Ultrasound or MR elastography of liver: which one shall I use?

Abstract

Liver stiffness is now a well-established noninvasive biomarker for assessing fibrosis in chronic liver disease. MRI-based and ultrasound-based dynamic elastography techniques have been introduced for assessment of liver stiffness and useful in clinical staging of hepatic fibrosis. Several different elastography techniques are now available with each method having inherent strengths and limitations. The published literature generally indicates that MR elastography has a higher diagnostic performance and fewer technical failures than ultrasound-based elastography techniques in assessing hepatic fibrosis. There is also significant potential to further develop elastography techniques to implement multiparametric methods that have promise for distinguishing between processes such as inflammation, fibrosis, venous congestion, and portal hypertension that can result in increased liver stiffness. In this commentary, we compare MR and ultrasound elastography methods and their utility in clinical practice.

Hepatic fibrosis is the common pathway of progressive hepatic damage resulting from many different causes of liver injury. Hepatic fibrosis can be clinically silent until it progresses to a high-mortality end stage of cirrhosis and associated complications. Practiced since the late nineteenth century, liver biopsy has been the gold standard to diagnose etiologies and assess fibrosis stage. However, this invasive method has the following drawbacks: risk of complications, sampling error [1], and subjective scoring system with considerable intra- and interobserver variability [2, 3]. Due to the growing concern about nonalcoholic fatty liver disease (NAFLD), which affects one-third of the adult population in the US, it can progress to steatohepatitis with concomitant inflammatory and fibrotic changes. It is critically important for clinical management to employ a safer, more comfortable, and less expensive alternative to liver biopsy for diagnosing hepatic fibrosis.

Soft tissues have complex and intricate structures and their mechanical properties are accounted by their microstructure and dynamic environment. Many disease processes cause marked changes in tissue mechanical properties, which can be associated with composition change, matrix deformation, and damage accumulation. Mechanical properties are therefore promising biomarkers for monitoring and characterizing various pathophysiologic conditions of tissues. In patient care, the innovative liver stiffness (LS) is beginning to see widespread clinical use for assessing hepatic fibrosis as an alternative biomarker to biopsy and is recommended in several guidelines for assessment of liver fibrosis [4–7]. In this emerging field of research, there are two major imaging methods of LS assessment for hepatic fibrosis: MR and ultrasound-based elastography.

Clinically, liver fibrosis staging is needed for management of chronic liver disease (CLD) patients. Patients without any fibrosis would be treated for the etiology of CLD, while patients with advanced fibrosis and/or cirrhosis would benefit from screening for esophageal varices and development of hepatocellular carcinoma. Patients in the middle category with mild–moderate fibrosis would be followed up for progression or improvement of fibrosis and would fall in one of the two categories described above. Elastography methods would therefore be most useful in clinical practice for ruling out any liver fibrosis and ruling in cirrhosis.

Methods

For both MR and ultrasound-based elastographic methods, there are two different approaches being explored. One is quasi-static elastography, in which the tissue deformation is produced by an external palpation with a probe or endogenous stress such as cardiovascular movements. In the resulting strain images, the least deformed regions are the stiffest, while the most deformed regions are the softest. However, the resulting strain images are not quantitative because the applied force can be neither controlled nor known with certainty. This method has been used in real-time ultrasound elastography [8] and in measuring liver strain secondary to cardiac motion using tagged MRI methods [9].

The other approach is with dynamic shear wave imaging that assesses quantitative stiffness by tracking dynamic shear wave propagation through the tissue. Shear waves can be generated by external mechanical vibration on the surface (abdominal wall) or by focusing acoustic radiation force impulses (ARFI) inside the tissue of interest. Stiffness is calculated by measuring complex shear modulus or shear wave velocity. Quantitative measurements are displayed as stiffness maps, which are called elastograms.

Quasi-static strain imaging has very limited applications for the liver as it is difficult to control the stress on the liver and quantify liver stiffness accurately. Therefore, we will focus on dynamic shear wave imaging techniques in this commentary. Both ultrasound and MR-based dynamic elastography methods have been introduced for clinical staging of fibrosis. Some of the methods are commercially available. Each has their inherent strengths and weaknesses, as shown in Table 1.

Table 1.

Assessment of liver fibrosis with biopsy, ultrasound, and MR elastography [1–3, 5, 6, 10–15, 19–33, 35–44]

	Volume	Acquisition time	Repeatabilityc	Reproducibilityc	Variationc	Bias	Success rate	Limita tions	Diagnostic performance (AUROC) F ≥ 2	Diagnostic performance (AUROC) F ≥ 4
Liver biopsy	0.1–0.5 cm3	6–15 min	0.33–0.76	0.33–0.76	24–33%	Sampling Subjective scoring System	80–97%	Invasive	N/A	N/A
Transient elastography (TE)	1 cm3	<1 s × 10a	0.60–0.95	0.86–0.96	12%	Sampling Diffraction Dispersion	75–94% (M. XL)	Ascites Obesity Narrow intercostal space	0.73–0.93	0.87–0.99
Point shear wave elastography (pSWE)	0.5–1 cm3	<1 s × 10a	0.83–0.96	0.83–0.93	15–21%		94–97%	Obesity Narrow intercostal space	0.80–0.89	0.93–0.95
Shear wave elastography (SWE)	20 cm3	<1 s × 10a	0.65–0.95	0.78–0.98	3–32%			90–99%	0.85–0.98	0.93–0.98
Magnetic resonance elastography (MRE)	≥250 cm3	16 s × 4b	0.88–0.96	0.90–0.99	6–10%	Inconsistent breath-hold	94–99%	Iron overloadd Claustrophobia	0.88–0.98	0.92–0.99

^aNumber of times for repeating measurements generally used in clinical applications of ultrasound-based elastography

^bNumber of 2D slices for liver coverage generally used in clinical acquisition of liver MRE

^cRepeatability/reproducibility/variation: Kappa value as in biopsy, interobserver correlation coefficient as in TE, pSWE, SWE & MRE (both within and between systems)

^dOnly unsuccessful in the liver of R2* > 400 s⁻¹ with an advanced liver MRE technique

Technical comparison

Ultrasound dynamic elastography has two main approaches that are well established in the clinical practice. They are transient elastography (TE) using an extrinsic vibrating source at 40–50 Hz and shear wave elastography (SWE) using an ARFI at 100–500 Hz.

TE uses a mechanical push to vibrate the skin and create a passing distortion in the tissue. Measurement of the speed of distortion as it passes deeper into the body is performed using a 1D ultrasound beam. TE measures LS in a 1D volume of a 5-mm-wide, 25–30-mm-long, 15–40-mm S1 probe or 20–50-mm S2 probe below the skin surface in pediatric patients; or a 10-mm-wide and 40-mm-long, 25–65-mm M probe or 35–75-mm XL probe below the skin surface in adult patients. TE gives a result of Young’s modulus (E = 3ρC_s², E: Young’s modulus; ρ: tissue density; C_s: velocity of shear wave) in units of kiloPascals (kPa). This technique was introduced in 2003 by the FibroScan system (Echosens, France).

SWE methods are incorporated into conventional ultrasound machines (sonoelastography) using an acoustic radiation force produced from a focused ultrasound beam. Point shear wave elastography (pSWE) is a one-dimensional (1D) method by creating a push point inside the tissue. Shear wave speed is then measured within a small region of interest in the shape of a rectangular box nearby (usually 5 × 10 mm² and 1.5 cm below the liver capsule). SWE has been extended to sequential multiple pushes and produces a two-dimensional (2D) map of stiffness measurement within a larger box with specialized hardware for ultrafast real-time imaging. Results are also expressed as Young’s modulus (E) in kPa or shear velocity (C_s) in m/s based on the measurement of the velocity of shear waves generated from ARFI.

Compared with MR elastography, ultrasound methods are fast to perform and relatively inexpensive. They can be readily used with an operator after a short training and a steep learning curve, and performed easily at each outpatient clinic visit in 5 min or less. However, ultrasound-based methods cannot evaluate the entire organ due to a limited accessible sampling size (TE, pSWE: 1 cm²; SWE: 20 × 20 cm²) considering their specific acoustic window/depth.

For TE, technical failure increases in patients with ascites, obesity, and narrow intercostal space. The reproducibility is reduced in patients with steatosis, obesity, and lower stages of hepatic fibrosis.

The assessment is performed in a tissue that lies about 6 cm below the skin surface and therefore is difficult to perform when the liver is further away from the skin surface as in obese individuals. Technical failures occur in patients with obesity, ascites, or narrow intercostal space (failure rate: 6–23%). The variability between machines and observers can vary on the order of 12% [10–12].

For SWE, technical failure increases in patients with poor sonic window, reverberation artifacts, pulsation artifacts, and poor breath-holding capacity (failure rate: 2–10%). BMI and the distance from the ultrasound probe to the liver surface are the factors associated with technical failures or unreliable measurements. Comparing SWE and TE for staging of hepatic fibrosis using ROC analysis, several studies demonstrated that real-time SWE was more accurate than TE for assessing substantial fibrosis (≥F2). The interobserver reproducibility is around 70–80% [13–16].

MR elastography (MRE) is a phase-contrast technique that estimates the shear stiffness of tissues by imaging propagating shear waves. These waves are generated from a standardized extrinsic vibrating source at 60 Hz for the commercially available method. The inversion algorithm calculates shear modulus or shear stiffness based on the wave speed of the propagating waves and generates a quantitative liver stiffness map (magnitude of the complex shear modulus in kPa |G*| = |G′ + iG″|, approximately 1/3 of E), which can cover an entire organ. This measurement is different from ultrasound elastography methods which assess Young’s modulus, but are related [Young’s modulus (E) = 3 × shear modulus (μ)]. MRE was introduced in 2007 for the clinical application of measuring LS. Many MRI vendors (GE, Siemens, and Philips) have commercially available MRE imaging sequences, standardized hardware (Resoundant, Inc.), and inversion software (MMDI). In these sequences, motion-sensitizing gradients are applied along one or several desired directions to record the shear displacement vectors in either 2D or 3D space. Complete displacement information of the generated wave field can be recorded for calculating multiple mechanical parameters at different frequencies for the clinically available options, which have the potential to distinguish other pathophysiologic statuses of the liver from hepatic fibrosis. Compared with ultrasound-based elastography, the advantages of MRE include its ability to analyze almost the entire liver, and its applicability to patients with obesity, ascites, or narrow intercostal spaces. It could be comparably inexpensive and fast if performed as a limited MR elastography-only exam. Technical failures occur in patients with iron-overloaded liver and claustrophobia (failure rate: 4.7–5.6%). However, modified MRE sequences are available for evaluation of LS in patients with mild to moderate iron overload [17, 18]. The overall difference between vendors, magnetic strength, and observers varies on the order of 10% [10–12].

Compared with ultrasound elastography techniques, MRE has the highest repeatability and reproducibility and thus provides the most reliable liver stiffness measurement [15, 19–24]. Ultrasound elastography systems use polychromatic (range of frequency) transient vibrations that generate complicated nonplanar shear wave fields. Wave diffraction and dispersion are inevitable. This leads to big variations between machines and observers [25–28]. Liver MRE uses monochromatic (single frequency) steady-state vibration for the shear wave generation, which gives a uniform wave field exactly at 60 Hz. Assuming that most people can hold their breath well for the duration of the MRE exam, and no biases from wave diffraction, dispersion, and sampling errors are observed, MRE can provide a more accurate shear stiffness measurement throughout the liver than that of ultrasound systems [29, 30]. Technical repeatability and reproducibility of elastography have also been rigorously evaluated in multiple studies for within-subject variability in a test–retest scenario and within-observer variability [31–33]. Results suggest that both ultrasound and MRE are reliable techniques with high repeatability and reproducibility, with some studies suggesting that MRE has superior reliability than ultrasound methods [15, 22–24]. To further improve interobserver reproducibility and remove the need for manual analysis in MRE, a fully automated segmentation algorithm has been developed for calculating liver stiffness. This automated method is highly consistent with the measurements manually performed by expert readers [34]. In addition, studies have shown that the liver MRE technique has the capability to serve as a “ground truth” to evaluate an ultrasound-based elastography technique for detecting fibrosis in a patient study [35].

Clinical comparison

A consensus conference held by the Society of Radiologists in Ultrasound [10, 36, 37] found that among all quantitative elastography methods, MRE has the closest performance to a properly performed biopsy in assessing hepatic fibrosis. MRE has significantly higher accuracy than TE [10]. These studies also suggest that patients can be grouped into three categories: those with normal elastography values who have a low likelihood of cirrhosis (stage F0 or F1) and may not require additional follow-up, those with high stiffness values who have a high likelihood of cirrhosis (F4), and those in between who have moderate to severe fibrosis (stages F2 and F3) and are at risk for progression of fibrosis, depending on the etiology of fibrosis [10–12]. As shown in the last two columns of Table 1, the areas under ROC curves (AUROC) were summarized from the selected worldwide published work. MRE has a higher technical success rate and a better diagnostic accuracy than TE in a prospective blind comparison study of 142 liver patients [38]. In meta-analysis studies, the diagnostic performance of TE is better than most serum markers [39, 40]. 2D SWE has at least equivalent or better diagnostic accuracy than that of TE [13, 41, 42]. MRE is more accurate than both ultrasound methods [43, 44]. In addition, MRE has a higher technical success rate as ultrasound-based methods have lower technical performance in patients with ascites, obesity, and narrow intercostal space.

In summary, the published literature generally indicates that MRE has higher diagnostic performance and fewer technical failures than ultrasound-based elastography in assessing hepatic fibrosis.

Results from elastography are almost always interpreted in conjunction with clinical profile, laboratory test results for liver function, and other imaging studies available for categorizing the patients into no fibrosis, some fibrosis, and cirrhosis (compensated that do not have symptoms or decompensated that have symptomatic complications such as jaundice, ascites, variceal hemorrhage, or hepatic encephalopathy). Liver stiffness may be useful as a prognostic tool for identifying patients with the highest risk for disease progression [45]. Table 2 summarizes the usefulness of elastography techniques for different clinical indications.

Table 2.

Usefulness of elastography techniques for different clinical indications

Indication	TE	PSWE	ARFI	MRE
Initial assessment for liver fibrosis and cirrhosis	X	X	X	X
Assessment for liver fat and fibrosis in NAFLD	Xa			Xb
Assessment during clinical follow-up, progression, or improvement	X	X	X	X
Assessment for cirrhosis and its complications (portal hypertension and HCC)			Xd	Xc

^aIn conjunction with controlled attenuation parameter (CAP) evaluation

^bIn conjunction with MRI of liver for assessment of fat signal fraction

^cIn conjunction with routine MRI of liver including dynamic contrast-enhanced studies

^dIn conjunction with routine US of liver including Doppler ultrasound

Future directions

Many investigations have demonstrated that LS has a static component that is mainly determined by extracellular matrix composites and liver structure (e.g., hepatic fibrosis, necrosis, loss of hepatocytes, regeneration, etc.), and a dynamic component that is affected by intrahepatic hemodynamic changes (e.g., perfusion, congestion, and inflammation). It has been well established that liver stiffness is associated with many pathologic processes: inflammation, fibrosis, postprandial state-induced hyperemia, passive hepatic congestion, and fibrosis-induced portal hypertension [46–49]. LS measured at any point of time may have contributions from these different processes, although hepatic fibrosis is the dominant factor. In early chronic liver disease, inflammation may dominate the effect on LS, and passive congestion may predominately determine the increased LS in early stages of congested livers due to chronic heart failure. There is a need to establish the relationships between mechanical properties (other than shear stiffness) in distinguishing different pathophysiologic processes of the liver. These quantities include the model-free properties and model-based viscoelastic parameters [50–59]. Among them, liver viscosity was found to be correlated with fibrosis, but not to steatosis or disease activity (inflammation) [16]. The dispersions of shear wave velocity and attenuation were found to be associated with the degree of steatosis [60]. The damping ratio and the loss modulus were found to increase significantly at the early onset of liver injury or necroinflammation [61]. This was apparent even with coexisting steatosis or before histologically detectable macrophage transformation or migration, but was not sensitive to the later progressive development of fibrosis. Being able to distinguish how these dynamic components contribute to tissue mechanical properties will provide valuable diagnostic information. Being able to describe how these contributions change with different pathologies and temporally over the course of disease development will have important prognostic implications and direct emerging therapeutic interventions designed to slow or halt the progression of liver diseases.

There are many other intriguing future directions for liver elastography. For example, both ultrasound and MRE can be extended from 2D to 3D. Shear wave propagation can be fully described in a displacement vector map along three directions by MRE. Currently, commercialized ultrasound methods track shear displacement along the axial direction only. 1D and 2D transducer arrays may provide the lateral and elevational displacement estimations. However, the trade-off between the lateral/elevational resolution and jitter/bias needs to be considered. Multiparametric liver elastography (e.g., MRE-assessed liver stiffness, damping ratio, and volumetric strain parameters) may be useful in extending the capability of this technique. Mechanical properties of spleen tissue could be helpful in assessing complications, such as portal hypertension, compensation, or decompensation. 3D multiparametric elastography can be used to evaluate focal lesions for distinguishing benign and malignant tumors and for noninvasive evaluation of grade of malignancy or invasiness to predict response/recurrence. Another potential application is slip interface imaging [62], which can be used to stage malignancy to predict how widespread the tumor is, and determine what the treatment or prognosis will be. To enhance the technical success of liver MRE, dedicated MRE sequences can be further developed for iron-overloaded livers. A free-breathing fast MRE imaging technique can also be developed to help patients with an inconsistent breath-hold.

Conclusion

Among dynamic elastography techniques, MRE has the strongest performance profile, superior to ultrasound-based techniques and with fewer technical failures. The higher diagnostic performance is most likely due to the larger volume of liver tissue that can be assessed with MRE, and the basic technical features such as the use of a narrow-band mechanical vibration spectrum, thereby avoiding the dispersion-mediated depth dependence that is seen with many ultrasound-based elastography techniques. Comprehensive MRI-based evaluation of liver disease allows for ready quantitative assessment of hepatic fat content, perfusion, and diffusion. For all elastography techniques, there is significant potential to implement multiparametric methods that show promise for distinguishing between processes such as inflammation, fibrosis, and venous congestion.