Focal Pancreatic Mass: Distinction of Pancreatic Cancer From Chronic Pancreatitis Using Gadolinium-Enhanced 3D-Gradient-Echo MRI

JOURNAL OF MAGNETIC RESONANCE IMAGING 26:313 322 (2007) Original Research Focal Pancreatic Mass: Distinction of Pancreatic Cancer From Chronic Pancreatitis Using Gadolinium-Enhanced 3D-Gradient-Echo MRI Jai K. Kim, MD, 1,2 Ersan Altun, MD, 1 Jorge Elias, Jr, MD, PhD, 1,3 Ertan Pamuklar, MD, 1 Hedrick Rivero, MD, 1 and Richard C. Semelka, MD 1 * Purpose: To determine the accuracy of MRI including T1- weighted gadolinium (Gd)-enhanced three-dimensionalgradient-echo (3D-GE) sequences to distinguish pancreatic cancer from chronic pancreatitis in patients with pancreatic mass or focal enlargement. Materials and Methods: The study included 22 patients (15 males and seven females; mean age SD, 56.2 11.5 years) with pancreatic mass or focal enlargement. Fourteen had pancreatic carcinoma and eight had chronic pancreatitis based on the histopathological and clinical findings. MRI examinations of all patients were retrospectively evaluated by two independent reviewers for the predetermined imaging findings of carcinoma and chronic pancreatitis. The accuracy of MRI for differentiating pancreatic carcinoma from chronic pancreatitis was determined. MRI findings of both entities were compared using t-tests, chi-squared tests, and logistic regression analyses for the differentiation of these two entities. The extent of agreement between two reviewers was determined with Kappa statistics. Results: The sensitivity and specificity of MRI including T1- weighted 3D-GE sequences for differentiating pancreatic carcinoma from chronic pancreatitis were 93% (13/14) and 75% (6/8), respectively. The most discriminative finding for pancreatic carcinoma was relative demarcation of the mass compared to background pancreas in contrast to chronic pancreatitis on post-gd 3D-GRE sequences (P 0.05). Conclusion: MRI including Gd-enhanced T1-weighted 3D-GE sequences can differentiate pancreatic carcinoma from chronic pancreatitis successfully in most cases. 1 Department of Radiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA. 2 Department of Radiology, Ajou University Medical School, Suwon-si, South Korea. 3 Department of Radiology, School of Medicine of Ribeirao Preto, University of Sao Paulo, Ribeirao Preto, Brazil. *Address reprint requests to: R.C.S., Department of Radiology, University of North Carolina at Chapel Hill, CB#7510 101 Manning Drive, Chapel Hill, NC 27599-7510. E-mail: richsem@med.unc.edu Contract grant sponsor: CNPq-Brasilia/Brazil. Presented in part at the 92nd Annual Meeting of the Radiological Society of North America, Chicago, 2006. Received August 31, 2006; Accepted February 7, 2007. DOI 10.1002/jmri.21010 Published online in Wiley InterScience (www.interscience.wiley.com). Key Words: pancreatic cancer; chronic pancreatitis; MRI; 3D-gradient-echo; comparative study J. Magn. Reson. Imaging 2007;26:313 322. 2007 Wiley-Liss, Inc. DIFFERENTIATING MALIGNANT FROM BENIGN DIS- EASE in a patient with a pancreatic mass can be difficult (1). Chronic pancreatitis and pancreatic carcinoma often show similar clinical and imaging appearances (2,3). The presence of a pancreatic mass is generally indicative of a neoplasm, but chronic pancreatitis may cause a focal pancreatic mass, usually in the pancreatic head. Furthermore, the incidence of pancreatic carcinoma is much higher in patients with chronic pancreatitis than in the general population (4). Since there is no single definitive diagnostic procedure (whether clinical, biochemical, or radiological) to distinguish carcinoma from chronic pancreatitis, uni- and multivariate approaches have been reported (1,5). The ability to detect pancreatic carcinoma on magnetic resonance imaging (MRI) including T1-weighted gadolinium (Gd)-enhanced three-dimensional gradient-echo (3D-GE) sequences was previously shown (6). The MRI features of a pancreatic mass originating from chronic pancreatitis have also been reported (7 9). A few studies reported that dynamic Gdenhanced MRI was not able to distinguish carcinoma from chronic pancreatitis solely on the basis of extent and time of enhancement of the lesion, because both conditions show abnormal pancreatic enhancement (7,10,11). To our knowledge, the ability of MRI including T1-weighted Gd-enhanced 3D-GE sequences to distinguish carcinoma from chronic pancreatitis in patients with pancreatic mass or focal enlargement has not been previously determined. The purpose of our study was to determine the accuracy of MRI including T1-weighted Gd-enhanced 3D-GE sequences to distinguish pancreatic carcinoma from chronic pancreatitis in patients with pancreatic mass or focal enlargement. 2007 Wiley-Liss, Inc. 313

314 Kim et al. MATERIALS AND METHODS Patients Institutional review board (IRB) approval was obtained for case review only. Signature waiver was also obtained in accordance with IRB regulations and the Health Insurance Portability and Privacy Act (HIPPA). This retrospective study included 22 consecutive patients (15 males and seven females; mean age SD 56.2 11.5 years) with a mass or focal enlargement in the pancreas who underwent MRI between January 2002 and January 2004. Prior to the MRI examination, all patients had clinical symptomatology that included but was not limited to weight loss, abdominal pain, jaundice, fatigue, and early satiety. All patients underwent MRI studies for pancreas evaluation on the basis of clinical indications. CA 19-9 levels were reached in all of the patients. According to histopathological, clinical evaluation, and follow-up results, 14 of the 22 patients (64%) had pancreatic carcinoma and eight (36%) had chronic pancreatitis. Pancreatic Cancer Group Patients All 14 pancreatic carcinoma patients (eight males and six females; mean age SD 60.3 11.8 years) were followed up for a mean time of 9.6 months (range 1 36 months). Nine of the 14 pancreatic carcinoma patients (64%) had associated acute and chronic pancreatitis. Pancreatic carcinoma was diagnosed histopathologically via surgery (4/14), intraoperative biopsy (6/14), and percutaneous biopsy (4/14). Serum CA19-9 was measured in 12 of the 14 patients at the time of initial diagnosis. Chronic Pancreatitis Group Patients All eight chronic pancreatitis patients (seven males and one female; mean age SD 49.1 6.7 years) were followed up for a mean time of 17.9 months (range 3 42 months). The diagnosis of chronic pancreatitis was confirmed histopathologically in five of eight patients via surgery (1/8) and percutaneous biopsy (4/8). The diagnosis was confirmed in the other three patients by clinical evaluation and follow-up. The mean follow-up time was 25.6 months (range 24 27 months) in these patients. Serum CA19-9 was measured in five of the eight patients at the time of initial diagnosis. MRI Technique MR examinations of the abdomen were performed on a 1.5-Tesla MRI scanner (Sonata, Siemens Medical System, Iselin, NJ, USA). All MRI examinations were performed using a set protocol. Gd was administered by power injector (Medrad, Pittsburgh, PA, USA) by bolus of 0.1 mmol/kg of Gd chelate (Omniscan, Nycomed, Princeton, NJ, USA) at 2 ml/second in all patients. MR cholangiopancreatography (MRCP) was performed for 15 of 22 patients (10/14 for pancreatic carcinoma and 5/8 for chronic pancreatitis). Detailed parameters of the protocols and sequences are displayed in Table 1. MRI Interpretation All MRI examinations were retrospectively and independently reviewed by two experienced radiologists for predetermined findings of pancreatic carcinoma and chronic pancreatitis. The patients clinical history and the original MR interpretation were withheld from the reviewing radiologists, who were aware that focal pancreatic disease was present. The independent final diagnoses, based on the overall evaluation of predetermined findings, were recorded. The differentiation between pancreatic carcinoma and chronic pancreatitis was evaluated qualitatively based on predetermined findings, which were as follows: demarcation of the lesion on the post-gd 3D-GE sequence, the presence or absence of duct obstruction or stenosis together with wall irregularity on MRCP or T2-weighted images, preservation or loss of marbled pancreatic texture, pancreatic atrophy, pseudocysts, peripancreatic fluid, lymphadenopathy, vascular encasement, and metastases (2,6,7,10,12,13). Table 1 Parameters of Protocols and Sequences Protocols (n/n) Planes Fat suppression IV contrast Repetition time (TR) (msec) Echo time (TE) (msec) Flip angle (degrees) Section thickness (mm) Conventional protocol (22/22) SGE (T1-weighted) Axial/coronal ( ) Pre and post a 140 2.2 4.4 80 8 10 128 256 SGE (T1-weighted) Axial ( ) Pre 147 4.1 80 6 144 256 3D-GE (T1-weighted) Axial ( ) Pre and post a 4.3 1.7 10 3.5 144 320 HASTE (T2-weighted) Axial/coronal ( ) and ( ) Pre 90 180 8 10 192 256 MRCP (15/22) b HASTE (T2-weighted) Axial/coronal oblique ( ) 99 150 4 240 256 TSE (T2-weighted) Coronal/coronal oblique ( ) 3100 1200 150 30 40 240 256 a Non-fat-suppressed SGE (2/22) and fat-suppressed 3D-GE (20/22) images were acquired 18 seconds (hepatic arterial-dominant phase), fat-suppressed 3D-GE (22/22) images 45 seconds (portal venous phase), and fat-suppressed SGE (22/22) images 120 seconds (interstitial phase) after contrast administration. b Thin section source images were reconstructed into maximum intensity projection (MIP) images. SGE spoiled gradient echo (breath-hold), 3DGE three-dimensional gradient echo (breath-hold), HASTE half-fourier rapid acquisition single-shot turbo spin echo (non-breath-hold), TSE turbo spin echo, MRCP magnetic resonance cholangiopancreatography. Matrix size

Pancreatic Mass Distinction Using 3D-GE 315 A relatively well-defined demarcation of the lesion with lower enhancement compared to background pancreas on post-gd 3D-GE sequences was accepted as the sign of pancreatic carcinoma, whereas an ill-defined demarcation of the lesion with heterogeneous signal and enhancement was accepted as the sign of chronic pancreatitis (7,10). This finding was evaluated by comparing hepatic arterial and portal venous phase enhancement patterns. The presence of high-grade duct obstruction with abrupt or irregular transition was accepted as the sign of pancreatic carcinoma, whereas the absence of duct obstruction or the presence of a tapered or regular transition in duct caliber was accepted as the sign of chronic pancreatitis (12). Loss of the marbled texture of the pancreas was accepted as the sign of pancreatic carcinoma, whereas preservation of the marbled texture of pancreas was accepted as the sign of chronic pancreatitis (6). This finding was evaluated on both hepatic arterial dominant and portal venous phases of the post-gd 3D-GE sequences. Vascular encasement was evaluated in all patients for celiac axis, common hepatic, splenic and superior mesenteric arteries, and portal, splenic, and superior mesenteric veins. Vascular encasement was accepted as positive when the perivascular fat appeared obliterated by a cuff tissue that encompassed 180 or more of the vessel circumference (14,15). The final MRI diagnoses were correlated with subsequent clinical, surgical, and pathology reports to determine the accuracy of MRI for differentiating pancreatic carcinoma from chronic pancreatitis. Additionally, the accuracy of relative demarcation of the lesion on post-gd 3D-GE sequences, the presence of high-grade duct obstruction with abrupt or irregular transition on MRCP or T2-weighted images, and the loss of pancreatic marble texture on post-gd 3D-GE sequences were calculated. Disagreements between the reviewers were determined and resolved during the consensus reading. The reviewers independently evaluated MRI sequences for lesion detection and determined the best sequence demonstrating the lesion. The reviewers rated sequences, including precontrast T1-weighted, T2- weighted, post-gd hepatic arterial dominant phase, and portal venous phase 3D-GE sequences, for lesion detection using a five-point scale that ranged from 1 representing poor to 5 representing excellent lesion detection. After independent MRI evaluations, the frequencies and percentages of predetermined findings were determined by the reviewers with a consensus reading. The reviewers also qualitatively determined lesion signal intensities (SIs) compared to background pancreas for all sequences by a consensus reading. Lesion SI differences compared to background pancreas were quantitatively determined as percentages for precontrast T1- weighted (spoiled gradient echo [SGE] and 3D-GE), hepatic arterial, and portal venous phase 3D-GE sequences. The SIs of the lesions and background pancreas were measured by using the standard region of interest (ROI), and the percentage of SI differences was calculated with the following formula: Relative SI of the Lesion SI (Lesion)/SI (Pancreas) 100. The change in enhancement [delta ( ) enhancement] from hepatic arterial to portal venous phase was calculated using the relative SI changes of lesions compared to background pancreas by the following formula: enhancement Relative SI (Portal venous phase) Relative SI (Hepatic arterial dominant phase)/relative SI (Hepatic arterial dominant phase) 100. The SI of background pancreas was measured from the parenchymal region that was the least involved with the inflammation and lesion. All quantitative analyses were performed by two reviewers in consensus. Statistical Analyses Our study was designed and analyzed as recommended by the Standards for Reporting Diagnostic Accuracy Steering Group (STARD). Lesion detection rates and quantitative lesion SI differences compared to background pancreas were analyzed with t-tests. Qualitative evaluation of lesion SIs for all sequences was analyzed with chi-squared tests. MRI findings of the two groups of patients were also compared with chi-squared tests. Logistic regression analyses were performed to determine the effects of different combinations of findings on the differentiation of pancreatic carcinoma from chronic pancreatitis. Kappa statistics were used to measure the extent of agreement between the final diagnoses of the two reviewers. Associations were considered statistically significant if two-tailed P 0.05. RESULTS Patients Pancreatic Cancer Group Patients The mean size of the masses in this group was 3.5 0.8 3.2 1 cm. Pancreatic carcinomas were located in 13 of 14 patients (93%) in the pancreatic head, and in one patient in the pancreatic tail. An increased level of serum CA 19-9 was detected in 10 of 12 patients (83%), whereas a normal level of serum CA 19-9 was found in two of 12 patients (17%). Chronic Pancreatitis Group Patients The mean size of the masses in this group was 4.4 1.3 3.7 1 cm. There was no statistically significant difference in the size of the masses between pancreatic carcinoma and chronic pancreatitis (P 0.05). The masses were located in the head in seven of eight patients, and in the body in one of eight patients. An increased level of CA 19-9 was detected in one of five patients (20%), whereas a normal serum level was detected in four of five patients (80%). MRI Findings and Accuracy of MRI On MRI, pancreatic carcinoma was correctly diagnosed in 13 of 14 patients (93%) and chronic pancreatitis was correctly diagnosed in six of eight patients (75%). The sensitivity, specificity, and positive and negative predictive values of MRI for the differentiation of pancreatic

316 Kim et al. Table 2 Frequencies and Proportions of MRI Findings MRI findings Pancreatic carcinoma (14/22) a (%) Chronic pancreatitis (8/22) (%) Well-defined demarcation of the lesion with relatively low signal and homogeneous enhancement on postgadolinium 3D-GE b 13/14 (93) c 2/8 (25) c Presence of pancreatic duct obstruction or presence of patent duct with wall irregularity 12/13 d (92) e 4/8 (50) e Loss of marbled pancreatic texture 14/14 (100) 8/8 (100) Pancreatic duct dilatation 11/14 (78.5) 7/8 (87.5) Pancreatic atrophy 6/14 (42) 5/8 (62.5) Pseudocyst 2/14 (14) 5/8 (62.5) Peripancreatic fluid 5/14 (36) 3/8 (38) Lymphadenopathy 4/14 (28) 2/8 (25) Vascular encasement 8/14 (57) 8/8 (100) Metastases 4/14 (28) 0/8 (0) a Numbers of patients. b 3D-GE is three-dimensional gradient echo. c The lesions with ill-defined demarcation with heterogeneous signal and enhancement were detected in the remaining patients. d Pancreatic duct could not be evaluated in 1 of 14 patient who had pancreatic tail mass. e Patent pancreatic ducts with regular walls were detected in the remaining patients. carcinoma from chronic pancreatitis were 93% (13/14), 75% (6/8), 87% (13/15), and 86% (6/7), respectively. The frequency and proportions of the MRI findings are displayed in Table 2. The agreement between two reviewers was substantial ( 0.79). There was no disagreement between the reviewers for the chronic pancreatitis group patients. However, there were disagree- Figure 1. A 72-year-old man with pancreatic carcinoma arising in the head. T2-weighted echo-train SE (a), fat-suppressed T1-weighted SGE (b), hepatic arterial dominant (c), and portal venous phase (d) post-gd fat-suppressed T1-weighted 3D-GE images. A small cancer is present in the pancreatic head (arrows, a d), which is isointense with background pancreas on precontrast sequences (a and b). The cancer is most clearly defined as a mass with demarcated borders on the hepatic arterial dominant phase (c). On the portal venous phase, the cancer preserves its demarcated borders compared to background pancreas despite its progressive enhancement (d).

Pancreatic Mass Distinction Using 3D-GE 317 ments for two patients in the pancreatic carcinoma group. The sensitivity and specificity of relatively well-defined demarcation of the lesion with low signal and homogeneous enhancement compared to background pancreas on post-gd 3D-GE sequences for the differentiation of pancreatic carcinoma (Fig. 1) from chronic pancreatitis (Fig. 2) were 93% (13/14) and 75% (6/8), respectively. The sensitivity and specificity of the presence of highgrade duct obstruction with abrupt or irregular transition for the differentiation of pancreatic carcinoma on MRCP or T2-weighted images were 92% (12/13) and 50% (4/8), respectively (Fig. 3). The sensitivity and specificity of the loss of pancreatic marble texture for the differentiation of pancreatic carcinoma on post-gd 3D-GE sequences were 100% and 0%, respectively. The accuracies of other findings were not calculated because they can be detected in both conditions commonly (except metastases), and therefore their diagnostic roles were supplementary rather than discriminative. A relatively well-defined demarcation of the lesion with low signal and homogeneous enhancement compared to background pancreas on post-gd 3D-GE sequences was the most discriminative finding for the differentiation of pancreatic carcinoma (P 0.05). The regression analyses of different combinations of MRI findings showed that only the combination of demarcation of the lesion on post-gd 3D-GE sequence with the presence or absence of duct obstruction or stenosis together with wall irregularity was effective for the differentiation of these two entities (P 0.05). The rating of sequences determined by two reviewers independently showed that the post-gd hepatic arterial dominant phase 3D-GE sequence had the highest rate for lesion detection (Table 3), which was statistically significant (P 0.05) compared to other sequences. Qualitative evaluation of lesion SIs compared to background pancreas showed that there was no statistically significant SI difference between pancreatic carcinoma and chronic pancreatitis on T1- or T2-weighted precontrast sequences and the hepatic arterial dominant phase 3D-GE sequence (P 0.05; Table 4). How- Figure 2. A 54-year-old man with focal inflammatory mass originating from chronic pancreatitis in the head of the pancreas. T2-weighted echo-train SE (a), fat-suppressed T1-weighted SGE (b), hepatic arterial dominant phase (c), and portal venous phase (d) post-gd fat-suppressed 3D gradient-echo images are shown. There is a focal enlargement of the pancreatic head (arrows, a and b) with heterogeneous SI on both T1- and T2-weighted images. On the hepatic arterial dominant phase the lesion shows a relatively well-defined demarcation (arrowheads, c) compared to background pancreas. On the portal venous phase the lesion demonstrates an ill-defined demarcation (arrowheads, d) and progressive heterogeneous enhancement, and has become less differentiated from the background pancreas.

318 Kim et al. Figure 3. MRCP images of a 74-year-old woman with pancreatic carcinoma originating from the head of the pancreas (a and b), and a 45-year-old man with focal inflammatory mass originating from chronic pancreatitis in the head of the pancreas (c and d). In the patient with pancreatic carcinoma, a thick-section TSE image (a) demonstrates the dilated common bile duct and pancreatic duct (arrows). A thin-section echo-train SE image (b) demonstrates the obstruction of both ducts (arrows). In the patient with chronic pancreatitis, a thick-section TSE image (c) demonstrates the dilated common bile duct and pancreatic duct (arrows). A thin-section echo-train SE image (d) demonstrates the obstruction of pancreatic duct (arrow). Pancreatic carcinoma and chronic pancreatitis demonstrate similar ductal changes on MRCP images in these cases. ever, there was a statistically significant SI difference between the pancreatic carcinoma and chronic pancreatitis groups on the portal venous phase 3D-GE sequence (P 0.05; Table 4). A quantitative evaluation of relative lesion SI differences compared to background pancreas (Table 5 and Fig. 4) showed that pancreatic carcinomas and chronic pancreatitis lesions did not have a statistically signifi- Table 3 Lesion Detectability Rates of the Reviewers For Sequences Sequences Reviewer 1 Reviewer 2 T1-weighted (precontrast SGE and 3D-GE) 2.77 0.61 2.59 0.59 T2-weighted (HASTE) 2.18 0.59 2.31 0.56 Postgadolinium 3D-GE hepatic arterial dominant phase 4.68 0.78 4.54 0.8 Postgadolinium 3D-GE portal venous phase 4.13 0.88 3.95 0.95 SGE spoiled gradient echo, 3D-GE three dimensional gradient echo, HASTE half-fourier rapid acquisition single-shot turbo spin echo.

Pancreatic Mass Distinction Using 3D-GE 319 Table 4 Distribution of Qualitative Lesion Signal Intensities Relative to Background Pancreas According to Sequences Sequences Pancreatic carcinoma (N 14) Hypointense Iso- or hyperintense Chronic pancreatitis (N 8) Hypointense Iso- or hyperintense T1-weighted (SGE) 7/14 7/14 5/8 3/8 T2-weighted (HASTE) 14/14 1/8 7/8 Fat-suppressed T1-weighted (SGE and 3D-GE) 5/14 9/14 1/8 7/8 Fat-suppressed T2-weighted (HASTE) 14/14 8/8 Postgadolinium T1-weighted (hepatic arterial dominant phase) (3D-GE) 11/12 a 1/12 a 5/8 3/8 Postgadolinium T1-weighted (portal venous phase) (3D-GE) 13/14 1/14 2/8 6/8 a 3D-GE sequence was not acquired in two patients on hepatic arterial phase. SGE spoiled gradient echo, HASTE half-fourier rapid acquisition single-shot turbo spin echo, 3D-GE three dimensional gradient echo. cant SI difference on precontrast sequences (P 0.05). On the hepatic arterial dominant phase, pancreatic carcinomas demonstrated a higher SI difference compared to chronic pancreatitis lesions because they demonstrated relatively less enhancement compared to chronic pancreatitis. However, this difference was not statistically significant (P 0.05). On the portal venous phase, pancreatic carcinomas demonstrated statistically significant (P 0.05) higher SI difference compared to chronic pancreatitis because chronic pancreatitis lesions demonstrated relatively more enhancement. The mean enhancements for pancreatic carcinomas and chronic pancreatitis were 28.3 37.1 and 33.6 30.3, respectively. Pancreatic carcinomas demonstrated lower enhancement compared to chronic pancreatitis, although it was not statistically significant (P 0.05). Discordant Cases Pancreatic carcinoma could not be detected in one of 14 patients in the pancreatic carcinoma group on MRI because the lesion was heterogeneously isointense on all sequences with a heterogeneous background pancreas. In the chronic pancreatitis group, two of eight patients were diagnosed as having pancreatic carcinoma. These lesions were part of the larger, heterogeneous focal pancreatic enlargements and showed a well-defined demarcation with low signal and homogeneous diminished enhancement on post-gd 3D-GE sequences (Fig. 5). Each reviewer incorrectly diagnosed pancreatic carcinoma as chronic pancreatitis in one each of these two patients. The evaluation of demarcation of the lesion on post-gd 3D-GE sequence (in one patient), and the presence or absence of duct obstruction or stenosis together with wall irregularity on MRCP images (in two patients) resulted in the incorrect diagnoses. DISCUSSION The present study shows that MRI including post-gd 3D-GE sequences has a high sensitivity and moderate specificity for the differentiation of pancreatic carcinoma from chronic pancreatitis. The most discriminative MRI finding for the differentiation of pancreatic carcinoma from chronic pancreatitis was a relatively well-defined demarcation of the lesion with low signal and homogeneous enhancement compared to background pancreas on post-gd 3D-GE sequences. This finding had the same sensitivity and specificity as the overall MRI interpretation. The reason for the moderate specificity of MRI was two false-positive discordant cases. These patients had focal pancreatic enlargements with ill-defined demarcation and heterogeneous enhancement on post-gd 3D-GE sequences, especially on the portal venous phase, which was indicative of chronic pancreatitis. However, because pancreatic carcinoma has an increased tendency to develop on the basis of chronic pancreatitis, small well-demarcated lesions located in these large lesions were assessed as pancreatic carcinoma. They probably represented proteinaceous cysts that were isointense with the background pancreas on T1- and T2-weighted sequences. Additionally, pancreatic carcinoma could not be described in one falsenegative patient on MRI. The lesion did not possess well-demarcated margins and was heterogeneously isointense on all sequences. Therefore, it was assessed as chronic pancreatitis. Infiltrating pancreatic carcinomas lack lesion demarcation and are therefore difficult to differentiate from chronic pancreatitis. Table 5 Distribution of Quantitative Lesion Signal Intensities Relative to Background Pancreas According to Sequences Sequences Lesion SI/background pancreas SI 100 (%) Pancreatic carcinoma (%) (mean SD) Chronic pancreatitis (%) (mean SD) Precontrast T1-weighted SGE or 3D-GE 94.7 20.6 85.1 34.7 Postgadolinium 3D-GE hepatic arterial dominant phase 64.2 16.4 81.7 30.3 Postgadolinium 3D-GE portal venous phase 79 22.4 103 24

320 Kim et al. Figure 4. A box-plot graph demonstrates the relative SIs of the lesions compared to background pancreas in pancreatic carcinoma (PCar) and chronic pancreatitis (CPan) patient groups on precontrast SGE or 3D-GE, hepatic arterial dominant phase, and portal venous phase 3D-GE sequences. There is less overlap of the relative SIs of the lesions on the portal venous phase 3D-GE sequence compared to the hepatic arterial dominant phase 3D-GE and precontrast SGE and 3D-GE sequences. Qualitative and quantitative SI analyses showed that it was not possible to differentiate pancreatic carcinoma and chronic pancreatitis on the basis of the SI characteristics of precontrast sequences. The lesions could not be successfully demarcated even in fat-suppressed T1-weighted sequences due to the resultant changes of the background pancreas parenchyma. Pancreatic carcinoma and chronic pancreatitis were demarcated from the background pancreas best shown on hepatic arterial dominant phase images because they demonstrated lower signal compared to background pancreas qualitatively and quantitatively. Additionally, although chronic pancreatitis showed more enhancement than pancreatic carcinomas on hepatic arterial dominant phase images qualitatively and quantitatively, the SI of chronic pancreatitis was still lower than that of background pancreas. On the portal venous phase, pancreatic carcinoma remained relatively demarcated from background pancreas qualitatively and quantitatively. Although pancreatic carcinoma showed progressive relatively homogeneous enhancement on portal venous phase images compared to the hepatic arterial dominant phase, the progression of increased enhancement was less than that observed for chronic pancreatitis qualitatively and quantitatively. Therefore, chronic pancreatitis had higher enhancement compared to pancreatic carcinoma. The qualitative and quantitative analyses supported the finding that pancreatic carcinomas demonstrated a well-defined demarcation compared to background pancreas in contrast to chronic pancreatitis. We would expect this feature of a relatively persistent lower signal for pancreatic cancer to be the most useful for larger cancers, whereas smaller cancers tend to show greater enhancement over time (16). Because the differentiation of pancreatic carcinoma from chronic pancreatitis may be very difficult even for larger tumors, we believe this finding may be very useful. The present study also indicates that post-gd 3D-GE is the best sequence for lesion detection, which concurs with prior studies (6) due to its known advantages (17,18). The results of our study show a superior ability to make the distinction between these two entities compared to a prior report that described quantitative analysis of postcontrast enhancement (10). The explanation of this likely reflects the importance of using the 3D-GE sequence, in which morphological and enhancement features are better evaluated. The present study indicates that the duct obstruction or the stenotic-dilated duct with irregular walls had high sensitivity (92%) but low specificity (50%) for the differentiation of pancreatic carcinoma from chronic pancreatitis. The specificity of the study was not in agreement with the high specificity (96 100%) reported by Ichikawa et al (12). Not all patients in our study underwent MRCP studies, which may be a contributing factor for lower specificity. We believe this discrepancy may largely reflect the fact that it may be difficult to differentiate between the abrupt or irregular transition of duct caliber in pancreatic carcinoma from the tapered or regular transition of duct caliber in chronic pancreatitis, and that overlap exists. Therefore, our impression is that although MRCP is a very useful tool for pancreatic duct evaluation, the role of MRCP in the diagnosis and differentiation of pancreatic carcinoma should be supplementary to conventional MRI examinations, and both examinations should be evaluated together (13). Our results suggest the value of combining duct findings with the demarcation of the lesion on post-gd 3D-GE sequences for the differentiation of these two entities. The loss of pancreatic marble texture could not differentiate pancreatic carcinoma from chronic pancreatitis in the present study because large chronic pancreatitis lesions also cause the destruction of marbled pancreatic texture. One limitation of our study is the small study population. Nonetheless, this represents our complete twoyear experience with MRI evaluation of pancreatic masses using this technique. Another limitation is that three of 22 patients did not have histopathologic correlation; however, we believe these patients had an adequately long-term clinical follow-up correlation. In conclusion, MRI including T1-weighted post-gd 3D-GE sequences can successfully differentiate pancreatic carcinoma from chronic pancreatitis. The discriminative feature of carcinoma was a relatively welldefined demarcation with relatively lower signal and decreased enhancement compared to background pancreas. In contrast, the discriminative feature of chronic pancreatitis was an ill-defined demarcation with relatively increased signal and enhancement. Overlap was found between these two entities, reflecting the facts that pancreatic carcinoma may be infiltrative and chronic pancreatitis can be destructive and sharply marginated. The comparison of hepatic arterial and

Pancreatic Mass Distinction Using 3D-GE 321 Figure 5. A 53-year-old man with focal inflammatory mass originating from chronic pancreatitis in the pancreatic head. T2-weighted echo-train SE (a), T1-weighted SGE (b), hepatic arterial dominant phase (c), and portal venous phase (d) post-gd fat-suppressed 3D-GE images are shown. There is an enlargement of the pancreatic head (white arrows, a d) that has heterogeneous SI on both T1- and T2-weighted images, and contains a few small cysts. No demarcation of the lesion edges is apparent. The mass shows diminished enhancement on the hepatic arterial dominant phase (arrowheads, c) but progress in intensity to a marked heterogeneous enhancement over time compared to normal adjacent pancreatic parenchyma. There is a focal well-demarcated area (black arrow, d) with very diminished enhancement especially on the portal venous phase, which is isointense to background pancreas on T2-weighted images. This lesion represented a high-protein-content cyst; however, it was evaluated as pancreatic carcinoma. portal venous phase enhancements ( enhancement) is useful for the diagnosis of these two entities. ACKNOWLEDGMENT Jorge Elias, Jr., is funded by CNPq-Brasilia/Brazil. REFERENCES 1. van Gulik TM, Moojen TM, van Geenen R, Rauws EA, Obertop H, Gouma DJ. Differential diagnosis of focal pancreatitis and pancreatic cancer. Ann Oncol 1999;10(Suppl 4):85 88. 2. Steer ML, Waxman I, Freedman S. Chronic pancreatitis. N Engl J Med 1995;332:1482 1490. 3. Freeny PC. Radiology of the pancreas: two decades of progress in imaging and intervention. AJR Am J Roentgenol 1988;150:975 981. 4. Lowenfels AB, Maisonneuve P. Risk factors for pancreatic cancer. J Cell Biochem 2005;95:649 656. 5. Cho SG, Lee DH, Lee KY, et al. Differentiation of chronic focal pancreatitis from pancreatic carcinoma by in vivo proton magnetic resonance spectroscopy. J Comput Assist Tomogr 2005;29:163 169. 6. Birchard KR, Semelka RC, Hyslop WB, et al. Suspected pancreatic cancer: evaluation by dynamic gadolinium-enhanced 3D gradientecho MRI. AJR Am J Roentgenol 2005;185:700 703. 7. Kim T, Murakami T, Takamura M, et al. Pancreatic mass due to chronic pancreatitis: correlation of CT and MR imaging features with pathologic findings. AJR Am J Roentgenol 2001;177:367 371. 8. Van Hoe L, Gryspeerdt S, Ectors N, et al. Nonalcoholic duct-destructive chronic pancreatitis: imaging findings. AJR Am J Roentgenol 1998;170:643 647. 9. Semelka RC, Shoenut JP, Kroeker MA, Micflikier AB. Chronic pancreatitis: MR imaging features before and after administration of gadopentetate dimeglumine. J Magn Reson Imaging 1993;3:79 82. 10. Johnson PT, Outwater EK. Pancreatic carcinoma versus chronic pancreatitis: dynamic MR imaging. Radiology 1999;212:213 218. 11. To o KJ, Raman SS, Yu CN, et al. Pancreatic and peripancreatic diseases mimicking primary pancreatic neoplasia. Radiographics 2005;25:949 965.

322 Kim et al. 12. Ichikawa T, Sou H, Araki T, et al. Duct penetrating sign at MRCP: usefulness for differentiating inflammatory pancreatic mass from pancreatic carcinoma. Radiology 2001;221:107 116. 13. Hanninen EL, Amthauer H, Hosten N, et al. Prospective evaluation of pancreatic tumors: accuracy of MR imaging with MR cholangiopancreatography and MR anjiography. Radiology 2002;224:34 41. 14. Sironi S, De Cobelli F, Zerbi A, et al. Pancreatic adenocarcinoma: assessment of vascular invasion with high-field MR imaging and a phased array coil. AJR Am J Roentgenol 1996;167:997 1001. 15. Li H, Zeng MS, Zhou KR, Jin DY, Lou WH. Pancreatic adenocarcinoma: the different CT criteria for peripancreatic major arterial and venous invasion. J Comput Assist Tomogr 2005;29:170 175. 16. Pamuklar E, Semelka RC. MR imaging of the pancreas. Magn Reson Imaging Clin N Am 2005;13:313 330. 17. Karabulut N, Martin DR, Yang M, Tallaksen RJ. MR imaging of the chest using a contrast-enhanced breath-hold modified three-dimensional gradient-echo technique: comparison with two-dimensional gradient-echo technique and multidetector CT. AJR Am J Roentgenol 2002;179:1225 1233. 18. Gohde SC, Toth J, Krestin GP, Debatin JF. Dynamic contrastenhanced FMPSPGR of the pancreas: impact on diagnostic performance. AJR Am J Roentgenol 1997;168:689 696.