Hepatocellular carcinoma (HCC) is the fastest growing cause of cancer-related death in the United States. Cirrhosis is the most important risk factor for HCC. Dynamic contrast-enhanced magnetic resonance (MR) imaging is the modality of choice for working up nodules detected at screening, for staging known HCC, and for follow-up. In cirrhotic livers, the combination of tumor arterial phase hyperenhancement plus washout and/or capsular enhancement is highly specific for HCC and can make biopsy unnecessary. Newer imaging techniques may further improve MR imaging sensitivity for HCC and help to characterize tumors with atypical dynamic enhancement patterns.
Key points
- •
Hepatocellular carcinoma (HCC) is the fastest growing cause of cancer-related death in the United States.
- •
Cirrhosis is the single most important risk factor for HCC, regardless of the cause of cirrhosis.
- •
Magnetic resonance (MR) imaging plays a key role in the diagnosis and staging of HCC.
- •
In cirrhosis, the combination of tumor arterial phase hyperenhancement plus washout and/or capsular enhancement is highly specific for HCC and can make biopsy unnecessary.
- •
Newer imaging techniques, including hepatocyte-specific contrast agents and diffusion-weighted imaging may further improve MR imaging sensitivity for HCC.
Introduction
Hepatocellular carcinoma (HCC) is an increasingly common disease. More than half a million new cases are diagnosed each year, making HCC the fifth most commonly diagnosed cancer worldwide. HCC is an aggressive neoplasm, being more lethal than common cancers such as prostate cancer and breast cancer, resulting in the second highest number of cancer deaths worldwide each year. Most cases occur in developing countries, but the incidence of HCC in the United States has been steadily increasing. HCC is now the fastest growing cause of cancer-related death in the United States.
The main risk factors for HCC are well understood, with cirrhosis being the single most important risk factor. More than 80% of HCCs arise in cirrhotic livers and the annual incidence of HCC is substantially higher in cirrhotic patients (2%–8%) compared with patients without cirrhosis (<0.5%). The major causes of cirrhosis and subsequent HCC vary greatly with geography. In developing nations such as sub-Saharan Africa and eastern Asia, chronic hepatitis B infection is the dominant cause, accounting for more than 50% of cases. In the United States and Europe, major causative factors include chronic hepatitis C infection (60% of cases) and alcohol abuse (45%), followed by hepatitis B (22%). The cited percentages sum to greater than 100%, suggesting that many patients with HCC have multiple risk factors. Synergistic interactions between risk factors markedly increase individual risk for HCC.
Until recently, increasing rates of HCC in the United States had been attributed primarily to an aging cohort of patients with chronic hepatitis C cirrhosis. However, over the past decade cirrhosis from nonalcoholic steatohepatitis has emerged as an important risk factor for HCC with an annual incidence estimated at 0.25% to 2.3%. In addition, metabolic conditions such as diabetes and obesity have also been implicated as independent risk factors for HCC. Less common risk factors for cirrhosis and HCC include genetic hemochromatosis, alpha1-antitrypsin deficiency, primary sclerosing cholangitis (PSC), primary biliary cirrhosis, and autoimmune hepatitis.
Introduction
Hepatocellular carcinoma (HCC) is an increasingly common disease. More than half a million new cases are diagnosed each year, making HCC the fifth most commonly diagnosed cancer worldwide. HCC is an aggressive neoplasm, being more lethal than common cancers such as prostate cancer and breast cancer, resulting in the second highest number of cancer deaths worldwide each year. Most cases occur in developing countries, but the incidence of HCC in the United States has been steadily increasing. HCC is now the fastest growing cause of cancer-related death in the United States.
The main risk factors for HCC are well understood, with cirrhosis being the single most important risk factor. More than 80% of HCCs arise in cirrhotic livers and the annual incidence of HCC is substantially higher in cirrhotic patients (2%–8%) compared with patients without cirrhosis (<0.5%). The major causes of cirrhosis and subsequent HCC vary greatly with geography. In developing nations such as sub-Saharan Africa and eastern Asia, chronic hepatitis B infection is the dominant cause, accounting for more than 50% of cases. In the United States and Europe, major causative factors include chronic hepatitis C infection (60% of cases) and alcohol abuse (45%), followed by hepatitis B (22%). The cited percentages sum to greater than 100%, suggesting that many patients with HCC have multiple risk factors. Synergistic interactions between risk factors markedly increase individual risk for HCC.
Until recently, increasing rates of HCC in the United States had been attributed primarily to an aging cohort of patients with chronic hepatitis C cirrhosis. However, over the past decade cirrhosis from nonalcoholic steatohepatitis has emerged as an important risk factor for HCC with an annual incidence estimated at 0.25% to 2.3%. In addition, metabolic conditions such as diabetes and obesity have also been implicated as independent risk factors for HCC. Less common risk factors for cirrhosis and HCC include genetic hemochromatosis, alpha1-antitrypsin deficiency, primary sclerosing cholangitis (PSC), primary biliary cirrhosis, and autoimmune hepatitis.
Surveillance
Surveillance for HCC is a controversial topic. Current surveillance recommendations are summarized later as an example of the role of magnetic resonance (MR) imaging.
Under the most recent American Association for the Study of Liver Disease (AASLD) practice guidelines, patients with cirrhosis of any cause should be screened for hepatocellular carcinoma every 6 months using ultrasonography. Noncirrhotic carriers of the hepatitis B virus who are Asian men more than 40 years old, Asian women more than 50 years old, African or North American black people, or people who have a strong family history of HCC should also be screened. Nodules measuring less than 1 cm at screening should be followed by ultrasonography every 3 months until either: (1) the nodule resolves, (2) the nodule is stable for 18 to 24 months, or (3) the nodule grows larger than 1 cm. Nodules larger than 1 cm are investigated further using dynamic contrast-enhanced (CE) multidetector computed tomography (MDCT) or dynamic CE-MR imaging. Many clinicians prefer CE-MR imaging because recent studies generally show superior performance compared with MDCT (CE-MR imaging: sensitivity 14%–82%, specificity 96%–100%, positive predictive value [PPV] 97%–100%, negative predictive value [NPV] 55%–84%. MDCT: sensitivity 40%–78%, specificity 93%–99%, PPV 86%–95%, NPV 60%–96% ). MR diagnostic performance can be further improved by using diffusion-weighted imaging (DWI) and hepatocyte-specific contrast agents (sensitivity 86%–95%, specificity 74%–98%, PPV 85%–92%, NPV 98%–99%). In limited situations MDCT or CE-MR imaging may be used for screening purposes, particularly for patients listed for liver transplantation or with very nodular livers that are difficult to screen by ultrasonography.
Technique
An in-depth discussion of liver MR imaging techniques is provided by Guglielmo and colleagues elsewhere in this issue; however, a few technical points are worth specific comment as they pertain to cirrhosis and HCC. Patients diagnosed with HCC are staged and treated according to guidelines developed by the Barcelona Clinic Liver Cancer group. Patients falling within the Milan criteria (single tumor ≤5.0 cm in diameter, or 2–3 tumors each ≤3.0 cm in diameter) may be candidates for liver transplantation. In the United States, donor organs are allocated via the Organ Procurement and Transplant Network (OPTN) under the direction of the United Network for Organ Sharing (UNOS). The latest OPTN/UNOS allocation policy prioritizes patients for transplant when tumors show a specific combination of imaging findings, even if biopsy has not been performed for histologic confirmation of HCC. However, the OPTN/UNOS policy does outline specific minimum technical requirements for MR imaging that must be met if MR imaging is used to diagnose HCC ( Table 1 ). If these requirements are not met, further steps may be required to confirm the diagnosis of HCC, such as biopsy or a repeat MR examination using approved technique. Therefore, it is imperative that all MR examinations in cirrhotic patients meet these minimum technical requirements to ensure appropriate patient management in a timely fashion.
| Feature | Specification |
|---|---|
| Field strength | 1.5 T or greater |
| Coil type | Phased-array multichannel torso coil |
| Sequences | Nonenhanced and enhanced dynamic gadolinium-enhanced T1-weighted GRE (three-dimensional preferable), T2-weighted (with and without fat saturation), T1-weighted in phase and opposed phase |
| Injector | Dual-chamber power injector |
| Contrast agent injection rate | For extracellular gadolinium chelate without dominant biliary excretion, 2–3 mL/s |
| Mandatory dynamic phases of CE-MR imaging | Nonenhanced T1 weighted, late arterial phase, portal venous phase, delayed phase |
| Dynamic phase timing | Use of bolus-tracking method for timing contrast agent arrival for late arterial phase imaging is preferable; portal venous phase (35–55 s after initiation of late arterial phase imaging); delayed phase (120–180 s after initial contrast injection) |
| Section thickness | For dynamic series, 5 mm or less; for other imaging, 8 mm or less |
| Breath holding | Maximum length of series requiring breath hold should be about 20 s, with minimum matrix 128 × 256 |
Cirrhosis
Regardless of the underlying cause, cirrhosis is the single most important risk factor for HCC. Cirrhosis results from chronic hepatic inflammation and is characterized by the replacement of the normal hepatic architecture by a mixture of parenchymal nodules and fibrosis. The MR imaging findings of cirrhosis mimic these histologic changes and include altered hepatic morphology, fibrosis, and cirrhotic nodules. Cirrhotic nodules are particularly important because they represent the early and intermediate stages in the stepwise progression of carcinogenesis leading to HCC.
Findings of Cirrhosis
Although 25% of cirrhotic patients show normal hepatic morphology on imaging examinations, most patients show at least one of several characteristic findings ( Fig. 1 ). Early signs of chronic liver disease and cirrhosis include enlargement of the hilar periportal space, a sharp posterior hepatic notch, and expansion of the gallbladder fossa. Enlargement of the hilar periportal space occurs because of medial segment atrophy and is defined as greater than 10 mm between the anterior aspect of the right portal vein and the posterior aspect of the medial segment left lobe (see Fig. 1 A). This sign is moderately sensitive (60%–93%) and specific (69%–92%) for cirrhosis. Expansion of the gallbladder fossa also occurs because of medial segment atrophy and is defined as a single axial image showing the gallbladder fossa bound laterally by the right hepatic lobe and medially by the lateral segment with no medial segment visible (see Fig. 1 B). An expanded gallbladder fossa is moderately sensitive (37%–68%) but highly specific (80%–98%) for cirrhosis. The posterior hepatic notch sign occurs because of posterior segment atrophy and appears as a sharp concave indentation along the medial posterior hepatic margin (see Fig. 1 C). The posterior notch sign is also moderately sensitive (47%–72%) but highly specific (83%–98%) for cirrhosis.
With time, obvious lobar and/or segmental atrophy occurs. In up to 35% of patients this results in diffuse hepatic atrophy. Approximately 40% of patients show a combination of atrophy in some areas with hypertrophy in others. Most commonly this takes the form of right lobe and medial segment atrophy with either lateral segment (viral hepatitis) or caudate lobe (alcoholic cirrhosis, PSC) hypertrophy (see Fig. 1 D). Therefore, an increased ratio of the caudate lobe width to the right lobe width measured in the transverse plane can be used to detect cirrhosis (ratio >0.90 using the right portal vein as the lateral caudate margin or >0.55–0.65 using the main portal vein).
The margin of a cirrhotic liver may appear smooth, nodular, or lobulated. In general, a smooth or finely nodular margin is more frequently seen in micronodular alcoholic cirrhosis in which the liver is replaced by many regenerative nodules (RNs) smaller than 3 mm. In contrast, macronodular cirrhosis caused by viral hepatitis features RNs larger than 3 mm, leading to a coarsely nodular margin. However, the hepatic margin appearance does not reliably diagnose the underlying cause of cirrhosis because of overlap between diseases.
Macroscopic hepatic fibrosis is classically described as a reticular or lacelike network of hypointense signal relative to the background liver on T1-weighted (T1W) images with corresponding moderately hyperintense signal on T2-weighted (T2W) images ( Fig. 2 ). In addition, large extravascular spaces within macroscopic fibrotic septa retain contrast, leading to progressive enhancement in the venous and delayed phases. Novel noninvasive methods for the identification and quantification of hepatic fibrosis have shown some promise for the diagnosis of early fibrosis, potentially at a reversible stage. Among the most promising techniques is hepatic MR elastography, which is discussed by Venkatesh and colleagues elsewhere in this issue.
RNs
RNs are benign and form during the normal response to a wide variety of liver injuries or altered circulation. On histology, RNs are composed of functioning hepatocytes resembling those of the background liver, organized around at least one portal tract. Adjacent RNs are separated by fibrous septa. Like normal hepatocytes, RNs are primarily supplied by the portal vein.
Because of their histologic similarity to normal hepatic parenchyma, typical RNs appear similar to background parenchyma on MR. RNs are typically isointense to the background liver on T1W images and isointense to hypointense on T2W images, primarily being visible where they are outlined by fibrosis. Because their vascular supply mirrors that of normal parenchyma, RNs are usually isointense on all postcontrast phases using extracellular contrast agents. Typical RNs are isointense on diffusion-weighted images, take up hepatocyte-specific contrast agents, and are isointense to hyperintense during the hepatobiliary phase.
RNs may show a variety of atypical imaging features, most commonly hyperintense T1W signal possibly caused by lipid, protein, or copper accumulation. Rare infarcted RNs may show hyperintense T2W signal. Nodules appearing hypointense on both T1Wand T2W sequences typically contain iron and are termed siderotic nodules ( Fig. 3 ). On histology, these nodules may be either regenerative or dysplastic; however, siderotic RNs and siderotic dysplastic nodules cannot be differentiated by imaging. The presence of siderotic nodules does not clearly increase the risk for HCC.
Dysplastic Nodules
Dysplastic nodules are detected by the pathologist on gross examination, based on identifying features that are dissimilar from the background pattern of nodules. The distinction between low-grade and high-grade dysplastic nodules is based on microscopic findings. Thus, dysplastic nodules consist of a spectrum of premalignant histopathologic abnormalities arising during the intermediate steps of hepatocarcinogenesis. They are seen in 15% to 20% of cirrhotic livers at imaging, but are often more common in pathologic specimens. Dysplastic nodules are histologically classified as either low-grade dysplastic nodules (LGDNs) or high-grade dysplastic nodules (HGDNs) depending on the level of cellular and structural atypia. LGDNs resemble RNs histologically and slowly progress to HCC (<3% at 1 year after diagnosis). In contrast, HGDNs resemble well-differentiated HCC and progress at a substantially higher rate (up to 46% at 1 year). Therefore, identification of HGDNs has significant prognostic implications. Some investigators even advocate treatment of HGDNs.
A few LGDNs and HGDNs, as well as some early HCC nodules, are hyperintense relative to the surrounding liver on T1W images but can be differentiated by their signal on T2W images. Dysplastic nodules usually have low signal intensity on T2W images, whereas early HCC is typically isointense or hyperintense signal. In addition, the vascular supply to LGDNs predominately arises from the portal vein with minor contributions from the hepatic artery, similar to normal hepatocytes and RNs. As a result, classic LGDNs enhance equally to the surrounding liver in the arterial, venous, and delayed phases. Classic LGDNs are isointense on DWI and take up hepatocyte-specific contrast agents, appearing isointense or hyperintense in the hepatobiliary phase. These nodules do not pose a diagnostic dilemma, because the lack of abnormality on postcontrast images clearly differentiates them from HCC.
As cellular atypia progresses toward malignancy, the normal portal venous supply diminishes, replaced by small unpaired arteries from neoangiogenesis. As a result, some LGDNs and many HGDNs show altered enhancement patterns including (1) isolated arterial phase hyperenhancement fading to isointensity in the venous and delayed phases and (2) arterial phase isointensity with isolated venous or delayed phase hypointensity. Interpretation and management strategies for these nodules are discussed later.
Hepatocellular carcinoma
HCCs are thought to arise from a stepwise progression of hepatocarcinogenesis, beginning with RNs, progressing to LGDNs and HGDNs, before becoming frankly malignant. However, from a radiological perspective, many HCCs arise de novo. The MR findings of HCC are discussed later. Arterial phase hyperenhancement, washout, and capsular enhancement are discussed together, because these 3 findings play a central role in liver transplant eligibility. Under the latest OPTN/UNOS liver allocation policy, a combination of tumor size, arterial phase hyperenhancement, washout, capsular enhancement, or threshold growth over a 6-month interval may be used to diagnose HCC by imaging alone , thereby increasing the patient’s transplant waiting list priority without the need for biopsy. The other imaging findings are often helpful in diagnosing HCC, but cannot replace biopsy under the current organ allocation system.
The dynamic contrast enhancement patterns described later were initially established using gadolinium agents that distribute into the intravascular and extracellular fluid space of the body, the so-called the extracellular (EC) agents (discussed elsewhere in this issue). Additional experience has been obtained using hepatobiliary agents and is discussed later.
Key imaging features of HCC
Arterial Phase Hyperenhancement
Arterial phase hyperenhancement relative to the background liver parenchyma is the single most important imaging finding of HCC ( Figs. 4 A and 5 A ). The progression from low-grade cellular dysplasia to high-grade dysplasia and eventually malignancy is accompanied by a shift in tumor supply from predominately portal venous to predominately from small arterial branches recruited during neoangiogenesis. This shift causes the tumor to hyperenhance markedly relative to the background liver in the arterial phase, with peak conspicuity around 30 seconds after injection. The tight correlation between the severity of cellular dysplasia and shift in tumor blood supply makes arterial phase hyperenhancement highly sensitive for HCC (sensitivity 82%–93%). Sensitivity may be lower (31%–69%) in smaller tumors less than 20 mm.
Other lesions, including hemangiomas, cholangiocarcinomas, metastases, and pseudolesions, can also hyperenhance in the arterial phase. Thus, arterial phase hyperenhancement by itself has low specificity and PPV for HCC, approximately 64% and 67%, respectively. However, the combination of arterial phase hyperenhancement and washout on later phases is extremely specific for HCC (>96%), thus forming the basis for the OPTN/UNOS guidelines.
Image acquisition in the late arterial phase is critically important in order to maximize visualization of arterial phase hyperenhancement. Optimal timing can be difficult because of significant variability in the time from start of contrast injection to the late arterial phase, especially in cirrhotic patients with altered fluid status and cardiovascular hemodynamics. However, several techniques, including test bolus and bolus tracking, help optimize timing of image acquisition. Furthermore, subtraction techniques can increase the conspicuity of arterial phase hyperenhancement, and are particularly helpful if the lesion has hyperintense signal on precontrast T1W images. Arterial phase hyperenhancement may be homogeneous or heterogeneous, particularly in tumors greater than 1.5 cm.
Washout
Washout refers to tumor hypointensity relative to the surrounding liver parenchyma during the venous or delayed postcontrast phases (see Figs. 4 B and 5 B). Because the tumor is supplied primarily by neovascular arteries and the liver is supplied primarily by the portal vein, the degree of tumor enhancement during the venous and delayed phases is substantially less than the surrounding parenchyma. Washout is a subjective observation. Efforts have recently been made to quantitatively define washout but no universally agreed on criteria yet exist. Washout shows high specificity for HCC, particularly in tumors greater than 20 mm (specificity 80%–100%). The specificity decreases to 62% to 100% in tumors less than 20 mm, perhaps because of interobserver variability. As previously discussed, the combination of arterial phase hyperenhancement plus venous or delayed phase washout yields a specificity greater than 96%, even in small tumors less than 20 mm.
Regardless of tumor size, the overall sensitivity of washout is only moderate, with up to 53% of tumors appearing isointense or hyperintense in the venous and delayed phases. Sensitivity improves in the delayed phase compared with the venous phase. Tumor washout is associated with histologic grade, because poorly differentiated tumors show greater hypointensity than well-differentiated tumors.
Capsular Enhancement
Capsular enhancement is defined as a peripheral hyperenhancing rim in the venous and/or delayed phases (see Fig. 5 B). The enhancing capsule seen on imaging may represent a true fibrous capsule or a pseudocapsule composed of mixed fibrous tissue and prominent sinusoids. The capsule shows typical characteristics of fibrotic tissue on MR imaging, with low signal on T1W and T2W sequences, first enhancing in the venous phase and increasing in intensity from the venous to the delayed phases. Capsular enhancement shows high specificity for HCC in cirrhotic patients, reportedly ranging from 83% to 96%. Sensitivity is only moderate, ranging from 43% to 55%. Most studies have found that encapsulated tumors have a better prognosis than nonencapsulated tumors, because encapsulation is associated with better cellular differentiation, lower rates of portal venous invasion, and higher survival rates including better outcomes after resection and transarterial chemoembolization.
Other findings of HCC
T1W Signal
HCC can have a variety of appearances on unenhanced T1W images, but is most commonly hypointense relative to the surrounding liver (sensitivity 21%–91%, specificity 70%–100%, PPV 49%–100%, NPV 20%–78% ). Increased lipid, glycogen, copper, melanin, protein, and hemorrhage within a tumor may cause T1W signal hyperintensity. Background parenchymal steatosis, fibrosis, iron deposition, or zinc accumulation may alternatively decrease signal around a tumor, leading to relative tumor isointensity or hyperintensity. Hyperintense T1W signal is more common in well-differentiated tumors. Poorly differentiated tumors are usually hypointense on T1W imaging.
Fatty Metamorphosis
Fatty components accumulated during hepatocarcinogenesis contribute to some HCCs appearing isointense or hyperintense relative to the surrounding liver on T1W images. Fat is more commonly present as intracellular lipid, although macroscopic fat can also occur. Using opposed-phase/in-phase gradient-recalled echo (GRE) technique, intracellular lipid can be visualized as intralesional signal loss on opposed-phase images relative to in-phase images ( Fig. 6 A, B). Macroscopic fat can be identified as signal loss from non–fat-suppressed to fat-suppressed T1W images. Most HCCs do not contain significant amounts of fat, and therefore the sensitivity for detecting HCC is low (12%–37%). In contrast, fat content is moderately specific for HCC (68%–100%) because other lesions containing intracellular lipid (adenomas, focal steatosis, atypical focal nodular hyperplasia, variant RNs) and macroscopic fat (adenoma, lipoma, angiomyolipoma, liposarcoma metastases) are uncommon in cirrhotic livers. For these reasons, any fat-containing tumor in a cirrhotic liver should be viewed with suspicion. Fat-containing masses larger than 1.5 cm are especially suspicious, because the combination of size, fat content, and hypointense signal on T1W in-phase imaging has been shown to have an 85% PPV for HCC.
Fatty metamorphosis occurs most commonly in well-differentiated tumors and is associated with improved outcomes compared with non–fat-containing tumors. In a recent study by Siripongsakun and colleagues, patients with fat-containing HCC showed lower rates of primary tumor progression, lower rates of distant metastases, and longer times to progression than controls with non–fat-containing tumors of similar stage who received similar treatment.
T2W Signal
HCC classically appears mild to moderately hyperintense relative to the background liver on T2W images (see Fig. 5 C), unlike cysts and hemangiomas, which are typically markedly hyperintense, and dysplastic nodules, which are almost always hypointense. As a result, moderately hyperintense T2W signal has high specificity (73%–100%) and PPV 72% to 100% for HCC. Note that other tumors in the cirrhotic liver, such as cholangiocarcinoma, may also display moderate T2W hyperintensity ( Fig. 7 A). However, the sensitivity of moderately hyperintense T2W signal tends to be low in cirrhotic patients (21%–75%) because of a variety of factors including fibrosis causing background liver heterogeneity and respiratory motion secondary to ascites. More well-differentiated tumors may be isointense or hypointense on T2W images, further decreasing sensitivity. Adding T2W images to dynamic gadolinium-enhanced fat-suppressed T1W images does not improve sensitivity compared with dynamic enhanced T1W images alone, leading some investigators to propose omitting T2W sequences from routine liver MR protocols in cirrhotic patients.
DWI
The DWI technique is based on proton mobility in tissue, primarily water-based protons. Compared with normal tissue, the extracellular space in tissues with high cellular density (ie, tumors) is compressed and tortuous, resulting in impeded or restricted water diffusion and relative signal hyperintensity on DWI ( Fig. 8 B).
In the past few years, numerous studies have shown the value of DWI sequences for qualitative HCC detection (sensitivity 14%–95%, specificity 83%–100% ), especially for small tumors less than 20 mm (sensitivity 57%–94%, specificity 87%–88%). A recent meta-analysis by Wu and colleagues found that DWI combined with conventional dynamic CE-MR performed significantly better (pooled sensitivity 93%, pooled specificity 84%) than either DWI alone or conventional dynamic CE-MR imaging alone. Several studies published after this meta-analysis have shown similar results.
In combination with hepatocyte-specific contrast agents, DWI is particularly useful for characterizing nodules that show atypical enhancement patterns on conventional dynamic enhanced MR imaging. This application of DWI is discussed in detail later.
DWI has also shown promising results as a method to predict microvascular invasion by HCC, which is also discussed in more detail later.
False-negative results with DWI may occur for several reasons. First, DWI signal increases in correlation with tumor grade; therefore, well-differentiated tumors may be isointense. Second, fibrosis in cirrhosis results in impeded water molecule motion, increasing DWI signal in the background liver and leading to decreased tissue contrast between tumor and liver. Third, hyperintense signal on DWI is not specific for HCC and can be seen with other tumors, such as intrahepatic cholangiocarcinoma.
Hepatocyte-specific Contrast Agents
Numerous recent studies have yielded promising results with the use of hepatocyte-specific contrast agents (HCAs) for detecting and diagnosing HCC. Two HCAs are currently approved for use in the United States: gadoxetate disodium (Bayer Healthcare, Wayne, NJ) and gadobenate dimeglumine (Bracco Diagnostics, Princeton, NJ). After injection, both drugs initially distribute to the extracellular space like non–liver-specific contrast agents, allowing typical dynamic arterial and venous postcontrast images to be obtained (see Fig. 6 C, D). The drugs are then taken up by hepatocytes and excreted into bile canaliculi, creating a hepatobiliary phase in which portions of the liver with functioning hepatocytes and bile ducts enhance, whereas areas lacking function are hypointense. Most HCCs do not contain functioning hepatocytes and bile ducts, resulting in hypointense signal relative to the surrounding liver in the hepatobiliary phase (see Fig. 6 E). As a result, hepatobiliary phase images are highly sensitive for HCC (79%–100%); however, specificity is poor (33%–92%) because a large number of other lesions are hypointense, including common benign lesions such as cysts and hemangiomas, and other malignant tumors such as cholangiocarcinoma ,168 .
The combination of CE-MR imaging and hepatobiliary phase imaging is accurate for the diagnosis of HCC, with recent meta-analyses showing pooled sensitivity of 91% and specificity of 93%. Performance did not change significantly when only tumors less than 20 mm were considered. These results compare favorably with CE-MR imaging alone (sensitivity 14%–82%, specificity 96%–100%) and MDCT (sensitivity 40%–78%, specificity 93%–99%). When characterizing an arterially hypervascular nodule using HCAs, it is important to distinguish between the washout feature that occurs in the venous phase, and the lack of enhancement with HCAs in the hepatobiliary phase (see Fig. 6 D, E). Although both of these features manifest as hypointensity of the lesion relative to liver parenchyma in their respective phases, they cannot be used interchangeably because they reflect different properties of the tumor. Washout reflects the extracellular spaces in the tumor relative to adjacent liver and is highly specific for HCC. Lack of gadoxetate uptake reflects the cellular properties of the tumor and is not as specific for HCC because this is seen with other masses. The use of the delayed (third) postcontrast phase of gadoxetate-enhanced MR imaging to assess for the washout feature is controversial because enhancement in this phase is often contaminated by hepatocyte uptake, which can start 90 seconds or earlier after contrast injection.
Nodules showing hypointense signal in the hepatobiliary phase but lacking diagnostic features of HCC in the earlier postcontrast phases are frequently encountered when using HCAs. These nodules may represent HGDNs or early HCC and they place patients at increased risk of progression to conventional hypervascular HCC. Such nodules are best evaluated using a combination of CE-MR imaging and unenhanced sequences including T1W opposed-phase/in-phase, T2W, and DWI as discussed later.
Several pitfalls specific to HCAs may lead to false-negatives. Tumor visualization in the hepatobiliary phase relies on differential contrast uptake between functioning hepatocytes and nonfunctional tumor cells. Patients with significant liver dysfunction may have decreased HCA uptake, reducing tissue contrast between tumor and surrounding liver, which is particularly problematic for small tumors. Up to 8% of HCCs are isointense or hyperintense relative to background liver in the hepatobiliary phase because of accumulation of contrast within the tumor, which is most commonly associated with well-differentiated HCC containing a pseudoglandular cellular architecture, but occasionally occurs in moderate or poorly differentiated HCC as well. In addition, gadoxetate may cause dyspnea after injection leading to degraded arterial phase images.
Mosaic Pattern
The mosaic pattern of HCC describes an encapsulated cluster of tumor nodules separated by fibrous septa, necrosis, fat, and/or hemorrhage. Because of their varied components, mosaic tumors are typically very heterogeneous on T1W and T2W sequences with multifocal arterial phase hyperenhancement and washout. Tumors with this appearance tend to be large (>3 cm) and generally do not cause a diagnostic dilemma.
Nodule Within a Nodule Appearance
The nodule within a nodule appearance describes an HCC arising within a dysplastic or siderotic nodule. Typical MR features include a T1W hyperintense, T2W hypointense dysplastic nodule containing a smaller T1W hypointense, mildly to moderately T2W hyperintense HCC. In the case of siderotic nodules the background nodule is hypointense on all sequences. After contrast administration, the dysplastic/siderotic nodule is isointense or hypointense in the arterial phase, whereas the HCC avidly hyperenhances. HCC arising as a nodule within a nodule may grow rapidly.
Vascular Invasion
Vascular invasion by HCC may be classified as macroscopic (macrovascular) or microscopic (microvascular). Macrovascular invasion occurs in 5% to 40% of cases, resulting in grossly visible tumor thrombus in the portal and/or hepatic veins; the incidence is higher in autopsy series. Tumor thrombus indicates advanced disease and patients are generally not candidates for transplant or chemoembolization (radioembolization may be still be possible). Bland portal vein thrombus is also common in cirrhosis (2%–26%). Although bland thrombus is not a contraindication for transplantation, studies have shown that patients with bland thrombus have a worse outcome after transplant than patients without thrombus. Therefore, identification of thrombus type is critical to optimizing patient management.
MR is excellent at depicting portal vein thrombosis. Several features help differentiate tumor and bland thrombus. Portal vein expansion is particularly helpful because expansion beyond 23 mm has moderate sensitivity (62%) and high specificity (100%) for tumor thrombus ( Fig. 9 ). When a parenchymal tumor is also present, tumor thrombus is typically contiguous with the parenchymal mass, whereas bland thrombus is often separate. Tumor thrombus signal characteristics mirror those of the parenchymal mass, showing neovascularity or arterial phase hyperenhancement, washout, hyperintense T2W signal, and hyperintensity on DWI sequences (see Figs. 8 and 9 ). In contrast, bland thrombus rarely enhances in the arterial phase, does not wash out in later phases, and is typically less hyperintense than the parenchymal tumor on T2W and DWI sequences. Acute thrombus can show increased T2W and DWI signal mimicking tumor thrombus. Therefore, in the absence of thrombus enhancement or marked portal vein expansion, these findings should be interpreted with caution.
