Computed Tomography Technology and Terminology

Three distinct generations of CT technology are in current use. In order of historical development and scanner speed, these are conventional CT, spiral (or helical) CT, and multislice (or multidetector) CT. Rapid scanning is a desirable feature in abdominal CT for two reasons. First, unlike the brain, the abdomen moves with respiration, and imaging during a single breath-hold removes the risk of respiratory misregistration (i.e., structures missed between slices because successive breath-holds are not identical). Second, many applications in abdominal imaging are affected by the time images are acquired relative to the bolus of intravenous contrast medium, and accurate diagnosis may require imaging precisely timed to one or more phases of enhancement. Godfrey Hounsfield invented CT in 1972, with initial clinical installations in 1974. The first CT scanner developed by Hounsfield in his laboratory at Electrical and Musical Industries required several hours to acquire the data for a single slice, and took several days to reconstruct the corresponding image. Data acquisition and image reconstruction became progressively faster during the 1970s and 1980s, although the speed of the scanners remained limited by the need for “stop-start,” slice-by-slice acquisition. That is, in conventional CT, an axial slice is generated by rotating an x-ray tube and detector array in a 360° circle around the patient. After a 360° rotation, the rotating gantry reverses direction to prevent disruption of the tethered cables that transfer the data from the detector array to the computer. Such sequential slice acquisition limits the speed of conventional CT, prevents volumetric data acquisition, results in slice misregistration, and limits temporal resolution such that multiphase volumetric scanning is not possible. The development of spiral (or helical) CT in the late 1980s represented a technologic breakthrough. In spiral CT, data is carried from the rotating gantry to the computer by slip rings, which allow continuous gantry rotation and data transfer. Scanning can be performed while the patient is moved slowly but continuously through the gantry. The ability to continuously scan allows for “non-stop” volumetric data acquisition. Data is gathered on a three-dimensional volume in a spiral fashion. Images are reconstructed from the data volume. Physical and clinical performance studies of spiral CT were first reported at the 1989 Radiological Society of North America meeting.¹ The first commercially available spiral CT scanner was released in 1990.²

Spiral CT was an important technologic advance, but many limitations and compromises remained. The rotational speed of the gantry was generally one rotation per second, and clinical results suggested that a pitch (ratio of longitudinal distance moved by the tabletop during one gantry rotation to beam collimation) greater than 2 was undesirable. As a result, either large volumes could be covered, or thin sections acquired, but not both. The Elscint CT Twin scanner employed a dual array of two side-by-side detector rows, and represented an early version of multislice technology.³ However, although this scanner was an improvement over single-detector spiral CT in terms of coverage,⁴ substantial limitations remained. It was not until 1998 that a scanner with four detector rows became commercially available. Modern multislice CT using four or more detector rows essentially abolishes the remaining obstacles to rapid isotropic volumetric imaging by using multiple side-by-side detectors simultaneously. During recent years, top-of-the-line scanners have used 64 detector rows, although machines with 256 rows or flat-panel detectors are being investigated and will likely become more widely available in the next few years. Such multislice CT systems can collect multiple slices of data in approximately 300 milliseconds and reconstruct a 512 × 512 matrix image from millions of data points in less than a second. An entire body cavity (brain, chest, or abdomen) can be scanned in 5 to 10 seconds using the most advanced multislice CT systems. Faster imaging with modern scanners is due not only to multiple detector rows, but also to increased gantry rotational speed. Many commercial scanners are now capable of a complete gantry rotation in 0.3 seconds or less. Multislice CT has been a major advance in several areas of abdominal imaging, particularly those requiring imaging in multiple phases of enhancement after intravenous contrast medium administration (Fig. 19-1) and rapid near-isotropic (i.e., equal spatial resolution in all three planes) volumetric acquisition, such as CT angiography (Fig. 19-2). The ability to acquire scans of the abdomen during multiple phases of enhancement is particularly helpful when both arteriographic and parenchymal phases of enhancement are required for complete evaluation. For example, the arteriographic phase is critical for the assessment of arterial encasement in pancreatic cancer, but parenchymal enhancement phases are required for demonstration of the primary tumor and detection of hepatic metastases.

FIGURE 19-1 • The ability to obtain multiple sequential images in different phases of enhancement is a major advantage of multislice computed tomography (CT), because different parameters are optimally evaluated in different phases of enhancement. A, Multislice CT performed during the arteriographic phase of enhancement in a 67-year-old patient with pancreatic cancer. The superior mesenteric artery (arrow) is well opacified, and is not encased by tumor. B, Image obtained in a slightly later phase of enhancement shows a greater degree of parenchymal enhancement. The tumor (asterisk) is well seen and is predominately hypovascular. C, Image obtained in the portal venous stage of enhancement and displayed at a liver window shows a small, hypodense lesion (arrow) in the liver, which may be a metastasis.

FIGURE 19-2 • Computed tomography (CT) arteriogram in a 26-year-old patient being evaluated as a potential living, related, right hepatic lobe donor. An aberrant right hepatic artery (vertical arrow) is seen arising from the superior mesenteric artery (horizontal arrow). With the advent of multislice CT, noninvasive arteriography can be performed routinely and with near-isotropic resolution (i.e., equal spatial resolution in all three planes), and the era of diagnostic catheter angiography is coming to an end.

A potential source of confusion that arises with respect to multislice CT is the definition of pitch. Pitch can be defined in one of two ways: either as the ratio of longitudinal distance moved by the tabletop during one gantry rotation to beam collimation, or the ratio of longitudinal distance moved by the tabletop during one gantry rotation to slice thickness. In single-detector spiral CT, these two definitions are the same, because the slice thickness is the beam collimation. In multislice CT, the collimated beam is split into several slices, and therefore the definitions of pitch are different. For example, a four-channel multislice scanner acquiring 2.5-mm thick slices (beam collimation of 10 mm) at a tabletop speed of 15 mm per 1-second gantry rotation could be defined as having a pitch of 6 (longitudinal distance moved by the tabletop during one gantry rotation ÷ slice thickness = 15 ÷ 2.5 = 6) or 1.5 (longitudinal distance moved by the tabletop during one gantry rotation ÷ beam collimation = 15 ÷ 10 = 1.5). There is no current consensus on which definition is preferable. Although most manufacturers use the ratio of longitudinal distance moved by the tabletop during one gantry rotation to slice thickness, there are arguments related to basic science and calculation of radiation dose that suggest the alternative definition is more appropriate. In any event, it is important to clarify which definition of pitch is being used when describing CT protocols and parameters for multislice scanners.

One of the unfortunate consequences of the increased ability to obtain multiphase imaging has been a corresponding increase in confusion with respect to the proper terminology for these studies.⁵ It has reasonably been suggested that the term phase should not be applied to noncontrast images.⁶ It also seems reasonable to use the term arteriographic phase for early arterial images in which contrast is primarily within the arteries, and the term parenchymal arterial phase for the subsequent period of early tissue enhancement when hypervascular masses are most prominent. Later periods of image acquisition can be referred to as portal venous or delayed, as appropriate. Until a universal terminology is established, it is probably best to avoid unclear terms such as dual or triple phase studies, and instead describe the specific phases acquired.

Cholangiocarcinoma

Imaging of Cholangiocarcinoma

The traditional standard of reference for the diagnosis and staging of cholangiocarcinoma has been the combination of conventional cholangiography and angiography.⁹ Modern cross-sectional imaging has largely replaced the need for these invasive tests, both with respect to diagnosis and staging. The diagnosis of cholangiocarcinoma should be suspected when US, CT, or MRI demonstrate dilated intrahepatic ducts with nondilated extrahepatic ducts in a patient with painless jaundice. Because of the infiltrative nature of cholangiocarcinoma, the primary tumor may or may not be seen, and visualization of a mass is not required to suggest the diagnosis. Potential differential diagnoses in this setting include postsurgical stricture (either iatrogenic bile duct injury or biliary enteric anastomotic stricture) or malignant masquerade (other hepatic masses, such as metastases, very rarely result in biliary dilatation). Postsurgical strictures are usually evident from the clinical history. Malignant masquerade is a rare nonneoplastic fibrous proliferation that occurs at the hilum of the liver and clinically and radiologically mimics perihilar cholangiocarcinoma.¹⁰ In a series of 94 patients with suspected obstructive malignancy at the hilum, 8 were found to have malignant masquerade at pathologic analysis.¹⁰ The 8 patients ranged in age from 37 to 66 years, and all presented with obstructive jaundice. One later developed cholangiographic changes of sclerosing cholangitis, and it may be that malignant masquerade represents a localized, masslike form of sclerosing cholangitis.¹¹

Conventional Cholangiography

Conventional cholangiography can be performed by endoscopic retrograde cholangiopancreatography (ERCP) or percutaneous transhepatic cholangiography (PTC). ERCP has the advantage of excluding ampullary pathologic conditions by direct endoscopic evaluation and allowing for simultaneous tissue sampling. Palliative endobiliary stenting can also be performed at the same time. ERCP is usually favored more than PTC, but the latter may be appropriate when attempts at ERCP have been unsuccessful. PTC may allow access to proximal lesions with obstruction of both right and left hepatic ducts. Tissue may be obtained for cytologic analysis and drainage performed. The typical appearance of a cholangiocarcinoma at conventional cholangiography is of a tight, irregular stricture of variable length associated with marked dilatation of the proximal biliary tract (Fig. 19-3).

FIGURE 19-3 • Endoscopic retrograde cholangiopancreatography in a 64-year-old woman with cholangiocarcinoma. A tight irregular stricture (arrow) is seen in the mid to upper common duct and is associated with moderate to marked dilatation of the intrahepatic biliary system.

Ultrasound

US is frequently the initial investigation requested in patients with suspected biliary obstruction. Klatskin tumors classically manifest at US as segmental dilatation and nonunion of the right and left hepatic ducts in the porta hepatis (Fig. 19-4). The papillary and nodular ductal variants of cholangiocarcinoma are relatively easy to see at US. Papillary tumors form polypoid intraluminal masses, whereas nodular cholangiocarcinoma manifests as a discrete smooth mass with associated mural thickening.¹² Smaller or infiltrative tumors may be missed, although experienced operators can demonstrate high sensitivity. In a retrospective study of 49 patients with pathologically proved cholangiocarcinoma, US demonstrated the tumor in 47 (96%) cases; a mass was demonstrated in 44 and focal or diffuse thickening of the bile duct wall in 3.¹³ In a study of 39 patients with Klatskin tumors undergoing preoperative US,¹⁴ ductal masses were detected in 34 patients (87%). Masses were isoechoic in 22 (65%), hypoechoic in 7 (21%), and hyperechoic in 5 (15%). Morphologic tumor patterns included nodular mural thickening (n = 19; 56%), irregular infiltration (n = 9; 26%), and intraductal polypoid masses (n = 6; 18%). Vascular involvement is an important determinant of irresectability, and portal vein involvement can be assessed with an accuracy of up to 91% by Doppler US.¹⁵ In a study of 22 patients with hilar cholangiocarcinoma, vascular involvement was evaluated by Doppler US and arteriography, using operative findings as the standard of reference. Vascular patency or involvement was correctly determined by Doppler US in 19 patients (86%), compared with 18 by arteriography (82%).¹⁶ EUS enables both bile duct visualization and nodal evaluation, and EUS-guided fine-needle aspiration biopsy may be diagnostic when other diagnostic tests are inconclusive.¹⁷

FIGURE 19-4 • Ultrasound in a 56-year-old man with cholangiocarcinoma. A subtle mass (asterisk) is seen in the central liver. Dilated bile ducts (horizontal arrows) are visible at the periphery of the mass. A plastic endobiliary stent (vertical arrow) traversing the tumor is also visible.

Computed Tomography

Depending on referral practices, CT may be the first investigation requested in patients with painless jaundice. The CT findings in Klatskin tumor resemble those seen at US, namely segmental dilatation and nonunion of the right and left hepatic ducts in the porta hepatis (Fig. 19-5). The reported frequency with which a primary tumor mass is seen varies from 40% to 91%.^18–21 Delayed enhancement is a relatively recently described finding that has been reported in 36% to 74% of cases (Fig. 19-6).^20,21 In one study, 3 of 47 cholangiocarcinomas were seen only in delayed images.²¹ The optimal time for acquisition of delayed images is 10 to 20 minutes after contrast media injection.²⁰ Lobar atrophy is a nonspecific but useful ancillary feature of cholangiocarcinoma that may be better appreciated at CT than US. Lobar atrophy usually indicates both biliary and portal venous obstruction in that lobe. Another advantage of CT is the ability to detect metastases, such as nodal or peritoneal deposits. Adenopathy in cholangiocarcinoma arising secondary to primary sclerosing cholangitis should be interpreted with caution, because nodal enlargement in this context may be reactive or metastatic. Occasionally, CT may suggest the diagnosis of peripheral intrahepatic cholangiocarcinoma when a patient with vague or nonspecific symptoms is found to have a hypodense hepatic mass with peripheral enhancement, biliary dilatation, and delayed enhancement.²²

FIGURE 19-5 • Contrast-enhanced computed tomography in a 72-year-old man with cholangiocarcinoma. A hypodense mass (black arrow) is present in the central liver, with associated marked biliary dilatation (white arrow) in the left hepatic lobe. The left lobe is also atrophic.

FIGURE 19-6 • A relatively recently recognized characteristic of cholangiocarcinoma is that the tumor often demonstrates delayed enhancement, and some cholangiocarcinomas may only be seen several minutes after administration of intravenous contrast. A, Contrast-enhanced computed tomography (CT) in a 63-year-old woman with cholangiocarcinoma. The tumor (arrow) is hypovascular during the portal venous phase of enhancement. B, CT obtained 10 minutes after the administration of intravenous contrast. The tumor (arrow) is now slightly hypervascular with respect to the adjacent hepatic parenchyma.

Magnetic Resonance Imaging

MRI findings in cholangiocarcinoma resemble those seen at CT: isolated intrahepatic biliary dilatation with or without a visible mass (Fig. 19-7). Although MRI is frequently reserved for patients with contraindications to contrast-enhanced CT, such as elevated creatinine or contrast allergy, the modality has several attractive imaging characteristics that may assist in tumor evaluation. MRI cholangiography provides similar information to conventional cholangiography, and may obviate the need for invasive biliary imaging.²³ MRI angiography can be used to assess vascular involvement and may replace catheter angiography for vascular evaluation. In a study of 27 preoperative patients with Klatskin tumors, evaluation of vascular involvement by MRI and US were compared with intraoperative findings.²⁴ Encasement, thrombosis, occlusion, and nonvisualization were considered to be evidence of vascular involvement. Of 16 involved veins, 12 were detected at MRI compared with 13 of 16 at US. There was one false-positive diagnosis of inferior vena cava involvement at both MRI and US. These results suggest that MRI and US provide comparable results for assessment of hepatic vein involvement by tumor. Central hypointensity on T2-weighted images appears to be a characteristic finding in intrahepatic cholangiocarcinoma (Fig. 19-8), and reflects severe fibrosis.²⁵

FIGURE 19-7 • Magnetic resonance imaging scan in a 65-year-old man with cholangiocarcinoma. A, T1-weighted image demonstrates a hypointense mass (asterisk) replacing much of the left hepatic lobe. B, After the administration of intravenous gadolinium, the mass demonstrates T1-weighted image demonstrates heterogeneous enhancement. C, T2-weighted image shows a linear region of central hypointensity (arrow) within the mass.

FIGURE 19-8 • Magnetic resonance imaging scan in a 55-year-old woman with cholangiocarcinoma. A, T2-weighted coronal image shows moderately dilated intrahepatic bile ducts (black arrow) to the level of an ill-defined mass (white arrow) in the central liver. B, The mass (arrow) is well seen on this T1-weighted image after the administration of intravenous gadolinium.

Positron Emission Tomography

Early results suggest PET scanning may have a role to play in the evaluation of cholangiocarcinoma,²⁶ because it may detect cholangiocarcinoma tumors as small as 1 cm in diameter and can simultaneously detect distant metastases. In a study of 54 patients, 26 with cholangiocarcinoma and 28 with no disease or benign disease, true-positive PET scans were obtained in 24 of 26 cholangiocarcinoma patients with only two false-positive results (sensitivity of 92% and specificity of 93%). Regional or hepatoduodenal lymph node metastases were detected with PET in only 2 of 15 cases, whereas distant metastases (peritoneal carcinomatosis, pulmonary metastases) were diagnosed in 7 of 10 cases.

Gallbladder Carcinoma

Imaging of Gallbladder Carcinoma

GBC is usually advanced at the time of diagnosis, and early-stage tumors are rarely encountered.²⁷ As a result, imaging findings are usually relatively obvious, and can be well assessed by US, CT, or MRI. Radiologic findings include focal or diffuse thickening of the gallbladder wall (49%), a solid mass replacing the gallbladder in the gallbladder fossa (37%), or an intraluminal mass in the gallbladder (14%). Associated findings include invasion of adjacent structures such as the liver (67%), coexistent gallstones (64%), biliary duct dilatation (38%), coexistent porcelain gallbladder (4%), and distant metastases outside the liver (3%).²⁸ Although gallbladder cancer may manifest as an intraluminal mass at imaging, it should be remembered that gallbladder polyps may be found in up to 2.5% of asymptomatic patients²⁹ and that most small, noncalculous gallbladder filling defects are benign. In a study of 143 gallbladder polyps less than 1.5 cm in diameter removed by laparoscopic cholecystectomy, the final diagnoses were cholesterol polyp (74%), adenoma (11%), adenomyomatosis (7%), adenocarcinoma (4%), and hyperplastic or inflammatory polyp (4%).³⁰ Historical reports have suggested the risk of GBC in patients with a porcelain gallbladder is between 12% and 61%,³¹ and such numbers are frequently quoted in clinical practice. Two recent series of 10,741 and 25,900 cholecystectomy specimens from the University of California, Los Angeles, and Massachusetts General Hospital, respectively, have suggested the true risk is 0% to 7%.^31,32

Ultrasound

Sonographic findings of GBC include variable degrees of gallbladder wall thickening and irregularity (Fig. 19-9). Concomitant gallstones are frequently seen. Gallbladder wall thickening can also be seen with chronic cholecystitis and adenomyomatosis, but the wall thickening of these inflammatory and hyperplastic conditions is usually more diffuse than the relatively focal changes typical of GBC. GBC may also manifest as one or more intraluminal masses that are typically of homogeneous echotexture without acoustic shadowing. The polyps are usually sessile and only rarely have a stalk. With more advanced disease, US may demonstrate a necrotic mass, direct hepatic invasion, biliary dilatation, or hepatic metastases. US is of limited utility in the evaluation of nodal, peritoneal, or more distant metastases. Occasionally tumefactive sludge may mimic an intraluminal mass, but mobility on changes in patient position and absence of flow on Doppler insonation help in the correct identification of sludge.³³ EUS has been used in some centers to study GBC, and early results suggest this modality may be helpful in assessing the depth of wall invasion.³⁴ Loss of the normal multiple-layer pattern of the gallbladder wall has been reported as the most specific finding at EUS in the diagnosis of GBC.³⁵

FIGURE 19-9 • Gallbladder cancer has three main gross morphologic patterns, as illustrated in these three different cases. A, Ultrasound (US) of the gallbladder in an 80-year-old woman shows irregular circumferential wall thickening (arrow). B, US of the gallbladder in a 75-year-old woman shows an irregular polypoid mass (arrow) extending into the lumen. C, US of the gallbladder in an 87-year-old woman shows complete replacement of the gallbladder by a large irregular mass (asterisk). The mass has infiltrated the adjacent common duct, which contains a plastic endobiliary stent (arrow).

Computed Tomography

CT is helpful in both the diagnosis and staging of GBC, because it is able to detect gallbladder masses, wall thickening, and hepatic invasion (Fig. 19-10). Spiral CT has an accuracy of 83% to 86% in the local staging of GBC.³⁶ The reported sensitivity for the detection of nodal metastases was more limited, ranging from 36% to 47%. CT has reasonably high positive predictive values for the detection of hepatic invasion of less than 2 cm (77%), hepatic invasion of more than 2 cm (100%), involvement of the common duct (90%), or invasion of the pancreas (100%).³⁷ Xanthogranulomatous cholecystitis is a rare inflammatory process that cannot be reliably distinguished from GBC on CT scans because of multiple overlapping features such as gallbladder wall thickening and involvement of the surrounding tissues, including portal lymph nodes, fat, and the liver.

FIGURE 19-10 • Contrast-enhanced computed tomography of the gallbladder in a 74-year-old man. The gallbladder is almost completely replaced by a large cancer (asterisk) that is also invading the adjacent liver (arrow).

Magnetic Resonance Imaging

MRI findings in GBC resemble those seen at CT (Fig. 19-11). MRI and magnetic resonance cholangiopancreatography (MRCP) can provide information relevant to preoperative staging of GBC. GBC manifests at MRI as focal gallbladder wall thickening with an eccentric mass in 76% of cases. The most common types of regional spread demonstrated are direct hepatic invasion (91%), lymphadenopathy (76%), and biliary tract invasion (62%). Sensitivity for direct hepatic invasion is reported as 100%, and sensitivity for lymph node metastasis as 92%.³⁸ In general, MRI findings are analogous to CT scanning. The tumor is usually poorly marginated and bright on T2-weighted images.

FIGURE 19-11 • Magnetic resonance imaging scan in a 85-year-old woman with gallbladder carcinoma. A, T2-weighted coronal image demonstrates a mass (asterisk) of intermediate intensity replacing the gallbladder and invading the liver (vertical arrow). The mass also encases the adjacent common duct, which contains a stent (horizontal arrow). B, A contiguous T2-weighted coronal image demonstrates a small peritoneal implant (arrow) adjacent to the hepatic flexure. Peritoneal spread of biliary malignancy is not unusual, and should be specifically sought out.

Pancreatic Cancer

Anatomy

The pancreas is an exocrine and endocrine gland, situated transversely in the retro peritoneum, at the level of the L1 and L2 vertebrae. The head lies within the duodenal sweep, and the body and tail extend to the gastric border of the spleen. Although several systems for describing the anatomic subdivisions of the pancreas have been proposed, the tumor-node-metastasis classification is straightforward and divides the pancreas into head, body, and tail.³⁹ According to this classification, the head lies to the right of the superior mesenteric vein, the body lies between the left border of the aorta and the superior mesenteric vein, and the tail lies to the right of the left border of the aorta (Fig. 19-12). The pancreatic head lies posterior to the stomach and transverse colon, and anterior to the inferior vena cava. The common bile duct courses through the pancreatic head to the ampulla of Vater. The pancreatic body lies anterior to the superior mesenteric vessels. The pancreatic tail extends laterally to abut the spleen and splenic flexure of the colon, and lies anterior to the left kidney. The blood supply of the pancreas is derived from the pancreaticoduodenal branches of the gastroduodenal and superior mesenteric arteries and from direct branches of the splenic artery. Pancreatic venous drainage is through the portal system to the liver. The pancreas has an extensive system of lymphatic drainage, primarily to superior and inferior pancreaticoduodenal, porta hepatis, and suprapancreatic nodes, but also to para-aortic nodes.

FIGURE 19-12 • The pancreas may be divided into the head, body, and tail. The heads lies to the right of the left border of the superior mesenteric vein (SMV). The tail lies to the left of the left border of the aorta (Ao). The body lies between the head and tail.

Imaging of Pancreatic Carcinoma

Imaging modalities for pancreatic carcinoma include transabdominal US, EUS, CT, MRI, PET, and conventional cholangiography.⁴³ Apart from distant metastases, involvement of the major vessels is the most important parameter for determining resectability in patients with pancreatic adenocarcinoma. Two common problems arise in imaging pancreatic carcinoma across all these modalities. First, pancreatic carcinoma is often infiltrative and the tumor may be poorly defined and difficult to appreciate. Even quite large tumors can be of similar appearance to the adjacent pancreatic parenchyma, and difficult to visualize. Such tumors may be diagnosed predominantly through the indirect findings of pancreatic or biliary dilatation (Fig. 19-13). Second, pancreatic carcinoma may be hard to distinguish from focal chronic pancreatitis.

FIGURE 19-13 • Computed tomography (CT) is the primary modality used to evaluate patients with suspected pancreatic cancer, but tumors may be infiltrative, isodense, and difficult to directly visualize. A, Contrast enhanced CT section in a 56-year-old man with painless jaundice and weight loss shows dilatation of the pancreatic duct (white arrow) with atrophy of the surrounding pancreas. These findings are indirect signs that strongly suggest the presence of a malignant obstruction. B, CT section at a more inferior level, where the pancreatic and common bile ducts are no longer dilated, shows the pancreatic head (arrow). No definite mass can be seen. At surgery, a 3-cm pancreatic carcinoma was found in the pancreatic head.

Ultrasound

Transabdominal US is of limited utility in the diagnosis of pancreatic carcinoma, because of suboptimal acoustic contrast and variable obscuration of the pancreas by overlying bowel. US may be helpful in the initial workup of patients, because it can distinguish obstructive from nonobstructive jaundice and thus guide further investigations. EUS is an evolving modality that shows great promise in the diagnosis and staging of pancreatic carcinoma. EUS remains the most sensitive and specific method to identify pancreatic masses. Early results suggest EUS is more accurate than CT for staging pancreatic malignancy, including the evaluation of vascular invasion and local resectability. In a study of 151 patients with pancreatic carcinoma, the overall accuracy for T and N staging was 85% and 72% for EUS compared with only 30% and 55% for CT, respectively. The ability to accurately predict vascular invasion was 93% for EUS compared with 62% for CT. EUS was 93% accurate for predicting local resectability versus 60% for CT.⁴⁴ In a study of 21 patients with pancreatic carcinoma, EUS demonstrated an accuracy of 81% in the assessment of vascular involvement compared with 38% for selective venous angiography.⁴⁵ EUS can also be combined with targeted fine-needle aspiration biopsy, and has shown a sensitivity of 93% and a specificity of 100% in establishing the correct diagnosis when employed in patients with pancreatic masses in whom prior biopsies have been nondiagnostic.⁴⁶

Computed Tomography

CT is the most frequently used modality for the evaluation of pancreatic carcinoma. CT-guided biopsy can be performed to provide a tissue diagnosis. The primary finding on contrast-enhanced CT is a focal hypodense mass (Fig. 19-14). The sensitivity of spiral CT for the detection of pancreatic carcinoma is between 89% and 97%.⁴⁷ The overall accuracy of spiral CT in the determination of resectability has been reported as 77%.⁴⁷ In a study of 80 evaluable major vessels in 25 patients, tumor contact with more than 50% of the vessel circumference demonstrated a sensitivity of 84% and specificity of 98% in the diagnosis of vascular invasion (Fig. 19-15).⁴⁸ The development of spiral CT has allowed imaging of pancreatic carcinoma in several different phases of enhancement, and each may provide useful information. Arteriographic images are helpful in the evaluation of arterial encasement by pancreatic carcinoma.⁴⁹ With respect to tumor detection, some studies have suggested tumor-to-background contrast is greater in the late arterial phase (scan delay of 40 to 70 seconds, sometimes called the pancreatic phase) than in the standard portal venous phase,^50,51 although other studies have suggested late arterial phase images do not improve pancreatic tumor detection when compared with standard portal venous phase images.^52,53 Accordingly, baseline studies for pancreatic carcinoma can reasonably be protocoled to include three phases: early arterial (arterial encasement), late arterial (tumor detection), and portal venous (hepatic evaluation).

FIGURE 19-14 • Contrast enhanced computed tomography in a 64-year-old woman with pancreatic carcinoma. A, Section through the pancreas demonstrates marked ductal dilatation (arrow) and atrophy of the surrounding gland. B, Section at a more inferior level shows the malignant mass (arrow) in the pancreatic head.

FIGURE 19-15 • Contrast enhanced computed tomography in a 58-year-old man with pancreatic carcinoma. A, Section through the origin of the celiac artery (black arrow) shows infiltrative tumor (white arrow) surrounding the vessel. Tumor contact with more than 50% of the circumference of the vessel is strongly predictive of vascular invasion and irresectability. B, Section at a more inferior level showing the celiac artery (black arrow) is both narrowed and encased by tumor.

Conventional Cholangiography

After CT, ERCP is perhaps the second most common investigation performed in patients with suspected pancreatic carcinoma (Fig. 19-16). Brush cytologic examination and forceps biopsy can be used to provide a tissue diagnosis. Endobiliary stents can also be inserted at the time of ERCP, although this may not always be appropriate in patients who are surgical candidates; two separate studies of periampullary tumors and proximal cholangiocarcinoma have shown that preoperative endobiliary stenting is associated with increased bacterial contamination of the biliary system, wound and intra-abdominal complications, and postoperative infections.^54,55 The principal disadvantage of conventional cholangiography, such as ERCP or PTC, is that these invasive procedures have a major complication rate on the order of 3%. Such complications include sepsis, bleeding, bile leak, and death.^56,57 EUS may be the most accurate test for the diagnosis of pancreatic cancer. Studies comparing EUS with CT have shown that EUS has a higher sensitivity and specificity for this diagnosis of pancreatic cancer, particularly for tumors less than 3 cm in diameter.⁵⁸ In addition, EUS is highly accurate for detecting local invasion and nodal metastases from pancreatic cancer, although results are similar when compared with dual-phase multislice CT. CT does provide additional information about hepatic metastases.⁵⁹ The side-viewing duodenoscope that delivers the US probe also permits the detection of ampullary and duodenal carcinomas.

FIGURE 19-16 • Endoscopic retrograde cholangiopancrea-tography in a 69-year-old man with pancreatic carcinoma. A tight irregular stricture (arrow) is seen in the distal common bile duct, with marked biliary dilatation above the stricture.

Magnetic Resonance Imaging

MRI is mainly used as a problem-solving modality in patients with suspected pancreatic carcinoma, typically in patients with an equivocal CT scan or with contraindications to iodinated contrast such as allergy or elevated creatinine. However, the ability of MRI to simultaneously image the pancreatic parenchyma and vasculature with gadolinium enhancement and the pancreaticobiliary system with MRCP provides an attractive combined assessment that may be superior to more traditional methods (Fig. 19-17).^60–62 For example, in a study of 124 patients with suspected pancreatic carcinoma in whom the standard of reference was biopsy or 1 year of clinical follow-up, the sensitivity and specificity of MRCP for the diagnosis of pancreatic cancer was 84% and 97%, respectively, compared with a sensitivity and specificity of 70% and 94% for ERCP.⁵⁸ Another study concluded that mangafodipir trisodium-enhanced MRI was as accurate as contrast-enhanced spiral CT for the detection and staging of pancreatic cancer but offers improved detection of small pancreatic metastases and of liver metastases.⁶¹ In a study of 33 patients with suspected pancreatic carcinoma, MRI was significantly better in the assessment of resectability than spiral CT.⁶³

FIGURE 19-17 • Magnetic resonance imaging scan in a 75-year-old man with pancreatic carcinoma. A, T2-weighted axial section shows a mildly hyperintense mass (arrow) in the pancreatic head. B, T1-weighted axial image after the administration of intravenous gadolinium demonstrates the mass (arrow) to be mildly hypointense.

Positron Emission Tomography

Initial studies suggest PET scanning may be helpful in the diagnosis of pancreatic carcinoma, detection of distant metastases, and evaluation of recurrent disease. In a study comparing PET and CT for tumor diagnosis in 73 patients with suspected pancreatic malignancy, PET had higher sensitivity (93% versus 80%) and specificity (93% versus 74%).⁶³ In another study of 65 patients with suspected pancreatic carcinoma, PET again demonstrated higher sensitivity and specificity than CT in the correct diagnosis of pancreatic carcinoma (92% and 85% versus 65% and 61%).⁶⁴ PET may also be helpful in the detection of small metastatic deposits that would preclude any attempt at surgical treatment (Fig. 19-18).

FIGURE 19-18 • Positron emission tomography/computed tomography (PET/CT) study in a 67-year-old woman with pancreatic carcinoma undergoing assessment for potential surgical resection, showing the benefit of PET in detecting and characterizing small tumor deposits that alter management. A, Axial PET image shows a large focus of increased fluorodeoxyglucose uptake in the primary tumor (black arrow), but also shows a second smaller separate focus of uptake (white arrow). B, Corresponding axial contrast-enhanced CT section demonstrates the large primary tumor (black arrow) and also confirms a second smaller metastatic tumor implant in the mesenteric root (white arrow) that precludes any attempt at surgical treatment.

Retroperitoneal Lymphadenopathy

Imaging of Retroperitoneal Adenopathy

Retroperitoneal adenopathy can be defined as an abnormal increase in the number or size of retroperitoneal nodes. This is a clinically useful operational definition, but includes an element of subjectivity as to what constitutes “abnormal.” As a result, retroperitoneal adenopathy is often defined in imaging studies as the presence of nodes with a short-axis diameter of 1 cm or more. Although this definition has the advantage of objectivity, it does not really reflect daily practice. It should be noted that both definitions are based on nodal size, with the implicit assumption that nodal size reflects pathologic condition. As we demonstrate, this assumption is of limited validity; normal-sized nodes can contain microscopic metastases and enlarged nodes may be reactive and not cancer-containing. Ideally, imaging evaluation would use a parameter that reflects node content rather than node size. Lymphangiography and PET are two modalities that at least partially reflect node content, although neither is routinely used for retroperitoneal nodal assessment, and CT and MRI remain the primary modalities in common use. US is of limited utility in evaluation of the retroperitoneum because of poor acoustic contrast and variable obscuration by overlying bowel gas, although large nodal masses may sometimes be visible.

Computed Tomography

Normal retroperitoneal lymph nodes are either invisible or appear on CT as small round or oval densities less than 1 cm in diameter (Fig. 19-19). Abnormal nodes are also seen as soft-tissue densities, but of greater size and number. Adenopathy can occasionally be confluent or calcified. The accuracy of CT evaluation of retroperitoneal nodes is partially disease-specific, and this is illustrated by describing nodal assessment in testicular, endometrial, ovarian, and renal cancers. Right-side testicular cancer typically spreads to interaortocaval or paracaval nodes at the level of the entry of the right renal vein into the inferior vena cava (Fig. 19-20). Left-sided testicular cancers typically spread to left para-aortic nodes at the level of the entry of the left renal vein into the inferior vena cava (Fig. 19-21). These nodal sites should be carefully scrutinized in patients with testicular cancer. A study of 70 patients with newly diagnosed clinical stage 1 testicular nonseminomatous germ cell cancer undergoing preoperative CT showed that a short-axis size threshold of 1 cm had a sensitivity of 37% and a specificity of 100% in the diagnosis of nodal metastases. Progressively lower-size thresholds demonstrated greater sensitivity but lower specificity, so that a 4-mm threshold had a sensitivity of 93% and a specificity of 58%. The same study also showed that nodes lying anterior to the center of the aorta were more likely to be metastatic than nodes of similar size lying more posteriorly.⁶⁸ The incidence of para-aortic nodal metastases in patients with clinical stages IA, IB, II, and III endometrial carcinoma is 2.5%, 8.5%, 15.7%, and 33.3%, respectively.⁶⁹ Deep myometrial invasion, cervical involvement, and lymphovascular space invasion also significantly increase the likelihood of nodal metastases. In a study of 56 women with endometrial carcinoma that had preoperative CT scans and lymph node sampling, the sensitivity and specificity of CT for detecting nodal metastases were 57% and 92%, respectively.⁷⁰ This study used a short-axis diameter of at least 1.5 cm to define adenopathy. Ovarian cancer can spread to pelvic and retroperitoneal nodes. In a study of 71 patients undergoing preoperative CT or MRI, the incidence of nodal metastases at surgery was 28% (20 of 71).⁷¹ CT showed a sensitivity of 50% and a specificity of 95% in the diagnosis of nodal metastases using a 1-cm short-axis diameter threshold. The evaluation of retroperitoneal adenopathy in renal cell carcinoma illustrates the importance of clinical context, because nodal enlargement in this setting is often reactive rather than metastatic. In a study of 163 patients undergoing preoperative CT and using a short-axis threshold of 1 cm to define adenopathy,⁷² lymphadenopathy was found in 43 patients, only 18 of whom had nodal metastases at surgery (positive predictive value of 43%). No adenopathy was detected in 120 patients, of whom 5 had metastases at surgery (negative predictive value of 96%).

FIGURE 19-19 • Contrast-enhanced computed tomography in a 42-year-old man illustrating normal retroperitoneal nodes. The nodal chains include left para-aortic (white arrow), interaortocaval (vertical black arrow), and right paracaval (horizontal black arrow) groups.

FIGURE 19-20 • Contrast-enhanced computed tomography in a 42-year-old man with a right-sided testicular cancer. An anterior interaortocaval nodal metastasis (arrow) is visible.

FIGURE 19-21 • Contrast-enhanced computed tomography in a 27-year-old man with a left-sided testicular cancer. A left para-aortic nodal metastasis (arrow) visible.

Magnetic Resonance Imaging

MRI is rarely employed as the primary imaging modality for assessment of retroperitoneal adenopathy, because CT, even without intravenous contrast, is usually adequate. In general, the same size criteria are used at MRI as are used at CT, although a recent study suggests different size criteria may be appropriate.⁷³ In a group of testicular cancer patients without metastases undergoing surveillance imaging, the 95th percentile for maximum short-axis diameter of retroperitoneal nodes at a variety of locations ranged from 3 to 5 mm.

Positron Emission Tomography

PET (Fig. 19-22) allows detection of small malignant nodes not identified or not meeting size criteria for malignancy by CT, or when lack of retroperitoneal fat or postsurgical change makes it difficult to identify retroperitoneal nodes with CT.⁷⁴ PET scanning is also gaining importance in diagnosing and staging of cervical cancers. In a study comparing PET and CT in 101 consecutive patients with carcinoma of the cervix treated by chemotherapy and radiotherapy as clinically indicated and followed for a median of 15.4 months, baseline CT and PET showed retroperitoneal adenopathy in 7 and 21 patients, respectively. The 2-year progression-free survival, based solely on retroperitoneal adenopathy, was 64% in patients with negative results in both modalities, 14% in patients with positive results in both modalities, and 18% in patients with negative results on CT but positive results on PET. Multivariate analysis demonstrated that the most significant prognostic factor for progression-free survival was the absence of positive retroperitoneal nodes on PET. These results suggest that PET detects abnormal lymph nodes more often than CT, and that PET findings are a better predictor of outcome.⁷⁵

FIGURE 19-22 • Positron emission tomography scan in a 51-year-old woman with endometrial cancer and retroperitoneal nodal metastases. The primary tumor (horizontal arrow) is seen above the bladder, and the nodal metastases (vertical arrow) are seen in the retroperitoneum.

Lymphangiography

Lymphangiography (Fig. 19-23) has largely fallen into disuse because it is an invasive and technically challenging procedure, because few practicing radiologists have the necessary training to perform or interpret lymphangiograms, and because the available data do not suggest any incremental clinical benefit. Even when lymphangiography was available, it was largely confined to a few specialist centers and reserved for evaluation of testicular and cervical cancer. A recent meta-analysis reviewed the performance of lymphangiography, CT, and MRI in nodal assessment in cervical cancer. The three modalities had similar accuracy, and the authors concluded that as CT and MRI are less invasive than lymphangiography and also assess local tumor extent, they should be considered the preferred adjuncts to clinical evaluation of invasive cervical cancer.⁷⁶ In a study comparing CT and lymphangiography in 50 patients with intra-abdominal metastatic testicular cancer, CT was more effective than lymphangiography. CT showed metastatic disease in 21 patients compared with 7 for lymphangiography. In the detection of relapse, lymphangiography showed nodal enlargement in 4 of 8 cases compared with 8 of 8 for CT.⁷⁷

FIGURE 19-23 • Lymphangiogram in a 24-year-old man with testicular cancer. Normal lymph nodes are opacified by contrast, which is injected into small lymphatics on the dorsum of the feet. The study is invasive, technically challenging, and is now rarely performed.

Gastric Cancer

Imaging of Gastric Carcinoma

Gastroscopy is the standard of reference for the diagnosis of gastric carcinoma, with a reported accuracy of more than 95%. EUS may be helpful to supplement endoscopy. CT and MRI are useful for detection of metastases in advanced disease. The double-contrast upper-gastrointestinal examination with air and barium has been largely replaced by gastroscopy, but may still be useful in selected cases. Direct endoscopy is used for areas of the stomach that require particular scrutiny. Even in Japan, where double-contrast upper-gastrointestinal examinations are still frequently performed as a first-line investigation, approximately 26% (7 of 27) of gastric cancers may be completely occult radiographically.⁷⁹

Endoscopic Ultrasound

EUS provides high spatial resolution images of the gastric wall, allowing visualization of all five layers, and also allows for evaluation of perigastric nodes. In a study of 119 patients with gastric cancer undergoing preoperative EUS, the modality demonstrated an accuracy of 70% and 65% in the prediction of depth of tumor invasion and nodal involvement, respectively.⁸⁰ In another study, the overall accuracy, sensitivity, and specificity of EUS in the diagnosis of lymph nodal metastases was reported as 79%, 79%, and 80%, respectively.⁸¹

Computed Tomography

CT is of limited value in the primary diagnosis of gastric cancer (Fig. 19-24). In the preoperative staging of gastric carcinoma the criteria to be assessed with CT includes extension of the tumor along the wall and adjacent areas, lymph node metastases, and distant metastases.⁸² The use of water as a negative contrast agent to produce gastric distension by water may assist tumor visualization and assessment. In a study of 71 patients with gastric cancer undergoing thin-section spiral CT, the sensitivity for early stage tumors was only 26% (12 of 46), although all 29 advanced tumors were seen.⁸³ The overall T staging was correct in only 66% (27 of 41). CT is more commonly used for detection of metastatic disease. In a study of 56 patients with node-positive gastric cancer undergoing spiral CT, the sensitivity, specificity, and accuracy in the detection of liver metastases were 89%, 98%, and 96%, respectively.⁸⁴ However, peritoneal dissemination was not detected in 15 of 56 patients (27%), and stage IV disease was not correctly diagnosed in 18 of 40 patients (45%).

FIGURE 19-24 • Contrast-enhanced computed tomography in a 77-year-old man with gastric cancer. The tumor is seen as an elevated ulcerated mass (arrow).

Magnetic Resonance Imaging

MRI is rarely used to evaluate gastric carcinoma, and only a few studies have been reported. Although the high tissue contrast and multiplanar capability of MRI are potentially attractive features for imaging gastric pathologic conditions, no convincing advantage has been shown for MRI versus CT. In a study of 30 patients with gastric carcinoma undergoing preoperative MRI and CT, MRI and CT were of similar accuracy in T staging (73% and 67%, respectively). Accuracy for nodal staging was also similar for MRI and CT (55% and 59%, respectively).⁸⁵

Positron Emission Tomography

The role of PET (Fig. 19-25) in the imaging evaluation of gastric cancer continues to evolve. PET has essentially no role for the detection of gastric cancer, because endoscopy is the key test for diagnosis. PET is also particularly limited in the assessment of adenopathy associated with gastric cancer, with a reported sensitivity and specificity of 23% and 100%, respectively.⁸⁶ Low sensitivity is likely due to a combination of confounding peristalsis and physiologic uptake. PET may have greater value in the prediction of prognosis and the early evaluation of therapeutic response. In a study of 65 patients with gastric cancer, those with a standardized uptake value (SUV) of less than 4 (n = 31) had a significantly better prognosis than those with a SUV more than 4.⁸⁶ In another study of 35 patients with stage T₃ or T₄ gastric cancer, those with a metabolic response (n = 13; defined as at least a 35% reduction in SUV) 14 days after starting chemotherapy had a significantly better outcome than those who did not have a metabolic response.⁸⁷ PET is also helpful in detecting recurrent disease. PET had an overall accuracy of 70% for the diagnosis of recurrent disease in a Belgian study of 33 patients with suspected recurrence after curative surgery using pathologic findings or follow-up as the standard of reference.⁸⁸ Interestingly, a negative PET scan in and of itself was associated with a better prognosis in this study. The mean survival after a negative PET scan was 18.5 months compared with 6.9 months in those with a positive scan.

FIGURE 19-25 • Positron emission tomography/computed tomography (PET/CT) study in a 62-year-old man with gastric carcinoma undergoing baseline staging prior to treatment. A, Axial contrast-enhanced CT section demonstrates extensive perigastric adenopathy (arrows). B, Corresponding axial PET image confirms markedly increased fluorodeoxyglucose uptake in the nodal metastases (arrows).