Dr. Shi is a third-year resident and Dr. Mitchell is a Professor of Radiology, Department of Radiology, Thomas Jefferson University Hospital, Philadelphia, PA.

Magnetic resonance imaging (MRI) of the liver has evolved over the past 2 decades to become an effective and widely used modality for diagnosing liver disease and evaluating its progress. This article presents the MRI techniques for evaluating the liver and the imaging characteristics of common diffuse liver diseases.

Technique

A comprehensive MR imaging examination of the liver should include T1-weighted images (T1WI), T2-weighted images (T2WI), and dynamic multiphasic contrast-enhanced images, preferably supplemented by chemical-shift images and fat-saturation images. A phased-array torso coil should be used to improve the signal-to-noise ratio (SNR).

Single-shot fast-spin echo (SSFSE) is recommended for coronal localizer images. Each of the images is obtained in <1 second, providing a rapid, motion-independent T2-weighted survey. These sequences not only give an overview of the liver, but also help depict fluid and differentiate solid lesions from cystic lesions or hemangiomas, particularly with the use of a heavily T2-weighted sequence (effective echo time [TEef] >180 msec). We routinely include a set of axial SSFSE heavily T2WI for this purpose.

At high field strength (>= 1 Tesla), gradient-echo sequences are generally used for T1-weighted sequences. Spoiled gradient-echo (SGE) sequences are advantageous; these allow acquisition of sufficient sections in one breath hold (about 20 sec) for complete liver coverage and provide good SNR. For multisection two-dimensional (2D) SGE techniques, repetition time (TR) is typically >100 msec, or as high as breath holding will allow, but always <500 msec to provide heavy T1-weighting. Flip angle is usually 90°, but is reduced if TR is <100 msec. Echo time (TE) is as short as possible. Usually, both in-phase and opposed-phase T1-weighted sequences are obtained, particularly for evaluation of cellular fat content, such as for the evaluation of fatty liver. For optimal in-phase effect, TE should be 4.6 msec, although 4.2 msec is adequate. For optimal fat-water opposed-phase effect, TE should be close to 2.3 msec.

For T2 weighting, echo-train spin- echo sequences are currently used because of the shorter sequence duration. Fat suppression is advantageous because it reduces motion-induced phase and chemical shift artifacts and improves the dynamic range of tissue signal intensities. Short tau inversion recovery (STIR) also may be used to acquire T2WI that are augmented by inverse T1-weighting; these images have high soft-tissue contrast and minimal signal from adipose tissue.

Contrast enhancement

The most commonly used contrast agents are gadolinium chelates, which are extracellular space agents. They are distributed initially in the intravascular compartment and diffuse rapidly throughout the extracellular space, analogous to water-soluble iodinated contrast agents for computed tomography (CT) scanning. A multiphasic dynamic contrast-enhanced series includes breath-hold SGE sequences acquired before the contrast injection, as well as postcontrast sequences in the arterial phase, portal venous phase, and delayed extracellular phase. If possible, fat suppression is applied to ensure the conspicuity of the enhancing tissue and vessels in relation to abdominal fat. Our current preference for dynamic multiphasic gadolinium-enhanced images are three-dimensional (3D) spoiled gradient-echo images using partial Fourier acquisition, segmented fat suppression through selective inversion, and zero-fill interpolation in the encoding select axis. Using this technique, it is possible to obtain fat-suppressed contiguous images at increments of ¾ 2.5 mm through the entire liver, during a single suspended respiration. These images are similar to those obtained for gadolinium-enhanced MR angiography, and, in fact, clearly depict the hepatic and abdominal arteries and veins.

Fatty liver

Fatty liver has many causes, including obesity, diabetes mellitus, malnutrition, or exposure to ethanol or other chemical toxins. In fatty liver, fat accumulates in hepatocytes, either focally or diffusely. On CT or ultrasound, focal fatty change can be difficult to differentiate from a neoplasm. ¹ With MRI, the confidence of diagnosing fatty liver is greatly improved by using chemical shift techniques. The most useful technique involves comparison of in-phase (water + fat) and opposed-phase (water ­ fat) images. ^2,3 Most tissues, including tumors, maintain similar signal intensity on in-phase and opposed-phase T1WI. Fatty liver, on the other hand, will have lower signal intensity on opposed-phase images because of cancellation of water and fat signals within voxels. Fat-suppressed images can also be helpful in cases of moderate to severe fatty infiltration. Fatty liver will have lower signal intensity with fat suppression, although the drop is not as much as on opposed-phase images (figure 1).

One must consider that certain neoplasms (such as some well-differentiated hepatocellular carcinomas [HCCs], adenoma, angiomyolipoma, or lipoma) contain lipid. The differentiation can be achieved by analyzing the morphology and other imaging characteristics of these tumors. Even though some well-differentiated HCCs contain lipid, most HCCs with high T1 signal intensity do not. Hepatocellular carcinomas are usually well defined, and often have capsules. ⁴ Adenomas have low signal intensity on opposed-phase T1WI, similar to fatty infiltration, but have a round shape and a characteristic early arterial phase blush postcontrast. Angiomyo-lipoma and lipoma have gross deposits of fat, which have decreased signal intensity on fat-suppressed images, but less on opposed-phase images. On opposed-phase T1WI, instead, there is cancellation of signal primarily at the edges of these tumors.

In the setting of diffuse fatty infiltration, neoplasms are more difficult to detect by CT or ultrasound. On the contrary, a diffuse fatty liver may help MRI detect a tumor. On non­fat-saturated T1WI, the fatty liver may have higher intensity than a normal liver, which increases the contrast relative to the low signal intensity of tumors. On fat-saturated T2WI, the fatty liver has lower intensity than a normal liver. Again, the contrast between the liver and high intensity tumors increases.

Iron overload

Iron accumulation in liver is seen in hemochromatosis, transfusion overload, hemolytic anemias, and cirrhosis.

Intracellular iron is usually superparamagnetic. It reduces T2 and T2* relaxation times, and results in low signal intensity on both T2WI and T2*-weighted images (T2*WI). This makes MRI a more sensitive and specific method for diagnosing iron overload; gradient recalled echo (GRE) T2*WI have greatest sensitivity. To achieve T2*-weighting, at 1.5 T, 9 msec or more of echo time is preferred. The flip angle should be ¾ 30° to minimize the T1 differences between tissues. TR is usually minimized to reduce acquisition time. Normally, liver has higher signal intensity than skeletal muscle on both T1WI and T2WI. The signal intensity of skeletal muscle serves as a good internal reference, because it does not vary over a wide range of iron overload or change with different causes. ⁵

Hemochromatosis is an autosomal recessive disorder with a homozygous prevalence of 0.25% to 0.50%. ⁶ In hemochromatosis, iron overload results from long-term increased iron absorption through the intestine. Reticuloendothelial (RE) function may be deficient. Iron accumulates in liver, pancreas, and heart, as well as the parenchyma of other organs. In liver, the iron is incorporated preferentially into the hepatocytes rather than Kupffer cells of the RE system. Hemochromatosis causes morbidity and mortality by means of cirrhosis, increased incidence of hepatocellular carcinoma, diabetes mellitus, and myocardial dysfunction. Diagnosis of hemochromatosis is difficult because of the nonspecific symptoms of hepatic, myocardial, and pancreatic dysfunction. If diagnosed early, normal life expectancy can be achieved with phlebotomy treatment. Magnetic resonance imaging detects low signal intensity on GRE T2*WI as evidence of excessive hepatic iron in preclinical hemochromatosis without cirrhosis, ⁷ when treatment is crucial before irreversible damage occurs. As the disease progresses, low T2* signal intensity can be seen in the pancreas and heart, but normal signal of the spleen is preserved (figure 2). Magnetic resonance imaging can be used longitudinally to monitor disease progress and to evaluate response to treatment. Liver iron concentration can be measured by calculating T2 relaxation time or intensity ratios. ⁸

In transfusional iron overload , iron deposits in RE cells of the liver, spleen, and bone marrow. In contradistinction to hemochromatosis, iron overload in RE is of little clinical significance. Magnetic resonance imaging shows low signal intensity of liver, spleen, and bone marrow on GRE T2*WI, sparing the heart and pancreas (figure 3). ⁹ This finding, as well as a clinical history of multiple transfusion, helps to differentiate transfusion overload from primary hemochromatosis. Iron deposition occurs in hepatocytes and pancreas only when RE cells are saturated (if a patient has received >100 units of blood).

Hemolytic anemia is also a cause
of hepatocellular iron overload. Thalassemia major patients have ineffective erythropoiesis, which triggers increased absorption and decreased excretion of oral iron, similar to that which occurs in genetic hemochromatosis. Without a history of transfusion, the pattern of iron overload is indistinguishable from hemochromatosis. Patients often develop coexisting transfusion iron overload with transfusion. In sickle cell anemia, patients have rapid hepatic iron turnover, and will not necessarily have hepatic low signal intensity, unless they recently received blood transfusion. Due to renal filtration and tubular absorption of free hemoglobin, they may have renal cortical low signal, which is independent of transfusion. Reduced signal intensity of the renal cortex is a characteristic finding in patients with intravascular hemolysis from other causes as well, such as paroxysmal nocturnal hemoglobinuria. ¹⁰

Cirrhotic liver from etiologies other than primary hemochromatosis also accumulates iron in mild degree. In some cirrhotic patients, iron is concentrated within some large regenerative nodules. ¹¹ Dominant iron-containing nodules are at risk for HCC. Magnetic resonance imaging serves as the imaging modality of choice in detecting early HCC in these regenerating nodules, since these foci of HCC do not contain iron. This produces a characteristic finding for a focus of HCC in an iron-containing regenerative nodule, the "nodule in nodule" sign. ¹² Developing high signal intensity within an iron-containing low-signal nodule on GRE T2*WI is highly suggestive of hepatocellular carcinoma. These nodules tend to grow quickly, doubling their diameter in about 2 months. ¹³ Therefore, ablative treatment or rapid follow-up is indicated.

Cirrhosis

Cirrhosis is a pathologically defined entity characterized by irreversible chronic injury of the hepatic parenchyma, extensive fibrosis, and regenerative nodules. ¹⁴ In North America, the leading cause of cirrhosis has been alcohol abuse, followed by chronic viral hepatitis, hemochromatosis, autoimmune chronic hepatitis, primary biliary sclerosis, drug toxicity, etc. However, the incidence of hepatitis C has increased dramatically in recent years. ¹⁵

Magnetic resonance imaging provides comprehensive evaluation of a cirrhotic liver, including diagnosis, assessment of disease progression, as well as detection of dysplastic nodules and hepatocellular carcinoma. The findings in cirrhosis are liver surface nodularity, regenerative nodules, change of hepatic morphology, fatty changes, and iron deposition, as well as ascites, splenomegaly, varices, and collaterals. These signs are usually seen in advanced cirrhosis.

The natural course of cirrhosis progresses from compensated stage (Child's class A) to decompensated stage (Child's class B or C), then to end-stage cirrhosis. Patients with compensated cirrhosis can develop features of decompensation at a rate of 10% per year; the survival rate is lower in patients with decompensated cirrhosis. In cases of compensated cirrhosis, hypertrophy of the lateral segment and caudate lobe are predominant findings on longitudinal MRI exam. Atrophy of the medial segment and the right lobe are the predominant findings in cases of progressive cirrhosis. ¹⁶ These MRI findings help predict a patient's clinical outcome. Although the cause of the segmental morphologic changes is unclear, it is thought to be related to altered blood flow in cirrhotic liver.

Many findings of early cirrhosis are caused primarily by preferential atrophy of the medial segment of the left lobe. As this segment atrophies, enlargement of the hilar portal space occurs. This has been described as an early MRI sign of cirrhosis ¹⁷ (figure 4). It has been reported that 98% of patients who had early cirrhosis have enlargement of the hilar periportal space before the presence of the conventional signs of cirrhosis. ¹⁷ Medial segment atrophy also causes an expanded gallbladder fossa, another MR sign of cirrhosis. This finding was reported to have 98% specificity, although the sensitivity was 68%. ¹⁸

Although there is no effective therapy to reverse the changes of cirrhosis, detection of the disease at an early stage is still important. There have been reports of effective interferon therapy, which reduces or prevents progression of hepatic damage, for patients with compensated cirrhosis due to hepatitis C. ¹⁵ Other imaging signs, such as ratio of the transverse diameter of the caudate to right lobe (C/RL > 0.65), have been used to determine cirrhosis. ¹⁹ However, this ratio is not increased in early cirrhotic patients. ¹⁷

Regenerative nodules are regenerated hepatocytes surrounded by fibrotic septa. Cirrhosis can be classified based on the size of the nodules as: micronodular (¾ 3 mm), macronodular (>3 mm) or mixed. Alcoholic cirrhosis is commonly micronodular, and viral-induced cirrhosis is typically macronodular. Regenerative nodules are hypointense in relation to hyperintense inflammatory septa on T2WI and gadolinium-enhanced MR images. ²⁰ They have variable intensity on T1WI and can be iso- or hypointense on T2WI, but are not hyperintense on T2WI, with the exception of those in chronic hepatic vein thrombosis. In contradistinction to HCC, regenerative nodules are not hyperintense on hepatic arterial-phase gadolinium-enhanced images.

Dysplastic nodules are premalignant and are found in 15% to 25% of cirrhotic livers. ²¹ These nodules are commonly hyperintense on T1WI and hypointense or isointense on T2WI. ^22-24 Like regenerative nodules, they are not hyperintense on T2WI. Occasionally, dysplastic nodules may be hyperintense on early post-gadolinium images.

Hepatocellular carcinoma is the most common primary cancer of liver. The 5-year survival rate is <5%, due to late presentation and large tumor burden. Early diagnosis with resection and transplantation can cure both the cancer and the cirrhosis. Magnetic resonance imaging is the most effective diagnostic imaging tool to detect small foci of HCCs, with the sensitivity and specificity ranging from 75% to 100%. ^23-28 Small HCCs have variable signal intensities on both T1WI and T2WI. Nevertheless, most well-differentiated HCCs are hyperintense on TIWI. The presence of moderate T2 hyperintensity is specific for distinguishing HCC from regenerative or dysplastic nodules, and homogeneous enhancement during arterial phase with rapid washout during the portal phase is characteristic ²³ (figure 5). Other features of HCC include presence of a capsule, size >3 cm, and rapid growth.

As cirrhosis progresses, portal venous flow is compromised secondary to liver parenchymal distortion. Portosystemic collaterals frequently develop as a result of portal hypertension. Gadolinium-enhanced axial thin-section MR images can depict the collateral vessels in these patients (figure 6). The vessels around the gastro-esophageal junction are the most visible and are the major cause of gastrointestinal bleed, as well as mortality. Other collateral vessels include collaterals to the left renal vein, paraumbilical vein, and abdominal wall. ²⁹

Budd-Chiari syndrome

Budd-Chiari syndrome results from obstruction of the hepatic venous outflow, causing portal hypertension, ascites, and hepatic failure. Most often there is atrophy of the peripheral liver, and hypertrophy of the caudate lobe and central liver. It is best diagnosed by multiphasic gadolinium-enhanced MR images, which show a thrombosed or absent hepatic vein and altered hepatic enhancement. In acute onset of the Budd-Chiari syndrome, the periphery of the liver has moderately low signal on T1WI and moderately high signal on T2WI relative to the central liver. Severe congestion causes decreased peripheral enhancement during the arterial and portal venous phase. ³⁰ In the subacute setting, the T1 and T2 signal characteristics are similar to those with acute disease. During early and late phases of contrast enhancement, there is heterogeneously increased enhancement at the periphery. In chronic disease, signal differences between the central and peripheral part of the liver on T1WI and T2WI and post-gadolinium contrast images are minimal. ³¹ Varices may be present in chronic Budd-Chiari syndrome. Curvilinear intrahepatic collaterals and capsular collaterals are characteristic. In patients with chronic Budd-Chiari syndrome, hyperplastic nodules may develop (figure 7). ³² These nodules are hyperintense on T1WI, variable on T2WI, and enhance early on post-gadolinium images. ³³ They resemble focal nodular hyperplasia in some respects and do not appear to have malignant potential.

Primary sclerosing cholangitis

Primary sclerosing cholangitis (PSC) is a disease of unknown etiology characterized by inflammation and fibrosis of the liver and the bile ducts. Endoscopic retrograde cholangiopancreatography (ERCP) is the standard reference for diagnosis of PSC, because the clinical, biochemical, and hepatic histologic findings are usually nonspecific. Good correlation between MR cholangiopancreatography (MRCP) and ERCP in diagnosing PSC has been demonstrated. ^34,35 Comparing MRCP with ERCP has potential advantages: 1) it depicts bile ducts that are not seen at ERCP; 2) it noninvasively detects progressive ductal abnormalities in asymptomatic patients; 3) it identifies the group of patients who might benefit from therapeutic ERCP; and 4) supplemental MR sequences detect early cholangiocarcinoma. The main criticism of MRCP is that the care may be delayed in patients who need therapeutic endoscopic or percutaneous intervention of obstructing bile ducts. The most common MRCP findings are intrahepatic ductal dilatation, stenosis, and a beaded appearance ^35-37 (figure 8). The key feature of PSC is randomly distributed annular strictures out of proportion to upstream dilatation. ³⁵ As the fibrosing process progresses, stricture increases and the peripheral ducts are obliterated, producing a characteristic pruned appearance. Wall-thickening and enhancement of the extrahepatic bile duct are frequent findings as well, which cannot be seen on MRCP. Instead, they are best seen on the contrast-enhanced dynamic images. Magnetic resonance images of the liver in patients with PSC often have patchy peripheral areas of atrophy and abnormal signal intensity and enhancement.

Conclusion

MR imaging is a superior modality for diagnosing diffuse liver disease andfollowing disease progress. It is particularly useful for the evaluation of fatty liver, iron overload, cirrhosis, hepatic venous thrombosis, and primary sclerosing cholangitis. AR