Материал: Advanced Imaging of the Abdomen - Jovitas Skucas

Tumor detection using a combined unenhanced, gadolinium-enhanced, and ferumoxides-enhanced MRI was similar to that obtained with a combined CT arterial portography and biphasic CT hepatic arteriography (93); an advantage of the MRI approach is that invasive CT angiography is not necessary.

Detection accuracy of malignant hepatic tumors (a mix of hepatocellular carcinomas and metastases) by SPIO-enhanced MR imaging is superior to unenhanced MRI and is similar to CT arterial portography (94). Relative tumor detection sensitivities depend on tumor size; CT and US detect more smaller tumors than DSA or even possibly iodized oil CT; but the latter are more sensitive with a larger tumor volume.

Magnetic resonance is probably more sensitive than CT in showing that a capsule is present and in identifying vascular involvement.

General Imaging Findings

A differentiation of hemangiomas from hepatocellular carcinomas (and metastases) was discussed above (see Differential Diagnosis of Focal Tumors).

About half of hepatocellular carcinomas occur as solitary, discrete tumors, with the other half consisting of several to multiple focal lesions or diffuse liver infiltration (Fig. 7.27). A rare one is extrahepatic, connected to a lobe by a pedicle. In children, signs pointing toward a hepatocellular carcinoma are the presence of underlying liver disease and significant venous

ADVANCED IMAGING OF THE ABDOMEN

Figure 7.27. Multifocal hepatocellular carcinoma.

invasion, findings somewhat uncommon with a hepatoblastoma. Imaging findings in children tend to mimic those seen with a hepatoblastoma (Fig. 7.28).

Most hepatocellular carcinomas are rather vascular, supplied primarily by hepatic artery branches.Vascular invasion more often involves portal vein branches rather than hepatic veins. Arterioportal shunting and hemorrhage are common. The less common well-differentiated hepatocellular carcinomas tend toward hypovascularity.

Computed tomography is not reliable in detecting bile duct invasion. Cholangiography

LIVER

can identify an intraductal tumor (Fig. 7.29), at times mimicking a stone. Also, a tumor involving the liver hilum may obstruct bile ducts and result in intrahepatic biliary dilation without direct bile duct invasion.

Aside from fibrolamellar carcinomas, gross calcifications in a hepatocellular carcinoma are uncommon. An occasional one contains a central scar, which is more often detected in a carcinoma developing in a noncirrhotic liver than in a cirrhotic liver.

Computed Tomography

A characteristic unenhanced CT appearance of hepatocellular carcinomas is that of a hypodense to isodense tumor. Some isodense tumors contain a thin hypodense rim, but the margin between tumor and adjacent parenchyma tends to be poorly defined. Typical postcontrast findings consist of marked neovascularity, enlarged feeding arteries, a dense tumor vascular blush, normal or delayed contrast washout, and arterioportal shunting. Nevertheless, considerable CT variability exists and an occasional one is even hypovascular (Fig. 7.30). Some reveal an enhancing rim during the portal

Figure 7.29. Hepatocellular carcinoma growing within hepatic duct (arrow), an uncommon presentation.

Figure 7.30. Hepatocellular carcinoma. Contrast-enhanced CT identifies a tumor replacing part of the right lobe (arrows). (Courtesy of Patrick Fultz, M.D., University of Rochester.)

phase. In general, CT contrast-enhancing regions represent viable tumor tissue, while hypodense regions are either necrotic or contain fibrosis and hemorrhage.

What is the relative value of arterial phase, portal venous phase, and delayed phase CT in detecting these tumors? Arterial phase imaging yields the highest tumor-to-parenchyma contrast and detects more tumors than the other phases. During the arterial phase they tend toward intense homogeneous enhancement, a finding also seen with other hypervascular tumors, both benign and malignant. Larger tumors have heterogeneous arterial-phase enhancement, probably due to interspersed necrosis. Tumors missed on the arterial phase tended to be well differentiated, do not have a fibrous capsule, and are relatively hypovascular. Although different publications provide different tumor detection rates, relative tumor detection between various imaging phases is remarkably constant. A combination of arterial and portal venous phase images detects significantly more tumors than a combination of unenhanced and portal venous phase images. Any combination of two phases that includes the arterial phase is superior to a combination of portal and delayed phases. Or, for maximum tumor detection, all three phases should be obtained because a small minority of tumors are detected only on unenhanced images.

A novel technique consists of both early and late arterial-phase images obtained serially

during a single breath-hold multidetector CT study. Such double arterial-phase imaging yielded detection sensitivities of 54% for the early arterial phase, 78% for the late arterial phase, and 86% for the double arterial phase (95); the number of false-positive tumors was also reduced with double arterial-phase imaging.

The place of 3D liver CT reconstruction in evaluating hepatocellular carcinomas is not yet clear in spite of some obvious advantages. It is useful to the surgeon in defining vascular and tumor anatomy in specific segments when planning resection, yet at present 3D reconstruction is not widely practiced; considerable operator time and experience is necessary to achieve meaningful results.

One technique consists of reconstructing maximum intensity projection images of intrahepatic portal venous branches and hepatic veins, plus shaded surface display images of hepatic tumors from postcontrast CT data; when superimposed, these two sets of images provided a 3D relationship between vessels and tumors, aiding tumor localization prior to resection.

Do CT arterial portography and CT hepatic arteriography add significant information beyond what is obtained with triple-phase CT during preoperative evaluation of hepatocellular carcinomas? Adding CT arterial portography to triple-phase CT increased sensitivity from 94% to 96% and the further addition of CT hepatic arteriography increased it further to 97% (96); a disadvantage is that CT arterial portography and CT hepatic arteriography increase the false-positive rates without a corresponding gain in sensitivity.

Computed tomography hepatic arteriography aids, in part, in differentiating dysplastic nodules from hepatocellular carcinomas. Thus CT hepatic arteriography reveals that although high-grade dysplastic nodules range from hypoto hyperdense, a majority of low-grade dysplastic nodules are isodense (97); carcinomas, on the other hand, are mostly hyperdense, especially moderately to poorly differentiated ones.

A limitation of CT hepatic arteriography is that in some patients the entire liver cannot be studied because of an aberrant blood supply; examples include the presence of an aberrant

ADVANCED IMAGING OF THE ABDOMEN

left hepatic artery originating from the left gastric artery and occlusion of a right hepatic artery; injection of contrast into the hepatic artery does not opacify these segments and thus misses tumors there.

Computed tomography arterial portography involves placing an angiographic catheter into the splenic artery or superior mesenteric artery distal to the origin of any hepatic arteries. During portal-phase imaging the normal liver is markedly enhanced, with little enhancement of neoplasms receiving their blood supply from the hepatic artery, thus accentuating contrast differences between normal parenchyma and a neoplasm. Potentially this technique allows detection of smaller tumors. Similarly to CT hepatic arteriography, CT arterial portography aids in differentiating dysplastic nodules from hepatocellular carcinomas. Thus 77% of low-grade dysplastic nodules were isodense, while only 32% of high-grade dysplastic nodules were isodense (the rest ranged from slightly to markedly hypodense) (97); all welldifferentiated carcinomas ranged from slightly to markedly hypodense and moderately to poorly differentiated carcinomas all were markedly hypodense. Practical application of such CT correlation with histologic findings is yet to be established.

While CT arterial portography is rather sensitive, its specificity is rather low. Hemangiomas, focal nodular hyperplasia, regenerating nodules, and liver parenchyma being fed by aberrant vessels are detected, and in many instances these false-positive lesions cannot be distinguished from neoplasms. Follow-up double-phase CT arteriography appears useful in differentiating hepatic tumors, especially malignancies, from nonspecific perfusion abnormalities (called pseudolesions by some). These nonneoplastic perfusion abnormalities tend to be more common adjacent to the falciform ligament, gallbladder, and posterior edge of the medial segment. For unknown reasons these abnormalities occur less often adjacent to the falciform ligament in patients with cirrhosis.

Hepatocellular carcinomas not identified by CT arterial portography (i.e., false negative) tend to be central in location or associated with segmental portal vein thromboses obstructing the flow of contrast to the liver periphery.

LIVER

Computed tomography arterial portography is of limited value in a setting of portal hypertension, regardless of the etiology; insufficient iodine is delivered via the portal vein to the liver to achieve adequate contrast differences for tumor visualization. Enhancement is also decreased if portosystemic shunting has developed. In these settings CT arteriography is a better choice.

Malignant tumor size tends to be overestimated with CT arterial portography. Reasons are due to parenchymal compression, portal vein obstruction, or possibly a siphoning effect by these hypervascular tumors.

Although an iodized oil CT study (also called Lipiodol CT) is often used as a gold standard for tumor detection, it has distinct limitations. It is generally assumed that iodized oil is retained only by malignant cells, although evidence suggests that oil is also retained by other tumors. Iodized oil can accumulate in a hemangioma. Iodized oil uptake tends to be homogeneous in tumors smaller than 2cm in diameter, becoming inhomogeneous in larger ones, presumably secondary to necrosis.

It would be naive to think that iodized oil CT detects all tumors. Iodized oil CT appears to be a tarnished gold standard when examining tumors in explanted livers, with pretransplantation iodized oil CT sensitivity being rather low. Nevertheless, contrary to some studies, it often detects more tumors than other imaging modalities.

Occasionally spasm is encountered during arterial injection of Lipiodol, resulting in proximal Lipiodol reflux. Intraarterial buflomedil has been used to decrease arterial spasm.

Ultrasonography

Sonographic findings of hepatocellular carcinomas vary and are nonspecific. Sonography detects only a minority of hepatocellular carcinomas in patients with advanced cirrhosis, in one study identifying only 27% of tumors prior to liver transplantation (98). In the West, even small tumors range from (mostly) hypoechoic to (less often) hyperechoic, while in parts of Africa multinodular, hyperechoic tumors are more common. A thin hypoechoic rim with a hyperechoic center is evident in some; such a

target sign is nonspecific and is also seen with metastases and some benign tumors. A complex pattern develops with growth, and these tumors range from heterogeneous, hyperechoic, hypoechoic, and nodular, to diffuse infiltration. An occasional one mimics an abscess.

Ultrasonography depicts “hemangioma-like” lesions in some cirrhotic patients (99); followup revealed that half were hyperechoic hepatocellular carcinomas.

Color Doppler signals are detected both at the periphery and within vascular tumors; they are detected in a majority of hepatocellular carcinomas and focal nodular hyperplasias. Power Doppler US is superior to color Doppler US in visualizing hepatocellular carcinoma blood flow and identifying tortuous intratumoral vessels. Numerous studies have confirmed that contrast-enhanced power Doppler US is superior to unenhanced power Doppler US in detecting hepatocellular carcinoma tumor vascularity. Contrast-enhanced color Doppler US of most hepatocellular carcinomas reveals intratumoral signals, afferent vessels, and peripheral vascularity.

Doppler US–generated hepatic artery velocity histograms show hepatocellular carcinoma artery flow tending toward turbulent. Occasionally color Doppler US detects reversed portal blood flow adjacent to a hepatocellular carcinoma; such flow reversal is nonspecific and is also identified with metastases, abscesses, and even subcapsular hematomas. Color Doppler US detects a feeding artery in hepatocellular carcinomas more often after contrast; a feeding artery, however, is not pathognomonic of a hepatocellular carcinoma.

Carbon dioxide–enhanced US is used to study blood flow patterns. In some hands US enhanced with intraarterial carbon dioxide microbubbles detects more hepatocellular carcinomas than contrast-enhanced CT or even iodized oil-CT. After carbon dioxide microbubble injection into the hepatic artery, real-time B-mode US will identify blood drainage into a portal vein branch through a hepatocellular carcinoma, thus confirming arterioportal shunting.

Pulse-inversion harmonic US with preand serial postintravenous microbubble contrast injection reveals dense tumor staining, enhancement ranging from homogeneous to

heterogeneous with rapid washout and a surrounding late pseudocapsule enhancement; centripetal fill-in, a characteristic finding with hemangiomas, is not found with these cancers (Fig. 7.31).

Using IV contrast, intermittent harmonic power Doppler US achieved a sensitivity of 93% and specificity of 100% for depicting tumor vessels and tumor stain (100). Postcontrast, conventional power Doppler US depicts significantly more signals from hepatocellular carcinomas than harmonic power Doppler US (101). On the other hand, in spite of relatively short effective enhancement duration with harmonic Doppler US, it produces fewer artifacts.

Intraoperative US detects tumors not imaged preoperatively and not localized by palpation or inspection during surgery. Not all nodules represent a malignancy. This modality thus is of value in planning specific segmental resection. In an occasional patient a planned resection based on prior CT arterial portography is modified by intraoperative US.

Magnetic Resonance Imaging

On both T1and T2-weighted MRI, hepatocellular carcinomas range from hypointense to hyperintense compared to normal liver parenchyma. Most often, however, a hyperin-

ADVANCED IMAGING OF THE ABDOMEN

tense tumor on T1-weighted images suggests a hepatocellular carcinoma or a benign tumor; most metastases are hypointense. A majority of hepatocellular carcinomas are hyperintense on T2-weighted images. Small tumors tend to be isoor somewhat hyperintense on T1and isointense on T2-weighted images. Larger tumors tend toward heterogeneity on both T1and T2-weighted images. Diffuse hepatocellular carcinoma infiltration ranges from subtle, often minimal precontrast findings, to a mottled hyperintense appearance on T2-weighted images; a similar heterogeneous pattern is often evident on immediate postcontrast images. A higher T1-weighted signal intensity appears related to a higher degree of tumor differentiation, more steatosis, and higher tumor copper content than surrounding liver parenchyma (102).

Chemical shift MR with both opposed-phase and in-phase FLASH imaging detects fat in a minority of hepatocellular carcinomas. T1weighted fat-suppressed images appear superior to T1and T2-weighted images in detecting small hepatocellular carcinomas (103).

Paramagnetic ions contributing to the MR signal include copper and iron both in the tumor and in surrounding liver parenchyma. Excess copper accumulation in some hepatocellular carcinomas may account for part of the hyperintense T1-weighted signal (103), yet the

LIVER

amount of divalent copper present appears to be insufficient to account for all of this signal. The often high iron content within dysplastic nodules leads to a hypointense appearance on T2-weighted images, while a carcinoma developing within this nodule will be hyperintense, an appearance of a nodule-within-nodule.

If a capsule is present, T1-weighted images reveal a hypointense rim. T2-weighted images show a capsule and surrounding structures as a double layer consisting of an inner hypointense and an outer hyperintense layer. Underlying cirrhosis influences tumor appearance and makes MR interpretation more difficult. Tumors in cirrhotic livers are smaller, less often solitary and less likely to contain a central scar.

Current evidence suggests that contrastenhanced MRI is superior to other imaging techniques in detecting and providing a viable differential diagnosis for hepatocellular carcinomas, yet the critical evidence does not instill confidence even in MR, especially in a setting of cirrhosis. Thus pretransplantation 2D or 3D gadolinium-enhanced gradient-echo MRI using arterial, portal venous, and equilibrium phases in cirrhotic patients detected cancers only in 55% of patients shown to have cancers (104); on a lesion-by-lesion basis, MRI achieved an overall sensitivity of only 55%, missing 50% of neoplasms 1 to 2cm in diameter and 66% of tumors <1cm.

The most optimal type of MR contrast agent, be it a primarily hepatobiliary, superparamagnetic, or other agent, is not clear. In patients with malignant hepatic tumors (mostly hepatocellular carcinomas) gadolinium-enhanced MRI achieved greater sensitivity (81%) than ferumoxides-enhanced MR (62%) for tumor detection, but specificity was comparable (94%) (105); superiority of gadolinium was enhanced in patients with underlying cirrhosis.

Considerable MR variability in appearance is a hallmark of postcontrast hepatocellular carcinomas. For most tumors, the tumor-to-liver contrast ratio increases after intravenous gadolinium; increasing the gadolinium dose from the recommended 0.1mL/kg of body weight does not improve the tumor-to-liver contrast.

A typical appearance of a small hypervascular hepatocellular carcinoma in cirrhosis is early but mild contrast enhancement, rapid washout and little delayed enhancement except for

peritumoral coronal enhancement (106). Larger tumors have a heterogeneous pattern. The more vascular poorly differentiated hepatocellular carcinomas show marked initial enhancement, become isointense during portal venous phase and even hypointense later on. Some tumors exhibit delayed enhancement. A small minority are hypointense during the arterial phase. Also, any superimposed liver disease modifies the MRI appearance.

Lesions <1cm and isointense on precontrast images are best identified (and often only) by their immediate homogeneous postcontrast enhancement. Of hepatocellular carcinomas < 3cm, roughly half are detected on T1-weighted images, half on T2-weighted images and half on post-gadolinium T1-weighted images. Although some authors have published greater detection rates, relative detection preand postcontrast as outlined above is typical. T2-weighted image hyperintensity is related to expansive growth, peliosis, and hypervascularity, while hypointense or undetected tumors tending to be well differentiated. On postgadolinium T1weighted images hyperintensity is related to peliosis, with undetected tumors being well differentiated and hypovascular. In general, adding dynamic imaging increases tumor detection by 20% or so over precontrast imaging.

After gadobenate dimeglumine, welldifferentiated hepatocellular carcinomas tend to have a rapid increase in signal intensity during the arterial phase, while poorly differentiated carcinomas have a more delayed rise in signal intensity. Postcontrast, some hepatocellular carcinomas are surrounded by a hypointense rim on early images, with this rim enhancing later on; this enhancing rim appears to be caused by contrast draining through a (pseudo)capsule into a surrounding layer containing portal venules; arterial blood in a hepatocellular carcinoma drains into adjacent portal veins (107). Rim enhancement is evident in a majority of hepatocellular carcinomas during the portal venous phase.

Typical findings, if present, help differentiate these carcinomas from hemangiomas; the latter have early peripheral enhancement, a more marked but delayed peak enhancement, and continued delayed enhancement. Metastases tend toward peripheral enhancement, a finding lacking with hepatocellular carcinomas, which enhance throughout, albeit heterogeneously.