Cirrhosis of the liver is the end stage of a complex process—resulting from hepatocyte injury and the response of the liver—that leads to partial regeneration and fibrosis of the liver.
Cirrhosis poses a difficult challenge for management, while the disease's prevention, detection, and therapy engender major health costs. Diagnostic imaging offers diverse modalities for use in the noninvasive evaluation of the liver, as well as in interventional techniques; the latter may be used to treat such complications as portal hypertension and neoplasia. The diagnosis, management, and treatment of cirrhosis are reviewed in this article.
Hepatic morphologic changes
Regardless of etiology, gross morphologic changes of cirrhosis are recognized by a variety of image techniques. Enlargement of the left lobe and caudate lobe, believed to be the result of lobar-relative regeneration rather than fibrosis, secondary to an accident of vascular supply, is recognized by any cross-sectional technique, such as computed tomography (CT) scanning, magnetic resonance imaging (MRI), or ultrasonography (US), each depicted in the images below.
Using MRI, Okazaki and colleagues determined that alcoholic cirrhosis is associated more frequently with caudate lobe enlargement and the presence of a right posterior hepatic notch than is virus-induced cirrhosis.  Harbin, Hess, Giorgio, Torres, and their coauthors have described a number of indices, including the ratio of transverse caudate lobe width to right lobe width, multidimensional caudate lobe indices that can be obtained by US or CT scanning, and volume analysis of each liver segment, based on cross-sectional area by CT scanning or MRI. [2, 3, 4, 5] Lafortune and colleagues suggested that a reduction in the medial segment of the left hepatic lobe diameter is a helpful adjunct finding of cirrhosis in ultrasonographic investigation. 
Another sign of cirrhosis, the expanded gallbladder fossa sign as depicted in the image below, has been described on MRI examination, based on an evaluation by Ito and coauthors of 190 patients with cirrhosis and of 123 control patients.  The authors' criterion was enlargement of the pericholecystic space (ie, gallbladder fossa)—which had to be demarcated laterally by the edge of the right hepatic lobe, medially by the edge of the lateral segment of the left hepatic lobe, or posteriorly by the anterior edge of the caudate lobe—in conjunction with nonvisualization of the medial segment of the left hepatic lobe on the same axial image. This achieved a sensitivity, specificity, accuracy, and positive predictive value for the MRI diagnosis of cirrhosis of 68%, 98%, 80%, and 98%, respectively.
On ultrasonographic examination, the liver contour may appear nodular, as in the first image below, although Ladenheim and colleagues have questioned the specificity of this sign. Similar contour deformities, depicted in the second image below, are evident on examination by CT scanning or MRI. The echo texture appears coarsened. Increase in echogenicity, shown in the third image below, is caused by fatty infiltration, which may be diffuse in hepatitis or focal in hepatitis or cirrhosis.
Intrahepatic vascular changes in cirrhosis
In cirrhosis, the dynamics of the hepatic arterial and portal venous circulation change as the degree of fibrosis progresses, as depicted in the image below.
In addition, the vessels appear to elongate and become more tortuous because of the underlying parenchymal architectural distortion. This is recognized classically in angiography as "corkscrewing" of vessels and can be appreciated on cross-sectional imaging, depicted below.
Secondary manifestations of cirrhosis may be seen as morphologic or physiologic evidence of the disease. The development of spontaneous shunts has been described in advanced cirrhosis and was initially demonstrated by angiography, although it is now demonstrable by noninvasive techniques, such as Doppler US as in the image below, at an incidence of up to 7%.
Multiphase CT scanning can demonstrate these shunts as early opacification of the intrahepatic veins during the early arterial phase-injection, portrayed below. The shunts are often accompanied by geographic, wedge-shaped perfusion abnormalities.
Extrahepatic manifestations of cirrhosis detectable by imaging techniques
Marshak, Karahan, and coauthors reported a higher frequency in the alteration in the thickness of the wall of the GI tract, depicted below, in patients with cirrhosis than in controls (64% vs 7%). [8, 9]
The development of splenomegaly and collaterals from portal hypertension is readily evident when any cross-sectional technique is used. Nodular iron deposition within the spleen, as seen on seen on MRI scans (Gamma-Gandy bodies), is highly suggestive of portal hypertension.
Functional imaging techniques, such as the use of technetium-99m (99m Tc)–labeled sulfur colloid, which is taken up by reticuloepithelial cells, and the presence of "colloid shift" to the bone marrow in cirrhosis, in addition to the recognition of hepatic morphologic changes and splenomegaly, have been helpful in confirming the presence and severity of cirrhosis, as in the image below.
Portal hypertension occurs once portal pressures reach 5-10 mm Hg above normal as a complication of cirrhosis. The well-recognized effect of increasing portal pressures is the development of splenomegaly, depicted below, and collateral portal-venous anastomoses, which occur at numerous sites, including gastroesophageal, paraumbilical, perirectal, and retroperitoneal locations.
Varices are not found when the portal vein pressure (indirectly measured as the hepatic vein pressure gradient [HVPG]) is less than 12 mm Hg. However, not all patients with elevated portal pressures develop variceal bleeding. Noninvasive diagnostic imaging methods, such as color flow Doppler US, contrast-enhanced CT scanning, and MRI, can be used to identify the presence of collaterals, as in the images below, but a major limitation is an inability to employ them in evaluating variceal pressures, which correlate more directly with the risk of hemorrhage.
Endoscopic evaluation provides a visual window on esophageal varices, which can be graded for prognosis. However, noninvasive techniques are useful in demonstrating collateral vessels beyond the reach of the endoscope. The images below depict 3-dimensional CT scanning, which is particularly helpful in demonstrating the development and pattern of collateral flow in portal venous hypertension.
Slow portal flow can mimic portal vein occlusion on cross-sectional imaging, and care must be taken when interpreting images in which not all diagnostic criteria are met. Long-standing thrombosis may be associated with cavernous transformation in which periportal collaterals re-establish flow to the liver, even in the setting of cirrhosis and elevated sinusoidal pressures. Neoplastic invasion of the portal veins must be differentiated from bland thrombus, as in the images below.
Imaging evaluation, initially, ultrasound at 6-month intervals, is recommended based on evidence that increased frequency of examination leads to detection of hepatocellular carcinoma (HCCA) at an earlier stage. It is recommended to screen patients with chronic hepatitis and/or biopsy-documented cirrhosis semiannually with these techniques, in accordance with the American Association for the Study of Liver Diseases (AASLD) guidelines. 
The following section is derived from the recommendations of Bruix et al and a synopsis or direct quotation is as follows: patients awaiting transplant in the United States should be screened for HCCA because development of HCCA increases priority for orthotopic transplantation, and, if unscreened, an undetected lesion may already be in an advanced stage, which would render the patient unsuitable for transplantation.
Nodules smaller than 1 cm detected in ultrasound surveillance should be followed with further ultrasound at intervals from 3-6 months for 2 years. If no growth occurs during that interval, a return to routine surveillance is recommended. However, “nodules >1 cm found on ultrasound screening of a cirrhotic liver should be investigated further with either 4-phase multidetector computed tomography (CT) scan or dynamic contrast-enhanced magnetic resonance imaging (MRI).“
The guidelines recommend that if the appearances are typical of HCCA (ie, hypervascular in the arterial phase with washout in the portal venous or delayed phase), the lesion should be treated as HCCA. “If the findings are not characteristic or the vascular profile is not typical, a second contrast-enhanced study with the other imaging modality should be performed or the lesion should be biopsied.” 
If biopsy including tissue markers such as CD34, CK7, glypican 3, HSP-70, and glutamine synthetase is not confirmatory, the lesion should be followed by imaging at 3- to 6-month intervals until the nodule disappears, enlarges, or displays diagnostic characteristics of HCCA, with repeat biopsy of any enlarging, atypical lesions.
Regarding imaging techniques used as screening tools, the accuracy of US, as with CT scanning and MRI, is more limited in the advanced stages of cirrhosis. An US study from Korea, with transplantation correlation in 52 patients, demonstrated a sensitivity for the detection of HCCA of only 33% (6 of 18) lesions.
CT scanning is believed to be equivalent in sensitivity to, and more specific than, US. However, there are disadvantages related to contrast risk and radiation exposure, particularly if the modality is used over a lifetime for screening. Thus, CT scanning as a screening modality should be reserved for further evaluation of patients with equivocal or technically limited ultrasound studies. For example, a heterogeneous appearance on US evaluation of the liver, as shown below, may mask malignant lesions and justify additional imaging.
Conversely, persistent lesions that are noted on US, even if not confirmed on a CT scan, probably should be biopsied, particularly in the setting of serologic abnormalities, as depicted in the image below. In end-stage patients who are destined for transplantation, the sensitivity of CT scanning is reduced (to as low as 37% in one series). 
MRI with gadolinium (or other contrast agents) is not recommended by the AASLD as an initial screening modality. Similar to US and CT scanning, a reduction in sensitivity to below 50% has been reported, particularly for lesions under 2 cm, in patients with end-stage disease. [13, 14]
Even more invasive and sophisticated techniques, such as CT scanning performed with a catheter in the hepatic artery, as well as angiography, are usually reserved for use in patients undergoing evaluation for transplantation at regional centers, where the goal is to exclude or establish the presence and multiplicity of malignant lesions for pretransplantation assessment. These techniques are not routinely used in the United States but are used extensively in Asia.
Unfortunately, for cultural reasons, orthotopic liver transplantation is not routinely performed in these countries; thus, pathologic correlation is limited to hepatic resections and biopsies. However, with the rapid increase in right lobe liver donation surgery, the pathologic correlation should be excellent because the entire explanted liver will be available.
Limitations of techniques
Real-time US is used extensively for screening, but biopsy or additional imaging modalities are required for confirmation. US is a nonspecific test and identifies many nodules, ranging from regenerative nodules, dysplastic nodules, and focal fat to benign neoplasms, such as hemangioma, many of which have no uniquely discriminating features on US.
Because these occur with significant frequency, they pose a diagnostic challenge. A significant minority of patients may have benign focal masses, such as hemangioma or focal fat, requiring further imaging evaluation, or focal lesions that may not be corroborated on other imaging studies or on subsequent US. The relatively high prevalence of benign lesions in patients with cirrhosis appears to be corroborated by a study by Horigome and colleagues, in which they established a prevalence of adenomatous hyperplasia in 30% of nodules less than 1 cm in diameter. 
Therefore, it is necessary to commit either to the biopsy of atypical persistent lesions, or to corroboration with other techniques, such as multiphase multidetector CT scanning or MRI, using dynamic imaging with contrast to obtain multiple vascular-phase images. Clinician preferences in the United States, as surveyed by Chalasani and coauthors, suggest an empirical trend toward routinely incorporating CT scanning in screening,  but increased awareness of radiation exposure may halt or slow this trend and lead to increasing use of MRI.
Some cause for optimism is warranted in terms of reduction in the incidence of HCCA in patients with hepatitis C following interferon therapy. A meta-analysis by Papatheodoridis and colleagues of 11 studies involving more than 2000 patients determined that the incidence of HCCA in patients who underwent interferon therapy was reduced to 8.2%, compared with 21.5% in untreated patients, and was even lower in sustained responders (0.9%). 
A major unresolved problem is the evaluation of the efficacy of screening and the economic consequences of aggressive screening. Bolondi and coauthors, in Italy, and Larcos and colleagues, in the United States, estimated that each case of HCCA that is detected costs $8000-24,000. [18, 19]
Despite the best efforts of the worldwide medical community in screening for HCCA, no evidence exists that mortality has been affected, because therapeutic options, although expanding, remain relatively limited. Survival in patients undergoing liver transplantation who have unsuspected HCCA is adversely affected by tumor recurrence (reported in a French series by Adam and colleagues as reaching 5%).  The presence of neoplasm is not a contraindication to transplantation, although survival in patients with tumors larger than 3 cm, multiple nodules, or portal invasion is sufficiently impacted to preclude consideration in this subgroup.
Of those patients with cirrhosis and varices, 25-40% experience bleeding. The management of portal hypertension and upper tract GI bleeding has been revolutionized by endoscopic and angiographic treatment. The use of the intravascularly placed transjugular intrahepatic portosystemic shunt (TIPS) has provided a second-line therapy for the management of portal hypertension, with reduced mortality and morbidity compared with that associated with the open surgical procedure.
Fischer, Kimura, and coauthors have used noninvasive modalities, such as Doppler US, portrayed below, to monitor shunt patency (with a reasonably high accuracy of >90%). [21, 22] Occasionally, however, conventional Doppler techniques fail to image signals. Thus, enhancement by US contrast agents appears promising in improving visualization, as shown below.
Radiography has a modest place in the diagnosis and management of patients with cirrhosis, being used, for example, in screening for ascites, shown below, seeking evidence of bowel perforation in patients with suspected bacterial peritonitis, and monitoring bowel distension in acutely ill patients admitted for treatment of decompensation or variceal hemorrhage.
Routine chest radiography in a patient with cirrhosis may demonstrate elevation of the diaphragms from ascites. Gynecomastia may be appreciated. The azygos vein may be enlarged because of collateral flow, as in the first image below, and pleural effusions, shown below, may occur from the presence of pleuroperitoneal fistulas.
Rarely, giant esophageal varices may be appreciated as a soft-tissue mass at the gastroesophageal junction, as in the image below.
An upper GI study can demonstrate varices at the gastroesophageal junction, depicted below, but for the most part, endoscopy has superseded fluoroscopic techniques.
Plain film findings are generally confirmed by other imaging modalities or by clinical evidence.
CT scanning (see the images below) is useful for demonstrating the morphologic evidence of cirrhosis within the liver and in showing mesenteric and GI tract abnormalities, as well as the development of collateral vessels in portal hypertension.
Splenomegaly and the presence of ascites, depicted in the second image above, are readily determined.
CT scanning is commonly used to evaluate acutely decompensated patients with suspected subacute bacterial peritonitis, in order to exclude other inflammatory causes. CT scanning is valuable in characterizing lesions shown by US techniques or in evaluating decompensated patients with cirrhosis. In addition, it is increasingly being incorporated into the management of stable patients undergoing screening to identify neoplastic lesions.
With improved technology, which permits rapid dynamic scanning using multi-slice CT scanners, scanning of the liver in multiple phases of contrast enhancement is now routinely recommended as the most sensitive method of detecting space-occupying lesions and evaluating vascular structures. However, substantial limitations remain in delineating small lesions (< 2 cm), particularly in patients with advanced cirrhosis.
Characteristic form of HCCA
The most characteristic form of hepatocellular carcinoma (HCCA) is a hyperattenuating nodule noted on arterial-phase imaging, with hyperattenuation and/or hypoattenuation developing on portal venous–phase imaging, shown below. On CT scanning, hyperattenuation in the arterial phase occurs in a variable proportion of cases, and in many instances, it is characteristic enough to permit confident diagnosis. The characterization of liver nodules is challenging when the findings are not “typical” of HCCA, and the Liver Imaging Reporting and Data (LI-RAD) classification system has recently been introduced to improve consistency and aid in management decisions. 
Five major features have been chosen, which, in combination, favor the diagnosis of HCCA: (1) masslike configuration, (2) arterial phase hyperenhancement, (3) portal venous phase or later phase hypoenhancement, (4) increase of 10 mm or more in diameter within 1 year, and (5) tumor within the lumen of a vein. The 5-point categorization scale is based on level of certainty of a benign, indeterminate lesion, or firm diagnosis of HCCA.
Nino-Murcia and colleagues described arterial enhancement with abnormal internal vessels or a variegated appearance.  In some instances, a single hyperattenuating focus may be the only evidence of HCCA, with no distinguishing characteristics on precontrast or portal venous-phase images. However, a proportion of lesions are hypoattenuating or isoattenuating on arterial-phase imaging. Dysplastic nodules also may be very similar to HCCA in their enhancement characteristics, shown below.
However, hypoattenuating nodules depicted on a CT scan have high malignant potential. In the series by Takayasu and coauthors, 36 (60%) of 60 such lesions converted to hyperattenuating lesions, with the cumulative attenuation conversion rates of these 60 lesions reaching 58.7% within 3 years of follow-up.  Thirteen of the lesions were biopsied immediately after attenuation conversion was observed to prove that they were HCCA. The presence of hepatitis C viral antibody and lesion size at detection were correlated with the attenuation conversion rate.
CT scanning is useful in documenting complications associated with HCCA, such as portal vein thrombosis, and can be used to identify malignant invasion with a high specificity, as portrayed in the image below.
The images below depict multifocal HCCA, which is commonly associated with extensive portal vein thrombosis and/or invasion at the time of diagnosis. Often, extensive portovenous shunts are present.
Degree of confidence
In terms of degree of confidence, multidetector CT is a robust imaging modality, with a reported sensitivity of 100% for tumors larger than 2 cm, but lower sensitivity (as high as 96%) for lesions in the 1- to 2-cm range.  It is incumbent on the radiologist to be vigilant in high-risk patients undergoing screening and pretransplantation evaluation and to suggest confirmation of suspected intrahepatic tumors by additional imaging modalities such as MRI, biopsy, or serial close surveillance if imaging characteristics are atypical, in accordance with American Association for the Study of Liver Diseases (AASLD) guidelines for lesions in the 1- to 2-cm range.  Use of the LI-RAD classification system is expected to enhance consistency in reporting and communication of abnormalities.
A major diagnostic dilemma for radiologists is the finding of a small, focal, transiently enhancing lesion (transient hepatic attenuation difference [THAD]) in a patient undergoing CT screening for hepatoma. Even if all other imaging phases are normal, statistically, on hepatic arterial phase images, a high-attenuation focal hepatic lesion of cirrhotic liver is usually HCCA.  The likelihood of a hypervascular tumor is of course higher if an area of relative hypoattenuation is present on portal venous–phase images.
THAD may also occur in other tumorous conditions, such as peripheral cholangiocarcinoma, and in nonmalignant neoplasms, such as small hemangiomas. THAD is associated with a change in the blood supply to the liver, which can occur with the development of arterioportal shunts; in perfusion changes related to venous thrombosis, congestion, or fatty infiltration following radiofrequency ablation; and in locations where auxiliary blood supply is present such as segment IV or the gallbladder fossa.
If the lesion has a wedge shape, has a straight margin, or if normal vessels can be seen passing through the lesion, the probability of nontumorous THAD becomes greater.
The presence of normal signal intensity on T1- and T2-weighted images excludes hypervascular tumor on MRI. 
The enlargement of the caudate lobe in cirrhosis, with other regions of retraction, is depicted in the image below and may be mimicked in patients with breast carcinoma metastatic to the liver who are undergoing chemotherapy. Young and colleagues suggest that the mechanism is through nodular regeneration. 
Concerns regarding the evaluation of patients with cirrhosis and HCCA for transplant are related to the likelihood of a successful outcome based on the stage of the carcinoma; solitary HCCAs that are smaller than 2 cm can be treated successfully with transplantation. Survival rates diminish with the presence of additional or larger lesions, with a 5-year survival rate of only 75% reported by Mazzaferro and coauthors for patients with up to 3 discrete lesions that are smaller than 3 cm or with solitary lesions of 2-5 cm. 
Recommendations by the AASLD for patients undergoing liver transplantation include the following statements  :
“Liver transplantation should be viewed as the treatment of choice for selected patients with hepatocellular carcinoma who are not candidates for surgical resection and in whom malignancy is confined to the liver (II-2).”
“Optimal results following transplantation are achieved in patients with a single lesion 2 cm or larger and less than 5 cm, or no more than three lesions, the largest of which is less than 3 cm, and no radiographic evidence of extrahepatic disease (II-2).”
“For ideal outcomes, patients who meet these criteria should receive a donor organ within 6 months of listing for transplantation (II-2).”
Liver transplantation can offer patients a chance for prolonged quality of life, with a 1-year survival of more than 80%. CT is useful pretransplantation in recipients for determining liver morphology and vascular anatomy, excluding advanced hepatocarcinoma, and in evaluating living donors. 
Transplantation is of no benefit when the lesions are diffuse or there are more than 3 lesions, as the likelihood of metastatic disease is high. The discovery of a lesion or multiple lesions in a patient with cirrhosis who is otherwise well compensated necessitates further imaging evaluation or biopsy to characterize the lesions accurately. This assists in deciding whether to refer the patient for transplantation or other therapy.
As understanding of tumor cytogenetics has advanced, tumor mutations, including TP53 and Ras homolog gene family mutations, are now understood to be associated with poorer prognosis and survival.  Future management will be enhanced by a better understanding of tumor biology and the development of techniques such as perfusion CT for monitoring of more specific therapy. 
HCCA has recently found to occur in patients with end-stage liver disease caused by nonalcoholic steatohepatitis,  for which potential carcinogenic mediators include insulin, lipid peroxidations, and oxidative stress associated with free radicals, but may be associated with well-differentiated tumors. 
Often, multiple abnormalities are present in a patient, including a spectrum of nodules in various stages of malignant transformation. One or more frank HCCAs may coexist with several dysplastic nodules and/or a multiplicity of regenerative nodules, making the likelihood of accurate pretransplant diagnosis minimal without highly refined imaging techniques.
In patients with advanced cirrhosis, peak enhancement of the liver is reduced and enhancement may appear heterogeneous, reducing the level of detection of focal lesions.
Distinguishing nodules from HCCA
Much effort has been devoted to distinguishing regenerative nodules from dysplastic ones, and dysplastic nodules from HCCA. CT scanning techniques, such as CT arterial portography ([CTAP], in which the liver is visualized following a superior mesenteric artery [SMA] injection in the portal phase) and CT arteriography (which utilizes direct injection into individual hepatic arteries, with imaging performed in the arterial phase), have been extensively investigated. Matsui and coauthors asserted that most low-grade dysplastic nodules have almost the same histopathologic and hemodynamic characteristics as those of regenerative nodules; therefore, they are isoattenuating to regenerative nodules at CT scanning and are not usually visualized on CT arterial portography.
High-grade dysplastic nodules may have a decreased portal supply and an increased arterial supply, but Lim and colleagues found the presence of such changes to be extremely variable; the grade of nodular dysplasia does not seem to affect the presence of the portal and arterial supplies.  Differentiating dysplastic nodules from HCCA is a difficult task because some low-grade dysplastic nodules lose the portal supply and gain an arterial supply, while some high-grade dysplastic nodules retain the portal supply without gaining an increased hepatic arterial supply.
Monzawa and colleagues reported that small HCCAs may be detected more accurately by combining the characteristics of arterial-phase, portal venous–phase, and delayed-phase images.  A receiver operating characteristic (ROC) analysis for combination 3-phase imaging was significantly higher than for arterial-phase and portal venous–phase imaging (Az = 0.940 vs 0.917).
In patients with advanced cirrhosis, a sensitivity of 88% for CT scanning in the detection of HCCA was obtained by Chalasani and coauthors, compared with 62% for alpha-fetoprotein (AFP) measurement and 59% for US.  However, the accuracy of 3-phase dynamic CT scanning in detecting tumors in cirrhotic livers smaller than 2 cm in diameter is in the range of 60%, with a somewhat higher accuracy for lesions of 2-5 cm (82%), based on a study by Lim and colleagues of 41 patients whose livers were explanted before transplantation.  This was corroborated by a study by Peterson and colleagues of 77 North American patients undergoing liver transplantation.  A prospective sensitivity of 59% for the presence of HCCA and a sensitivity per lesion of 37% were found.
A subsequent study of 88 patients undergoing orthotopic liver transplantation, evaluated pre-operatively with multidetector CT (MDCT) scanning by Ronzoni and coauthors, confirms that despite improved detector technology and scan speed, no real improvement in sensitivity can be expected from MDCT.  Observers detected 89 of 139 hepatocellular carcinomas in the 88 patients, with a sensitivity of 64%. However, the specificity was only 75% because many regenerative or dysplastic nodules were present, leading the authors to caution against overestimation of disease that might preclude patients from liver transplantation.
Characterizing nonmalignant lesions
CT scanning is useful in the characterization of nonmalignant lesions, such as hemangiomata, which occur with relatively high frequency in patients with cirrhosis. Other, more invasive forms of CT scanning have been evaluated over the past few years; Jang and colleagues reported evidence that CTAP and CT hepatic arteriography add little or no additional information to that obtained using triple-phase, helical CT scanning in the detection of HCCA. 
As noted, false negatives may occur as a result of technical limitations or because an advanced degree of liver fibrosis impacts enhancement characteristics. The process of the liver’s response to injury results in regenerative nodules and dysplasia, which occur prior to frank transformation into HCCA. Imaging characteristics may not be typical of HCCA and lead to a lower level of confidence and lower LI-RAD classification, which may then result in additional contrast imaging or further surveillance for lesion growth.
Dysplastic nodules may mimic HCCA. Nontumorous arterioportal shunting in livers with cirrhosis has been reported by Kim and colleagues as mimicking hypervascular tumor. 
A higher injection rate may increase the number of small, false-positive, hypervascular lesions. Ichikawa and coauthors studied 60 patients with suspected HCCA  ; they reported in 2006 that the use of an iodinated contrast injection rate of 5 mL/sec resulted in an 18% reduction in specificity, from 67% to 48%, with no significant change in sensitivity (88% vs 80%) compared with a 3-mL/sec injection rate. 
Magnetic Resonance Imaging
MRI offers an alternative noninvasive method of imaging the liver based on tissue-specific characteristics. [42, 43] In addition to demonstrating morphologic changes in cirrhosis, MRI is suited for the evaluation of vascular structures for patency or tumor invasion. T1-weighted images are valuable in providing anatomic detail, and T2-weighted images are more sensitive in detecting mass lesions and characterizing cysts and hemangiomata. MRI technology continues to evolve rapidly, with the development of techniques, such as the use of gradient-echo, fast spin-echo (SE), and diffusion-weighted sequences, that permit the rapid acquisition of images required in association with paramagnetic contrast use. Tumor enhancement patterns following administration of gadolinium-cased contrast agents are incorporated into Liver Imaging Reporting and Data (LI-RAD) criteria.
Gadolinium, through its paramagnetic properties, reduces T1 and T2 relaxation times, with an improved signal-to-noise ratio. Gadolinium is chelated to organic compounds in order to form extracellular contrast agents which, having entered the liver, become distributed from the intravascular to interstitial spaces.  Agents have now been developed with characteristics of extracellular contrast agents combined with hepatocyte-selective and blood-pool characteristics. The combined agents, such as gadobenate dimeglumine and gadoxetic acid, can be used for dynamic-phase imaging for liver lesion detection and characterization with sensitivity similar to that of extracellular contrast agents. Their uptake into hepatocytes from the blood and excretion into the bile via the organic anion transport protein are analogous to bilirubin uptake. In their hepatocyte-selective phase, these agents provide prolonged opacification of liver parenchyma. 
Earlier work evaluating contrast-enhanced images of the liver alone did not appear to be very sensitive; a 2001 study using gadopentetate dimeglumine (Magnevist) in explanted livers for confirmation found an overall sensitivity of only 54% in the detection of hepatocellular carcinoma (HCCA), with sensitivity achieving 80% for lesions larger than 2 cm, 50% for lesions of 1-2 cm, and 33% for lesions smaller than 1 cm. The sensitivity for dysplastic nodules was only 15%. 
Use of diffusion-weighted sequences appears to be able to predict development of HCCA in dysplastic nodules. When used with a more recently developed chelate, gadoxetic acid, a hepatobiliary specific agent, nodules that showed hyperintensity on diffusion-weighted images and were hypovascular and hypointense on hepatobiliary-phase imaging were more likely to progress to hypervascular HCCA.  The introduction of new, more specific hepatic agents such as gadobenate dimeglumine, ferucarbotran, and gadoxetic acid have also improved accuracy, which is now reported to exceed 95%. 
Gadoxetate was recently found superior to multiphase multidetector CT imaging in a population of 58 patients with 87 HCCAs (mean size ± standard deviation, 1.8 cm ±1.5; range, 0.3–7 cm) in a multireader study.  Regardless of lesion size, the average diagnostic accuracy and sensitivity were significantly greater with gadoxetate disodium–enhanced MRI (average diagnostic accuracy, 0.88; 95% confidence interval [CI], 0.80-0.97; average sensitivity, 0.85; 95% CI, 0.74-0.96) than with multidetector CT (average diagnostic accuracy, 0.74; 95% CI, 0.65-0.82; average sensitivity, 0.69; 95% CI, 0.59-0.79) (P < .001 for each), with inter-reader agreement good to excellent.
The use of newer hepatobiliary-specific agents remains controversial, because they do not provide increased sensitivity with respect to the extracellular agents. Sensitivity may be limited by compromised uptake in patients with advanced cirrhosis and poor liver function. As well-differentiated HCCA may accumulate hepatobiliary-specific agents on delayed hepatocyte imaging, these lesions may be indistinguishable from benign hepatocyte-containing lesions. 
MRI has been studied extensively in diffuse liver disease. Tani and coauthors reported that focal and diffuse steatosis are recognized as increased signal intensity on T1-weighted MRIs and as diffuse low signal intensity on opposed-phase, T1-weighted images.  Regenerative nodules are seen as small, masslike structures and are hypointense on T2-weighted images.
In one prospective study, liver T1 mapping showed a higher diagnostic accuracy than liver and spleen DWI and T2 mapping and had a significant correlation with Child-Pugh score (Pearson's correlation coefficient of 0.46), MEDL score (0.30), and liver stiffness measurement (0.52). 
Hemochromatosis is particularly suited to evaluation by MRI; iron has a superparamagnetic effect on signal intensity that is best appreciated on T2-weighted images. The deposition of iron in siderotic nodules, which can be readily evaluated using MRI techniques, has been suggested by Ito and colleagues as an indicator of risk for malignant degeneration in patients with cirrhosis.  Increased iron deposition associated with hemochromatosis is one of the well-recognized risk factors for HCCA.
Ernst and coauthors described a statistically significant correlation between the hepatic iron concentration revealed at MRI with T1- or T2-weighted sequences,  and the hepatic iron concentration value measured at biopsy.
To quantify the amount of hepatic iron, Ito and colleagues used T2-weighted SE or fast SE images and gradient-recalled echo (GRE) images (echo time 6.0 ms), which they had determined were sensitive to the paramagnetic effects of hepatic iron among the MRI scans obtained at routine abdominal examination.  At MRI, hepatic parenchymal iron deposition was seen in 79 patients (40%) and iron deposition in regenerative nodules was seen in 71 patients (36%).
The mean signal-intensity ratio of GRE images in 125 patients with hepatic iron deposition was significantly lower than that in patients without such deposition (P < .001). The frequency of HCCA in patients with iron deposition in regenerative nodules was 52%, significantly higher (P = .015) than that in patients without iron in regenerative nodules (34%). Ito and coauthors suggested a role for MRI in monitoring patients undergoing phlebotomy, because this may reduce iron deposition in regenerative nodules and potentially decrease the risk of HCCA.
Patients without hemochromatosis
In patients without hemochromatosis, no association has been shown between frequency of siderotic nodules and an increased incidence of HCCA. 
MRI may be useful in identifying intratumoral fat, tumor encapsulation, portal or hepatic vein invasion, and arterial-portal venous shunting, depicted below, all of which are characteristic of HCCA and have been extensively described in papers by Ebara, Kadoya, Winter, and colleagues. [52, 53, 54]
Interest has focused on the use of gadolinium with currently available MRI sequences that incorporate gradient-echo (GE) imaging, recognizing that the above-mentioned tumor characteristics are not always present. Lauenstein and colleagues evaluated 115 patients undergoing MRI with gadolinium-enhanced, 3-dimensional, gradient-echo sequences in arterial, venous, and delayed phases.  Using various imaging criteria, including arterial-phase enhancement, delayed-phase hypointensity, and capsular enhancement, the authors detected 28 of 36 pathologically confirmed HCCAs in 27 patients (with the lesion-based sensitivity being 77.8%; the patient-based sensitivity, 88.9%; and the specificity, 97.7%). The diagnosis of HCCA when tumors are under 2 cm remains a challenge; among 18 of these smaller lesions, only 10 were diagnosed, and another 2 had characteristics of dysplastic nodules.
The use of superparamagnetic iron oxide (SPIO) in HCCA diagnosis appears promising. In a ROC study by Reimer and coauthors of patients destined for transplantation or biopsy, 27 patients with MRI-scan features of dysplastic nodules and/or HCCA were examined.  T2-weighted, spin-echo imaging and T1-weighted, GE imaging were performed before and after SPIO administration and were followed by T1-weighted, GE imaging at 10, 40, and 120 seconds after bolus injection of a gadolinium-based contrast material.
SPIO-enhanced MRI (mean accuracy, 0.76) was more accurate than was unenhanced MRI (mean accuracy, 0.64) (P < .04), while double-contrast MRI (mean accuracy, 0.86) was more accurate than SPIO-enhanced MRI (P < .05). The sensitivity for detecting lesions smaller than 1 cm was greater with double-contrast MRI than with SPIO-enhanced or unenhanced MRI. Adjacent fibrosis was found to be the most frequent cause of false-positive findings on SPIO-enhanced and unenhanced images, but this could be reduced by the evaluation of postgadolinium-enhanced images obtained during the arterial and portal phases of enhancement.
Gadolinium-based contrast agents (gadopentetate dimeglumine [Magnevist], gadobenate dimeglumine [MultiHance], gadodiamide [Omniscan], gadoversetamide [OptiMARK], gadoteridol [ProHance]) have been linked to the development of nephrogenic systemic fibrosis (NSF) or nephrogenic fibrosing dermopathy (NFD). For more information, see Nephrogenic Fibrosing Dermopathy. The disease has occurred in patients with moderate to end-stage renal disease after being given a gadolinium-based contrast agent to enhance MRI or magnetic resonance angiography (MRA) scans.
NSF/NFD is a debilitating and sometimes fatal disease. Characteristics include red or dark patches on the skin; burning, itching, swelling, hardening, and tightening of the skin; yellow spots on the whites of the eyes; joint stiffness with trouble moving or straightening the arms, hands, legs, or feet; pain deep in the hip bones or ribs; and muscle weakness. Great caution should therefore be exercised in evaluating patients with cirrhosis and renal insufficiency.
Magnetic resonance spectroscopy does not involve the use of gadolinium; a novel use for31 P MR spectroscopy has been proposed in evaluating the response to antiviral therapy in patients with hepatitis-C related liver disease. Lim and colleagues detected a decreased ratio of phosphomonoester (PME) to phosphodiester (PDE) in 25 of 32 responders to interferon and ribavirin, suggesting that PME and PDE can be used as biomarkers to evaluate treatment response.  (Patients with severe hepatitis C have an increased PME/PDE ratio.) There was no decrease in the ratio in 15 nonresponders.
Degree of confidence
MRI has an increasing role in screening particularly at specialized transplantation centers. The level of confidence in MRI, particularly when newer contrast agents such as hepatobiliary contrast are used, appears to be equivalent to, or exceed, the level of confidence in dual-phase, spiral CT scanning or triple phase MDCT scanning. In early reporting, overall sensitivity of MRI was reported by Bartolozzi and colleagues to be 86% for a prospective assessment of precontrast and postcontrast images. [58, 59]
Similar sensitivity has been reported by Kondo and coauthors, who retrospectively analyzed images of the liver from 33 patients on a segment-by-segment basis. A total of 261 segments, which included 39 HCCAs and 21 metastases, were independently reviewed by 3 radiologists. Unenhanced and gadolinium-enhanced MRI scans were reviewed first, and then ferumoxides-enhanced MRI scans were added for a combined review. CTAP images and biphasic CT hepatic angiography (CTHA) scans were reviewed together.
It was determined in the study that the sensitivity for the detection of hepatic tumors was equivalent for combined unenhanced, gadolinium-enhanced, and ferumoxides-enhanced MRI scans (86%) and for combined CTAP images and biphasic CTHA scans (87%). Specificity was higher with MRI scans (95%, P < .01) than with CT scans (91%), with improved performance achieved by combining ferumoxides-enhanced MRI scans with unenhanced and gadolinium-enhanced MRI scans (Az = 0.9 vs 0.950, P = .0502). The radiologists' preoperative ability to detect malignant hepatic tumors using combined unenhanced, gadolinium-enhanced, and ferumoxides-enhanced MRI scans was analogous to that when combined CTAP images and biphasic CTHA images (Az = 0.959) were used.
The introduction of new, more specific hepatic agents such as gadobenate dimeglumine, ferucarbotran, and gadoxetic acid have improved accuracy, now reported to exceed 95%.  Therefore, it appears that MRI has a diagnostic accuracy similar to or higher than that of CT scanning for lesions over 1 cm.
However, MRI still has significant limitations in the specificity of small-tumor detection, which further development of tissue-specific contrast agents may overcome. MRI does appear to enable the distinction of arterioportal shunts associated with tumor from spontaneous shunts associated with cirrhosis alone. When a superparamagnetic agent (iron oxide) is used, Mori and coauthors noted that tumorous shunts have reduced signal loss, whereas nontumorous shunts resemble normal liver parenchyma in the degree of signal loss, particularly on T2-weighted gradient-echo images. 
Regenerative nodules can resemble hypovascular HCCA; Kim and colleagues recognized that infarcted regenerative nodules can pose particular problems.  The liver of patients with nodular regenerative hyperplasia, also known as idiopathic portal hypertension or hepatoportal sclerosis, has morphologic features that are indistinguishable from cirrhosis.
However, the histologic features of these livers demonstrate nodules but not evidence of fibrosis, which is the hallmark of cirrhosis.
Dysplastic nodules occurring in cirrhotic patients, as already noted, may share imaging characteristics of HCCA. Differentiation may be difficult. Hyperintensity on diffusion-weighted images in hypovascular, hypointense nodules on hepatobiliary-phase gadoxetic acid has been found to be strongly associated with progression to hypervascular HCCA,  but this finding awaits validation in a population that did not have a history of prior treatment for HCCA.
False negatives may occur in small lesions, related to observer limitations. In a 2013 study tracking 17 patients with elevated alpha-fetoprotein (AFP) and initially negative MRI scans, 10 (59%) of 17 of patients developed HCCA over the next months after a mean of 138 days (range, 41-247 d). Of 10 HCCAs detected at follow-up MRI, 5 were identifiable in retrospect at initial MR studies (mean diameter, 1.4 cm). Fifty-percent of the lesions detected on subsequent MRI were visible in retrospect. Serum AFP levels in patients with HCCAs were significantly higher than those in patients without HCCAs and progressively increased over time (P = .012). 
Real-time US, in combination with color flow Doppler US, is currently the most frequently used diagnostic imaging modality worldwide in the screening and evaluation of patients with cirrhosis (see the images below). [63, 64, 65, 66, 67] In addition to demonstrating the morphologic characteristics of cirrhosis, including hepatic contour, texture, and the presence of portal collaterals, Doppler US provides useful information on portal hemodynamics. [68, 69, 70, 71]
Real-time US can be used to detect ascites and splenomegaly, to differentiate intrahepatic or extrahepatic causes of jaundice, and to detect portal vein thrombosis in patients who have decompensated (see the images below).
Portal blood flow
Doppler evaluation in a patient with cirrhosis can demonstrate high-velocity blood flow in the enlarged hepatic artery, which becomes tortuous as the underlying degree of fibrosis increases (see the image below).
PI, a measure of hepatic arterial vascular resistance, is elevated in patients with cirrhosis, and Schneider and colleagues have determined that it correlates quite well with the HVPG. 
The normal direction of portal blood flow is maintained initially, but as the degree of cirrhosis progresses, damping of the usual triphasic signal in the intrahepatic veins and loss of respiratory variation in the portal venous system occur. Flow within the main portal vein gradually diminishes; bidirectional and (subsequently) reversal of flow may be seen, usually with accompanying development of collateral vessels, as shown in the images below.
These collaterals are most frequently detected in the splenorenal region (21%), or as patent paraumbilical collaterals (14%) (see the images below).
In a study by von Herbay and coauthors of 109 patients with cirrhosis, the presence of collaterals correlated significantly with the presence of ascites, esophageal varices, and the inversion of portal flow, but not with splenomegaly. 
Doppler US continues to be used in the noninvasive physiologic evaluation of the portal tract in patients who, in an attempt to reduce the risk of GI hemorrhage, undergo pharmacologic modulation of portal pressures. However, Doppler US does not correlate well with intrahepatic pressures or with the portal systemic pressure gradient. For example, the evaluation of systemic flow in the femoral or brachial artery also has been studied, but only a 50% correlation is observed in reduction of femoral blood flow and portal pressure in response to propranolol treatment.
Effects of pharmacologic agents
The potentially confounding effects of pharmacologic agents on portal and systemic blood flow and resistance, coupled with a wide range of variability in individual response and observer measurements, continue to make this a perplexing area of investigation. A direct correlation of multiple flow parameters to the HVPG remains elusive.
Arterial vascular impedance can be estimated as the RI, which represents the ratio of the difference between the peak systolic and end-diastolic velocities to the peak systolic velocity. This can be measured directly in the superior mesenteric or hepatic artery. In addition to pharmacologic agents, however, numerous factors on the capillary and venous side can affect the RI. These include alteration of blood flow in the portal veins following a meal and the extent of development of collateral vessels, in addition to increased resistance from fibrosis or hepatic congestion because of fatty infiltration or right-sided heart failure.
Screening for focal hepatic masses
US has an established role in screening for focal hepatic masses, despite rather low specificity (see the images below).
Demonstration of shunt vascularity by Doppler US enables a diagnosis to be made with high specificity (see the images below), but neovascularization occurring in small lesions may be below the threshold of detection of even sophisticated US systems.
Multifocal lesions occasionally may be obscured, but in general, the lesions can be appreciated as tumor masses that either have vascularity or are avascular, but displacing, vessels (see the images below).
Pulse Doppler US is useful in demonstrating shunt velocity, as shown below, which in the author's population has been found to be highly specific for HCCA when in excess of 2.4 kHz.
Portal vein thrombosis
Portal vein thrombosis, shown below, can be diagnosed with relative confidence if a distended portal vein is found that contains echogenic material in the absence of Doppler signal.
Malignant invasion of the portal vein may be detectable as neovascularity within the thrombus, occasionally with direct contiguity with an intrahepatic lesion. (See the image below).
Low flow within the portal vein may be misinterpreted as thrombus, and careful attention to technique is necessary to ensure that the sensitivity of the Doppler signal is optimized.
Harmonic imaging/intravascular contrast agents
The development of intravascular contrast agents (which have little or no toxicity) initiated a re-evaluation of ultrasonographic sensitivity and specificity, which early investigations have suggested are greatly improved. The technical performance of ultrasonographic systems concomitantly has been modified to insonate tissue optimally, as well as to detect and process vascular and parenchymal signals from contrast agents. Techniques that are used include harmonic imaging, which is designed to capture nonlinear resonant frequencies from tissue and microbubbles with enhanced signal compared to background noise.
The microbubbles can be disrupted by insonation at a high mechanical index (MI), which represents the peak negative pressure of the transmitted ultrasonographic pulse, and this produces a strong, very brief echo. The microbubbles can then be visualized at a lower MI intensity (< 0.5) without causing further disruption. The bubbles can be seen within vessels and are detectable within capillaries in which conventional Doppler techniques cannot detect flow.
HCCAs have variable enhancement patterns on contrast-enhanced harmonic US. Homogeneous and heterogeneous enhancement have been described by Kim and colleagues, correlating with CT-scan enhancement patterns.  Three of 8 patients in this study also had linear tumor vessels in the lesions, but globular or peripheral enhancement seen in hemangioma and metastases, respectively, were not shown. Wilson and colleagues described perilesional and intralesional vessels. In a pilot study of 3 patients with biopsy-proven HCCA, the authors found variable characteristics, including identification of tumor vessels within the lesion and increased echogenicity within the center of the tumor.
The use of harmonic power Doppler US remains in the investigational phase, as researchers study the impact of technical parameters, such as pulse repetition frequency, wall filter settings, and injection rates on lesion detection. The decreased sensitivity of harmonic power Doppler US, in comparison with conventional power Doppler US on precontrast, is more than compensated for on contrast-enhanced imaging.
Through the evaluation of characteristics related to contrast-enhanced US, including portal-phase enhancement, negative washout (also called negative enhancement), arterial-phase peripheral nodularity and fill-in, and degree of arterial enhancement, algorithms have been developed that allow a logical and accurate differentiation of HCCA from other lesions, such as hemangioma or focal nodular hyperplasia. 
In the foreseeable future, the use of contrast in US is expected to reduce the necessity for additional corroborative imaging studies and to increase reliance on this already widely available, reasonably economical, and adaptable modality.
Real-time elastography is a promising technique for the noninvasive evaluation of the severity of hepatic fibrosis. This technique has been commercially developed by Hitachi Medical Systems and was used by Friedrich-Rust and colleagues to assess liver fibrosis in 79 patients with chronic viral hepatitis. Using a stepwise logistic regression analysis in patients and controls to define a tissue elasticity score, diagnostic accuracy was 0.75 for significant fibrosis, 0.73 for severe fibrosis, and 0.69 for cirrhosis, with a highly significant correlation (Spearman's correlation coefficient = 0.48) between the elasticity scores and the histologic fibrosis stage. 
Elastography is now widely used as a noninvasive test for staging fibrosis and is now being used in place of liver biopsy to investigate the natural history of chronic liver diseases; however, wide-scale outcome studies are not yet published. Its use is restricted in patients with acute hepatitis, obstructive cholestasis, and passive congestion, which can also alter liver stiffness. 
Role of US in biopsy and ablative therapy
The fact that US is readily available and can be used in the guidance of percutaneous biopsy and of ablative ethanol or acetic acid injection of focal lesions, as well as the fact that it can be employed in conjunction with radiofrequency (RF) probes for thermal ablation, means that selected patients to be evaluated, diagnosed, and treated using 1 modality. The use of US contrast agents (SonoVue®) has been helpful in differentiating viable from necrotic tissue, thereby improving diagnostic accuracy, particularly for lesions under 2 cm. Wu and coauthors achieved a better diagnostic accuracy for lesions that were evaluated by contrast enhancement prior to biopsy than for those that were evaluated with unenhanced US (97.1% vs 78.8%). 
In a representative study by Livraghi using percutaneous ethanol ablation, 5-year survival rates for patients with HCCA lesions that were smaller than 5 cm and who suffered from Child C, B, or A cirrhosis were 0%, 29%, and 47%, respectively.  Poorer results were obtained for multiple tumors or in the presence of portal thrombosis.
Degree of confidence
The presence of portal hypertension can be inferred based on the measurement of portal vein diameter; a sensitivity of 75% and a specificity of 100% for a diameter greater than 1.3 cm have been claimed. As previously noted, however, measurements of flow and vessel diameter are only indirectly related to portal pressure, and the degree and level of intrahepatic obstruction (presinusoidal or postsinusoidal), arterial flow to the liver, and capacitance of the collateral flow may affect flow parameters. Other findings, such as loss of respiratory variation in the diameter of the main portal vein or the presence of collaterals, are considered by Zimmerman and coauthors to be approximately 80% sensitive. 
Such a wide range of variability exists among patients that measurements of this nature should be considered useful only in research settings. If no other corroborative evidence has been obtained, caution should be used in interpreting these measurements as determinants of the presence of portal hypertension.
The ultrasonographic characteristics of HCCAs are variable, reflecting the diversity of neoplastic differentiation. However, certain pathologic characteristics occur with greater frequency and are helpful in characterizing hepatic lesions on ultrasonographic examination. For example, a pseudocapsule may be identified as a halo on ultrasonographic imaging. Neovascularity with arterial-venous shunting, the hallmark of malignant transformation, can be identified by current ultrasonographic systems once a lesion has reached approximately 2 cm. Contrast agents that increase the signal-to-noise ratio enable tumor vascularity to be detected with greater sensitivity.
In patients with cirrhosis attributed to multiple risk factors, Fasani and colleagues report that, compared with CT scanning, US appears to understage patients with multinodular lesions.  The sensitivity of US is also reduced in patients with heterogeneous livers. This understaging may be significant when considering patients for transplantation or ablative therapy, indicating that corroborative imaging with MRI or CT scanning may be of benefit in patients with advanced cirrhosis or a multifactorial etiology.
US combined with intravascular contrast agents
US appears to be very promising, particularly when combined with intravascular microbubble contrast agents, in assessing the effectiveness of tumor ablation. Choi studied the tumor characteristics of 40 patients with 45 nodular HCCA lesions 1-3.8 cm in diameter.  The patients were undergoing US-guided, percutaneous RF ablation with power Doppler US before and after intravenous injection of a microbubble contrast agent. In 33 of the 45 HCCAs, intratumoral flow was seen at power Doppler US before the administration of a contrast agent. After administration of the contrast agent, an increase in the degree of visualized flow was observed.
After RF ablation, none of the ablated tumors showed intratumoral flow signals at unenhanced power Doppler US, whereas 6 showed marginal intratumoral flow signals at contrast agent–enhanced power Doppler US. This correlated with enhancing foci that were suggestive of viable tumor in corresponding areas, as found at immediate follow-up with contrast-enhanced CT scanning. Thus, these preliminary data suggest that contrast-enhanced power Doppler US can be a promising noninvasive technique for assessing therapeutic response.
Regenerative nodules, dysplastic nodules, focal fat, and fatty sparing may mimic focal HCCA. Other nonmalignant hepatic neoplasms, such as hemangioma, may appear similar to HCCA, although arteriovenous (AV) shunts are uncommon. Focal nodular hyperplasia and liver cell adenoma may have extensive AV shunting, with this occurring most often in females.
The development of US contrast agents should further increase sensitivity; evidence suggests that the combination of advanced ultrasonographic imaging techniques (harmonic imaging) can increase the conspicuity of liver lesions (hence, the sensitivity of US when combined with microbubble contrast).
Functional imaging techniques using99m Tc-labeled sulfur colloid provide some indication of hepatic function. The agent is taken up by reticuloepithelial (RE) cells, and colloid shift to the other RE organs (bone marrow, spleen) provides indirect evidence of portal hypertension. In addition, heterogeneous uptake enables recognition of underlying hepatic dysfunction, as in the image below. Volumetric estimates of the liver can be made but have been superseded by other imaging techniques.
Fluorine-18 fluorodeoxyglucose (18F-FDG) is taken up by tumor cells, but the use of this agent in conjunction with positron emission tomography (PET) scanning appears to be more suited to larger, better-differentiated lesions. Therefore, at present, Trojan and colleagues believe that 18F-FDG PET is unlikely to replace the other techniques.  Sensitivity appears in the range of only 55%, compared with the 90% sensitivity of CT scanning, and Khan and coauthors report that better-differentiated tumors tend to have a lower level of uptake.  The prognostic implications of this finding have not been elucidated. In an investigation, Kim and colleagues expressed hope that functional imaging techniques may be able to predict tumor response to chemotherapy. 
Degree of confidence
Since the sensitivity of PET is relatively low, this modality is not at present recommended as a clinical screening tool for HCCA, and its use remains investigational. Kurtaran and coauthors report that it may be helpful in discriminating benign hepatic lesions, such as focal nodular hyperplasia (FNH), from malignant lesions, because there is reduced uptake in FNH. 
Angiography has evolved from an invasive modality used in the diagnostic evaluation of tumors and other complications of cirrhosis (in the decades prior to the introduction of noninvasive, cross-sectional imaging modalities) to an imaging method with a far more sophisticated interventional and therapeutic use. The angiographic characteristics of the hepatic circulation in cirrhosis (see the image below) and of tumor vascularity, including the demonstration of AV shunting characteristic of hepatocellular carcinoma (HCCA), were described several decades ago, and the knowledge of these characteristics now forms the cornerstone of our understanding of dynamic hepatic imaging by US, CT scanning, and MRI.
The angiographic demonstration of hepatic perfusion remains essential in transplant assessment, given the remarkable variability of the liver's arterial and venous drainage, although Smith and colleagues suggest that the development of computer-based, 3-dimensional rendering techniques may make this obsolete.  It is accepted that angiography, although useful in demonstrating vascular anatomy, is not the most sensitive technique available for the diagnosis of small, occult HCCAs.
Patients with asymptomatic noninvasive multifocal HCCA confined to the liver and preserved liver function are candidates for transcatheter arterial chemoembolization (TACE). Drug-eluting beads reduce liver toxicity and systemic drug exposure compared with conventional regimens. Clinical trials are in progress to investigate the potential value of molecular-targeted drugs and antiangiogenic agents such as sorafenib. 
An accurate assessment of portal hemodynamics of the patient with cirrhosis is necessary for prognostic purposes. Such assessment requires the measurement of hepatic wedge pressures, the direct measurement of right atrial pressure, and the measurement of inferior vena caval pressures above, within, and caudal to the liver (see the images below).
Degree of confidence
Escorsell and colleagues determined that portal pressure (evaluated as the HVPG) is an independent prognostic factor in variceal bleeding risk. Their evidence indicated that such risk is very low in patients in whom a reduction in variceal pressure of 20% or greater is achieved.  The measurement of portal pressures is necessary in patients who are undergoing pharmacologic intervention designed to reduce portal pressure gradients.
Wide individual variation was described by Bosch and Garcia-Pagan in response to pharmacologic agents administered to reduce portal pressures, which means that it is important to obtain follow-up measurements during the course of therapy.  As yet, no noninvasive substitutes for the HVPG have been identified.