Early diagnosis with subsequent transplantation before the development of large or infiltrative tumors offers the most effective treatment for hepatocellular carcinoma, because it can cure both the cancer and underlying cirrhosis. In one study, patients transplanted with either one lesion measuring 5 cm or less without vascular invasion, or three lesions measuring 3 cm or less, had a four-year survival rate of 92%.1
As cadaveric organs are scarce, it is critical to detect nodules that contain HCC at an early size both to control tumor burden with nonsurgical techniques while awaiting transplantation and to stratify which patients might have favorable long-term results with transplantation.
One of the most confusing aspects of the MRI literature on cirrhotic nodules has been the terminology. Until recently, most lesions had more than one name, which made it difficult to compare radiologic and pathologic studies performed at different centers. The need to standardize the nomenclature of hepatocellular nodules was recognized in 1994 by the World Congress of Gastroenterology, where an "international language" was proposed that greatly simplified the interpretation of the radiologic and pathologic literature. Ordinary cirrhotic nodules are termed regenerative (RN). The term dysplastic nodule (DN) replaces older terms like adenomatous hyperplasia and macroregenerative nodule.
DNs are now considered neoplastic, clonal lesions that represent an intermediate step in the pathway of hepatocarcinogenesis in cirrhotic livers.2,3
Above: Axial source image from breath-hold 3-D fat-suppressed gradient-echo pulse sequence (TR 4.7/TE 1.8/14¡ flip angle) demonstrates small HCC (curved arrow) with homogeneous enhancement.
Below: The mass was not seen on any other T1-weighted sequence but was visualized on fat-suppressed T2-weighted HASTE image (TR infinite/ TE effective 62) (curved arrow).
No single MRI technique is optimal for imaging the cirrhotic liver, and individual preferences are usually based on the hardware and software available. We strongly encourage use of a body phased-array coil for the best signal-to-noise ratio. We prefer a 1 to 1.5-tesla magnet with a gradient rise time of at least 600 msec.
T2-weighted (or STIR, for short tau-inversion recovery) images are an essential component of the MR exam. While conventional spin-echo pulse sequences provide excellent image contrast, long acquisition times may result in substantial motion artifact. Many centers have abandoned this technique for faster echo-train fast spin-echo pulse sequences.4 These may be performed with respiratory gating or respiratory ordered phase-encoding, depending on the manufacturer.
By decreasing the repetition time (TR) and using a relatively long echo-train, T2-weighted images can be acquired in the time frame of a single breath-hold. Frequency-selective fat suppression may be used to augment image contrast, which is especially helpful if the liver is fatty. Alternatively, fat-suppressed images with additive T1 and T2 contrast can be performed with STIR methods when inversion times are selected to null fat. This technique is less dependent on a homogeneous magnetic field, and it results in fewer artifacts than seen with frequency-selective fat suppression techniques.
New developments that provide for long echo-train length, short echo spacing, and half-Fourier computations provide for single-shot imaging that can be performed in patients unable to suspend respiration. However, these methods may result in a decrease in T2 contrast for detecting solid lesions such as HCC.4,5
Breath-hold, T1-weighted images should be performed using a short TR/short TE gradient-echo pulse sequence with a large flip angle. The authors acquire these at two separate echo times for in-phase and out-of-phase imaging (4.4 msec in-phase and 2.2 msec out-of-phase at 1.5 tesla). This improves the detection of either diffuse or focal fatty infiltration and nodular lesions that contain fatty components.
Imaging at short echo times has both benefits and disadvantages. It minimizes the magnetic susceptibility artifacts from transjugular intrahepatic portosystemic shunts (TIPS) and certain embolization coils, which can severely degrade image quality. However, it also decreases the ability to detect susceptibility effects of iron within the liver, whether diffuse or within siderotic nodules. A third gradient-echo pulse sequence with increased echo time and decreased flip angle can be used to increase the magnetic susceptibility effects of iron. This sequence can also be used to determine direction of portal venous flow when saturation bands are placed above and then below the imaging volume.
For MRI of the cirrhotic liver, dynamic, breath-hold, gadolinium-enhanced imaging performed with either two- or three-dimensional gradient-echo pulse sequences is essential.6-8 Contrast-enhanced, frequency-selective fat-suppressed sequences can be used to improve conspicuity of liver lesions, particularly in the setting of fatty infiltration. Fat suppression is also helpful when evaluating the extrahepatic manifestations of cirrhosis, including varices and bowel edema. Other contrast agents are available, including agents that are specifically taken up by hepatocytes or by reticuloendothelial cells. A discussion of these agents is beyond the scope of this article, and the reader is referred to a recent review.9
MR images reveal degenerative nodules and small HCC. |
Above: Equilibrium phase from breath-hold gadolinium-enhanced T1-weighted image demonstrates regenerative nodules outlined by the enhancing septations of cirrhosis. Short TR/short TE gradient-echo with large flip angle is recommended.
Below: Small HCC with a thick enhancing capsule is characterized in another contrast-enhanced T1-weighted image.
Contrast-enhanced images should be acquired in at least three phases: the hepatic arterial, portal venous, and equilibrium phases. A successful arterial phase is critical for the detection of small HCC.6,7 Traditionally, hepatic arterial phase images have been obtained following a fixed delay after intravenous bolus of contrast. However, use of a fixed delay often results in a suboptimal hepatic arterial phase in the cirrhotic cohort because of the wide range in circulation times that result from hemodynamic alterations intrinsic to the disease. In these patients, recent developments suggest that using a test bolus to determine an individual's circulation time, or fluoroscopic "real-time" triggering, can help obtain arterial phase imaging more reliably. In addition, an MR-compatible power injector is helpful for injections of the test bolus and main bolus of contrast material, as it provides a precise infusion rate (typically, 2 mL/sec) and dose (0.1 mmol/kg) and allows a technician to perform the entire examination outside of the magnet.
When near-isotropic pixel size (such as lesser than or equal to 2 mm in all three dimensions) can be achieved, 3-D gradient-echo imaging has clear advantages over conventional 2-D. Image sets can be reformatted using multiplanar reconstructions without loss of in-plane resolution. Moreover, angiographic reconstructions such as maximum intensity projections can be obtained for each contrast-enhanced acquisition, resulting in a "free" angiogram and portogram.8
The equilibrium-phase, contrast-enhanced study can be helpful for distinguishing HCC from RN and DN, as some HCC will have a discernable capsule at this stage while RN and DN should not.
RN of cirrhosis have characteristic features on MR imaging that usually allow distinction from HCC, but not always from DN. They invariably have a portal venous blood supply with minimal or no contribution from the hepatic artery. RN are usually isointense with other background nodules on both T1-weighted and T2-weighted images. Less commonly, they may be hyperintense on T1-weighted images and hypointense on T2-weighted images. However, unlike some HCC, RN are almost never hyperintense on T2-weighted images, with the noted exception of those that occur in the setting of chronic Budd-Chiari syndrome10 or those that have undergone infarction.
DN are neoplastic, premalignant nodules2,3 and are found in 15% to 25% of cirrhotic livers.3 Longitudinal studies have documented the development of HCC within a DN in as little as four months.11 On gross pathologic examination, DN can usually be distinguished by color, texture, or the degree to which they bulge from the cut surface of the liver.5 DNs are divided into low and high-grade nodules.
The signal intensity characteristics of DN overlap substantially with small HCC.12 A common pattern is homogeneously hyperintense on T1-weighted images and hypointense on T2-weighted images.12 However, this pattern is seen with HCC as well.12,13 Like RN, DN can be siderotic and, therefore, hypointense on both T1- and T2-weighted images. One very helpful distinction between HCC and DN is that the latter are almost never hyperintense on T2-weighted images.
Because DN and HCC may be hyperintense on T1-weighted images, it may be difficult to discern if they enhance in the arterial phase. By subtracting the precontrast image from the arterial phase image, the presence and degree of enhancement can be visualized as increased signal intensity on the subtracted image. The same technique, when performed with multiple phases, can be used to assess for tumor necrosis in the response to percutaneous or interventional therapy for HCC.
HCC may arise within a high-grade DN. The classic MR appearance of these lesions is a "nodule within nodule," consisting of a high signal intensity focus within a low signal intensity nodule on T2-weighted images.13 The central nodule of high signal intensity may also demonstrate enhancement during the hepatic arterial phase and represents the focus of HCC.
When a patient with cirrhosis presents with a large, infiltrative tumor with vascular invasion and arterioportal shunting, the diagnosis of HCC can be made with certainty by almost any imaging modality. Unfortunately, these patients are incurable.
Small HCCs demonstrate variable patterns of signal intensity on T1- and T2-weighted images. However, the hepatic arterial blood supply of most HCCs, except for well-differentiated lesions, which retain some portal venous supply, can facilitate diagnosis. Thus, not only are dynamic gadolinium-enhanced images essential for the detection of small HCCs,6,7 but they also may help to distinguish them from RN and most DN. The most common enhancement pattern of HCC lesser than or equal to 1.5 cm in North America is diffuse, homogeneous enhancement during the hepatic arterial phase, with rapid washout during the portal venous phase.13 In contrast, hemangiomas should not wash out but remain isointense to the hepatic vasculature for multiple phases.
Larger HCCs tend to display a heterogeneous or mosaic pattern of enhancement in the hepatic arterial phase and the presence of a capsule in later phases. The former pattern should be distinguished from the transient hepatic intensity difference (THID), which is often identified as a focal peripheral wedge of increased intensity seen only on hepatic arterial phase imaging.14 However, these lesions are almost always subcapsular and do not cause bulging of the capsule. They are usually due to arterioportal shunts or aberrant venous drainage.14
While THIDs may appear nodular on 2-D imaging, multiplanar reformations from 3-D data sets can usually identify their characteristic wedge shape.14 It is important to remember that large THIDs (>2 cm) in a cirrhotic liver may be caused by a number of entities: malignant obstruction of the portal vein, previous percutaneous interventional therapy, biliary disease, a siphon effect from hypervascular HCC, arterioportal shunting from HCC, and even benign lesions such as hemangiomas.15
The detection and characterization of nodular lesions in cirrhotic livers on MRI remain a challenge for radiologists. Further research with meticulous radiologic-pathologic correlation and improved MR techniques that exploit advances in hardware and pulse sequence implementation should help advance our knowledge and enable us to diagnose HCC when it is still potentially curable. But to compete with multidetector CT, high spatial resolution and short examination times (<20 minutes) are necessary.
Dr. Krinsky is an associate professor of radiology at New York University Medical Center in New York City.