coilCSI TR 3000 ms NA 128 Voxel 8 cm375 patients 15 normalViralBiopsyFibrosis stageGlx, PME, Glu spectral ratios increased with fibrosis stage
Taylor-Robinson et al.199731P MRS1.5 T 15 cm 1H/31P surface coilCSI TR 500 ms NA 80 30 mm slice width31 cirrhosisVariousClinical and BiopsyCTP scoreIncreased PME and reduced PDE with worsening cirrhosis
Menon et al.199531P MRS1.5 T 15 cm 1H/31P surface coilCSI TR 500 ms NA 80 30 mm slice width85 cirrhosis 16 normalVariousBiopsyCirrhosisIncreased PME/PDE, PME/ATP and lower PDE/ATP in cirrhosis Higher Pi/ATP in viral disease
Van-Wassenar et al.199531P MRS1.5 T 15 cm 1H/31P surface coilISIS TR 1500 ms NA 256 Voxel 200-500 ml38 patients 22 normalVariousClinical and BiopsySerum enzymes, histologyPME/P correlation with AST, and inflammation No relationship with fibrosis.
Jalan et al.199531P MRS1.5 T 15 cm 1H/31P surface coilCSI TR 500 ms NA 80 30 mm slice width23 cirrhosis 16 normalPBCClinicalCTP ScoreSpectral ratios higher in cirrhosis PME/Pi correlates with CTP score
Munakata et al.199331P MRS1.5 T 15 cm 1H/31P surface coilCSI TR 500 ms14 cirrhosis 7 normalVariousClinicalCirrhosisIncreased PME with worsening cirrhosis

Reconciling the biologic plausibility of MR spectroscopy results with histological manifestations of chronic liver disease has been attempted. Changes in phospholipid metabolites seen in liver disease are believed to represent regenerative activity.50 An increased PME signal may represent extensive membrane remodeling that occurs in liver disease.44 Reduced PDE may be because of a decrease in the rate of cell membrane breakdown. Reduced ATP levels have also been discovered in association with cirrhosis. However, the presence of ATP and other phosphorus-based compounds have yet to be clearly associated with clinically relevant outcomes.45

Several limitations with current MR spectroscopy approaches, however, are observed when comparisons between existing studies are made. Most studies contain small numbers of patients from heterogeneous populations assessed by varying MR spectroscopy methods.43 Scan time and volume of liver examined in terms of voxel number are also different between studies. In addition, the variation in reproducibility of data acquisition from healthy volunteers can range between 4% and 20% for both subject and examination. The appearance of individual signals can be affected by variations in data acquisition and subsequent analysis.45 The overall specificity of calculated signals may also be reduced based on contributions from nonhepatocyte cells (macrophages, endothelial cells) and overlap between certain spectral profiles.44, 45, 47 Ultimately, the role of in vivo MR spectoscopy for detecting hepatic fibrosis requires assessment in larger diagnostic accuracy studies among patients with various hepatobiliary disorders.

Magnetic Resonance Elastography.

Disease processes such as malignancy create increased tissue stiffness, which may be identified by manual palpation of selected organs on physical examination. Based on ex vivo and intraoperative studies, the elasticity (or stiffness) of hepatic parenchymal tissue may also be influenced by the presence of fibrosis.56 Furthermore, the resulting stiffness of hepatic matrix may be stimulatory for continued hepatic stellate cell transdifferentiation and fibrosis production.57 With recent advances in biomedical imaging, the ability to quantify in vivo stiffness by detecting wave propagation velocity through specific human tissue has now become possible.56, 58 For example, the measurement of in vivo liver stiffness by ultrasound-based transient elastography appears feasible and valid to predict the severity of hepatic fibrosis for selected patients with advanced histological disease.59–62

Recently, the technique of MR elastography has also been described for assessing shear stiffness in various tissue types.58, 63, 64 MR elastography uses a modified phase-contrast imaging sequence to detect propagating shear waves within the organ of interest.64 The technique can be implemented on a conventional MRI system with additional hardware and software required for examination performance. Following a standard MRI examination, a pneumatic or electromagnetic transducer is placed on the anterior abdomen in a supine patient and used to generate propagating mechanical waves in the liver at frequencies between 40 and 120 Hz.65–68 In our experience, the pneumatic driver can be implemented as a simple passive drumlike device that is placed over the anterior abdominal wall and activated using air pressure waves from a remotely located speakerlike device.66, 68 (Fig. 2). A single breath-hold of 10 to 15 seconds is required to allow for one measurement of wave propagation.68

Figure 2.

Setup for applying shear waves to the abdomen for MR elastography of the liver. Acoustic pressure waves (at 60 Hz) are generated by an active audio driver, located away from the magnetic field of the MRI unit, and transmitted via a flexible tube to a passive pneumatic driver placed over the anterior body wall. The left diagram is a coronal illustration of the location of the passive pneumatic driver (circle) with respect to the liver. (Adapted from Yin M, Talwalkar JA, Glaser KJ, Manduca A, Grimm RC, Rossman PJ, et al. Assessment of hepatic fibrosis with magnetic resonance elastography. Clin Gastro Hep 2007;5:1207-1213.e2.)

A specialized phase-contrast MRI sequence is then used to image the propagating waves in the liver. This sequence uses motion-encoding gradients that are oscillated in precise synchronization with the applied vibrations, allowing waves with amplitudes in the micron range to be readily imaged. Each MR elastographic acquisition provides an image that represents the displacement caused by shear wave propagation in the medium.63, 64 The wave images are then processed using a specially developed inversion algorithm software to generate quantitative images that depict tissue stiffness called elastograms (Fig. 3A,B).58 Quantitative mean elasticity values (from triplicate measurement) within a region of interest just below the initial point of wave propagation into the liver can also be calculated. The unit of measurement for elasticity is kiloPascals, as it is with ultrasound-based transient elastography.

Figure 3.

(A) MR elastography of the liver in a healthy volunteer and a patient with cirrhosis. The middle column of images shows wave image data in the liver and spleen, superimposed on the corresponding anatomical images. The resulting elastograms are shown in the far right column. Elastograms show a higher mean stiffness of the fibrotic liver compared with the normal liver (12.1 ± 1.2 kPa versus 1.8 ± 0.3 kPa, respectively). (Adapted from Yin M, Talwalkar JA, Glaser KJ, Manduca A, Grimm RC, Rossman PJ, et al. Assessment of hepatic fibrosis with magnetic resonance elastography. Clin Gastro Hepatol 2007. In press). (B) MR elastography of the liver in patients with obesity and ascites. The top row demonstrates a patient with obesity (BMI = 36) and stage 2 fibrosis on liver biopsy with a mean liver stiffness of 3.2 ± 0.8 kPa. The bottom row illustrates a patient with ascites. Excellent shear wave illumination of the liver was obtained, and the mean liver stiffness was 11.3 ± 2.8 kPa. (Adapted from Yin M, Talwalkar JA, Glaser KJ, Manduca A, Grimm RC, Rossman PJ, et al. Assessment of hepatic fibrosis with magnetic resonance elastography. Clin Gastro Hepatol 2007;5:1207-1213.e2.)

Preliminary studies have confirmed the feasibility of MR elastography assessment in phantom models and human subjects (Table 4).65–67 The presence of overlying ribs does not interfere with the quality of wave propagation and calculation of stiffness values.66 Reliability assessed by coefficients of variation within subjects is reported at 7% to 9%.65 From 2 experiences published in manuscript form,65, 66 the mean liver shear stiffness in patients with chronic liver disease was found to be significantly higher (P < 0.001) compared with the mean value for healthy volunteers. Initial results suggest a quadratic relationship between histological fibrosis stage versus elasticity measurements as observed in studies with ultrasound-based transient elastography.59–61 Issues with biopsy quality, sampling error, and observer variation may, in part, explain this relationship. Moreover, it was observed that elastogram values in the liver demonstrated heterogeneous patterns, further suggesting the validity of its relationship with fibrosis.66, 68 Recently, the reproduction and validation of MR elastography was performed in an independent population of 35 healthy individuals and 48 patients with varying degrees of chronic liver disease. In this experience, a sensitivity of 86% and specificity of 85% were noted for the detection of stages 2 to 4 fibrosis compared with liver histology from biopsy. No relationship between liver stiffness and degree of hepatic steatosis was observed. High negative predictive value (97%) for excluding the presence of fibrosis was also noted, suggesting that MR elastography could have a role for improving the ability to risk-stratify patients for liver biopsy to exclude occult advanced fibrosis.68

Table 4. Summary of MR Elastography Investigations for Detecting Hepatic Fibrosis in Human Subjects
AuthorYearMethodMRI HardwareMRI TechniqueNMajor EtiologiesRef StdOutcomeResults
  1. Abbreviations: MRI, magnetic resonance imaging; Ref Std, reference standard; MRE, magnetic resonance elastography; T, Tesla; GRE, gradient echo; MPG, motion probing gradient; N/A, not available; CTP, MRE, magnetic resonance elastography; MEG, motion encoding gradient; SSFP, steady-state free precision; GRE, gradient echo.

Huwart et al.2006MRE1.5 T Body coilMEG Spin echo TE 61 ms TR 431 ms Wave frequency = 65 Hz25 patients 5 normalViral AlcoholBiopsyFibrosis stageMean liver elasticity increased with fibrosis; mean liver viscosity higher in cirrhosis
Klatt et al.2006MRE1.5 T Body coilMEG SSFP TR 980 ms Wave frequency = 51 Hz2 patients 12 normaln/aBiopsyFibrosisMean elastic moduli was higher in patients
Rouviere et al.2006MRE1.5 T Body coilMEG GRE TE 18 ms TR 33 ms Wave frequency =90 Hz12 patients 12 normalHepatitis CBiopsyFibrosis stageMean liver stiffness greater in patients

Several unique qualities are observed with MR elastography with respect to its potential broad application in heterogeneous populations. These include (1) a freely oriented field of view, (2) no acoustical window requirement, (3) operator independence, (4) insensitivity to body habitus, (5) potential assessment of the entire hepatic parenchyma as the region of interest, and (6) the ability to obtain a conventional MR examination at the same time.6, 65, 66 As with the other techniques, efforts to standardize the equipment and techniques used for MR elastography should be pursued to maximize diagnostic accuracy and facilitate comparison of results in different settings. Reproducibility appears good from initial studies (coefficient of variation <10%) but requires additional study for verification. Further prospective evaluation is required for characterizing the diagnostic performance characteristics of MR elastography, including its longitudinal reproducibility among patients with stable and progressive liver disease.

Conduct of Prospective MR Studies in Human Subjects

Issues in Study Design.

The design and conduct of high-quality diagnostic accuracy studies is essential for ongoing validation of emerging noninvasive techniques for hepatic fibrosis. Specific recommendations have been published to improve the reporting of study results while minimizing the effects of bias as much as possible.69 Particular features of note include (1) the use of consecutive patients, (2) the study of a wide spectrum of disease severity, (3) the application of test and reference standards in all patients, (4) interpreter blinding of results for all tests, and (5) the appropriate measurement and reporting of diagnostic test parameters including measurements declared to be indeterminate or incomplete. The majority of published studies in hepatology to date have not explicitly described the use of formal methods to prevent observer error.70

As with other investigations of noninvasive methods to detect hepatic fibrosis, the inclusion of small patient numbers and lack of independent assessment for test results is also observed with MRI techniques. In addition to considering the criteria listed above when designing future diagnostic accuracy studies, there should also be an emphasis placed on rationale and justification for power and sample size calculations. The decision to employ a continuous or dichotomous outcome variable will require an explicit description of the goal to be sought. A binary primary endpoint (such as the detection of advanced versus early fibrosis) will be useful when this knowledge can facilitate a change in clinical management strategy (in other words, institution of surveillance procedures associated with cirrhosis). Greater statistical complexity arises, however, when designing clinical trials to examine multiple categorical stages of fibrosis, which is needed to assess treatment candidacy. In this instance, there will be greater demands for larger samples sizes to preserve study power. It is generally recognized that measurement of a continuous outcome would be an ideal situation to provide a means for assessing prognosis and treatment response as antifibrotic therapies emerge in the future. Substantial work will be required, however, to verify the surrogate or biological properties of such an endpoint with conduct of validation studies of sufficient length given the slowly progressive nature of diseases such as chronic hepatitis C and nonalcoholic steatohepatitis.

Issues with Reference Standard.

Regardless of acquisition method and specimen length, the persistence of sampling variability associated with liver biopsy remains a significant obstacle in considering this as a true reference standard. However, the use of liver biopsy will still be required unless alternate methods are recognized as reliable and valid.

Recent interest has focused the role of portal venous pressure as a potential reference standard representing the pathophysiological effects of hepatic fibrosis. A recent analysis verified the strong correlations between septal fibrosis thickness and regenerative nodule size with portal venous hypertension.71 In turn, the use of HVPG measurement could also serve as a reliable and valid reference standard for diagnostic accuracy studies given its ability to assess prognosis.72 This has already been demonstrated in the setting of recurrent hepatitis C after liver transplantation as liver stiffness by transient elastography was more strongly correlated with HVPG measurement as compared with histology from transjugular biopsy methods.73 However, further clarification of potential limitations with HVPG will likely be required. For example, hemodynamic alterations within the intrahepatic vasculature may occur independently of architectural changes related to fibrosis. Therefore, additional evidence is still required to demonstrate the sensitivity of HVPG to mild changes in fibrosis. Using multiple or objective end points such as death and clinical hepatic decompensation may address these issues. Finally, the use of HVPG as a corroborative reference standard will most likely be successful within the context of clinical trials because physician and patient acceptance of this technique remains questionable in practice. Further assessments incorporating physiological endpoints such as HVPG appear inevitable as interest in developing noninvasive detection techniques continues to grow.74

Defining the Role of MR Imaging for Hepatic Fibrosis in Clinical Practice

Even with further advances in MRI, a number of issues remain unsettled before these methods can be effectively translated into clinical practice. The ability to perform multiple examinations without concern over the cumulative risk of ionizing radiation is advantageous for MRI.75 In turn, patient acceptance rates of image-based assessments for hepatic fibrosis should be higher when compared with liver biopsy, especially if these methods can provide serial assessments in detecting fibrosis progression.

The clinical applicability of any MRI-based technology, however, will require endorsement by practitioners for use in the examination of their patients. As with any emerging technology, cost and available expertise will play important roles in determining the utility of MRI-based technologies for detecting hepatic fibrosis. Moreover, the pragmatic advantages of less expensive and, perhaps, more readily available methods (for example, serum markers) needs to be balanced against any incremental benefit in diagnostic accuracy that imaging would produce. Furthermore, the sophisticated nature of emerging MRI techniques must not obstruct its diffusion from tertiary medical centers to the community, where limited availability of required equipment may exist in specific geographic areas. Finally, there will need to be alternative strategies for patients who are unable or unwilling to undergo MRI examination (for example, those with claustrophobia requiring conscious sedation). Establishing the cost-effectiveness of MRI-based techniques would assist greatly in facilitating the rate of translation of these techniques into clinical practice.

Although the number of patients screened for hepatic fibrosis could increase using MRI, proof will be required that early detection and intervention can reduce morbidity and resource utilization associated with the clinical sequelae of advanced disease. This is particularly important in patients with advanced fibrosis, where surveillance for hepatocellular carcinoma could also be performed assuming its value is established in clinical trial settings.76


The development of a reliable and valid noninvasive method to assess hepatic fibrosis could result in comparable or, perhaps, improved accuracy in terms of staging. Several MR techniques (or combination of approaches) are being refined to meet these requirements and demonstrate early promise. As described previously,77 the absence of robust noninvasive markers now provides the most significant barrier to clinical trial development for novel pharmacological strategies to combat hepatic fibrosis. Should quantitative assessment by novel MRI techniques be predictive of clinical outcomes of hepatic fibrosis, then perhaps additional study to demonstrate reliable longitudinal assessment in the setting of therapy may be possible. Nevertheless, the emergence of MRI techniques (singly or in combination with other methods) could result in the performance of true functional hepatic imaging.76