M R C h a r a c t e r i z a t i o n of Focal Liver Lesions Pearls and Pitfalls Evan S. Siegelman, MD*, Anil Chauhan, MD KEYWORDS Focal nodular hyperplasia Hepatic adenoma Hepatic steatosis Hepatocellular carcinoma Magnetic resonance imaging Siderotic regenerative nodules
KEY POINTS Focal liver lesions that are isointense to hyperintense to liver on in-phase T1-weighted images are usually hepatocellular in origin. Focal liver lesions that lose signal intensity on an opposed-phase image compared with the matched in-phase image contain lipid and are usually hepatocellular in origin. Focal liver lesions that lose signal intensity on an in-phase image compared with an opposed-phase image are most often iron-containing siderotic nodules. Focal liver lesions that are isointense to spleen on T2-weighted images are solid and often malignant, whereas focal liver lesions that are hyperintense to spleen on heavily T2-weighted images are usually nonsolid benign cysts or hemangiomas.
The first 3 imaging pearls discussed in this review are the 3 instances when focal liver lesion characterization is possible with T1-weighted gradient echo images. The 3 most common focal liver lesions encountered in clinical practice are cysts, hemangiomas, and metastatic disease. Nonsolid benign hepatic lesions (cysts and hemangiomas) and almost all metastatic lesions are hypointense relative to liver on T1-weighted images (Fig. 1). Normal liver has relative high signal intensity on T1-weighted images that has been attributed to high concentrations of protein, rough endoplasmic reticulum, and paramagnetic substances, such as manganese and copper.1–3 One study calculated the
T1 relaxation times of liver at 1.5 T as 547 ms and that of solid lesions, hemangiomas, and cysts to be 1004, 1337, and 3143 ms, respectively. Thus, most liver lesions are initially detected on T1-weighted images as being hypointense to liver.4 The authors use other pulse sequences besides T1-weighted images in order to differentiate among cysts, hemangiomas, and solid liver lesions. If a focal liver lesion is isointense to hyperintense to liver on a T1-weighted image, then it is most commonly hepatocellular in origin. The 5 most common focal hepatocellular lesions encountered in clinical practice are regenerative nodules (RN), hepatocellular carcinoma (HCC), focal nodular hyperplasia (FNH), hepatocellular adenoma (HCA), and focal steatosis. In this section, the authors discuss FNH and the inflammatory subtype of HCA (IHCA). The reader is referred to the articles by Barr and Hussain and Sirlin in this issue of the
Department of Radiology, Perelman School of Medicine, University of Pennsylvania, 34th and Spruce Streets, 1st Floor Silverstein, Philadelphia, PA 19104-4283, USA * Corresponding author. E-mail address: [email protected] Magn Reson Imaging Clin N Am 22 (2014) 295–313 http://dx.doi.org/10.1016/j.mric.2014.04.005 1064-9689/14/$ – see front matter Ó 2014 Elsevier Inc. All rights reserved.
mri.theclinics.com
PEARL 1: THE T1 PEARL: A FOCAL LESION THAT IS ISOINTENSE TO HYPERINTENSE TO LIVER ON T1-WEIGHTED IMAGES IS HEPATOCELLULAR IN ORIGIN
Siegelman & Chauhan
296
Fig. 1. Magnetic resonance demonstration of hepatic hemangioma and metastatic breast cancer in a 51-year-old woman. Metastases, cysts, and hemangiomas are almost all hypointense relative to liver on T1-weighted images. (A) Axial in-phase T1-weighted image shows low signal intensity hemangioma (white arrow) and infiltrative multifocal metastases (smaller black arrows). The metastases are isointense and the hemangioma is hypointense to spleen (S). However, the authors prefer to use T2-weighted images, diffusion-weighted images, and/or enhanced imaging to differentiate solid masses, such as metastatic disease, from nonsolid hepatic cysts and hemangiomas. (B) Corresponding opposed-phase image shows similar relative signal intensities of the hemangioma and metastatic disease. Geographic regions of lower signal intensity within the left lobe of liver (black arrows) represent steatosis. (C, D) T2-weighted fast-spin-echo images obtained with effective echo times of 90 (C) and 180 ms (D). The hemangioma is hyperintense to spleen, whereas the metastases are isointense. As the echo time increases, the contrast between the hemangioma and the adjacent metastases improve. This improvement is not because the hemangioma lights up or enhances; rather, the improved image contrast is because the hemangioma loses less signal as the echo time increases compared with liver, spleen, and metastases. (E) Opposedphase T1-weighted imaging performed 2 years prior shows that the hemangioma (arrow) is hyperintense to the surrounding liver (L). When trying to establish the hepatocellular origin of a focal liver lesion by showing isointensity or hyperintensity to liver, one should use an in-phase image as steatotic liver can be of low signal intensity on opposed-phase imaging and can confound relative signal assessment. (F) Corresponding in-phase T1-weighted image shows that the hemangioma (black arrow) is hypointense to the steatotic liver.
Magnetic Resonance Imaging Clinics of North America concerning the evaluation of the cirrhotic liver and how to differentiate RN from HCC. Lipid and fat-containing liver lesions are discussed in “Pearl 2.”
FNH FNH is the second most common benign hepatic tumor in adults after hemangioma. FNH composes approximately 8% of all primary liver tumors and has an estimated prevalence between 0.3% and 3.0%.5,6 FNH is not considered a neoplasm but instead is hypothesized to develop as a hyperplastic response of hepatic parenchyma around a central developmental vascular malformation.7 Individuals with FNH are more likely to have coexistent hepatic hemangiomas (20%) than would be expected by chance8; both hemangioma and FNH involve focal abnormalities in the hepatic blood supply.
FNH is typically detected in women aged 20 to 50 years and is uncommon in men (female-tomale ratio 5 10:1).5 Unlike hepatic adenomas, there is no proven association of oral contraceptive use or pregnancy with the development or growth of FNH.9,10 Although up to 15% of FNH lesions can grow when followed longitudinally,11 this should not cause clinical concern. The authors are skeptical of reports of malignant transformation to fibrolamellar hepatoma12 or HCC,13 as most investigators think that malignant transformation of FNH does not occur.14 The magnetic resonance (MR) features of FNH can be considered in 2 parts (Fig. 2). The first component is the vascular nidus that forms the central scar of FNH, and the second is the surrounding hyperplastic response of adjacent liver. The vascular scar is hypointense to liver on both in-phase and opposed-phase imaging and is hyperintense to liver on T2-weighted imaging. The higher T2-weighted signal intensity is
MR Characterization of Focal Liver Lesions
Fig. 2. MR findings of FNH and lipid-containing hepatocyte nuclear factor-1a inactivated hepatic adenoma in a 25-year-old woman. (A, B) Axial in-phase (A) and opposed-phase (B) T1-weighted images show 2 lesions: a larger FNH in segment 8 and a smaller adenoma in segment 2. The FNH has outer components (thick arrow in A) that are isointense to surrounding liver suggesting that it is hepatocellular in origin. There is a lower signal intensity central scar (thin arrow in A). The adenoma is isointense to liver and not perceptible on in-phase imaging; however, it loses signal and becomes recognizable on opposed-phase imaging (arrow in B). In-phase isointensity and loss of signal on opposed-phase image indicate that the lesion is lipid containing and hepatocellular in origin. (C, D) T2-weighted fast-spin-echo images (effective echo time 5 95 ms) shows that both the FNH and adenoma (thick arrows) are hyperintense to liver and relatively isointense to spleen. The central scar of the FNH (thin arrow) is hyperintense to spleen. On the fat-suppressed image (D), there is similar contrast with the exception of lower signal intensity of the adenoma (thick arrow in D) because of the suppressed lipid. (E) Hepatobiliary-phase enhanced fat-suppressed T1-weighted image obtained after the intravenous administration of gadoxetate disodium shows hyperenhancement of the nonscar components of the FNH (thick arrow) and hypoenhancement of the adenoma (thin arrow). The adenoma showed hyperenhancement compared with adjacent liver on dynamic enhanced imaging (not shown).
hypothesized to be secondary to slow flow within vessels. The scar does not enhance during the dynamic administration of gadolinium but does show delayed enhancement during the interstitial phase of enhancement when a conventional extracellular gadolinium contrast agent is used.15,16 The surrounding lesion of FNH itself is usually isointense to liver on both in-phase and opposedphase imaging. Cases of lipid-containing FNH are uncommon and reportable17; many are associated with diffuse hepatic steatosis.18 On T2-weighted imaging, FNH is isointense to minimally hyperintense relative to liver. On arterial-phase dynamic gadolinium-enhanced imaging, FNH shows marked homogeneous enhancement, often with rapid washout during the portal and interstitial phases of enhancement.16,19 If a gadolinium contrast agent with hepatobiliary excretion is used (eg, gadoxetate disodium), FNH will typically have components that are isointense to hyperintense relative to the surrounding liver on delayed hepatobiliary-phase imaging, which can be useful for distinguishing FNH from HCA for some lesions (see Fig. 2E).19–24 In one study of 30
FNH lesions, only 2 showed homogenous low signal intensity on delayed hepatobiliary-phase images.25 If one uses gadoxetate disodium–enhanced MR to characterize FNH, the central scar may not show enhancement during the interstitial phase of contrast enhancement.26 The absent enhancement of the central scar is hypothesized to be from the more rapid removal of gadoxetate disodium from the circulation compared with other gadolinium contrast agents. The reader is referred to the article by Bashir for a detailed discussion concerning the use of the various MR contrast agents used for liver imaging. Diffusion-weighted imaging and apparent diffusion coefficient (ADC) values alone should not be used to differentiate benign from malignant focal liver lesions as there can be substantial overlap. For example, many benign FNH and hepatic adenomas can show restricted diffusion similar to metastatic disease.27 For a review concerning the performance and interpretation of diffusionweighted imaging of the liver, the reader is referred to the article by Taouli in this issue of the Magnetic Resonance Clinics of North America.
297
298
Siegelman & Chauhan IHCA Although pathologists had previously classified HCA into a single pathologic entity, recent molecular and immunohistochemical markers have determined that there are 4 distinct subtypes of HCAs.28–31 These subtypes include IHCAs and hepatocyte nuclear factor 1a inactivated (HNF-1a), b-catenin–activated, and unclassified HCAs. MR can often detect and characterize the subtype of HCA based on specific imaging features.32 Independent of the subtype, larger (>4 cm) adenomas are at risk of developing intralesional hemorrhage33 and are usually treated with surgical resection or with nonsurgical radiofrequency ablation or bland embolization.5,29,34,35 Smaller adenomas can be managed conservatively and followed by imaging.35 Women who take oral contraceptives should discontinue their use as some HCAs may subsequently decrease in size.5,31 Obesity and metabolic syndrome are associated with HCA36,37; patients are encouraged to modify their diet, develop an exercise program, and lose weight as part of a treatment plan.
Hepatic adenomatosis is a distinct entity that was originally described and defined in 1985 as the presence of greater than 10 HCAs (Fig. 3).38 The HCA associated with adenomatosis is not limited to any one particular subtype,39,40 although patients with multiple adenomas are more likely to have steatosis.41 As with single lesions, management is based on lesion size and patients’ symptoms.42,43 The IHCA is the most common subtype of HCA and accounts for 40% to 60% of lesions in reported series. IHCA is associated with obesity34 and metabolic syndrome.36 In one study of 32 women with IHCA, the median body mass index was 32.5.44 In this same group of 32 women, steatosis was present in 59% on liver biopsies obtained distant from the tumor. In another series comparing 63 IHCA versus 46 HNF-1a adenomas, there was significantly less intralesional steatosis in the IHCA subgroup (43% vs 82%).34 Ten of the 63 patients with IHCA had findings of either intralesional or peritumoral hemorrhage, which was not significantly different than those found in the HNF-1a subtype.
Fig. 3. MR imaging findings of pathology-proven hepatic steatosis and hepatic adenomatosis secondary to IHCA in a 43-year-old woman. The patient stopped taking oral contraceptives at the time of diagnosis and has been followed conservatively for the next 7 years. (A) Axial in-phase T1-weighted gradient echo image shows subtle foci of increased signal intensity relative to liver (black arrows). (B) Corresponding opposed-phase image shows loss of signal intensity within the liver confirming the presence of hepatic steatosis. There are multiple lesions that are hyperintense to the steatotic liver (arrows). Given that these lesions were isointense to liver on inphase imaging, this pattern suggests they are hepatocellular in origin. (C) Subtraction image created by subtracting the opposed-phase image from the in-phase image (in-phase minus opposed-phase) depicts those voxels that contain both lipid and water protons. The highest signal intensity on the subtraction images occurs at fat-water interfaces (eg, junction of left kidney and perirenal fat [thin arrows]). This image confirms the presence of hepatic steatosis and the absence of lipid within the focal liver lesions (thick arrows). (D, E) Precontrast and arterial phase–enhanced fat-suppressed T1-weighted images show hyperenhancement of the adenomas (arrows) relative to the surrounding liver. The presence of arterial phase enhancement assists in differentiating IHCAs from masslike regions of focal sparing of steatosis.
MR Characterization of Focal Liver Lesions On T1-weighted images, IHCAs are usually isointense to slightly hyperintense to surrounding liver.28,29 On opposed-phase imaging, the surrounding liver is more likely to lose signal intensity because of steatosis than the inflammatory adenoma itself (see Fig. 3; Fig. 4).34,44 On T2weighted images, most IHCAs are minimally hyperintense relative to liver.28,29 In one series, 13 or 30 IHCAs revealed a specific atoll sign, consisting of a hyperintense rim that enhances on delayed gadolinium-enhanced imaging (see Fig. 4C).29 It is hypothesized that the atoll sign is secondary to dilated sinusoids within the periphery of the adenoma. Like FNH, HCAs will enhance after gadolinium contrast. However, unlike FNH, IHCAs do not have a central scar and almost all IHCAs do not have components that are isointense to hyperintense to liver on delayed hepatobiliaryphase gadolinium-enhanced imaging.23 Thus, in a noncirrhotic woman with hepatic steatosis who has a solid enhancing mass that is isointense to liver on precontrast T1-weighted images and hypointense on hepatobiliary-phase enhanced images, one should consider an IHCA as the most likely cause.
THE EXCEPTIONS Nonhepatocellular Focal Liver Lesions in a Liver Containing Background Moderate or Marked Steatosis On opposed-phase T1-weighted images, liver that is involved with moderate or severe steatosis will have very low signal intensity. Thus, if one were to evaluate the relative signal intensity of a focal liver lesion with surrounding steatotic liver on an opposed-phase T1-weighted image alone, then one may come to a false conclusion that the lesion is hepatocellular. For example, a benign hemangioma (see Fig. 1E) or even a metastasis can appear hyperintense to steatotic liver on opposed-phase imaging. Therefore, one should use the in-phase image as the T1-weighted reference when using pearl 1.
Hemorrhagic Metastases to the Liver Rare hepatic metastases can show T1 hyperintensity secondary to the T1 shortening properties of methemoglobin within subacute intralesional hemorrhage.45 This hyperintensity has been described in cases of metastatic renal, neuroendocrine, and
Fig. 4. MR illustration of the atoll sign in a 26-year-old woman with a body mass index of 39 with a surgically proven IHCA. Obesity and metabolic syndrome are associated with hepatic adenomas. (A, B) Axial in (A) and opposed-phase (B) T1-weighted images show a subcapsular hyperintense hepatic mass (arrow) indicating it is hepatocellular in origin. Mild steatosis, which was confirmed at surgery, is most pronounced and best revealed as lower signal intensity within segment 4 on the opposed-phase image (small arrows). (C) Axial fast-spin-echo T2-weighted image (effective echo time 5 93 ms) shows that the adenoma is hyperintense to liver and isointense to spleen. Not every focal lesion that is isointense to spleen is malignant, especially in individuals who do not have cirrhosis or history of a primary malignancy. A high-signal-intensity rim around the adenoma (small arrows) has been termed the atoll sign and is hypothesized to be secondary to dilated sinusoids. (D) Arterial-phase enhanced fat-suppressed T1-weighted image shows hyperenhancement of the adenoma and hypoenhancement of the peritumoral sinusoids.
299
300
Siegelman & Chauhan lung cancers as well as choriocarcinoma.2,46 Focal liver lesion hyperintensity can also be secondary to the paramagnetic metals attached to melanin contained within metastatic melanoma (Fig. 5).2,47,48 In these instances, patients’ primary tumor and the presence of metastatic disease are often known or apparent at the time of imaging. After radiofrequency ablation or chemoembolization of liver malignancies, intratumoral T1 hyperintensity can result from hemorrhagic and coagulative necrosis.49 Once again, the cancer history and prior procedures are usually known at the time of image interpretation. The reader is referred to the article by Kamel concerning the use of MR to evaluate treated liver metastases.
Hemorrhagic Cysts of the Liver Most liver cysts have simple fluid content and very low signal intensity on T1-weighted images. Occasionally, intralesional hemorrhage can occur in simple hepatic cysts that appear hyperintense on T1-weighted images secondary to methemoglobin and or intracystic proteins.50 Hemorrhage into simple hepatic cysts has been described in autosomal dominant polycystic liver disease51 and large bile duct hamartomas.52 The absence of a
thick fibrous capsule, septa, nodules, and solid enhancing components differentiates the benign hemorrhagic hepatic cyst from biliary cystadenoma-biliary cystadenocarcinoma.51
PEARL 2: THE CHEMICAL SHIFT PEARL: FOCAL LIVER LESIONS THAT LOSE SIGNAL INTENSITY ON OPPOSED-PHASE IMAGING CONTAIN LIPID AND ARE MOST OFTEN HEPATOCELLULAR IN ORIGIN When performing dual-phase in and opposedphase T1-weighted gradient echo imaging of the liver, the echo time of the opposed-phase image should be shorter than the in-phase echo time. This timing will ensure that any loss of signal intensity on the opposed-phase image compared with the in-phase image will be caused by the presence of lipid and water protons within the same voxel.53,54 If the opposed-phase echo time is longer than the in-phase echo time, then signal loss could also be secondary to T2* susceptibility effects.55 Some manufacturers’ 3-T MR systems come with configured dual-phase gradient echo sequences (eg, echo times of 2.2 and 3.3 ms) that could result in potential diagnostic confusion.
Fig. 5. MR depiction of T1 hyperintense melanoma metastases to the liver in a 64-year-old woman. Some hemorrhagic metastases can be hyperintense to liver and should not necessarily be assumed to be hepatocellular in origin. (A) Axial in-phase T1-weighted gradient echo image shows multiple hypointense liver metastases. Two of the lesions have central high signal intensity (arrows). (B) Opposed-phase T1-weighted image shows persistent high internal signal intensity within 2 of the metastases (arrows). There is no loss of internal signal intensity to suggest intracellular lipid. Had the high signal intensity been secondary to macroscopic fat, one would have expected an etching artifact at the interface of the high signal intensity with the adjacent water-containing tissue (see Fig. 11B). (C) Fat-suppressed T2-weighted image (effective echo time [TE] 5 90 ms) shows that most of the metastases (arrows) are isointense to spleen (S). The hemorrhagic components of the metastases present within the liver and spleen (small arrows) are isointense to hypointense to liver secondary to intracellular methemoglobin and/or melanin.
MR Characterization of Focal Liver Lesions The reader is referred to the article by Wells to ensure that an appropriate echo pair is used at 3 T. Cysts, hemangiomas, and almost all liver metastases do not contain lipid and, therefore, do not lose signal intensity when comparing the opposed-phase image with the corresponding inphase image. Pearl 2 is similar to pearl 1 in that it uses T1-weighted pulse sequences to characterize focal lesions as hepatocellular in origin. Next is the list of those hepatocellular lesions that contain microscopic lipid and lose signal intensity on opposed-phase images.
RN and HCC Both RN and HCC may contain lipid and lose signal intensity on chemical shift imaging (Figs. 6 and 7). The presence of intralesional lipid is one of the ancillary features that favor HCC over RN in the liver imaging reporting and data system classification.5,56 However, intralesional lipid is not 100% specific for HCC. Lipid-containing RN can occur and should be favored when multiple and less than 10 mm.57 Investigators have shown that lipid-containing HCCs tend to have better prognosis58 and better differentiation59 compared
Fig. 6. Lipid-containing HCC in a 59-year-old man with cirrhosis secondary to hepatitis C infection. (A) Axial inphase T1-weighted image shows a slightly hypointense liver dome mass (arrow). (B) Opposed-phase T1-weighted image shows subtle loss of signal intensity (arrow) indicating the presence of intracellular lipid. (C) Subtraction image (similar technique as described in Fig. 3C) demonstrates intralesional signal intensity (arrow) confirming the presence of lipid. (D) Axial fat-suppressed T2-weighted image shows that the mass has components isointense to spleen (S). In the setting of cirrhosis, this is specific for HCC. (E) Hepatobiliary-phase enhanced fat-suppressed T1-weighted image obtained after the intravenous administration of gadoxetate disodium shows that the mass is hypointense relative to the surrounding liver, also suggesting that it represents HCC as opposed to a benign RN. (F) Repeat axial in-phase T1-weighted image obtained after transarterial chemoembolization shows decrease in size of the mass and subtle increase in signal intensity, especially anteriorly (arrow) where components of the treated lesion approach the signal intensity of the adjacent liver. (G) Opposed-phase image shows moderate to marked loss of signal intensity within the mass (arrow). In the patient, the signal loss could be either from intracellular lipid or intratumoral ethiodized oil. (H) Subtraction image obtained after gadolinium enhancement shows no internal enhancement. No viable tumor was identified at subsequent transplant. (I) Unenhanced computed tomography (CT) examination shows dense ethiodized oil accumulation. The CT and chemical shift MR images illustrate complementary techniques of showing ethiodized oil uptake; the CT shows the presence of intratumoral iodine, whereas the MR shows the lipid contained within the ethiodized oil.
301
302
Siegelman & Chauhan
Fig. 7. MR depiction of a lipid-containing, iron-spared HCC, siderotic nodules, and Gamna-Gandy bodies in a 66-year-old man with hepatitis C infection and cirrhosis. (A) Axial in-phase image obtained at 3 T (echo time 2.48 ms) shows a high-signal-intensity lesion within segment 7 of the liver (arrow). The liver is diffusely of low signal intensity secondary to iron deposition. Focal low signal intensity within the liver and spleen (thin arrows) represent siderotic RN and Gamna-Gandy bodies, respectively. (B) Opposed-phase image (echo time 1.24 ms) shows higher signal intensity within those tissues that contain iron, including the diffuse liver parenchyma and both the focal siderotic nodules and Gamna-Gandy bodies. There is loss of signal intensity within the posterior aspect of the segment 7 lesion (arrow) indicating the presence of intracellular lipid. (C) Subtraction image (in-phase minus opposed-phase) confirms the presence of intralesional lipid (arrow). The presence of lipid is specific for hepatocellular tissue but is not specific for malignancy. However, the absence of intralesional iron in cirrhotic patients with diffuse and focal iron deposition is suspicious for malignancy. A 1.9-cm HCC was confirmed at surgery.
with non–lipid-containing HCCs. Imaging features that would favor HCC over an RN would include larger size, restricted diffusion, isointensity to spleen on T2-weighted images, arterial phase enhancement, and portal venous washout with a pseudocapsule. The reader is referred to the articles by Barr and Hussain and Sirlin in this issue of the Magnetic Resonance Clinics of North America for further discussion concerning lipidcontaining HCC.
HNF-1a Hepatic Adenoma In “Pearl 1,” the authors discuss that the IHCA was associated with hepatic steatosis and lacked intracellular lipid. The HNF-1a adenoma has opposite qualities; the lesions themselves contain lipid, but the surrounding liver usually does not have hepatic steatosis (see Fig. 2; Fig. 8). The accumulation of intracellular lipid within HNF-1a adenomas is hypothesized to be secondary to the aberrant promotion of lipogenesis secondary to gene inactivation.60 Some investigators have suggested that HNF-1a subtype adenomas are less hyperintense on T2-weighted images and show greater contrast washout on interstitial gadolinium-enhanced
imaging compared with IHCA.28,34,39 However, van Aalten and colleagues29 did not confirm these findings. It is probably more clinically important to establish a diagnosis of HCA and exclude a diagnosis of intralesional hemorrhage or HCC than it is to be able to establish the specific subtype of HCA. There are unusual instances of FNH that contain intracellular lipid.17,61 Fortunately, most of these FNH demonstrate a central scar. However, in those lipid-containing FNH lesions whereby no scar is depicted, differentiation from an HNF1-a adenoma may be difficult. In such cases, performing a gadolinium-enhanced study with a hepatobiliary contrast agent should be helpful in separating these entities. If an HCA undergoes intralesional hemorrhage, high T1 signal intensity that persists on opposedphase and fat-suppressed images can be present secondary to the T1 shortening effects of methemoglobin (see Fig. 8).45
Nodular Steatosis Chemical shift imaging methods can readily both qualify and quantify the presence and degree of hepatic steatosis.56,62 One of the causes of
MR Characterization of Focal Liver Lesions
Fig. 8. MR findings of a surgically proven lipid-containing HNF-1a hepatic adenoma with intralesional hemorrhage in a 40-year-old man who presents with 3 days of severe left upper quadrant pain. (A) Axial in-phase T1-weighted gradient echo image shows an 11-cm left lobe liver mass that has central components that are both hypointense and hyperintense to surrounding liver and peripheral tissue that is minimally hyperintense to liver. (B) Corresponding opposed-phase image shows loss of signal intensity within the periphery of the mass (arrow), establishing the presence of intracellular lipid. There is no loss of signal intensity within the central high signal intensity components, which are secondary to methemoglobin within subacute hemorrhage (H). It is the presence of lipid and not hemorrhage that is specific for characterizing the mass as hepatocellular in origin. (C) Subtraction image (in-phase minus opposed-phase; see Fig. 3 legend) confirms the presence of intracellular lipid not only within the segment 1 and 2 nodules (arrow) but also in the remainder of the liver to a variable degree.
steatosis is chemotherapy.63 In oncology patients, hepatic steatosis can either mask or mimic liver metastases on computed tomography (CT).64 Some patients have undergone liver biopsy or even wedge resections based on suspicious imaging findings.65,66 Fat has shorter T1 values than liver at both 1.5 and 3.0 T.67 Thus, nodular steatosis should be hyperintense to surrounding liver on in-phase T1-weighted imaging (Fig. 9), suggesting it is of hepatocellular origin based on pearl 1. The loss of signal on opposed-phase imaging establishes the presence of lipid. On all other pulse sequences, nodular steatosis should follow the expected behavior of steatotic non-neoplastic liver and, thus, should not be confused with an HNF-1a adenoma or well-differentiated lipid-containing HCC.
THE EXCEPTIONS Clear Cell Renal Cell Carcinoma Clear cell renal cell carcinomas were named because of the transparency of the cytoplasm on
hematoxylin and eosin staining. Intracellular glycogen and lipids are removed during the staining process. Chemical shift MR can detect and characterize the lipid content of primary clear cell renal cell carcinomas68 as well as metastatic disease to the adrenal gland69 and pancreas.70 Although there are no published case reports of chemical shift imaging of hepatic metastatic clear cell renal cell carcinomas, the authors have seen examples (Fig. 10). Metastatic clear cell renal cell carcinoma to the liver should not be misdiagnosed as a benign lipid-containing hepatocellular lesion, as these patients often have a known primary renal cancer and other sites of metastatic disease.
Liposarcoma and Germ Cell Neoplasms Retroperitoneal and extremity liposarcomas can rarely metastasize to the liver. Liposarcoma metastatic to the liver can have variable amounts of fat that can be characterized with chemical shift imaging and fat-suppressed imaging techniques. Examples of these metastases are rare and
303
304
Siegelman & Chauhan
Fig. 9. Nodular steatosis and a hepatic cyst in a 49-year-old woman with acute alcoholic hepatitis. (A) Enhanced CT examination detected nodular and regional foci of decreased attenuation (arrows) for which MR was requested for further evaluation. (B) Axial in-phase image shows that lesions of the caudate and posterior aspect of segment 2 (black arrows) are hyperintense to the remainder of the left lobe of the liver suggesting hepatocellular origin. An 8-mm cyst of the subcapsular portion of segment 4 (small white arrow) is hypointense to both liver and spleen. (C, D) Corresponding opposed-phase image (C) and subtraction image (in-phase minus opposed-phase; see Fig. 3 legend) (D) not only confirms the presence of intracellular lipid within the caudate and segment 2 nodules (arrows) but also shows mild regional and diffuse steatosis of the remainder of the liver. (E, F) Axial T2-weighted fast-spin-echo images with effective echo times of 94 (E) and 180 ms (F) show that the nodular lesions are isointense to spleen (S), whereas the 8-mm cyst (arrow) is hyperintense to spleen and isointense to cerebrospinal fluid and ascites. The higher signal intensity of the nodular steatosis compared with surrounding liver is from the T2 lengthening effects of fat. (G) Fat-suppressed T2-weighted image with an effective echo time of 94 ms shows a change in contrast; the hepatic nodules (arrows) are now moderately hypointense to spleen (S). One would have expected a lipid-containing HCC to have components that were isointense to spleen on this pulse sequence (see Fig. 6C). These foci of nodular steatosis did not restrict diffusion or show arterial phase enhancement and did enhance on hepatobiliary-phase enhanced imaging (not shown), all of which supported a benign diagnosis.
reportable.71,72 There are also rare reports of hepatic metastases from malignant mature cystic teratoma or from an immature ovarian teratoma.73,74 Like peritoneal extension of other ovarian neoplasms, these lesions are centered on the peritoneal surface of the liver as opposed to intrahepatic lesions from hematogenous or lymphatic spread. Primary hepatic teratomas are even rarer than peritoneal spread of an ovarian teratoma, with only 25 reported cases.18,75
Ethiodized Oil Ethiodized oil (e.g., lipiodol) is a compound that contains both iodine and long-chain fatty acids.76 Hepatic malignancies treated with ethiodized oil chemoembolization can show loss of signal intensity on chemical shift imaging secondary to the lipid components of the agent. Unenhanced CT can be used to evaluate the distribution and accumulation of high-attenuation iodine within treated
MR Characterization of Focal Liver Lesions
Fig. 10. Chemical shift MR findings of metastatic clear cell renal cell carcinoma in a 52-year-old man. (A) Axial inphase T1-weighted image shows hypointense liver dome (arrow) and left chest wall (large arrows) masses. (B) Opposed-phase T1-weighted image shows subtle signal loss within both lesions indicating the presence of intracellular lipid. (C) Subtraction image (similar technique as described in Fig. 3C) shows signal within both lesion (arrows) confirming the presence of lipid. (D) T2-weighted fast-spin-echo image (effective echo time [TE] 5 79 ms) shows that the liver metastasis (arrow) is isointense to spleen (S), whereas the chest wall mass is slightly hyperintense to spleen. A small focus of increased signal intensity (small arrow) is present within the metastasis. (E) Heavily T2-weighted image (effective TE 5 179 ms) shows that the liver lesion remains isointense to spleen with continued punctuate central hyperintensity (arrow). This focus did not enhance after contrast (not shown) and is in keeping with central necrosis. Note the decreased contrast between the liver metastases with both the surrounding liver and spleen. This pulse sequence is ideal for characterizing cysts and nonsolid lesions but should not be relied on for solid lesion detection.
lesions (see Fig. 6I).77 Although the loss of signal intensity within malignancies after chemoembolization has not been formally reported, the authors have observed this phenomenon in HCC and metastatic disease after chemoembolization (see Fig. 6F–H).
Hepatic Angiomyolipoma Hepatic angiomyolipomas (HAML) are rare mesenchymal neoplasms composed of variable amounts of fat, smooth muscle, and blood vessels.78 Although HAML is a primary hepatic lesion, it does not contain hepatocytes and, thus, is not hepatocellular in origin. HAMLs with greater than 70% fat content have been termed lipomatous HAMLs that compose 20% of all HAMLs.79 Chemical shift MR can categorize these lipomatous HAMLs by depicting an etching artifact at the interface of the HAML and the adjacent liver on opposed-phase imaging (Fig. 11).80 The presence of macroscopic fat can then be confirmed by showing marked loss of internal signal intensity on a corresponding fat-suppressed T1-weighted image, similar to how one diagnoses fat-containing renal AMLs. Opposed-phase imaging can detect smaller amounts of lipid, so-called microscopic
fat, present within the nonlipomatous HAMLs. The signal loss within these HAMLs on opposed-phase imaging can appear similar to the examples of focal steatosis and HNF-1a adenomas as discussed earlier.81 Lipid-poor HAMLs may not be definitely characterized by imaging and may require tissue sampling. For example, in a single-institution experience of 178 resected nonlipomatous HAMLs, less than 10% of the lesions were considered HAML at preoperative imaging.78 The resected HAMLs were most commonly misdiagnosed as HCC. Other less common fat-containing lesions, such as lipoma and pseudolipoma of Glisson capsule, are discussed and illustrated elsewhere.18,72,82
PEARL 3 (IN 3 PARTS): THE IRON PEARLS A. Focal liver lesions that diffusely lose signal intensity on an in-phase image compared with the corresponding opposed-phase image are siderotic nodules. B. A siderotic nodule that contains an iron-spared region within it is an HCC until proven otherwise. C. In an iron overloaded cirrhotic liver, an ironspared lesion is suspicious for HCC.
305
306
Siegelman & Chauhan
Fig. 11. Chemical shift imaging illustration of a primary HAML in a 37-year-old woman with tuberous sclerosis. Not every T1 hyperintense or lipid-containing liver lesion is hepatocellular in origin. (A, B) Axial in (A) and opposed-phase (B) T1-weighted images show a focal liver lesion (arrows) that is hyperintense to liver at both echo times. The lesion can be characterized as containing macroscopic fat because of the presence of an etching artifact (arrows) at the interface between the mass and the surrounding liver indicating a fat-water interface. As the liver is almost never composed of more lipid than water, one can conclude the mass is composed of mostly fat and that the residual signal with the lesion is from fat protons. (C) Subtraction image (in-phase minus opposedphase; see Fig. 3 legend) confirms the presence of both lipid and water protons within the HAML and the absence of hepatic steatosis. Even though this HAML is composed mostly of fat, the amount of signal intensity present on this subtraction image is minimal. The maximal signal on the subtraction image corresponds to those voxels that have maximum signal loss on the opposed-phase image (eg, at the interface of the HAML with the adjacent liver [arrows]); this corresponds to those voxels that have similar signal contributions from lipid and water protons.
Loss of signal intensity within the liver on an inphase image when compared with the corresponding opposed-phase image is usually from the susceptibility effects of excess iron, assuming the in-phase image is acquired at a longer echo time as was suggested in “Pearl 2.” The 2 most common causes of diffuse excess liver iron are from prior transfusions, where liver iron accumulates primarily in the Kupffer cells, or genetic hemochromatosis, where iron selectively deposits within hepatocytes.83,84 Loss of signal intensity on an in-phase image within a focal liver lesion in cirrhotic patients is almost always secondary to a siderotic RN (see Fig. 7; Fig. 12). Gradient echo images obtained with a longer echo time (eg, >10 ms) and a lower flip angle (eg,
KEY POINTS Focal liver lesions that are isointense to hyperintense to liver on in-phase T1-weighted images are usually hepatocellular in origin. Focal liver lesions that lose signal intensity on an opposed-phase image compared with the matched in-phase image contain lipid and are usually hepatocellular in origin. Focal liver lesions that lose signal intensity on an in-phase image compared with an opposed-phase image are most often iron-containing siderotic nodules. Focal liver lesions that are isointense to spleen on T2-weighted images are solid and often malignant, whereas focal liver lesions that are hyperintense to spleen on heavily T2-weighted images are usually nonsolid benign cysts or hemangiomas.
The first 3 imaging pearls discussed in this review are the 3 instances when focal liver lesion characterization is possible with T1-weighted gradient echo images. The 3 most common focal liver lesions encountered in clinical practice are cysts, hemangiomas, and metastatic disease. Nonsolid benign hepatic lesions (cysts and hemangiomas) and almost all metastatic lesions are hypointense relative to liver on T1-weighted images (Fig. 1). Normal liver has relative high signal intensity on T1-weighted images that has been attributed to high concentrations of protein, rough endoplasmic reticulum, and paramagnetic substances, such as manganese and copper.1–3 One study calculated the
T1 relaxation times of liver at 1.5 T as 547 ms and that of solid lesions, hemangiomas, and cysts to be 1004, 1337, and 3143 ms, respectively. Thus, most liver lesions are initially detected on T1-weighted images as being hypointense to liver.4 The authors use other pulse sequences besides T1-weighted images in order to differentiate among cysts, hemangiomas, and solid liver lesions. If a focal liver lesion is isointense to hyperintense to liver on a T1-weighted image, then it is most commonly hepatocellular in origin. The 5 most common focal hepatocellular lesions encountered in clinical practice are regenerative nodules (RN), hepatocellular carcinoma (HCC), focal nodular hyperplasia (FNH), hepatocellular adenoma (HCA), and focal steatosis. In this section, the authors discuss FNH and the inflammatory subtype of HCA (IHCA). The reader is referred to the articles by Barr and Hussain and Sirlin in this issue of the
Department of Radiology, Perelman School of Medicine, University of Pennsylvania, 34th and Spruce Streets, 1st Floor Silverstein, Philadelphia, PA 19104-4283, USA * Corresponding author. E-mail address: [email protected] Magn Reson Imaging Clin N Am 22 (2014) 295–313 http://dx.doi.org/10.1016/j.mric.2014.04.005 1064-9689/14/$ – see front matter Ó 2014 Elsevier Inc. All rights reserved.
mri.theclinics.com
PEARL 1: THE T1 PEARL: A FOCAL LESION THAT IS ISOINTENSE TO HYPERINTENSE TO LIVER ON T1-WEIGHTED IMAGES IS HEPATOCELLULAR IN ORIGIN
Siegelman & Chauhan
296
Fig. 1. Magnetic resonance demonstration of hepatic hemangioma and metastatic breast cancer in a 51-year-old woman. Metastases, cysts, and hemangiomas are almost all hypointense relative to liver on T1-weighted images. (A) Axial in-phase T1-weighted image shows low signal intensity hemangioma (white arrow) and infiltrative multifocal metastases (smaller black arrows). The metastases are isointense and the hemangioma is hypointense to spleen (S). However, the authors prefer to use T2-weighted images, diffusion-weighted images, and/or enhanced imaging to differentiate solid masses, such as metastatic disease, from nonsolid hepatic cysts and hemangiomas. (B) Corresponding opposed-phase image shows similar relative signal intensities of the hemangioma and metastatic disease. Geographic regions of lower signal intensity within the left lobe of liver (black arrows) represent steatosis. (C, D) T2-weighted fast-spin-echo images obtained with effective echo times of 90 (C) and 180 ms (D). The hemangioma is hyperintense to spleen, whereas the metastases are isointense. As the echo time increases, the contrast between the hemangioma and the adjacent metastases improve. This improvement is not because the hemangioma lights up or enhances; rather, the improved image contrast is because the hemangioma loses less signal as the echo time increases compared with liver, spleen, and metastases. (E) Opposedphase T1-weighted imaging performed 2 years prior shows that the hemangioma (arrow) is hyperintense to the surrounding liver (L). When trying to establish the hepatocellular origin of a focal liver lesion by showing isointensity or hyperintensity to liver, one should use an in-phase image as steatotic liver can be of low signal intensity on opposed-phase imaging and can confound relative signal assessment. (F) Corresponding in-phase T1-weighted image shows that the hemangioma (black arrow) is hypointense to the steatotic liver.
Magnetic Resonance Imaging Clinics of North America concerning the evaluation of the cirrhotic liver and how to differentiate RN from HCC. Lipid and fat-containing liver lesions are discussed in “Pearl 2.”
FNH FNH is the second most common benign hepatic tumor in adults after hemangioma. FNH composes approximately 8% of all primary liver tumors and has an estimated prevalence between 0.3% and 3.0%.5,6 FNH is not considered a neoplasm but instead is hypothesized to develop as a hyperplastic response of hepatic parenchyma around a central developmental vascular malformation.7 Individuals with FNH are more likely to have coexistent hepatic hemangiomas (20%) than would be expected by chance8; both hemangioma and FNH involve focal abnormalities in the hepatic blood supply.
FNH is typically detected in women aged 20 to 50 years and is uncommon in men (female-tomale ratio 5 10:1).5 Unlike hepatic adenomas, there is no proven association of oral contraceptive use or pregnancy with the development or growth of FNH.9,10 Although up to 15% of FNH lesions can grow when followed longitudinally,11 this should not cause clinical concern. The authors are skeptical of reports of malignant transformation to fibrolamellar hepatoma12 or HCC,13 as most investigators think that malignant transformation of FNH does not occur.14 The magnetic resonance (MR) features of FNH can be considered in 2 parts (Fig. 2). The first component is the vascular nidus that forms the central scar of FNH, and the second is the surrounding hyperplastic response of adjacent liver. The vascular scar is hypointense to liver on both in-phase and opposed-phase imaging and is hyperintense to liver on T2-weighted imaging. The higher T2-weighted signal intensity is
MR Characterization of Focal Liver Lesions
Fig. 2. MR findings of FNH and lipid-containing hepatocyte nuclear factor-1a inactivated hepatic adenoma in a 25-year-old woman. (A, B) Axial in-phase (A) and opposed-phase (B) T1-weighted images show 2 lesions: a larger FNH in segment 8 and a smaller adenoma in segment 2. The FNH has outer components (thick arrow in A) that are isointense to surrounding liver suggesting that it is hepatocellular in origin. There is a lower signal intensity central scar (thin arrow in A). The adenoma is isointense to liver and not perceptible on in-phase imaging; however, it loses signal and becomes recognizable on opposed-phase imaging (arrow in B). In-phase isointensity and loss of signal on opposed-phase image indicate that the lesion is lipid containing and hepatocellular in origin. (C, D) T2-weighted fast-spin-echo images (effective echo time 5 95 ms) shows that both the FNH and adenoma (thick arrows) are hyperintense to liver and relatively isointense to spleen. The central scar of the FNH (thin arrow) is hyperintense to spleen. On the fat-suppressed image (D), there is similar contrast with the exception of lower signal intensity of the adenoma (thick arrow in D) because of the suppressed lipid. (E) Hepatobiliary-phase enhanced fat-suppressed T1-weighted image obtained after the intravenous administration of gadoxetate disodium shows hyperenhancement of the nonscar components of the FNH (thick arrow) and hypoenhancement of the adenoma (thin arrow). The adenoma showed hyperenhancement compared with adjacent liver on dynamic enhanced imaging (not shown).
hypothesized to be secondary to slow flow within vessels. The scar does not enhance during the dynamic administration of gadolinium but does show delayed enhancement during the interstitial phase of enhancement when a conventional extracellular gadolinium contrast agent is used.15,16 The surrounding lesion of FNH itself is usually isointense to liver on both in-phase and opposedphase imaging. Cases of lipid-containing FNH are uncommon and reportable17; many are associated with diffuse hepatic steatosis.18 On T2-weighted imaging, FNH is isointense to minimally hyperintense relative to liver. On arterial-phase dynamic gadolinium-enhanced imaging, FNH shows marked homogeneous enhancement, often with rapid washout during the portal and interstitial phases of enhancement.16,19 If a gadolinium contrast agent with hepatobiliary excretion is used (eg, gadoxetate disodium), FNH will typically have components that are isointense to hyperintense relative to the surrounding liver on delayed hepatobiliary-phase imaging, which can be useful for distinguishing FNH from HCA for some lesions (see Fig. 2E).19–24 In one study of 30
FNH lesions, only 2 showed homogenous low signal intensity on delayed hepatobiliary-phase images.25 If one uses gadoxetate disodium–enhanced MR to characterize FNH, the central scar may not show enhancement during the interstitial phase of contrast enhancement.26 The absent enhancement of the central scar is hypothesized to be from the more rapid removal of gadoxetate disodium from the circulation compared with other gadolinium contrast agents. The reader is referred to the article by Bashir for a detailed discussion concerning the use of the various MR contrast agents used for liver imaging. Diffusion-weighted imaging and apparent diffusion coefficient (ADC) values alone should not be used to differentiate benign from malignant focal liver lesions as there can be substantial overlap. For example, many benign FNH and hepatic adenomas can show restricted diffusion similar to metastatic disease.27 For a review concerning the performance and interpretation of diffusionweighted imaging of the liver, the reader is referred to the article by Taouli in this issue of the Magnetic Resonance Clinics of North America.
297
298
Siegelman & Chauhan IHCA Although pathologists had previously classified HCA into a single pathologic entity, recent molecular and immunohistochemical markers have determined that there are 4 distinct subtypes of HCAs.28–31 These subtypes include IHCAs and hepatocyte nuclear factor 1a inactivated (HNF-1a), b-catenin–activated, and unclassified HCAs. MR can often detect and characterize the subtype of HCA based on specific imaging features.32 Independent of the subtype, larger (>4 cm) adenomas are at risk of developing intralesional hemorrhage33 and are usually treated with surgical resection or with nonsurgical radiofrequency ablation or bland embolization.5,29,34,35 Smaller adenomas can be managed conservatively and followed by imaging.35 Women who take oral contraceptives should discontinue their use as some HCAs may subsequently decrease in size.5,31 Obesity and metabolic syndrome are associated with HCA36,37; patients are encouraged to modify their diet, develop an exercise program, and lose weight as part of a treatment plan.
Hepatic adenomatosis is a distinct entity that was originally described and defined in 1985 as the presence of greater than 10 HCAs (Fig. 3).38 The HCA associated with adenomatosis is not limited to any one particular subtype,39,40 although patients with multiple adenomas are more likely to have steatosis.41 As with single lesions, management is based on lesion size and patients’ symptoms.42,43 The IHCA is the most common subtype of HCA and accounts for 40% to 60% of lesions in reported series. IHCA is associated with obesity34 and metabolic syndrome.36 In one study of 32 women with IHCA, the median body mass index was 32.5.44 In this same group of 32 women, steatosis was present in 59% on liver biopsies obtained distant from the tumor. In another series comparing 63 IHCA versus 46 HNF-1a adenomas, there was significantly less intralesional steatosis in the IHCA subgroup (43% vs 82%).34 Ten of the 63 patients with IHCA had findings of either intralesional or peritumoral hemorrhage, which was not significantly different than those found in the HNF-1a subtype.
Fig. 3. MR imaging findings of pathology-proven hepatic steatosis and hepatic adenomatosis secondary to IHCA in a 43-year-old woman. The patient stopped taking oral contraceptives at the time of diagnosis and has been followed conservatively for the next 7 years. (A) Axial in-phase T1-weighted gradient echo image shows subtle foci of increased signal intensity relative to liver (black arrows). (B) Corresponding opposed-phase image shows loss of signal intensity within the liver confirming the presence of hepatic steatosis. There are multiple lesions that are hyperintense to the steatotic liver (arrows). Given that these lesions were isointense to liver on inphase imaging, this pattern suggests they are hepatocellular in origin. (C) Subtraction image created by subtracting the opposed-phase image from the in-phase image (in-phase minus opposed-phase) depicts those voxels that contain both lipid and water protons. The highest signal intensity on the subtraction images occurs at fat-water interfaces (eg, junction of left kidney and perirenal fat [thin arrows]). This image confirms the presence of hepatic steatosis and the absence of lipid within the focal liver lesions (thick arrows). (D, E) Precontrast and arterial phase–enhanced fat-suppressed T1-weighted images show hyperenhancement of the adenomas (arrows) relative to the surrounding liver. The presence of arterial phase enhancement assists in differentiating IHCAs from masslike regions of focal sparing of steatosis.
MR Characterization of Focal Liver Lesions On T1-weighted images, IHCAs are usually isointense to slightly hyperintense to surrounding liver.28,29 On opposed-phase imaging, the surrounding liver is more likely to lose signal intensity because of steatosis than the inflammatory adenoma itself (see Fig. 3; Fig. 4).34,44 On T2weighted images, most IHCAs are minimally hyperintense relative to liver.28,29 In one series, 13 or 30 IHCAs revealed a specific atoll sign, consisting of a hyperintense rim that enhances on delayed gadolinium-enhanced imaging (see Fig. 4C).29 It is hypothesized that the atoll sign is secondary to dilated sinusoids within the periphery of the adenoma. Like FNH, HCAs will enhance after gadolinium contrast. However, unlike FNH, IHCAs do not have a central scar and almost all IHCAs do not have components that are isointense to hyperintense to liver on delayed hepatobiliaryphase gadolinium-enhanced imaging.23 Thus, in a noncirrhotic woman with hepatic steatosis who has a solid enhancing mass that is isointense to liver on precontrast T1-weighted images and hypointense on hepatobiliary-phase enhanced images, one should consider an IHCA as the most likely cause.
THE EXCEPTIONS Nonhepatocellular Focal Liver Lesions in a Liver Containing Background Moderate or Marked Steatosis On opposed-phase T1-weighted images, liver that is involved with moderate or severe steatosis will have very low signal intensity. Thus, if one were to evaluate the relative signal intensity of a focal liver lesion with surrounding steatotic liver on an opposed-phase T1-weighted image alone, then one may come to a false conclusion that the lesion is hepatocellular. For example, a benign hemangioma (see Fig. 1E) or even a metastasis can appear hyperintense to steatotic liver on opposed-phase imaging. Therefore, one should use the in-phase image as the T1-weighted reference when using pearl 1.
Hemorrhagic Metastases to the Liver Rare hepatic metastases can show T1 hyperintensity secondary to the T1 shortening properties of methemoglobin within subacute intralesional hemorrhage.45 This hyperintensity has been described in cases of metastatic renal, neuroendocrine, and
Fig. 4. MR illustration of the atoll sign in a 26-year-old woman with a body mass index of 39 with a surgically proven IHCA. Obesity and metabolic syndrome are associated with hepatic adenomas. (A, B) Axial in (A) and opposed-phase (B) T1-weighted images show a subcapsular hyperintense hepatic mass (arrow) indicating it is hepatocellular in origin. Mild steatosis, which was confirmed at surgery, is most pronounced and best revealed as lower signal intensity within segment 4 on the opposed-phase image (small arrows). (C) Axial fast-spin-echo T2-weighted image (effective echo time 5 93 ms) shows that the adenoma is hyperintense to liver and isointense to spleen. Not every focal lesion that is isointense to spleen is malignant, especially in individuals who do not have cirrhosis or history of a primary malignancy. A high-signal-intensity rim around the adenoma (small arrows) has been termed the atoll sign and is hypothesized to be secondary to dilated sinusoids. (D) Arterial-phase enhanced fat-suppressed T1-weighted image shows hyperenhancement of the adenoma and hypoenhancement of the peritumoral sinusoids.
299
300
Siegelman & Chauhan lung cancers as well as choriocarcinoma.2,46 Focal liver lesion hyperintensity can also be secondary to the paramagnetic metals attached to melanin contained within metastatic melanoma (Fig. 5).2,47,48 In these instances, patients’ primary tumor and the presence of metastatic disease are often known or apparent at the time of imaging. After radiofrequency ablation or chemoembolization of liver malignancies, intratumoral T1 hyperintensity can result from hemorrhagic and coagulative necrosis.49 Once again, the cancer history and prior procedures are usually known at the time of image interpretation. The reader is referred to the article by Kamel concerning the use of MR to evaluate treated liver metastases.
Hemorrhagic Cysts of the Liver Most liver cysts have simple fluid content and very low signal intensity on T1-weighted images. Occasionally, intralesional hemorrhage can occur in simple hepatic cysts that appear hyperintense on T1-weighted images secondary to methemoglobin and or intracystic proteins.50 Hemorrhage into simple hepatic cysts has been described in autosomal dominant polycystic liver disease51 and large bile duct hamartomas.52 The absence of a
thick fibrous capsule, septa, nodules, and solid enhancing components differentiates the benign hemorrhagic hepatic cyst from biliary cystadenoma-biliary cystadenocarcinoma.51
PEARL 2: THE CHEMICAL SHIFT PEARL: FOCAL LIVER LESIONS THAT LOSE SIGNAL INTENSITY ON OPPOSED-PHASE IMAGING CONTAIN LIPID AND ARE MOST OFTEN HEPATOCELLULAR IN ORIGIN When performing dual-phase in and opposedphase T1-weighted gradient echo imaging of the liver, the echo time of the opposed-phase image should be shorter than the in-phase echo time. This timing will ensure that any loss of signal intensity on the opposed-phase image compared with the in-phase image will be caused by the presence of lipid and water protons within the same voxel.53,54 If the opposed-phase echo time is longer than the in-phase echo time, then signal loss could also be secondary to T2* susceptibility effects.55 Some manufacturers’ 3-T MR systems come with configured dual-phase gradient echo sequences (eg, echo times of 2.2 and 3.3 ms) that could result in potential diagnostic confusion.
Fig. 5. MR depiction of T1 hyperintense melanoma metastases to the liver in a 64-year-old woman. Some hemorrhagic metastases can be hyperintense to liver and should not necessarily be assumed to be hepatocellular in origin. (A) Axial in-phase T1-weighted gradient echo image shows multiple hypointense liver metastases. Two of the lesions have central high signal intensity (arrows). (B) Opposed-phase T1-weighted image shows persistent high internal signal intensity within 2 of the metastases (arrows). There is no loss of internal signal intensity to suggest intracellular lipid. Had the high signal intensity been secondary to macroscopic fat, one would have expected an etching artifact at the interface of the high signal intensity with the adjacent water-containing tissue (see Fig. 11B). (C) Fat-suppressed T2-weighted image (effective echo time [TE] 5 90 ms) shows that most of the metastases (arrows) are isointense to spleen (S). The hemorrhagic components of the metastases present within the liver and spleen (small arrows) are isointense to hypointense to liver secondary to intracellular methemoglobin and/or melanin.
MR Characterization of Focal Liver Lesions The reader is referred to the article by Wells to ensure that an appropriate echo pair is used at 3 T. Cysts, hemangiomas, and almost all liver metastases do not contain lipid and, therefore, do not lose signal intensity when comparing the opposed-phase image with the corresponding inphase image. Pearl 2 is similar to pearl 1 in that it uses T1-weighted pulse sequences to characterize focal lesions as hepatocellular in origin. Next is the list of those hepatocellular lesions that contain microscopic lipid and lose signal intensity on opposed-phase images.
RN and HCC Both RN and HCC may contain lipid and lose signal intensity on chemical shift imaging (Figs. 6 and 7). The presence of intralesional lipid is one of the ancillary features that favor HCC over RN in the liver imaging reporting and data system classification.5,56 However, intralesional lipid is not 100% specific for HCC. Lipid-containing RN can occur and should be favored when multiple and less than 10 mm.57 Investigators have shown that lipid-containing HCCs tend to have better prognosis58 and better differentiation59 compared
Fig. 6. Lipid-containing HCC in a 59-year-old man with cirrhosis secondary to hepatitis C infection. (A) Axial inphase T1-weighted image shows a slightly hypointense liver dome mass (arrow). (B) Opposed-phase T1-weighted image shows subtle loss of signal intensity (arrow) indicating the presence of intracellular lipid. (C) Subtraction image (similar technique as described in Fig. 3C) demonstrates intralesional signal intensity (arrow) confirming the presence of lipid. (D) Axial fat-suppressed T2-weighted image shows that the mass has components isointense to spleen (S). In the setting of cirrhosis, this is specific for HCC. (E) Hepatobiliary-phase enhanced fat-suppressed T1-weighted image obtained after the intravenous administration of gadoxetate disodium shows that the mass is hypointense relative to the surrounding liver, also suggesting that it represents HCC as opposed to a benign RN. (F) Repeat axial in-phase T1-weighted image obtained after transarterial chemoembolization shows decrease in size of the mass and subtle increase in signal intensity, especially anteriorly (arrow) where components of the treated lesion approach the signal intensity of the adjacent liver. (G) Opposed-phase image shows moderate to marked loss of signal intensity within the mass (arrow). In the patient, the signal loss could be either from intracellular lipid or intratumoral ethiodized oil. (H) Subtraction image obtained after gadolinium enhancement shows no internal enhancement. No viable tumor was identified at subsequent transplant. (I) Unenhanced computed tomography (CT) examination shows dense ethiodized oil accumulation. The CT and chemical shift MR images illustrate complementary techniques of showing ethiodized oil uptake; the CT shows the presence of intratumoral iodine, whereas the MR shows the lipid contained within the ethiodized oil.
301
302
Siegelman & Chauhan
Fig. 7. MR depiction of a lipid-containing, iron-spared HCC, siderotic nodules, and Gamna-Gandy bodies in a 66-year-old man with hepatitis C infection and cirrhosis. (A) Axial in-phase image obtained at 3 T (echo time 2.48 ms) shows a high-signal-intensity lesion within segment 7 of the liver (arrow). The liver is diffusely of low signal intensity secondary to iron deposition. Focal low signal intensity within the liver and spleen (thin arrows) represent siderotic RN and Gamna-Gandy bodies, respectively. (B) Opposed-phase image (echo time 1.24 ms) shows higher signal intensity within those tissues that contain iron, including the diffuse liver parenchyma and both the focal siderotic nodules and Gamna-Gandy bodies. There is loss of signal intensity within the posterior aspect of the segment 7 lesion (arrow) indicating the presence of intracellular lipid. (C) Subtraction image (in-phase minus opposed-phase) confirms the presence of intralesional lipid (arrow). The presence of lipid is specific for hepatocellular tissue but is not specific for malignancy. However, the absence of intralesional iron in cirrhotic patients with diffuse and focal iron deposition is suspicious for malignancy. A 1.9-cm HCC was confirmed at surgery.
with non–lipid-containing HCCs. Imaging features that would favor HCC over an RN would include larger size, restricted diffusion, isointensity to spleen on T2-weighted images, arterial phase enhancement, and portal venous washout with a pseudocapsule. The reader is referred to the articles by Barr and Hussain and Sirlin in this issue of the Magnetic Resonance Clinics of North America for further discussion concerning lipidcontaining HCC.
HNF-1a Hepatic Adenoma In “Pearl 1,” the authors discuss that the IHCA was associated with hepatic steatosis and lacked intracellular lipid. The HNF-1a adenoma has opposite qualities; the lesions themselves contain lipid, but the surrounding liver usually does not have hepatic steatosis (see Fig. 2; Fig. 8). The accumulation of intracellular lipid within HNF-1a adenomas is hypothesized to be secondary to the aberrant promotion of lipogenesis secondary to gene inactivation.60 Some investigators have suggested that HNF-1a subtype adenomas are less hyperintense on T2-weighted images and show greater contrast washout on interstitial gadolinium-enhanced
imaging compared with IHCA.28,34,39 However, van Aalten and colleagues29 did not confirm these findings. It is probably more clinically important to establish a diagnosis of HCA and exclude a diagnosis of intralesional hemorrhage or HCC than it is to be able to establish the specific subtype of HCA. There are unusual instances of FNH that contain intracellular lipid.17,61 Fortunately, most of these FNH demonstrate a central scar. However, in those lipid-containing FNH lesions whereby no scar is depicted, differentiation from an HNF1-a adenoma may be difficult. In such cases, performing a gadolinium-enhanced study with a hepatobiliary contrast agent should be helpful in separating these entities. If an HCA undergoes intralesional hemorrhage, high T1 signal intensity that persists on opposedphase and fat-suppressed images can be present secondary to the T1 shortening effects of methemoglobin (see Fig. 8).45
Nodular Steatosis Chemical shift imaging methods can readily both qualify and quantify the presence and degree of hepatic steatosis.56,62 One of the causes of
MR Characterization of Focal Liver Lesions
Fig. 8. MR findings of a surgically proven lipid-containing HNF-1a hepatic adenoma with intralesional hemorrhage in a 40-year-old man who presents with 3 days of severe left upper quadrant pain. (A) Axial in-phase T1-weighted gradient echo image shows an 11-cm left lobe liver mass that has central components that are both hypointense and hyperintense to surrounding liver and peripheral tissue that is minimally hyperintense to liver. (B) Corresponding opposed-phase image shows loss of signal intensity within the periphery of the mass (arrow), establishing the presence of intracellular lipid. There is no loss of signal intensity within the central high signal intensity components, which are secondary to methemoglobin within subacute hemorrhage (H). It is the presence of lipid and not hemorrhage that is specific for characterizing the mass as hepatocellular in origin. (C) Subtraction image (in-phase minus opposed-phase; see Fig. 3 legend) confirms the presence of intracellular lipid not only within the segment 1 and 2 nodules (arrow) but also in the remainder of the liver to a variable degree.
steatosis is chemotherapy.63 In oncology patients, hepatic steatosis can either mask or mimic liver metastases on computed tomography (CT).64 Some patients have undergone liver biopsy or even wedge resections based on suspicious imaging findings.65,66 Fat has shorter T1 values than liver at both 1.5 and 3.0 T.67 Thus, nodular steatosis should be hyperintense to surrounding liver on in-phase T1-weighted imaging (Fig. 9), suggesting it is of hepatocellular origin based on pearl 1. The loss of signal on opposed-phase imaging establishes the presence of lipid. On all other pulse sequences, nodular steatosis should follow the expected behavior of steatotic non-neoplastic liver and, thus, should not be confused with an HNF-1a adenoma or well-differentiated lipid-containing HCC.
THE EXCEPTIONS Clear Cell Renal Cell Carcinoma Clear cell renal cell carcinomas were named because of the transparency of the cytoplasm on
hematoxylin and eosin staining. Intracellular glycogen and lipids are removed during the staining process. Chemical shift MR can detect and characterize the lipid content of primary clear cell renal cell carcinomas68 as well as metastatic disease to the adrenal gland69 and pancreas.70 Although there are no published case reports of chemical shift imaging of hepatic metastatic clear cell renal cell carcinomas, the authors have seen examples (Fig. 10). Metastatic clear cell renal cell carcinoma to the liver should not be misdiagnosed as a benign lipid-containing hepatocellular lesion, as these patients often have a known primary renal cancer and other sites of metastatic disease.
Liposarcoma and Germ Cell Neoplasms Retroperitoneal and extremity liposarcomas can rarely metastasize to the liver. Liposarcoma metastatic to the liver can have variable amounts of fat that can be characterized with chemical shift imaging and fat-suppressed imaging techniques. Examples of these metastases are rare and
303
304
Siegelman & Chauhan
Fig. 9. Nodular steatosis and a hepatic cyst in a 49-year-old woman with acute alcoholic hepatitis. (A) Enhanced CT examination detected nodular and regional foci of decreased attenuation (arrows) for which MR was requested for further evaluation. (B) Axial in-phase image shows that lesions of the caudate and posterior aspect of segment 2 (black arrows) are hyperintense to the remainder of the left lobe of the liver suggesting hepatocellular origin. An 8-mm cyst of the subcapsular portion of segment 4 (small white arrow) is hypointense to both liver and spleen. (C, D) Corresponding opposed-phase image (C) and subtraction image (in-phase minus opposed-phase; see Fig. 3 legend) (D) not only confirms the presence of intracellular lipid within the caudate and segment 2 nodules (arrows) but also shows mild regional and diffuse steatosis of the remainder of the liver. (E, F) Axial T2-weighted fast-spin-echo images with effective echo times of 94 (E) and 180 ms (F) show that the nodular lesions are isointense to spleen (S), whereas the 8-mm cyst (arrow) is hyperintense to spleen and isointense to cerebrospinal fluid and ascites. The higher signal intensity of the nodular steatosis compared with surrounding liver is from the T2 lengthening effects of fat. (G) Fat-suppressed T2-weighted image with an effective echo time of 94 ms shows a change in contrast; the hepatic nodules (arrows) are now moderately hypointense to spleen (S). One would have expected a lipid-containing HCC to have components that were isointense to spleen on this pulse sequence (see Fig. 6C). These foci of nodular steatosis did not restrict diffusion or show arterial phase enhancement and did enhance on hepatobiliary-phase enhanced imaging (not shown), all of which supported a benign diagnosis.
reportable.71,72 There are also rare reports of hepatic metastases from malignant mature cystic teratoma or from an immature ovarian teratoma.73,74 Like peritoneal extension of other ovarian neoplasms, these lesions are centered on the peritoneal surface of the liver as opposed to intrahepatic lesions from hematogenous or lymphatic spread. Primary hepatic teratomas are even rarer than peritoneal spread of an ovarian teratoma, with only 25 reported cases.18,75
Ethiodized Oil Ethiodized oil (e.g., lipiodol) is a compound that contains both iodine and long-chain fatty acids.76 Hepatic malignancies treated with ethiodized oil chemoembolization can show loss of signal intensity on chemical shift imaging secondary to the lipid components of the agent. Unenhanced CT can be used to evaluate the distribution and accumulation of high-attenuation iodine within treated
MR Characterization of Focal Liver Lesions
Fig. 10. Chemical shift MR findings of metastatic clear cell renal cell carcinoma in a 52-year-old man. (A) Axial inphase T1-weighted image shows hypointense liver dome (arrow) and left chest wall (large arrows) masses. (B) Opposed-phase T1-weighted image shows subtle signal loss within both lesions indicating the presence of intracellular lipid. (C) Subtraction image (similar technique as described in Fig. 3C) shows signal within both lesion (arrows) confirming the presence of lipid. (D) T2-weighted fast-spin-echo image (effective echo time [TE] 5 79 ms) shows that the liver metastasis (arrow) is isointense to spleen (S), whereas the chest wall mass is slightly hyperintense to spleen. A small focus of increased signal intensity (small arrow) is present within the metastasis. (E) Heavily T2-weighted image (effective TE 5 179 ms) shows that the liver lesion remains isointense to spleen with continued punctuate central hyperintensity (arrow). This focus did not enhance after contrast (not shown) and is in keeping with central necrosis. Note the decreased contrast between the liver metastases with both the surrounding liver and spleen. This pulse sequence is ideal for characterizing cysts and nonsolid lesions but should not be relied on for solid lesion detection.
lesions (see Fig. 6I).77 Although the loss of signal intensity within malignancies after chemoembolization has not been formally reported, the authors have observed this phenomenon in HCC and metastatic disease after chemoembolization (see Fig. 6F–H).
Hepatic Angiomyolipoma Hepatic angiomyolipomas (HAML) are rare mesenchymal neoplasms composed of variable amounts of fat, smooth muscle, and blood vessels.78 Although HAML is a primary hepatic lesion, it does not contain hepatocytes and, thus, is not hepatocellular in origin. HAMLs with greater than 70% fat content have been termed lipomatous HAMLs that compose 20% of all HAMLs.79 Chemical shift MR can categorize these lipomatous HAMLs by depicting an etching artifact at the interface of the HAML and the adjacent liver on opposed-phase imaging (Fig. 11).80 The presence of macroscopic fat can then be confirmed by showing marked loss of internal signal intensity on a corresponding fat-suppressed T1-weighted image, similar to how one diagnoses fat-containing renal AMLs. Opposed-phase imaging can detect smaller amounts of lipid, so-called microscopic
fat, present within the nonlipomatous HAMLs. The signal loss within these HAMLs on opposed-phase imaging can appear similar to the examples of focal steatosis and HNF-1a adenomas as discussed earlier.81 Lipid-poor HAMLs may not be definitely characterized by imaging and may require tissue sampling. For example, in a single-institution experience of 178 resected nonlipomatous HAMLs, less than 10% of the lesions were considered HAML at preoperative imaging.78 The resected HAMLs were most commonly misdiagnosed as HCC. Other less common fat-containing lesions, such as lipoma and pseudolipoma of Glisson capsule, are discussed and illustrated elsewhere.18,72,82
PEARL 3 (IN 3 PARTS): THE IRON PEARLS A. Focal liver lesions that diffusely lose signal intensity on an in-phase image compared with the corresponding opposed-phase image are siderotic nodules. B. A siderotic nodule that contains an iron-spared region within it is an HCC until proven otherwise. C. In an iron overloaded cirrhotic liver, an ironspared lesion is suspicious for HCC.
305
306
Siegelman & Chauhan
Fig. 11. Chemical shift imaging illustration of a primary HAML in a 37-year-old woman with tuberous sclerosis. Not every T1 hyperintense or lipid-containing liver lesion is hepatocellular in origin. (A, B) Axial in (A) and opposed-phase (B) T1-weighted images show a focal liver lesion (arrows) that is hyperintense to liver at both echo times. The lesion can be characterized as containing macroscopic fat because of the presence of an etching artifact (arrows) at the interface between the mass and the surrounding liver indicating a fat-water interface. As the liver is almost never composed of more lipid than water, one can conclude the mass is composed of mostly fat and that the residual signal with the lesion is from fat protons. (C) Subtraction image (in-phase minus opposedphase; see Fig. 3 legend) confirms the presence of both lipid and water protons within the HAML and the absence of hepatic steatosis. Even though this HAML is composed mostly of fat, the amount of signal intensity present on this subtraction image is minimal. The maximal signal on the subtraction image corresponds to those voxels that have maximum signal loss on the opposed-phase image (eg, at the interface of the HAML with the adjacent liver [arrows]); this corresponds to those voxels that have similar signal contributions from lipid and water protons.
Loss of signal intensity within the liver on an inphase image when compared with the corresponding opposed-phase image is usually from the susceptibility effects of excess iron, assuming the in-phase image is acquired at a longer echo time as was suggested in “Pearl 2.” The 2 most common causes of diffuse excess liver iron are from prior transfusions, where liver iron accumulates primarily in the Kupffer cells, or genetic hemochromatosis, where iron selectively deposits within hepatocytes.83,84 Loss of signal intensity on an in-phase image within a focal liver lesion in cirrhotic patients is almost always secondary to a siderotic RN (see Fig. 7; Fig. 12). Gradient echo images obtained with a longer echo time (eg, >10 ms) and a lower flip angle (eg,
