Magnetic resonance imaging of focal liver lesions: Approach to imaging diagnosis


  • Potential conflict of interest: Nothing to report.


This article is a review of magnetic resonance imaging (MRI) of incidental focal liver lesions. This review provides an overview of liver MRI protocol, diffusion-weighted imaging, and contrast agents. Additionally, the most commonly encountered benign and malignant lesions are discussed with emphasis on imaging appearance and the diagnostic performance of MRI based on a review of the literature. (HEPATOLOGY 2011)

The incidence of incidentally detected focal liver lesions (FLL) parallels growth in imaging utilization. The majority of FLL arising in noncirrhotic livers are benign. Hemangiomas, focal nodular hyperplasias (FNH), and adenomas (HCA) are the most commonly encountered solid benign lesions.1-3 The most commonly encountered malignant lesions in noncirrhotic livers are metastases. Hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICC) occur in the setting of chronic liver disease.

Maximizing specificity and accuracy of cross-sectional imaging in the context of these incidental liver lesions is paramount in avoiding unnecessary biopsies, which may portend a postprocedural morbidity of 2.0% to 4.8% and mortality of 0.05%.4-6 Ultrasound, computed tomography (CT), and magnetic resonance imaging (MRI) are the main liver imaging modalities. A meta-analysis comparing contrast-enhanced ultrasound, CT, and MRI in evaluating incidental FLLs demonstrated similar diagnostic performance with specificities ranging from 82%-89% and no significant difference in the summary receiver operating characteristic between modalities.7 Given the lack of ionizing radiation and relative nonavailability of ultrasound contrast in the U.S., MRI is the imaging test of choice for FLL characterization, demonstrating similar if not superior performance to CT. This review focuses on the diagnostic performance of MRI in evaluating the most common FLL in noncirrhotic livers with additional discussion of HCC and ICC, which, although highly associated with chronic liver disease, are important differential considerations.


AASLD, American Association for the Study of Liver Disease; ADC, apparent diffusion coefficient; DWI, diffusion-weighted imaging; ECA, extracellular contrast agents; FLL, focal liver lesion/s; FNH, focal nodular hyperplasia/s; HAS, hepatocyte specific contrast agents; HCA, hepatocellular adenoma/s; HCC, hepatocellular carcinoma/s; ICC, intrahepatic cholangiocarcinoma/s; MRI, magnetic resonance imaging; OPTN, Organ Procurement and Transplantation Network; UNOS, United Network for Organ Sharing.

Liver MRI

Basic Protocol.

A comprehensive liver protocol evaluates the parenchyma, vasculature, and biliary system. This is accomplished by way of a combination of single-shot T2-weighted fast spin-echo, gradient echo T1-weighted in- and opposed-phase, fat suppressed T2-weighted, dynamic pre- and postcontrast T1-weighted imaging and potentially subtraction of pre- from postcontrast image sets.8 High-quality images require compromise between achievable resolution and the need for breath-holding, which limits each sequence to 20 seconds or less. Breath-holding is not always possible in sick patients. As a result, modifications to the basic protocol may include the addition of free-breathing sequences, respiratory-gating, motion correction techniques (i.e., BLADE or PROPELLER or radial acquisition of k-space).

MRI quality can be variable due to differences in sequences, gradient, and magnetic field strength. In recognition of this variability, a recent publication on behalf of the American Association for the Study of Liver Disease (AASLD), under the auspices of the Organ Procurement and Transplantation Network (OPTN)/United Network for Organ Sharing (UNOS), describes minimum technical specifications for liver MRI.9 Although devised for HCC imaging in cirrhosis patients, the specifications provide useful guidance for liver MRI in general with suggestions on minimum sequences, injection rates, timing of dynamic imaging, slice thickness, and imaging matrix.

Diffusion-Weighted Imaging (DWI).

DWI is a measure of the ability of water molecule protons to diffuse freely within intra- and extracellular environments. DWI of FLL therefore reflects cellular density of the lesion. Apparent diffusion coefficient (ADC) values are calculated from tridirectional gradients (b-values), providing a quantifiable variable reflecting both diffusion and perfusion within imaged tissue.10 The b-values utilized in liver imaging range from 0-800, with b0 serving as a T2-weighted sequence used for lesion conspicuity and anatomic correlation. Higher b-values reflect true impedance. Lesions with the lowest ADC value, i.e., impeding diffusion to the greatest degree, are more likely to be malignant, although there is overlap with benign lesions.11-14 Several authors have suggested ADC thresholds for differentiating malignancy from benignity, with values ranging from ≤1.2 to 1.6 (1.2 × 10−3 mm2/s to 1.6 × 10−3 mm2/s), yielding specificities from 80 to >90%.13-19 In a study of 68 patients with 192 liver lesions, representing both metastases and benign lesions, DWI combined with dynamic contrast-enhanced MRI demonstrated a diagnostic accuracy of approximately 93%.17 Although initial studies show promise in differentiating benign from malignant lesions,18-22 these results often included cysts and hemangiomas, known to demonstrate high ADC values. Taouli et al.23 in a study of 68 patients found benign lesions, excluding cysts and hemangiomas, demonstrated intermediate ADC values, making the use of a set threshold of questionable utility. Hence, further investigation is needed to better delineate the role of DWI in characterizing FLL.

Contrast Agents.

Intravenous MR contrast agents can be divided into extracellular (ECA) and hepatocyte-specific agents (HSA). ECA equilibrate with the extracellular fluid space after intravenous injection and are excreted by glomerular filtration, similar to CT agents. This permits multiphase dynamic postcontrast imaging during late arterial, portal venous, and equilibrium phases, allowing assessment of enhancement kinetics, a reflection of both vascularity and permeability.

HSA have dual elimination, with a portion of the dose distributed extracellularly and eliminated by the kidneys; the remainder is taken up by hepatocytes and excreted into the bile. The two HSAs available in the U.S. are Eovist (gadoxetate disodium, Bayer HealthCare Pharmaceuticals, marketed as Primovist outside the U.S.) and Multihance (gadobenate dimeglumine, Bracco). To date, no large studies compare diagnostic accuracy of the two HSAs for FLL characterization. With Eovist, 50% of the dose is taken up by hepatocytes and eliminated by way of biliary excretion, compared to 3%-5% with MultiHance. This results in greater hepatobiliary phase parenchymal enhancement with Eovist. Hepatobiliary phase images are acquired 20-40 minutes after Eovist injection, compared to 1-2 hours after Multihance injection.

HSAs possess some properties of ECA, yielding dynamic postcontrast imaging with the added benefit of hepatobiliary phase imaging. The Food and Drug Administration (FDA)-approved dose of Eovist is 0.025 mmol/kg, which is one-fourth that of other approved agents. Although Eovist has greater T1 relaxivity, this reduced dose may lead to less robust arterial enhancement, prompting some radiologists to double the dose or acquire multiple arterial phases.24, 25 Additionally, hepatocyte uptake may begin as early as the portal venous phase, potentially confounding evaluation of enhancement kinetics. Given these issues, in our practice Eovist is the contrast of choice when evaluating suspected FNH and staging metastatic disease. However, for routine problem-solving MRI and in patients with suspected HCC, Eovist is reserved for special cases.

FLL Characterization


Hepatic hemangiomas are the most common benign FLL, with an incidence of 2%-20%.1-5 Classic MRI features include round or lobular margins, marked T2 hyperintensity (referred to as light-bulb bright), and characteristic enhancement pattern.26-30 Three distinct patterns of enhancement have been described on ECA-enhanced MRI, with reported specificities of 100% and diagnostic accuracies of 95%.31 Smaller lesions (<1.5 cm) may demonstrate uniform arterial enhancement, referred to as flash-filling. Larger, cavernous hemangiomas demonstrate either nodular peripheral interrupted enhancement coalescing centripetally to uniform enhancement (Fig. 1) or nodular peripheral interrupted enhancement that coalesces centripetally but with persistent central hypointensity.26 Giant hemangiomas may have regions of fibrosis and/or thrombosis, resulting in a central scar with strands of T2 hypointensity.26

Figure 1.

Hemangiomas. Image set 1. (A-C) Classic hemangioma T2 “light bulb” bright hepatic lesion (A), which displays classic interrupted peripheral enhancement that fills in from arterial to equilibrium phases (B,C) on this ECA-enhanced MRI. Image set 2. (D-F) Giant hemangioma T2 “light bulb” bright hepatic lesions (D), which display classic interrupted peripheral enhancement that incompletely coalesces with persistent areas of hypointensity on this ECA-enhanced MRI (E,F).

Caution should be exercised in differentiating hemangiomas from hypervascular metastases, such as neuroendocrine tumors, which can be markedly T2-hyperintense and arterially enhanced.32-35 Small flash-filling hemangiomas may require MR follow-up, as differentiation from metastases can be difficult. Metastases may demonstrate a continuous targetoid rim of enhancement compared to the discontinuous rim displayed by hemangiomas. With metastases, the arterial enhancing rim may washout, or become hypointense relative to the liver during the portal venous phase.

With HSA, hemangiomas demonstrate expected enhancement during the dynamic phase images and are hypointense during the hepatocyte phase, mirroring the signal intensity of the portal veins. This imaging appearance has been referred to as “pseudo-washout.”30, 36, 37 This hypointensity during the hepatobiliary phase is expected given the lack of hepatocytes within the lesion. Although the imaging appearance on T2-weighted and dynamic postcontrast sequences should allow for accurate diagnosis, HSA may not be the best option for suspected hemangiomas.

Focal Nodular Hyperplasia

FNH, common in asymptomatic patients, pathologically consists of nonneoplastic hepatocytes in a disorganized array surrounding a central scar with anomalous vessels. As FNH are composed of hepatocytes, they are relatively stealthy (barely discernable from normal parenchyma) on noncontrast images and show a characteristic enhancement pattern.38-43 A typical enhancement pattern with ECA is early nodular arterial enhancement, which equilibrates, or becomes isointense, with the background liver on portal venous phase images (Fig. 2). Some lesions contain a T2 hyperintense central scar. The scar may be hypointense during the arterial phase and show delayed enhancement with ECA.

Figure 2.

Focal nodular hyperplasia. Image set 1. (A-C) Typical FNH. Large arterially enhancing (A), lobulated mass (arrows) with central scar becomes isointense to the liver during portal venous phase (B) and retains contrast on the hepatobiliary phase sequence (C). Image set 2. (D-F) Typical FNH in two different patients. Delayed enhancement of the central scar on ECA-enhanced MRI (arrowheads) is shown in (D,E) (arterial and equilibrium phases). In a second patient typical peripheral and confluent retention of contrast within two hepatic FNH (arrows) is shown on the hepatobiliary phase image on Multihance MRI (F).

HSA-enhanced MRI is the study of choice for FNH. On hepatobiliary phase images, FNH are iso- or hyperintense to the background liver, reflecting uptake of contrast by lesional hepatocytes. A multicenter study of 550 consecutive patients with FLL characterized on Multihance MRI demonstrated that 95% (289/302) of FNHs were iso- or hyperintense on hepatobiliary phase images.43 In the same study, the overall diagnostic performance of hepatobiliary MRI in differentiating benign from malignant lesions demonstrated sensitivity of 96.6%, specificity 87.6%, and positive predictive value of 85%.43 Zech et al.39 demonstrated hepatobiliary MRI with Eovist yielded confident diagnosis of FNH in 88% of patients. Graziolo et al.44 in a study of Multihance MRI in differentiating HCA from FNH found 97% sensitivity and 100% specificity in diagnosing FNH.

Although HSA yields reliable results in diagnosing FNH, some caution may be warranted. As experience with Eovist accumulates, there are several reports of paradoxical hepatobiliary phase enhancement in a minority of HCC and HCA. Although HCC typically arises in underlying cirrhosis, this is a potential pitfall of hepatobiliary imaging. It is hypothesized that some HCCs overexpress the cellular receptor, OATP1B3, which facilitates uptake of Eovist.45-51 Additionally, extracellular pooling of contrast within tumors may explain hyperintensity on hepatobiliary phase images, potentially seen with Eovist or Multihance in the setting of fibrotic or necrotic metastases.

Hepatocellular Adenoma

HCA is an uncommon benign lesion linked to exogenous hormone exposure. Recent advances have facilitated identification of three distinct subtypes of HCA with different biological behavior and potentially different imaging features (Figs. 3, 4).52-56 These subtypes include: hepatocyte nuclear factor-1 alpha (HNF-1 alpha)-mutated, beta-catenin, and inflammatory HCA.

Figure 3.

Hepatic adenomas. Image set 1. (A-C) Hemorrhagic adenoma. T2 inversion recovery, in- and opposed-phase images show a heterogeneous hemorrhagic mass containing fat. The presence of hemorrhage is indicated by the region of T2 hypointensity (A) and T1 hyperintensity (B). The loss of signal on the opposed-phase image (C) indicates presence of fat within the lesion. This lesion was pathologically proven to be an adenoma. Image set 2. (D,E) Steatotic adenoma (HNF-1 alpha). In- and opposed-phase images nicely show drop in signal of this fat containing adenoma.

Figure 4.

Inflammatory adenoma. Image set 1. (H-L) Inflammatory adenoma. Coronal T2 HASTE, transaxial precontrast, arterial, portal venous, and 1-hour delayed Multihance hepatocyte phase images show a large T2 hyperintense/T1 mildly hypointense, arterial enhancing lesion which remains mildly hyperintense to the remainder of the liver during portal venous phase imaging. This lesion did not retain contrast on the Multihance delayed image (L). Biopsy of the mass was consistent with an inflammatory adenoma, previously called telangiectatic variant of FNH.

The majority of HCA are HNF-1 alpha and inflammatory subtypes. Inflammatory HCA has been described as T2 hyperintense with arterial hyperenhancement that persists on portal venous phase (Fig. 4). HNF-1 alpha HCA may show nonpersistent arterial enhancement and intralesional lipid, the presence of which can be diagnosed on in- and opposed-phase imaging. HNF-1 alpha are also called steatotic HCA due to diffuse lipid content. Inflammatory HCA may contain lipid; however, they potentially have only a small component or demonstrate heterogeneous signal loss on opposed-phase images compared to diffuse signal loss described with steatotic HCA (Fig. 3).56 Studies evaluating the diagnostic performance of MRI in subtyping HCA show high specificity in identifying steatotic and inflammatory subtypes, with specificities ranging from 88%-100%.56, 57 Further investigation is needed to validate the MR criteria for HCA subtyping, especially in defining the imaging features of beta-catenin lesions, which portend the highest potential risk of malignant transformation.54, 55

Hepatobiliary phase imaging may distinguish HCA from FNH, a common differential in young asymptomatic women. Unlike FNH, HCA with rare exceptions do not retain contrast on hepatobiliary phase images.46

Hepatocellular Carcinoma

HCC occurs almost exclusively in the setting of chronic liver disease.58 The classic HCC appearance on ECA-enhanced MRI, described in the setting of cirrhosis, is tumoral arterial enhancement, subsequent “washout” during portal venous or equilibrium phases, and delayed enhancing pseudocapsule (Fig. 5).59, 60 It is believed that loss of portal venous blood supply, occurring by way of a multistep carcinogenesis pathway, may explain the imaging appearance of “washout.”61 The presence of “washout” and delayed enhancing pseudocapsule are highly specific features of HCC when using ECA.60, 62, 63 Caution should be exercised in describing a pseudocapsule in the setting of prior hepatic-directed treatment, as granulation tissue can form a ring of peripheral enhancement in this scenario. In the setting of cirrhosis/fibrosis, lacy parenchymal enhancement surrounding a regenerative nodule may mimic the appearance of a pseudocapsule and/or potentially confound assessment of “washout.” Other supportive findings of HCC include vascular invasion, restricted diffusion, and T2 hyperintensity.

Figure 5.

Hepatocellular carcinoma. Image set 1. (A-C) Typical features of hepatocellular carcinoma in two patients. Images (A,B) (arterial and equllibrium phase ECA-enhanced MRI) show nicely arterial enhancement and washout with pseudocapsule. Image (C) (portal venous phase) from a second patient shows a large lesion that demonstrates washout and pseudocapsule on ECA-enhanced MRI. Image set 2. (D-F) Typical features of hepatocellular carcinoma in a single patient. Coronal postcontrast T1-weighted image shows a large mass in the inferior right lobe with scalloping of the portal vein (D), neovascularity on arteriogram DSA coronal image (E), and restricted diffusion on the ADC map MR image (F) demonstrated by the dark central region of the tumor.

Initial Eovist studies demonstrate a possible role in differentiating arterial pseudolesions from small HCC.64-71 However, Eovist remains controversial, with reports of paradoxical enhancement of HCC, nonretention by dysplastic nodules or fibrosis, and the potential diagnostic dilemma of small lesions (<1-2 cm) only seen on hepatobiliary phase images.45-51, 72

Arterial enhancement, although nonspecific, is an essential diagnostic feature of HCC and currently the only criterion required by UNOS in cirrhosis patients.73 With rising incidence and growing demand for liver transplantation, the AASLD/UNOS/OPTN and, separately, the American College of Radiology have proposed revised guidelines to improve the specificity of HCC diagnosis to best allocate the limited supply of organs.9, 73, 74 The revised guidelines rely on multiple features (i.e., arterial enhancement and washout or growth) with more stringent requirements for smaller 1-2 cm lesions. Neither system recognizes <1 cm nodules as HCC or describes a role for HSA. In an effort to validate the OPTN criteria, ACRIN 6690, a multicenter center study of MRI versus CT is currently enrolling subjects in the U.S.

The tradeoff of higher specificity at the expense of sensitivity is unavoidable, especially when dealing with HCC <2 cm and hypovascular HCC, the latter accounting for up to 5%-10% of cases.75-77 Consequently, if the new guidelines are adopted there is risk of increased biopsy-related morbidity and the potential for more advanced stage HCC prior to initiation of treatment. This potential downside may be balanced in effect by more appropriate organ allocation. However, additional large-scale investigation is needed to validate these new guidelines and determine potential impact.


ICC represents 10% of primary hepatic malignant tumors and tends to arise in the background of chronic liver disease such as cholangitis, hepatitis, nonalcoholic chronic liver disease, and obesity.3, 78 The MR appearance of ICC consists of irregular T1 hypointense, T2 hyperintense heterogeneous mass with early rim enhancement followed by progressive centripetal heterogeneous enhancement of the remainder of the lesion with ECA.79, 80 The initial peripheral rim enhancement of ICC is usually continuous and should not be mistaken for interrupted peripheral enhancement of hemangiomas. The rim of arterial enhancement in ICC may show peripheral washout, a feature that is never seen with hemangiomas. The more specific features of cholangiocarcinoma, although not frequently present, include T2 hypointense scar (potentially reflecting central fibrosis), capsular retraction, and peripheral biliary dilation (Fig. 6).79-84 Although delayed enhancement is the most commonly described pattern for ICC, this overlaps in appearance with metastatic disease and hence biopsy is often warranted.

Figure 6.

Intrahepatic cholangiocarcinoma. Image set 1 (A-C) Classic enhancement pattern for intrahepatic cholangiocarcinoma on ECA-enhanced MRI. Precontrast (A), arterial (B), and 5-minute delayed phase (C) images show the typical arterial enhancing rim and delayed central enhancement with washout of the rim in a lesion showing the typical enhancement pattern of cholangiocarcinoma. Image set 2. (D-F) Intrahepatic cholangiocarcinoma in two different patients. Precontrast (D) and portal venous phase (E) on ECA-enhanced MRI in a patient show typical enhancement and peripheral biliary dilation, which is classic for intrahepatic cholangiocarcinoma. Image (F) (equilibrium phase on ECA-enhanced MRI) in a different patient shows nicely capsular retraction, which is another typical finding in intrahepatic cholangiocarcinoma.


Hepatic metastases have variable appearances depending on the primary tumor and are characterized as hypervascular or hypovascular, enhancing more or less than surrounding parenchyma (Fig. 7). Hypervascular metastases are seen with neuroendocrine tumors, renal cell carcinoma, thyroid carcinoma, melanoma, and sarcoma. Metastases from other primaries tend to be hypovascular. Internal hemorrhage may occur with metastases from renal cell carcinoma, melanoma, and lung cancer, often demonstrating T1 hyperintensity (Fig. 8). Hepatobiliary imaging with Eovist and DWI can be useful for detection of small hepatic metastases, demonstrating improved sensitivity over traditional MRI and CT.17-22, 85, 86

Figure 7.

Metastases. Image set 1. (A-E) Hepatic metastasis (breast cancer). T1-weighted precontrast, arterial and equilibrium phase postcontrast, and T2-weighted images (A-D) demonstrate continuous rim of enhancement with progressive filling in a T2-hyperintense lesion. The continuous rim of enhancement is most consistent with metastases, as was the case in this woman with breast cancer. This lesion demonstrated atypical features on the 1-hour delayed image with Multihance (E). The central retention of contrast was felt to be consistent with a fibrous component to the metastasis, not to be confused with a hepatocyte containing lesion such as a focal nodular hyperplasia.

Figure 8.

Metastases, continued. Image set 1. (F-I) Hepatic metastasis (carcinoid). Diffusion-weighted image (F), T1 precontrast (G), arterial (H), and equilibrium phase (I) images from an ECA-enhanced MRI show typical features of a hypervascular carcinoid tumor metastasis. Note the washout on equilibrium phase image due to the vascular nature of the lesion. Image set 2. (J-K) Hepatic metastasis (ocular melanoma). T2 inversion recovery and T1 precontrast images show a T2 hypointense, T1 hyperintense metastatic lesion. The atypical signal characteristics in this case are due to melanin content of the lesion.


MRI is a highly specific and accurate modality for FLL characterization. An experienced MR radiologist is essential to maintain high-quality liver MR protocols, determine appropriate indications for hepatocyte versus extracellular contrast agents, and guide management. Although many hepatic lesions have characteristic imaging features, consideration of the clinical context, in particular the presence or absence of underlying liver disease when considering HCC or ICC, is essential to confidently diagnose and direct management in these patients.