MR Imaging Findings of Early Bile Duct Cancer

Transcription

1 JOURNAL OF MAGNETIC RESONANCE IMAGING 28: (2008) Original Research MR Imaging Findings of Early Bile Duct Cancer Dong Ho Lee, MD, Jeong Min Lee, MD,* Kyung Won Kim, MD, Hee Sun Park, MD, Se Hyung Kim, MD, Jae Young Lee, MD, Joon Koo Han, MD, and Byung Ihn Choi, MD Purpose: To retrospectively evaluate the MR imaging features of early bile duct cancer and to correlate them with the clinicopathologic findings. Materials and Methods: This retrospective study was approved by our institutional review board, and informed consent was waived. Seventeen patients with surgically proven early bile duct cancer who had undergone preoperative MR cholangiopancreatography with gadolinium-enhanced MR imaging, were included in this study. Two, experienced radiologists evaluated the MR images in consensus regarding the following findings: tumor number and morphology; signal intensity of the tumor; sharpness of the outer border of the bile duct wall; enhancement pattern of the tumor; and the presence of enlarged peribiliary lymph nodes. Another radiologist measured the SNR of the tumor and bile duct wall on gadolinium-enhanced MRI during the dynamic phases to evaluate the tumor enhancement degree. Results: In all patients, MR imaging demonstrated single or multiple intraluminal bile duct masses showing a sharply defined outer margin. The most common enhancement pattern of the biliary lesions showed heterogeneous amorphous enhancement or heterogeneous enhancement with central, dot-like structures or vascular structures (76.5%, 13/17 patients). The difference of SNR between bile duct and tumor was greatest in the equilibrium phase (P 0.05). Conclusion: MRCP combined with dynamic contrast-enhanced MRI can be useful for detecting early bile duct cancers. Common MR findings of early bile duct cancer include one or more inhomogeneously enhancing intraductal masses with clear outer margins and preservation of the bile duct wall. Key Words: early bile duct cancer; magnetic resonance imaging (MRI); magnetic resonance cholangiography (MRCP); papillary bile duct carcinoma J. Magn. Reson. Imaging 2008;28: Wiley-Liss, Inc. Department of Radiology and Institute of Radiation Medicine, Seoul National University College of Medicine, Seoul, Korea. *Address reprint requests to: J.M.L., Department of Radiology and Institute of Radiation Medicine, Seoul National University College of Medicine, 28 Yeongon-dong, Jongno-gu, Seoul , Korea. leejm@radcom.snu.ac.kr Received April 21, 2008; Accepted August 20, DOI /jmri Published online in Wiley InterScience ( BILE DUCT CANCER is the second most common primary hepatobiliary cancer after hepatocellular carcinoma, with approximately 2500 cases diagnosed annually in the United States There are several histologic types, the most common of which are adenocarcinoma, papillary carcinoma, and mucinous carcinoma (1). Bile duct cancer is frequently diagnosed in the advanced disease stage, and more than 50% of patients with advanced bile duct cancer are not amenable to surgical treatment (2). Although survival after surgical treatment of bile duct cancer has improved during the past yr, its prognosis is poor even after curative resections (3,4). Patients who undergo a margin-negative resection have a reported 5-yr survival rate of up to 37% (4). More recently, there have been a few sporadic reports indicating that patients with early bile duct carcinoma showed significantly better 5-yr survival rates (80 100%) after surgical treatment than patients with cancer invasion beyond the fibromuscular layer (46%) (5 8). Therefore, detection of early bile duct cancer using imaging studies may provide the opportunity for a better prognosis for patients with bile duct cancer. Although there have been no worldwide accepted criteria for early bile duct cancer, it is defined as a carcinoma whose invasion is confined within the fibromuscular layer of the bile duct and without distant metastasis irrespective of possible lymph node involvement (5,9). Considering that the prognosis of bile duct cancer is generally poor, early detection using noninvasive cross-sectional imaging is very important in improving the prognosis. In fact, the only successful approach for achieving long-term survival in patients with bile duct cancer is to detect the tumor at an early stage when therapy can be effective. Although there have been several literature reports of surgical and histopathologic findings of early bile duct cancer, only a few radiologic findings have been reported (10,11). According to the study by Lim et al (11), early bile duct carcinoma features presenting on imaging studies include the presence of a tumor mass in the bile duct lumen and the preserved integrity of the tumor-bearing bile duct wall with no spread outside the wall. Currently, various invasive and noninvasive imaging modalities, such as ultrasound, CT, MR imaging with/ without MR cholangiopancreatography (MRCP), endoscopic retrograde cholangiography, and endoscopic ultrasound, can be used to diagnose bile duct cancer (2,12). Among these modalities, contrast-en Wiley-Liss, Inc. 1466

2 MRI Findings of Early Bile Duct Cancer 1467 hanced MR imaging combined with MRCP using stateof-the-art MR scanners, can facilitate viewing longitudinal cancer spread by simultaneously assessing bile ducts upstream and downstream and can permit assessment of the vessels and the depth of invasion (9,13 17). Therefore, the purposes of this study are to retrospectively evaluate the imaging features of early bile duct carcinoma on unenhanced and contrast-enhanced MR images as well as on MR cholangiography, and to correlate them with the clinicopathologic findings. MATERIALS AND METHODS Patients This retrospective study was approved by our institutional review board, and informed patient consent was waived. Using a database maintained by the Department of Pathology, we identified patients with pathologically proven early bile duct cancer who had undergone surgical treatment at our hospital between January 2003 and March Inclusion criteria for this study were as follows: patients with bile duct cancer who had undergone surgical exploration at our hospital; preoperative MRCP including contrast-enhanced dynamic MR imaging obtained within one month before surgery; and a diagnosis of early bile duct cancer on pathologic examination of a surgical specimen. Pathologic diagnosis of early bile duct cancer was made when the bile duct carcinoma was confined to the bile duct wall and its depth of invasion was confined within the fibromuscular layer of the bile duct. Seventeen patients comprised our study population, that is, 10 men and seven women. The mean patient age was 65.3 years, and the age range was from 58 to 70 years. The mean interval period between imaging and surgery was 12.4 days. Pylorus preserving pancreaticoduodenectomy was performed on seven patients, Whipple surgery on one patient, common bile duct excision with hepaticojejunostomy on four patients, right lobectomy on two patients, and left lobectomy on two patients. As the remaining patient was diagnosed on surgery as having an unresectable tumor, palliative surgery (hepaticojejunostomy) was performed. The cause for unresectability in these patients was bilateral tumor involvement of the secondary biliary confluence (Bismuth type IV). Initial symptom presentation and initial serum total bilirubin and CA19-9 level were recorded based on the patient chart review performed by one radiologist. MR Examination All MR imaging was performed on 1.5Tesla (T) superconducting systems (Signa HDi, GE Medical system, n 8; Sonata, n 9, Siemens Medical solution, Erlangen, Germany) using a phased-array torso coil. MRCP examinations were performed with thick-slab T2- weighted (T2W) turbo spin echo and thin-slab multisection half-fourier acquired single-shot turbo spinecho (HASTE) sequences. Thick-slab T2W TSE MRCP images were obtained in the coronal plane, and thinslab T2W HASTE MRCP images were obtained in both the coronal and transverse planes. At least five thickslab T2W TSE MRCP images were obtained, and the imaging parameters were as follows: coronal and 15 and 30 oblique coronal angle; repetition time (TR) ms/ effective echo time (TE) ms, 2800/1100; echo train length, 240; flip angle, 150 ; slab thickness, 60mm; field of view (FOV), 220 mm; and matrix, Thin-slab T2W HASTE MRCP images were obtained with the following parameters: TR/TE /110; echo train length, 240; refocusing flip angle, 150 ; section thickness, 3- to 4-mm-thick consecutive sections; FOV, 220 mm; 15 sections acquired per breath hold (volumes of coverage, 60 mm); and matrix for Sonata and for Signa HDi. Three-dimensional (3D) MRCP imaging was performed on all 17 study patients. For 3D-MRCP using TSE on the Sonata scanner, the following parameters were used: TR(ms)/TE(ms)/flip angle, / ; matrix size, ; section thickness 2.0 mm; FOV, mm; 60 partitions with interpolated partition thickness 1 mm; typical voxel size 1mm 1mm 1mm; coronal orientation; and minimum scan time 5 min (range, 5 10 min). Parallel imaging (msense,siemens Medical Solution) with a reduction factor of two was applied in an in-plane phaseencoding direction of 3D-MRCP imaging. The msense allowed acquisition of folded images in each receiver channel by reduced k-space sampling and unwrapped them in the reconstruction process from a reference scan. In addition, for 3D-MRCP using a 3D fast-recovery fast spin echo (FR-FSE) sequence with an array spatial sensitivity technique (ASSET, GE Medical System) on the Signa HDi scanner, the following parameters were used: TR(ms)/TE(ms)/flip angle, 4000/693; matrix size, ; FOV, mm; section thickness 1.4 mm; 120 partitions with interpolated partition (ZIP2; GE Medical system) thickness 0.7 mm; typical voxel size 0.8 mm 1.4 mm 0.7mm, coronal orientation; parallel acquisition technique factor, 2 and minimum scan time, 4 min (range, 4 10 min). Technologists were instructed to obtain slabs oriented parallel to the portal vein bifurcation, in a straight coronal plane, covering at least a third of the biliary confluence of both intrahepatic bile ducts, and centered on the middle portion of the common bile duct to cover the biliary tree. 3D-MRCP images were then reconstructed using a maximum intensity projection algorithm to produce oblique images rotating about the z-axis in 10 increments. Transverse T2W HASTE images were obtained with the following parameters: TR/TE, /110; echo train length, 240; refocusing flip angle, 150 ; section thickness, 4 mm thick consecutive sections; field of view, 220 mm; and matrix, Nonenhanced T1- weighted (T1W) imaging was performed using in-phase and opposed-phasespoiled gradient-recalled echo (GRE) techniques in the transverse plane using the following imaging parameters: 150/2.3 for opposedphase images; 150/5.7 for in-phase images; a flip angle of 70 ; matrix; a receive bandwidth of plus or minus 32 khz; slice thickness, 6 mm; one signal acquired and scan time, 18 s. Dynamic images were obtained using a fat-suppressed, three-dimensional GRE

3 1468 Lee et al. sequence (VIBE, Siemens Medical Solution; LAVA, GE Medical System) before and after administration of 0.1 mmol gadobenate dimeglumine (MultiHance, Gd- BOPTA; Bracco SpA, Milan, Italy) per kilogram of body weight and with an injection rate of 2 ml/s (18). The imaging parameters were as follows: TR/TE 4.87/2.35 and a flip angle of 10 for Sonata; and TR/TE 4.4/2.0 and a flip angle of 12 for HDi. There were partitions which interpolated to with a partition thickness of 5 mm. The matrix was Arterial, venous, and equilibrium-phase images were obtained serially at s, s, and 3 min, respectively, following contrast injection. For timing of the contrast injection, we used an MR fluoroscopy technique which allows real-time visualization of the heart and aorta during repetitive measurements at the same coronal position with a T1-weighted gradient-echo sequence. The arterial phase images were obtained 8 seconds after the arrival of the gadolinium-bolus was detected in the aorta. Image Analysis Qualitative Analysis All images were reviewed on a Picture Archiving and Communications System (PACS) workstation monitor (m-view, Marotech, Seoul, Korea) by two, board-certified radiologists. These physicians served mainly as hepatobiliary radiologists and interpreted abdominal MR images of the liver as part of their daily clinical and research practice (12 yr and 4 yr of clinical experience, respectively). The radiologists reviewed the images using the stack-view mode on the PACS viewer with an optimal window setting adjustment. Before image interpretation, these reviewers were made aware that the patients had early bile duct carcinoma, but they did not know the detailed pathologic, clinical or operative findings. Decisions regarding imaging features were determined by consensus. On MR images, loss of continuity of the bile duct or ductal obstruction, abrupt and irregular narrowing of the distal segment by prestenotic biliary dilatation, irregularly shaped intraluminal filling defects, and segmental enhancing wall thickening of the bile duct were considered to be findings indicating biliary involvement (15). MR cholangiograms with heavily T2-weighted axial and coronal images were first evaluated for tumor size, shape, and location, extent of bile duct involvement by the tumor, the presence of bile duct stricture or obstruction, and dilation of the duct proximal or distal to the tumor. Dilatation of the bile duct was classified as mild or moderate to severe. Mild bile duct dilatation was defined as duct dilatation confined to the central portion of the bile duct, and moderate to severe dilatation was defined as bile duct dilatation extending to the peripheral portion of the bile duct. And then, the two radiologists then evaluated the unenhanced MR images and the dynamic MR images regarding the signal intensity of the tumors, bile duct tumor growth patterns, degree of contrast enhancement, depth of wall invasion, that is, preservation of the enhancing thin wall on dynamic MR images, and outer margin of the bile duct wall, that is, presence of periductal tumor infiltration (11). The signal intensity on T2-WI and T1-WI relative to the liver was then analyzed as low, iso, or high. The reviewers also evaluated the Gd-BOPTA-enhanced, 3D dynamic GRE T1-weighted images during the portal venous phase to analyze the degree of enhancement and sharpness of the outer margin. The degree of enhancement of the thickened ductal wall or the intraluminal mass was classified as low, iso, or high relative to the liver parenchyma. If determining the tumor signal intensity relative to the liver parenchyma was difficult, the reviews used ROI measurement for accurate analysis of the signal intensity. The outer margin of the bile ductal wall of the narrowed segment was classified as indistinct or distinct. Indistinct was defined as having marginal blurring, speculation, or irregularity of the outer margin of the strictured ductal wall. Regional and retroperitoneal lymph nodes were also evaluated, and lymph node metastasis was considered to be present on MR images when the short-axis diameter was greater than 10mm or when central necrosis was present regardless of lymph node size or when its signal intensity was higher than that of liver parenchyma on the portal venous phase (14,16). Quantitative Analysis To assess the degree of tumor and bile duct wall enhancement, quantitative image analysis was performed by a single radiologist who knew the patients had early bile duct carcinoma and who was not involved in the qualitative analysis, measured the normal bile duct wall signal intensity, tumor signal intensity, and the standard deviation of background noise using operatordefined regions of interest (ROIs) for arterial, portal venous, and equilibrium-phase images. For bile duct wall and tumors too small to place in an ROI, the image was magnified as much as three to four times. For signal intensities of the bile duct wall, ROIs were drawn in a location not involved by the tumor and at least 2 cm distant from the tumor. Background noise was measured along the phase-encoding direction outside the body just ventral to the anterior abdominal wall and included respiratory or motion-related artifacts. The shape and size of the ROIs (23 mm 2 ) were identical for all images and contained 5 to 20 pixels (mean, 10 pixels). The signal-to-noise ratio of the bile duct wall and tumor and the contrast-to-noise ratio of the tumor relative to that of the normal bile duct wall were calculated from the signal intensity and the standard deviation (SD) of the background noise. The signal-to-noise ratio (SNR) and the contrast-to-noise ratio (CNR) of tumor relative to the normal bile duct wall were then calculated according to the following formulas: SNR signal intensity of the bile duct wall (tumor) / SD of background noise and CNR (signal intensity of the bile duct wall signal intensity of the tumor) / SD of the background noise Statistical Analysis Paired t-tests were used for SNR analysis (GraphPad InStat, version 3.05; GraphPad Software, San Diego,

4 MRI Findings of Early Bile Duct Cancer 1469 Table 1 Clinicopathologic Features of 17 Patients With Early Bile Duct Cancer Case no. Age (yrs) / Sex Symptoms T-Bila (mg/dl ) CA.19-9 Gross type Location Bismuth Type Multiplicity 1 67/M Abdominal pain 1 7 Papillary IHD and EHD II Yes 2 62/ F Abdominal pain Papillary EHD NA No 3 59/M Abdominal pain Papillary EHD NA Yes 4 63/M Jaundice Papillary IHD and EHD IIIa No 5 58/ F Jaundice Papillary IHD and EHD II Yes 6 60/M Abdominal pain Papillary IHD and EHD II Yes 7 68/ F No symptom Nodular EHD NA Yes 8 69/M Jaundice Papillary IHD and EHD IV Yes 9 69/M Abdominal pain Papillaryb IHD NA Yes 10 63/ F Jaundice Nodular EHD NA No 11 63/ F Abdominal pain 1 7 Nodular EHD NA No 12 68/ F Vomiting Papillary EHD NA No 13 63/M Jaundice Papillary EHD NA No 14 64/M No symptom Papillary EHD NA No 15 60/M No symptom Papillary* IHD NA No 16 67/ F Abdominal pain Papillary IHD and EHD IIIa Yes 17 70/M Indigestion Papillary IHD and EHD II Yes a T-Bil total bilirubin level at initial manifestation (normal range: mg/dl). b Intraductal mucin producing papillary carcinoma. IHD intrahepatic bile duct; EHD extrahepatic bile duct, NA not available. CA). A P-value less than 0.05 were considered to indicate a statistically significant difference. Correlation of the MR Findings and the Clinicopathologic Features After analysis of the MR imaging features, correlation of the MR findings with the clinicopathologic features was performed. The presence and location of the intraductal mass as seen on MRCP and on the contrast-enhanced dynamic sequence of MR imaging, were compared with the surgical and pathology findings. The inner surface of the tumor as seen on T2-WI MRCP images, was also compared with the gross subtypes of tumor. The degree of bile duct dilatation seen on MR images was also evaluated regarding the preoperative serum total bilirubin and the CA19-9 level. RESULTS Clinicopathologic Features The clinicopathologic findings of 17 patients with early bile duct cancer are summarized in Table 1. Among these 17 patients, jaundice was the initial presenting symptom in 5 patients and nonspecific symptoms such as abdominal pain, vomiting, and indigestion were the initial symptoms in 9 patients. Three patients had no initial symptoms; however, a bile duct tumor was suspected on routine ultrasonography or computed tomography screening. According to the preoperative laboratory test, five patients had elevated serum total bilirubin levels. CA19-9 level was markedly increased in two patients (9,300 and 220,000, respectively). According to the operative and pathologic findings, seven patients were diagnosed with both intrahepatic and extrahepatic cholangiocarcinoma classified as follows: Bismuth type II (n 4); type IIIa (n 2); and type IV (n 1). Eight patients had extrahepatic cholangiocarcinomas located in the following areas: common hepatic duct (n 1); proximal common bile duct (n 2); mid common bile duct (n 1); and distal common bile duct (n 4). Two intrahepatic cholangiocarcinomas were located in the left lobe of the liver. Nine patients had multiple tumors. Examination of the surgical specimens confirmed that there were intraductal tumors in all patients. Tumors were manifested as a nodular mass in three patients (3/17, 17.6%) and a papillary mass in the remaining patients (14/17, 82.4%). Among the 14 papillary tumors, there were two mucin-producing papillary carcinomas. No patients had periductal infiltration or invasion into the adjacent organs or vascular structures. In 15 patients, no lymph node metastasis was demonstrated in the surgical specimens. Two patients had gastric cancer concurrently and had lymph node metastasis at the greater and lesser curvature of the stomach. Qualitative Imaging Analysis MR cholangiography The MR imaging features of the 17 patients with early bile duct carcinoma are summarized in Table 2. In all patients, MRCP images revealed one or multiple intraductal masses in the biliary tree with upstream biliary tree dilatation. In eight patients, there was a single intraductal mass, but in nine patients, there were multiple intraductal tumors. The biliary tumors were located in the intrahepatic bile duct (IHD) in two patients, in the extrahepatic bile duct (EHD) in eight patients, and in both the EHD and IHD locations in seven patients. In 6 patients, the upstream bile duct dilatation was mild and was confined to the central portion of the bile duct, and in 11 patients, bile duct dilatation was

5 1470 Lee et al. Table 2 MR Imaging Features of 17 Patients With Early Bile Duct Cancer Case no. Intraluminal mass Degree of upstream duct dilatation Wall thickening Inner surface Outer margin T2 signal 1 Present Mild No Papillary Distinct High 2 Present Moderate to severe No Papillary Distinct High 3 Present Mild No Papillary Distinct High 4 Present Moderate to severe No Papillary Indistinct High 5 Present Moderate to severe No Papillary Distinct High 6 Present Moderate to severe No Papillary Distinct High 7 Present Moderate to severe No Irregular Distinct High 8 Present Moderate to severe No Smooth Distinct High 9 Present Moderate to severe No Papillary Distinct High 10 Present Mild No Irregular Distinct High 11 Present Mild No Irregular Distinct High 12 Present Moderate to severe No Papillary Distinct High 13 Present Moderate to severe No Papillary Distinct High 14 Present Mild No Papillary Distinct High 15 Present Moderate to severe No Papillary Distinct High 16 Present Moderate to severe No Papillary Distinct High 17 Present Mild No Papillary Distinct High moderate to severe and extended to the peripheral portion of the bile duct. Unenhanced MR Images On T2-WI, all tumors showed high signal intensity compared with the liver parenchyma. In 15 patients, the tumors showed low signal intensity on T1-WI, but in two patients the tumors showed high signal intensity. The outer margin of the bile duct wall bearing the tumor was sharply defined in 16 of 17 patients (94.1%) on T2-WI. On T2-WI, the inner surface of the lesion had a papillary appearance in 13 patients, a smooth margin in one patient, and an irregular surface in three patients. There was no visible bile duct wall thickening in any of the patients. Dynamic Imaging The signal intensities of the bile duct tumors in the dynamic sequence are summarized in Table 3. During the arterial phase, 5 tumors (5/17, 29.4%) showed high signal intensity compared with the liver parenchyma, 4 tumors showed iso-signal intensity (4/17, 23,5%), and 8 tumors (8/17, 47.1%) had low signal intensity. During the portal and equilibrium phases, 11 (11/17, 64.7%) and 12 tumors (12/17, 70.6%), respectively, had low signal intensity compared with the liver parenchyma. In addition, during the arterial phase, the most common enhancing feature of biliary tumors was heterogeneous enhancement (n 13, 76.5%), which consisted of an amorphous pattern (n 5), central Table 3 Relative Enhancement Degree* and Enhancement Pattern of Early Bile Duct Cancer Case no. Age (yrs) / Sex SI on AP SI on PVP SI on DP Enhancement pattern on AP Enhancement pattern on PVP 1 67/M hypo Hypo Hypo Minimal Minimal 2 62/ F Iso Iso Iso Heteroa, central dot-like Hetero, central dot-like 3 59/M Hyper Hypo Hypo Hetero, amorphous Hetero, amorphous 4 63/M Iso Hypo Hypo Hetero, central dot-like Hetero, amorphous 5 58/ F Hypo Hypo Hypo Hetero, central dot-like Hetero, amorphous 6 60/M Hypo Iso Hypo Hetero, central dot-like Hetero, amorphous 7 68/ F Hypo Hyper Hypo Hetero, amorphous Hetero, amorphous 8 69/M Hyper Hypo Hypo Hetero, central vascular structure Hetero, central dot-like 9 69/M Hyper Hypo Hyper Hetero, central dot-like Minimal 10 63/ F Iso Iso Iso Homogeneous Homogeneous 11 63/ F Hypo Hypo Hypo Minimal Minimal 12 68/ F Hyper Hyper Hyper Hetero, central vascular structure Hetero, amorphous 13 63/M Iso Hypo hypo Hetero, amorphous Hetero, amorphous 14 64/M Hypo Hypo Hypo Hetero, amorphous Minimal 15 60/M Hypo Hypo Hypo Hetero, central dot-like Minimal 16 67/ F Hyper Hyper Iso Hetero, amorphous Hetero, amorphous 17 70/M Hypo Hypo Hypo Minimal Minimal *Relative enhancement degree: compared with liver parenchyma. SI signal intensity; AP arterial phase; PVP portal venous phase; DP delayed image; Hypo hypointensity; Iso isointensity; Hyper hyperintensity; a Hetero heterogeneous enhancement.

6 MRI Findings of Early Bile Duct Cancer 1471 Table 4 Result of Quantitative Analysis of Enhancement Degree of Early Bile Duct Cancers and Adjacent Bile Duct Wall SNR(mean 2 SD) Precontrast AP PVP DP Tumor Bile duct wall P value ( 0.05)* CNR(mean 2 SD) SNR signal-to-noise ratio; CNR contrast-to-noise ratio of tumor relative to normal bile duct wall; AP arterial phase; PVP portal venous phase; DP delayed phase. dots (n 6) or a central vascular structure (n 2) (Table 3). During the portal phase, 10 tumors (58.8%) showed heterogeneous amorphous enhancement with/without central dot-like enhancement, but six tumors only showed minimal enhancement (Table 3). The outer margin of the bile duct wall bearing the tumor was sharply defined in 16 patients (16/ 17, 94.1%) on the delayed phase image. There was no demonstrable lymph node metastasis or metastatic lesion on MR images in any of the study patients. Quantitative Analysis The mean SNR value of the bile duct wall was greater than that of the tumor (Table 4), but the difference in the mean SNR value between the bile duct wall and tumor was statistically significant in only the delayed phase image (P-value: ). The CNR value of tumor relative to normal bile duct wall was highest on the delayed phase images. Correlation of the MR Findings and Clinicopathologic Features In all 17 cases, intraductal masses were found in the surgical specimens. MR cholangiographic images showed well-demonstrated intraductal masses in all cases (Fig. 1). The tumor location on MRCP images was also well-correlated with the pathology findings. Gross Figure 1. Images of a 68-year-old woman with papillary carcinoma in the mid-portion of the common bile duct (case no. 12). a,b: T2-WI HASTE axial (a) and coronal (b) MR images shows an intraluminal mass (arrows) with a papillary appearance in the mid-portion of the common bile duct. The outer margin of the bile duct wall is sharply defined at the site of tumor involvement. c: MRCP image demonstrates an intraluminal mass lesion in the mid-cbd (arrow) with diffuse dilatation of the upstream bile duct. d: Contrast-enhanced arterial phase image shows heterogeneous enhancement of an intraluminal mass with a central vascular-like structure (arrow).

7 1472 Lee et al. Figure 2. Images of a 63-year-old woman with nodular bile duct carcinoma in the distal common bile duct (case no. 10). a,b: T2 HASTE axial (a) and coronal (b) MR images show an intraluminal mass in the distal common bile duct (arrow). The inner surface of the intraluminal mass has an irregular shape. The outer surface of the bile duct wall is relatively well-defined on the axial image. c: MRCP image demonstrates an intraluminal mass lesion in the distal CBD with mild dilatation of the upstream bile duct. d: Contrast-enhanced arterial phase image shows minimal enhancement of the intraluminal mass and the welldefined bile duct wall (arrows). examination of the surgical specimens showed 3 tumors that were nodular-shaped and 14 tumors with a papillary appearance. Regarding the inner surface of the lesions on T2-WI and MRCP images, all 3 nodular tumors detected on pathology had an irregular surface on MR imaging (Fig. 2) and 13 papillary tumors detected on pathology demonstrated papillary surfaces (Fig. 1). The remaining papillary tumors detected on pathology showed a smooth inner surface along both the IHD and the EHD (Fig. 3). Among the study 17 patients, 2 had an elevated serum CA19-9 level. MR cholangiogaphic images of these two patients showed marked dilatation of the bile ducts extanding to the periphery of the liver with multiple intraductal tumors noted in both the IHD and the EHD; these tumors occupied almost the entire ductal lumen (Fig. 3). On histopathologic examination, two patients were shown to have mucin-producing papillary adenocarcinomas. In these two patients, MRCP images demonstrated intraductal masses and upstream bile duct dilatation. The MRCP images also showed bile duct dilation downstream from the tumor (Fig. 4). DISCUSSION In our study, axial T1-WI, T2-WI, and dynamic imaging showed that although the early bile duct tumors frequently demonstrated an irregular or papillary inner surface, they showed a sharp outer margin in 16 of 17 patients (94.1%). In addition, there was no definite wall thickening of any of the tumor-bearing bile ducts shown on MR images. Furthermore, these image findings of early bile duct cancer, including a lack of bile duct wall thickening and/or the presence of a sharp outer bile duct wall margin, would be compatible with the pathologic features of early bile duct cancer in which invasion is confined to the fibromuscular layer of the bile duct. In our study, 14 of the 17 patients (82.4%) had papillary carcinomas. This high prevalence of papillary carcinomas among the early bile duct cancers can be explained by the strong tendency of papillary tumors to grow and spread superficially along the bile duct mucosa (5 9). Our study results are well-correlated with those of previous reports on early bile duct cancer (9,11) which indicated that the most common gross and microscopic types of early bile duct cancer were the

8 MRI Findings of Early Bile Duct Cancer 1473 Figure 3. Images of a 70-year-old man with jaundice and elevated CA Because the surgical exploration showed type IV cholangiocarcinoma, palliative resection of the bile duct with hepaticojejunostomy was performed. Pathology specimen reveals papillary cholangiocarcinomas (case no. 8). a,b: T2 HASTE axial (a) and coronal (b) images show intraluminal masses in both the intrahepatic and extrahepatic bile ducts. The inner surface of the intraluminal mass is smooth, and the outer surface of the bile duct wall is well-defined on the axial image (arrow). The upstream bile duct is severely dilated, and the tumors occupy the entire ductal lumen. c,d: Contrast-enhanced arterial phase axial (c) and coronal (d) images demonstrate heterogeneous enhancement of an intraluminal mass with a distinct bile duct wall (arrows). Note that the tumors show less intense contrast enhancement compared with the bile duct walls and appear as a hypointense soft tissue mass along the bile ducts. intraductal-growing type (47 100%) and papillary carcinomas (31 90%). Indeed, our study results of early bile duct cancers are opposite to the results of previous study regarding biliary malignant stricture by Kim et al (17) in which periductal infiltrating, that is, sclerotic type, or massforming types were common. In their study, strictures due to advanced bile duct carcinoma were significantly thicker and longer than benign strictures and frequently showed an indistinct outer margin. Indeed, in our study, the MR features of early bile duct cancers reflect the nature of intraductal papillary tumor, that is, that tumor cells grow and spread superficially along the bile duct mucosa and do not invade deeply into the fibromuscular layer (19). To the contrary, as sclerosing or mass forming type tumors tend to grow in a radial direction and involve perirductal tissue at an early stage, they commonly show a long biliary stricture with wall thickening and an indistinct outer margin, both of which suggest periductal invasion. In our study, the enhancement pattern of early bile duct cancers during both the arterial and portal phases, was predominantly heterogeneous in nature. Heterogeneous amorphous enhancement or heterogeneous enhancement with central dot-like structures or a vascular structure, was demonstrated on arterial phase images in 13 patients with papillary-type tumors (76.5%, 13/17). The central, dot-like structures or the vascular structure was identified in eight patients with papillary tumors, but it was not found in nodular tumors. Slender fibrovascular stalks supported by connective tissue from the lamina propria in papillary tumors (19,20), seem to be responsible for the central dot-like or vascular structure seen in the arterial phase. This enhancement pattern may be the characteristic finding of papillary tumors and could provide additional information regarding the nature of the tumor. In addition, on quantitative analysis of the enhancement characteristics of these tumors and of the bile duct walls, intraductal tumors showed less enhancement than the adjacent normal bile duct wall. This result of our study is in agreement with the results of the previous study regarding enhancement characteristics of cholangiocarcinoma on helical CT (21). The difference between bile duct wall enhancement and tumor enhancement was greatest on the equilibrium phase compared with

9 1474 Lee et al. Figure 4. Images of a 60-year-old man with mucin producing papillary carcinoma of the left IHD (case no. 15). a,b: T2 HASTE axial (a) and coronal (b) images show the papillary appearance of an intraluminal mass in the left intrahepatic bile duct (arrow). Note that the surface of the bile duct wall of the tumor-bearing segment is sharply defined. c: MRCP image clearly demonstrates an intraluminal mass (arrow) in the dilated left intrahepatic bile duct. Note that both the upstream and downstream ducts to the tumor are dilated. d: Contrast-enhanced portal phase coronal image shows a heterogeneous enhancing intraluminal mass in the left intrahepatic bile duct (arrow). the arterial and portal venous phases, and the difference between the phases was statistically significant (P 0.05). Although jaundice is a common initial presenting symptom in patients with bile duct cancer (12), only 5 of our 17 patients (5/17, 29.4%) exhibited jaundice as the initial presenting symptom. However, the MR images showed bile duct dilatation as well as intraluminal masses in all 17 patients. The low degree of obstruction in early bile duct carcinoma and the fragile nature of papillary tumors (19) may be responsible for this finding. In the clinical setting, extensive use of diagnostic imaging is widely accepted as part of the diagnostic workup for jaundice or biliary dysfunction as most bile duct obstructions or strictures must be considered malignant until proven otherwise in patients with no history of biliary stones or biliary surgery. The diagnostic work-up could include ultrasound, CT, MR imaging, MRCP, endoscopic retrograde cholangiography or endoscopic ultrasound. Although endoscopic examinations are generally successful for detecting bile duct obstruction with ductal narrowing, additional CT or MRI would be necessary for preoperative evaluation of tumor staging (22). Although MDCT is widely used for biliary tumor staging, intraductal papillary tumors are easily mistaken for biliary stones and show poor contrast between tumor and bile on CT (19). To the contrary, MR imaging and MRCP may more effectively reveal intraluminal filling defects than CT due to its inherent high soft-tissue contrast. In our study, MR imaging combined with MRCP was able to provide a noninvasive evaluation of not only the diagnosis of early bile duct cancer but also tumor extent and resectability. Indeed, the preoperative radiologic diagnosis of early bile duct cancer is important for improving the patient prognosis as in many instances a better outcome can be obtained when tumor resection is performed (6,8,9). To the contrary, for patients with carcinoma infiltrating beyond the bile duct, the prognosis is poor even after extensive radical surgery (3 8,23). However, although patients with early bile duct carcinomas show a favorable prognosis after radical resection, this does not always indicate that early bile duct carcinomas are curable. Given that papillary cholangiocarcinomas spread along the bile duct mucosa, it is not uncommon for papillary tumors to involve both intrahepatic and extrahepatic bile ducts. When they involve intrahepatic bile ducts in both lobes of the liver, com-

10 MRI Findings of Early Bile Duct Cancer 1475 plete tumor resection is not possible. In one of our cases, as the patient had a carcinoma confined within the bile duct wall, that extended to the bilateral second level confluence of the intrahepatic bile duct, curative resection could not be performed. Therefore, one of the reasons to include imaging as part of the preoperative workup for patients with cholangiocarcinoma is to evaluate the longitudinal extent of the carcinoma to determine its curative resectability (14). Several limitations to our study should also be mentioned. First, because of the nature of the retrospective study design, a selection bias might have been unavoidable. Second, the relatively small sample size also limits this study; however, as detection of bile duct cancer at an early stage is quite difficult, our small study population is therefore understandable. Nevertheless, the identification of characteristic MR features of early bile duct cancer resulting from this study, should be very helpful. Third, we did not evaluate the diagnostic accuracy of MR images for biliary tumor staging. Therefore, future studies should include large patient populations with advanced bile duct carcinoma as well as early bile duct carcinoma to evaluate the diagnostic performance of MR imaging for this purpose. Fourth, as the MR examinations were performed on two MR scanners, this may have affected the qualitative analysis results. However, the overall results did not differ on the two scanners. In conclusion, early bile duct cancer appears as single or multiple intraluminal masses with irregular or papillary inner margins and clear outer margins, showing inhomogeneous enhancement on MR imaging. In addition, MRCP and the dynamic sequence can successfully visualize these morphologic tumor characteristics. ACKNOWLEDGEMENT We thank Bonnie Hami, M.A. (USA) for her help in manuscript editing. REFERENCES 1. de Groen PC, Gores GJ, LaRusso NF, et al. Biliary tract cancers. N Engl J Med 1999;341: Slattery JM, Sahani DV. What is the current state-of-the-art imaging for detection and staging of cholangiocarcinoma? Oncologist 2006;11: Nimura Y, Kamiya J, Kondo S, et al. Aggressive preoperative management and extended surgery for hilar cholangiocarcinoma: Nagoya experience. J Hepatobiliary Pancreat Surg 2000;7: Jarnagin WR, Fong Y, DeMatteo RP, et al. Staging, resectability, and outcome in 225 patients with hilar cholangiocarcinoma. Ann Surg 2001;234: Mizumoto R, Ogura Y, Kusuda T. Definition and diagnosis of early cancer of the bile duct. Hepatogastroenterology 1993;40: Kurosaki I, Tsukada K, Watanabe H, Hatakeyama K. Prognostic determinants in extrahepatic bile duct cancer. Hepatogastroenterology 1998;45: Tamada K, Tomiyama T, Wda S, et al. Cholangiographic findings of early-stage extrahepatic bile duct carcinoma. J Gastroenetrol 2001; 36: Cha JM, Kim MH, Lee SK, et al. Clinicopathological review of 61 patients with early bile duct cancer. Clin Oncol (R Coll Radiol) 2006;18: Cha JM, Kim MH, Jang SJ. Early bile duct cancer. World J Gastroeneterol 2007;13: Tamada K, Kanai N, Ueno N, et al. Limitations of intraductal ultrasonography in differentiating between bile duct cancer in stage T1 and stage T2: in-vitro and in-vivo studies. Endoscopy 1997;29: Lim JH, Jang KT, Choi D, Lee WJ, Lim HK. Early bile duct carcinoma: comparison of imaging features with pathologic findings. Radiology 2006;238: Rosch T, Meining A, Fruhmorgen S, et al. A prospective comparison of the diagnostic accuracy of ERCP, MRCP, CT and EUS in biliary strictures. Gastrointest Endosc 2002;55: Zandrino F, Curone P, Benzi L, Ferretti ML, Musante F. MR versus multislice CT cholangiography in evaluating patients with obstruction of the biliary tract. Abdom Imaging 2005;30: Park HS, Lee JM, Kim SH, et al. Preoperative evaluation of bile duct cancer: comparison of MRI combined with MRCP versus MDCT with direct cholangiography. AJR Am J Roentgenol 2008;190: Hanninen EL, Pech M, Jonas S, et al. Magnetic resonance imaging including magnetic resonance cholangiopancreatography for tumor localization and therapy planning in malignant hilar obstructions. Acta Radiol 2005;46: Kim MJ, Mitchell DG, Ito K, Outwater EK. Biliary dilatation: differentiation of benign from malignant causes value of adding conventional MR imaging to MR cholangiopancreatography. Radiology 2000;214: Kim JY, Lee JM, Han JK, et al. Contrast-enhanced MRI combined with MR cholangiopancreatography for the evaluation of patients with biliary strictures: differentiation of malignant from benign bile duct strictures. J Magn Reson Imaging 2007;26: Kim YK, Kim CS, Kwak HS, Lee JM. Three-dimensional dynamic MR imaging using sensitivity encoding for detection of hepatocellular carcinomas: comparison with superraramagnetic iron oxideenhanced MR imaging. J Magn Reson Imaging 2007;20: Lim JH, Yoon KH, Kim SH, et al. Intraductal papillary mucinous tumor of the bile ducts. Radiographics 2004;24: Kawakatsu M, Vilgrain V, Zins M, Vullierme M, Belghiti J, Menu Y. Radiologic features of papillary adenoma and papillomatosis of the biliary tract. Abdom Imaging 1997;22: Kim NR, Lee JM, Kim SH, et al. Enhancement characteristics of cholangiocarcinomas on mutiphasic helical CT: emphasis on morphologic subtypes. Clin Imaging 2008;32: Domagk D, Wessling J, Conrad B, et al. Which imaging modalities should be used for biliary strictures of unknown aetiology? Gut 2007;13: Pichlmayr R, Lamesch P, Weimann A, Tusch G, Ringe B. Surgical treatment of cholangiocellular carcinoma. World J Surg 1995;19: