- Email: [email protected]
CT for evaluation of pancreatic lesions
European Journal of Radiology 81 (2012) 2527–2532
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
MR imaging versus PET/CT for evaluation of pancreatic lesions Sara Belião a,∗ , Alexandra Ferreira b,1 , Irina Vierasu c,2 , Didier Blocklet c,2 , Serge Goldman c,2 , Thierry Metens d,3 , Celso Matos d,3 a
Department of Radiology Hospital S. Francisco Xavier, Estrada do Forte do Alto do Duque, 1495-005 Lisbon, Portugal Department of Radiology, Hospital D. Estefânia, Rua Jacinta Marto, 1169-045 Lisbon, Portugal Service de Médecine Nucléaire, Route de Lennik 808, 1070 Brussels, Belgium d Service de Radiologie – Imagerie par Resonance Magnétique, Route de Lennik 808, 1070 Brussels, Belgium b c
a r t i c l e
i n f o
Article history: Received 19 April 2011 Received in revised form 23 November 2011 Accepted 24 November 2011 Keywords: Pancreatic cancer Chronic pancreatitis MRI FDG-PET/CT
a b s t r a c t Purpose: To retrospectively determine the diagnostic accuracy of magnetic resonance imaging (MRI) and combined positron emission tomography/computed tomography (PET/CT) in the differential diagnosis of benign and malignant pancreatic lesions. Materials and methods: Twenty-seven patients (15 women/12 men, mean age 56.5 years) with MR imaging and PET/CT studies performed to differentiate benign and malignant pancreatic lesions were identified between October 2008 and October 2010. Both MR and PET/CT data sets were retrospectively and blindly evaluated by two independent readers (4 readers total) with different degrees of experience, using a visual five-point score system. The results were correlated with final diagnosis obtained by histopathology. Results: 17 patients had malignant diseases and 10 patients had benign diseases. Depending on the observer, the sensitivity, specificity, positive predictive value and negative predictive value of MRI varied between 88–94%, 50–80%, 75–89% and 71–89% respectively. Sensitivities, specificities, positive predictive values and negative predictive values of PET/CT were 73%, 56%, 73% and 56% respectively. The diagnostic accuracy of MR for the differential diagnosis of pancreatic lesions was 74–89%, compared with 67% for PET/CT. The weighted Cohen’s kappa coefficient was 0.47 at MR and 0.53 at PET/CT. Conclusion: MRI achieved higher sensitivity and specificity in the differential diagnosis of pancreatic lesions. © 2011 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Pancreatic cancer is usually diagnosed at advanced stages, when the disease is no longer resectable. This is related with the late onset and non-specificity of the symptoms, which determines an overall survival rate of less than 4% [1]. Despite all the imaging modalities available, detection of pancreatic cancer is still challenging and the differential diagnosis with focal pancreatitis is difficult [2,3]. To complicate this issue, chronic pancreatitis may develop into pancreatic carcinoma, and also pancreatic carcinoma may develop obstructive chronic pancreatitis secondary to pancreatic ductal obstruction [4]. This is clinically relevant because the impossibil-
∗ Corresponding author at: R. Prof. Jorge da Silva Horta, n◦ 8, B2, 2◦ FTE, 1500 Lisbon, Portugal. Tel.: +351 964765625; fax: +351 210083643. E-mail addresses: [email protected] (S. Belião), [email protected] (A. Ferreira), [email protected] (I. Vierasu), [email protected] (D. Blocklet), [email protected] (S. Goldman), [email protected] (T. Metens), [email protected] (C. Matos). 1 Tel.: +351 966518734; fax: +351 213126667. 2 Tel.: +32 02 555 47 11; fax: +32 02 555 47 01. 3 Tel.: +32 02 555 43 25; fax: +32 02 555 66 70. 0720-048X/$ – see front matter © 2011 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2011.11.045
ity of obtaining a preoperative diagnosis is associated with adverse outcomes including reoperative pancreaticoduodenectomies and resections of benign disease [5]. The development of more sensitive and specific imaging modalities may avoid these adverse outcomes and improve the diagnostic accuracy of pancreatic lesions [6]. Positron emission tomography (PET) is a functional imaging technique that provides unique molecular and metabolic information. By using the radiotracer 18 F-FDG, PET allows direct evaluation of glucose metabolism and is a validated imaging modality for a variety of oncologic and non-oncologic conditions [7]. A working group concluded in 2008 that PET benefits the differentiation of benign from malignant lesions in the diagnostic work-up of patients with suspected pancreatic lesions. The greatest benefit was attributed to the possibility of excluding cancer without the need for biopsy or surgery [1]. Limitations of PET include unspecific tracer uptake in inflammatory changes such as acute pancreatitis and low spatial resolution however, the development of PET/CT holds great promise as an imaging technique for the detection and staging of pancreatic cancer [8,9]. MR imaging is being increasingly used as a noninvasive diagnostic modality of pancreatic diseases. Improvement in the quality of MR sequences, single shot T2-weighted and three-dimensional
2528
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532
unenhanced and contrast-enhanced T1-weighted gradient-echo sequences facilitated successful depiction of pancreatic lesions [10]. In addition, magnetic resonance colangiopancreatography (MRCP) allows the evaluation of pancreatic and bile ducts noninvasively and has been reported as accurate in the depiction of duct dilation, strictures, tumors and stones [2]. The abilities of MRI in conjunction with MRCP are useful in establishing resectability of the lesions as well as preventing unnecessary endoscopic retrograde cholangiopancreatography (ERCP) and stenting [11]. Diffusion-weighted MR imaging has been increasingly used to characterize pancreatic lesions and may depict small solid masses better than routine MR sequences due to its greater image contrast, despite its relatively poorer resolution [10]. Diffusion-weighted imaging may also be useful in the detection of metastases in the liver and lymph nodes [10]. Although MR imaging provides highresolution images with excellent anatomic delineation of lesions, differentiating benign from malignant pancreatic diseases remains challenging. In the other way, PET allows only limited anatomic localization of an abnormal uptake spot [12]. Despite the considerable attention that has been paid to study the diagnostic accuracy of these techniques, there are still few studies comparing MRI and PET/CT in the same group of patients [13]. The purpose of this study was to retrospectively assess the accuracy of MR and PET/CT imaging in differentiating benign and malignant pancreatic lesions. 2. Materials and methods 2.1. Patients Twenty-seven patients with suspected pancreatic lesions between October 2008 and October 2010 were retrospectively reviewed. Only patients with histopathologic analysis of the lesions and who had undergone both MR imaging and PET/CT examinations within a short interval (mean of 20 days) were included in this study. This group of patients ranged in age from 30 to 84 years (mean age, 56.5 years) and comprised 15 women and 12 men.
anatomical correlation were carried out by the means of a low dose CT (4.4 mGy/slice, 40 mAs/slice) obtained without injection of contrast media. 2.4. MR analysis Two radiologists, with 5 and more than 15 years of experience interpreting MR examinations independently reviewed the MR examinations without knowledge of patient’s identity, clinical information, findings of other imaging examinations and histopathologic results. Observer 1 occasionally interprets body MR examinations, and observer 2 interprets body MR examinations on a daily basis. For the assessment of the lesions, TSE T2-weighted images, MRCP images and DW images with different b values and pre and postcontrast T1-weighted gradient echo images were interpreted. Signal intensity of the lesions on T1, on T2-weighted images, and on DW imaging was qualitatively assessed without using a threshold for hyperintensity. On DW sequences images with b = 0 and 150–1000 s/mm2 were compared looking for areas that did not suppress on the 1000 s/mm2 DW images. Focal nodular or mass like areas of hyperintensity were recorded as tumor. The location and size of pancreatic lesions were reported. The evaluation also comprised the extension of the lesions to the peripancreatic fat or adjacent organs. The presence/absence of lymph nodes and their diameters were assessed. Caliber of the common bile duct and pancreatic duct was measured using MRCP images to assess the presence of dilation or stenoses. Vascular anatomy and assessment of vascular infiltration were registered. Evaluation of vascular infiltration included identification of the celiac axis, the splenic artery, the hepatic artery, the superior mesenteric artery, the portal vein and the superior mesenteric vein. Focal liver lesions were evaluated and classified as benign or metastatic. Based on these parameters, the probability of malignancy was assessed by each reader with a five-point visual scoring system: score 1 indicated no malignancy; score 2 indicated malignancy probably absent; score 3, undetermined; score 4, malignancy probably present; score 5 indicated definitely malignant.
2.2. MR imaging 2.5. PET/CT analysis The examinations were performed using 1.5 T and 3 T magnets (Philips Achieva, Best, the Netherlands) equipped with sense compatible abdominal coils. Sequences included single shot respiratory triggered turbo spin echo (TSE) T2-weighted in the axial and coronal orientations with TE = 80 ms, axial diffusion-weighted (DW) including at least b = 0, b = 150 and b = 1000 s/mm2 values with diffusion sensitization in 3 orthogonal directions. All images had 5 mm thickness. Breath hold MRCP 2D projections with TE = 1000 ms in several oblique orientations were also acquired. Axial 3D T1weighted fat suppressed gradient echo images with an isotropic resolution of 1.5–2 mm were obtained dynamically before and during the arterial, portal venous and equilibrium phases after injection of 0.1 mmol/kg body weight contrast material (Gadovist, Bayer-Schering® , Germany). 2.3. Whole-body 18 F-FDG PET/CT All the patients were injected intravenously with a bolus of 10 mCi (370 MBq) after a fasting period of at least 6 h. Glucose blood level had to be less than 7.5 mmol/L. 18 F-FDG PET/CT were performed on a Gemini-16 PET/CT camera (Philips Medical Systems, Cleveland, OH, USA) in full-3D acquisition mode. Static images from skull base to mid-thigh were acquired starting 80 ± 10 min after FDG injection (3 min per bed position). The images were iteratively reconstructed by means of the Philips-supplied 3D Raw Action Maximum-Likelihood Algorithm (3D-RAMLA) with scatter correction through single scatter simulation. Attenuation correction and
18 F-FDG PET/CT were interpreted by two nuclear medicine practitioners with 5 and 15 years of experience, blinded to tumor location, morphological imaging results, and clinical status of the patients. The coregistered PET and CT images were analyzed on a clinical workstation (Brightview, Philips Medical Systems, Cleveland, OH, USA). The nuclear medicine practitioners analyzed all images independently and each of their evaluation was reported (no consensus reading). Evaluation was primarily based on visual interpretation but standardized uptake value (SUV) values were also considered in the diagnostic process by comparison with values in other organs (liver, spleen). No specific threshold was used to differentiate malignant from benign lesions. The evaluation was not restricted to the upper abdomen; any abnormality detected in the whole-body examination was considered in the global classification of the patient. Criteria used to classified local abnormalities in the pancreatic area included intensity, homogeneity and extension of abnormal FDG uptake. The probability of malignancy was assessed with the same five-point visual score-system used for MR imaging.
2.6. Statistical analysis Based on the histopathologic analysis of the lesions, the accuracy, sensitivity, specificity, positive predictive value and negative predictive value were calculated for both techniques.
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532 Table 1 Distribution of pathology. Malignant disease Pancreatic adenocarcinoma Neuroendocrine tumor Ampulloma Metastasis from melanoma GIST in duodenum Benign disease Chronic pancreatitis Adenopathy Cyst Benign alteration Focal tuberculosis Tumor localization Head Body Tail
2529
diagnose malignancy). The results of the interobserver grading and corresponding weighted kappa’s are given in Fig. 1a (MRI) and b (PET/CT). N = 10 N=4 N=1 N=1 N=1
4. Discussion
N=5 N=1 N=1 N=2 N=1 N=5 N=4 N=6
Grades 1 and 2 were accepted as negative for the presence of malignancy. Grades 4 and 5 were accepted as positive for the presence of malignancy. Regarding Grade 3, it was considered malignant since the main concern of the authors was the ability of excluding malignancy by both techniques. The inter-observer agreement was assessed with the Cohen’s kappa statistics K for the agreement about the diagnosis of malignancy. Additionally, the linear weighted Cohen’s kappa statistics was performed to assess the agreement between scores. The standard error on kappa (SE) was reported as well. The kappa statistic was categorized as follows: 0–0.2, slight agreement; 0.21–0.4, fair agreement; 0.41–0.6, moderate agreement; 0.61–0.8, substantial agreement; 0.81–1, almost perfect agreement [14]. Calculations were performed using the software Medcalc (MedCalc Software, Mariakerke, Belgium). 3. Results All 27 patients were fully evaluated by the MR observers. Given that PET/CT data archived for 3 patients could not be properly retrieved in order to obtain adequate images reconstructions, only 24 patients were fully evaluated by nuclear medicine observers. Among the 17 patients with malignancies, there were 10 adenocarcinomas, 4 neuroendocrine tumors (NET), 1 ampulloma, 1 melanoma metastasis and 1 gastrointestinal stromal tumor (GIST) of the duodenum. Benign findings included chronic pancreatitis in 5 cases, other benign pancreatic lesions (2 patients), focal tuberculosis (1 patient), cystic lesion (1 patient) and adenopathy in the context of HIV (1 patient). Histopathologic results were confirmed in all patients (Table 1). It should be noted that three of the patients with proved adenocarcinoma had background clinical history of chronic pancreatitis. The most common location of pancreatic tumors was the tail of pancreas (n = 6). Other locations included the head (n = 5) and body (n = 4). The size of pancreatic tumors ranged from 0.7 to 3 cm. Open surgery was performed in 6 patients, laparoscopy was performed in 2 patients and biopsy/cytology of the lesions under endoscopic guidance was obtained in 19 patients. The results about diagnosis of malignancy are summarized in Table 2 (Grades 3–5 to
In this study, MR imaging was the most accurate, sensitive and specific method for diagnosing benign and malignant lesions. Depending on the observer, the values of sensitivity and specificity of MRI varied between 88–94% and 50–80% respectively. Overall accuracy was calculated at 74% and 89%. A prospective evaluation of patients with suspected solid pancreatic tumors evaluated with a combination of non-enhanced and contrast-enhanced MRI, MRCP, and contrast-enhanced MR angiography reported a sensitivity of 95% and a specificity of 82% for MRI in the diagnosis of malignancy with an overall accuracy of 90% [2]. The accuracy achieved in our study by the most experienced observer was similar to accuracy of the aforementioned study. Within the group of malignant lesions, three lesions were classified as benign by the MR observers. Two of these patients had background clinical history of chronic pancreatitis. As previously described, the presence of chronic pancreatitis represents an extremely difficult challenge in the differential diagnosis of pancreatic cancer. In the same way, the majority of false-positive MR studies corresponded to patients with chronic pancreatitis. Because pancreatic tumors with equivocal imaging findings are potentially resectable, the observers are more likely to make the diagnosis of cancer in the patients with focal pancreatitis. This tendency explains the relatively high number in false-positive diagnoses of cancer in this study. For both readers, the values of sensitivity and specificity of PET/CT were 73% and 56% respectively. The sensitivity and specificity achieved by PET/CT was lower than the sensibility (89%) and specificity (64–74%) reported on recent literature by Kauhanen et al. [15] and Schick et al. [16]. One possible explanation can be the fact that in those studies the observers were not blinded to clinical information or to imaging findings in previous examinations. It can also be explained by the five-visual point score system used in this study, which can be difficult to apply to PET/CT examinations. There were 4 false-negatives for both readers at PET/CT. Among them, 3 were common to the two readers. These results might be explained with the appearance and small size of the tumors in our population. Given the diffuse pancreatic uptake seen, some tumors were erroneously interpreted at PET/CT as inflammatory activity (Fig. 2). A reduced sensitivity for small pancreatic cancers has also been described for the standard radiotracer 18 F-FDG, presumably because of partial-volume effects [17]. In this study, the pancreatic cancers ranged in size from 0.7 to 3 cm. Additionally, one of the patients had a duodenal GIST treated with imatinib that showed no uptake on PET/CT. The difference in metabolic activity after even a single dose of imatinib can be shown on 18 F-FDG PET scans and is often dramatic [18]. However, it should be noted that unlike the MR results, all the patients with chronic pancreatitis complicated with pancreatic cancer were classified as malignant by the most experienced
Table 2 PET/CT and MRI diagnosis of malignancy (Grades 3–5 to diagnose malignancy). K is the Cohen’s kappa coefficient for diagnosing malignancy. SE – standard error. Accuracy (%) MR Observer 1 Observer 2 PET-CT Observer 1 Observer 2
Sensitivity (%)
Specificity (%)
Positive predictive value (%)
Negative predictive value (%)
K (SE)
74 89
88 (15/17) 94 (16/17)
50 (5/10) 80 (8/10)
75 (15/20) 89 (16/18)
71 (5/7) 89 (8/9)
0.47 (0.19)
67 67
73 (11/15) 73 (11/15)
56 (5/9) 56 (5/9)
73 (11/15) 73 (11/15)
56 (5/9) 56 (5/9)
0.64 (0.16)
2530
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532
Fig. 1. (a) Interobserver grading variability at MRI. The weighted kappa was 0.47 (standard error 0.13). (b) Interobserver grading variability at PET/CT. The weighted kappa was 0.53 (standard error 0.15).
Fig. 2. 62-year-old man with pancreatic adenocarcinoma. T2-weighted MR image shows a mass (arrow) in the head of the pancreas (a). Peripancreatic adenopathies (arrow) are also seen (b). Corresponding axial high-b value (b = 1000 s/mm2 ) diffusion-weighted image (c). Coronal MR cholangiopancreatography image demonstrates the doubleduct sign (d). Positron emission tomography axial and coronal images show intense and diffuse FDG uptake in pancreatic head (e and f). This lesion was classified as malignant by MR readers and considered indetermined by both PET/CT readers.
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532
2531
Fig. 3. 55-year-old woman with chronic pancreatitis complicated of pancreatic adenocarcinoma. A periampullary mass (arrow) is shown in axial T2-weighted and coronal MRCP images (a and b). MRCP also demonstrates dilation of the pancreatic duct with stigmas of chronic pancreatitis. The common bile duct is not dilated (b). PET axial and coronal images show a focal FDG uptake in pancreatic head region (c and d). MR readers classified the imaging findings as benign and the most experienced PET/CT reader considered this lesion malignant.
PET/CT observer (Fig. 3). These results are consistent with previous studies, which have shown that in this group of patients PET-scans are usually positive [19]. Regarding the 5 cases with chronic pancreatitis, one the PET/CT observers classified 3 lesions as benign and one lesion as malignant with score 3; the other observer classified 3 lesions as benign and one lesion as malignant with score 5. One of the cases was not assessed due to technical problems. Therefore, the majority of false-positive cases registered at PET/CT were not related with patients with chronic pancreatitis. Falsepositive PET-scans have been noted in the presence of acute and chronic inflammatory reactions, including granulomatous disease, osteomyelitis and abdominal abscesses [5]. Three of the patients with false-positive PET/CT studies in this series had peri-pancreatic adenopathies (one patient with HIV, one patient with focal tuberculosis and one patient with resected benign disease of the common bile duct). All readers (MR and PET/CT studies) misinterpreted focal tuberculosis as malignant. In fact, the radiologic findings of focal tuberculosis are not specific and usually mimic pancreatic malignancy. The interobserver agreement calculated with kappa statistics for both MR and PET-CT readers was moderate (weighted kappa of 0.47 at MRI and 0.53 at PET/CT). This aspect can be due to the great variability of cases included in our series and to the different
experience degrees between observers. Fig. 1a and b shows a higher agreement at MRI in the attribution of scores 4 and 5, however score 3 is less reproducible. This explains the best kappa value obtained for diagnosis of malignancy based on scores 3–5 at PET/CT. This study was limited by its retrospective design and by the low number of patients. A small sample size is due to the relatively low number of patients with histopathologic confirmation who had undergone MRI and PET/CT within a short interval of time. A case selection bias must also be considered. At our institution, the majority of patients submitted to both techniques are those where the differential diagnosis between pancreatic cancer and chronic pancreatic is difficult. Apparent diffusion coefficient (ADC) values were not used to assess the probability of malignancy because of the reported overlap between malignant and benign lesions [10]. In the same way, no specific SUV threshold was used to differentiate malignant from benign lesions because a previous study has shown that malignant and inflammatory lesions present with similar SUV values and are better differentiated based on the FDG uptake distribution [20]. The results from this study indicate that MR is able to depict and characterize pancreatic lesions, especially if performed by experienced readers. The major drawbacks were noted in patients with chronic pancreatitis and in patients with chronic pancreatitis complicated of pancreatic cancer. Once our results suggest a better
2532
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532
performance of PET/CT in this specific group of patients, the combination of both MR and PET/CT may yield the highest overall value for the diagnosis of pancreatic lesions. In fact, negative predictive value for malignancy seems higher at MRI while PET-CT could help in identifying chronic pancreatitis. In our study, information provided by DW-imaging was especially helpful for the detection of the lesions and less helpful for their characterization. Like conventional MR imaging, it is difficult to differentiate pancreatic adenocarcinoma from mass-forming pancreatitis at DW-imaging [10]. Furthermore, DW images can be fused with conventional MR images, like the fusion images obtained with PET/CT scanners, to achieve better anatomic resolution. In our study, MR readers did not have access to this kind of fusion images, but some studies have reported their advantages [6]. 5. Conclusion The results of this study show that MRI has higher performance than PET/CT for the characterization of pancreatic lesions. Nevertheless, both techniques presented diagnostic limitations when compared with previous studies. Further studies in larger clinical settings that compare diagnostic performance of each technique with the accuracy of fully integrated PET/MRI systems are needed to support our findings. References [1] Fletcher J, Djulbegovic B, Soares H, et al. Recommendations on the use of 18 FFDG PET in oncology. J Nucl Med 2008;49(3):480–508. [2] Hanninen EH, Amthauer H, Hosten N, et al. Prospective evaluation of pancreatic tumors: accuracy of MR imaging with MR cholangiopancreatography and MR angiography. Radiology 2002;224:34–41. [3] 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;10:85–8. [4] Tajima Y, Kuroki T, Tsutsumi R, Isomoto I, Uetani M, Kanematsu T. Pancreatic carcinoma coexisting with chronic pancreatitis versus tumor-forming pancreatitis: diagnostic utility of the time-signal intensity curve from dynamic contrast-enhanced MR imaging. World J Gastroenterol 2007;13(6): 858–65.
[5] Rose DM, Delbeke D, Beauchamp MDRD, et al. 18 Fluorodeoxyglucose-positron emission tomography in the management of patients with suspected pancreatic cancer. Ann Surg 1999;229(5):729. [6] Ichikawa T, Erturk SM, Motosugi U, et al. High-b value diffusion-weighted MRI for detecting pancreatic adenocarcinoma: preliminary results. AJR 2007;188:409–14. [7] Kwee T, Takahara T, Ochiai R, et al. Complementary roles of whole-body diffusion-weighted MRI and 18 F-FDG PET: the state of the art and potential applications. J Nucl Med 2010;51:1549–58. [8] Sahani DV, Zarine KS, Catalano OA, Boland GW, Brugge WR. Radiology of pancreatic adenocarcinoma: current status of imaging. J Gastroenterol Hepatol 2008;23:23–33. [9] Kalra M, Maher M, Boland G, Saini S, Fischman A. Correlation of positron emission tomography and CT in evaluating pancreatic tumors: technical and clinical implications. AJR 2003;181:387–93. [10] Wang Y, Miller F, Chen Z, et al. Diffusion-weighted MR imaging of solid and cystic lesions of the pancreas. Radiographics 2011;31:E47–65. [11] Kalra M, Maher M, Mueller P, et al. State of the art imaging of pancreatic neoplasms. Br J Radiol 2003;76:857–65. [12] Balci NV, Semelka RC. Radiologic diagnosis and staging of pancreatic ductal adenocarcinoma. Eur J Radiol 2001;38:105–12. [13] Bipat S, Phoa S, Delden O, et al. Ultrasonography, computed tomography and magnetic resonance imaging for diagnosing and determining resectability of pancreatic adenocarcinoma, a meta-analysis. J Comput Assist Tomogr 2005;29:438–45. [14] Svanholm H, Starklint H, Gundersen HJ, Fabricius J, Barlebo H, Olsen S. Reproducibility of histomorphologic diagnoses with special reference to the kappa statistic. AP-MIS 1989;97:689–98. [15] Kauhanen S, Komar G, Seppänen M, et al. A Prospective diagnostic accuracy study of 18 F-fluorodeoxyglucose positron emissiontomography/computed tomography multidetector rowcomputed tomography, and magnetic resonance imaging in primary diagnosis and staging of pancreatic cancer. Ann Surg 2009;250:957–63. [16] Schick V, Franzius C, Beyna T, et al. Diagnostic impact of 18 F-FDG PET-CT evaluating solid pancreatic lesions versus endosonography, endoscopic retrograde cholangio-pancreatography with intraductal ultrasonography and abdominal ultrasound. Eur J Nucl Med Mol Imaging 2008;35:1775–85. [17] Herrmann K, Eckel F, Schmidt S, et al. In vivo characterization of proliferation for discriminating cancer from pancreatic pseudotumors. J Nucl Med 2008;49(9):1437–44. [18] Holdsworth CH, Badawi RD, Manola JB, et al. CT and PET: early prognostic indicators of response to imatinib mesylate in patients with gastrointestinal stromal tumor. AJR 2007;189:W324–30. [19] Kouwen M, Jansen J, Goor H, Castro S, Oyen W, Drenth J. FDG-PET is able to detect pancreatic carcinoma in chronic pancreatitis. Eur J Nucl Med Mol Imaging 2005;32(4):399–404. [20] Lee TY, Kim MH, Park do H, et al. Utility of 18 F-FDG PET/CT for differentiation of autoimmune pancreatitis with atypical pancreatic imaging findings from pancreatic cancer. Am J Roentgenol 2009;193:343–8.
Contents lists available at SciVerse ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
MR imaging versus PET/CT for evaluation of pancreatic lesions Sara Belião a,∗ , Alexandra Ferreira b,1 , Irina Vierasu c,2 , Didier Blocklet c,2 , Serge Goldman c,2 , Thierry Metens d,3 , Celso Matos d,3 a
Department of Radiology Hospital S. Francisco Xavier, Estrada do Forte do Alto do Duque, 1495-005 Lisbon, Portugal Department of Radiology, Hospital D. Estefânia, Rua Jacinta Marto, 1169-045 Lisbon, Portugal Service de Médecine Nucléaire, Route de Lennik 808, 1070 Brussels, Belgium d Service de Radiologie – Imagerie par Resonance Magnétique, Route de Lennik 808, 1070 Brussels, Belgium b c
a r t i c l e
i n f o
Article history: Received 19 April 2011 Received in revised form 23 November 2011 Accepted 24 November 2011 Keywords: Pancreatic cancer Chronic pancreatitis MRI FDG-PET/CT
a b s t r a c t Purpose: To retrospectively determine the diagnostic accuracy of magnetic resonance imaging (MRI) and combined positron emission tomography/computed tomography (PET/CT) in the differential diagnosis of benign and malignant pancreatic lesions. Materials and methods: Twenty-seven patients (15 women/12 men, mean age 56.5 years) with MR imaging and PET/CT studies performed to differentiate benign and malignant pancreatic lesions were identified between October 2008 and October 2010. Both MR and PET/CT data sets were retrospectively and blindly evaluated by two independent readers (4 readers total) with different degrees of experience, using a visual five-point score system. The results were correlated with final diagnosis obtained by histopathology. Results: 17 patients had malignant diseases and 10 patients had benign diseases. Depending on the observer, the sensitivity, specificity, positive predictive value and negative predictive value of MRI varied between 88–94%, 50–80%, 75–89% and 71–89% respectively. Sensitivities, specificities, positive predictive values and negative predictive values of PET/CT were 73%, 56%, 73% and 56% respectively. The diagnostic accuracy of MR for the differential diagnosis of pancreatic lesions was 74–89%, compared with 67% for PET/CT. The weighted Cohen’s kappa coefficient was 0.47 at MR and 0.53 at PET/CT. Conclusion: MRI achieved higher sensitivity and specificity in the differential diagnosis of pancreatic lesions. © 2011 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Pancreatic cancer is usually diagnosed at advanced stages, when the disease is no longer resectable. This is related with the late onset and non-specificity of the symptoms, which determines an overall survival rate of less than 4% [1]. Despite all the imaging modalities available, detection of pancreatic cancer is still challenging and the differential diagnosis with focal pancreatitis is difficult [2,3]. To complicate this issue, chronic pancreatitis may develop into pancreatic carcinoma, and also pancreatic carcinoma may develop obstructive chronic pancreatitis secondary to pancreatic ductal obstruction [4]. This is clinically relevant because the impossibil-
∗ Corresponding author at: R. Prof. Jorge da Silva Horta, n◦ 8, B2, 2◦ FTE, 1500 Lisbon, Portugal. Tel.: +351 964765625; fax: +351 210083643. E-mail addresses: [email protected] (S. Belião), [email protected] (A. Ferreira), [email protected] (I. Vierasu), [email protected] (D. Blocklet), [email protected] (S. Goldman), [email protected] (T. Metens), [email protected] (C. Matos). 1 Tel.: +351 966518734; fax: +351 213126667. 2 Tel.: +32 02 555 47 11; fax: +32 02 555 47 01. 3 Tel.: +32 02 555 43 25; fax: +32 02 555 66 70. 0720-048X/$ – see front matter © 2011 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ejrad.2011.11.045
ity of obtaining a preoperative diagnosis is associated with adverse outcomes including reoperative pancreaticoduodenectomies and resections of benign disease [5]. The development of more sensitive and specific imaging modalities may avoid these adverse outcomes and improve the diagnostic accuracy of pancreatic lesions [6]. Positron emission tomography (PET) is a functional imaging technique that provides unique molecular and metabolic information. By using the radiotracer 18 F-FDG, PET allows direct evaluation of glucose metabolism and is a validated imaging modality for a variety of oncologic and non-oncologic conditions [7]. A working group concluded in 2008 that PET benefits the differentiation of benign from malignant lesions in the diagnostic work-up of patients with suspected pancreatic lesions. The greatest benefit was attributed to the possibility of excluding cancer without the need for biopsy or surgery [1]. Limitations of PET include unspecific tracer uptake in inflammatory changes such as acute pancreatitis and low spatial resolution however, the development of PET/CT holds great promise as an imaging technique for the detection and staging of pancreatic cancer [8,9]. MR imaging is being increasingly used as a noninvasive diagnostic modality of pancreatic diseases. Improvement in the quality of MR sequences, single shot T2-weighted and three-dimensional
2528
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532
unenhanced and contrast-enhanced T1-weighted gradient-echo sequences facilitated successful depiction of pancreatic lesions [10]. In addition, magnetic resonance colangiopancreatography (MRCP) allows the evaluation of pancreatic and bile ducts noninvasively and has been reported as accurate in the depiction of duct dilation, strictures, tumors and stones [2]. The abilities of MRI in conjunction with MRCP are useful in establishing resectability of the lesions as well as preventing unnecessary endoscopic retrograde cholangiopancreatography (ERCP) and stenting [11]. Diffusion-weighted MR imaging has been increasingly used to characterize pancreatic lesions and may depict small solid masses better than routine MR sequences due to its greater image contrast, despite its relatively poorer resolution [10]. Diffusion-weighted imaging may also be useful in the detection of metastases in the liver and lymph nodes [10]. Although MR imaging provides highresolution images with excellent anatomic delineation of lesions, differentiating benign from malignant pancreatic diseases remains challenging. In the other way, PET allows only limited anatomic localization of an abnormal uptake spot [12]. Despite the considerable attention that has been paid to study the diagnostic accuracy of these techniques, there are still few studies comparing MRI and PET/CT in the same group of patients [13]. The purpose of this study was to retrospectively assess the accuracy of MR and PET/CT imaging in differentiating benign and malignant pancreatic lesions. 2. Materials and methods 2.1. Patients Twenty-seven patients with suspected pancreatic lesions between October 2008 and October 2010 were retrospectively reviewed. Only patients with histopathologic analysis of the lesions and who had undergone both MR imaging and PET/CT examinations within a short interval (mean of 20 days) were included in this study. This group of patients ranged in age from 30 to 84 years (mean age, 56.5 years) and comprised 15 women and 12 men.
anatomical correlation were carried out by the means of a low dose CT (4.4 mGy/slice, 40 mAs/slice) obtained without injection of contrast media. 2.4. MR analysis Two radiologists, with 5 and more than 15 years of experience interpreting MR examinations independently reviewed the MR examinations without knowledge of patient’s identity, clinical information, findings of other imaging examinations and histopathologic results. Observer 1 occasionally interprets body MR examinations, and observer 2 interprets body MR examinations on a daily basis. For the assessment of the lesions, TSE T2-weighted images, MRCP images and DW images with different b values and pre and postcontrast T1-weighted gradient echo images were interpreted. Signal intensity of the lesions on T1, on T2-weighted images, and on DW imaging was qualitatively assessed without using a threshold for hyperintensity. On DW sequences images with b = 0 and 150–1000 s/mm2 were compared looking for areas that did not suppress on the 1000 s/mm2 DW images. Focal nodular or mass like areas of hyperintensity were recorded as tumor. The location and size of pancreatic lesions were reported. The evaluation also comprised the extension of the lesions to the peripancreatic fat or adjacent organs. The presence/absence of lymph nodes and their diameters were assessed. Caliber of the common bile duct and pancreatic duct was measured using MRCP images to assess the presence of dilation or stenoses. Vascular anatomy and assessment of vascular infiltration were registered. Evaluation of vascular infiltration included identification of the celiac axis, the splenic artery, the hepatic artery, the superior mesenteric artery, the portal vein and the superior mesenteric vein. Focal liver lesions were evaluated and classified as benign or metastatic. Based on these parameters, the probability of malignancy was assessed by each reader with a five-point visual scoring system: score 1 indicated no malignancy; score 2 indicated malignancy probably absent; score 3, undetermined; score 4, malignancy probably present; score 5 indicated definitely malignant.
2.2. MR imaging 2.5. PET/CT analysis The examinations were performed using 1.5 T and 3 T magnets (Philips Achieva, Best, the Netherlands) equipped with sense compatible abdominal coils. Sequences included single shot respiratory triggered turbo spin echo (TSE) T2-weighted in the axial and coronal orientations with TE = 80 ms, axial diffusion-weighted (DW) including at least b = 0, b = 150 and b = 1000 s/mm2 values with diffusion sensitization in 3 orthogonal directions. All images had 5 mm thickness. Breath hold MRCP 2D projections with TE = 1000 ms in several oblique orientations were also acquired. Axial 3D T1weighted fat suppressed gradient echo images with an isotropic resolution of 1.5–2 mm were obtained dynamically before and during the arterial, portal venous and equilibrium phases after injection of 0.1 mmol/kg body weight contrast material (Gadovist, Bayer-Schering® , Germany). 2.3. Whole-body 18 F-FDG PET/CT All the patients were injected intravenously with a bolus of 10 mCi (370 MBq) after a fasting period of at least 6 h. Glucose blood level had to be less than 7.5 mmol/L. 18 F-FDG PET/CT were performed on a Gemini-16 PET/CT camera (Philips Medical Systems, Cleveland, OH, USA) in full-3D acquisition mode. Static images from skull base to mid-thigh were acquired starting 80 ± 10 min after FDG injection (3 min per bed position). The images were iteratively reconstructed by means of the Philips-supplied 3D Raw Action Maximum-Likelihood Algorithm (3D-RAMLA) with scatter correction through single scatter simulation. Attenuation correction and
18 F-FDG PET/CT were interpreted by two nuclear medicine practitioners with 5 and 15 years of experience, blinded to tumor location, morphological imaging results, and clinical status of the patients. The coregistered PET and CT images were analyzed on a clinical workstation (Brightview, Philips Medical Systems, Cleveland, OH, USA). The nuclear medicine practitioners analyzed all images independently and each of their evaluation was reported (no consensus reading). Evaluation was primarily based on visual interpretation but standardized uptake value (SUV) values were also considered in the diagnostic process by comparison with values in other organs (liver, spleen). No specific threshold was used to differentiate malignant from benign lesions. The evaluation was not restricted to the upper abdomen; any abnormality detected in the whole-body examination was considered in the global classification of the patient. Criteria used to classified local abnormalities in the pancreatic area included intensity, homogeneity and extension of abnormal FDG uptake. The probability of malignancy was assessed with the same five-point visual score-system used for MR imaging.
2.6. Statistical analysis Based on the histopathologic analysis of the lesions, the accuracy, sensitivity, specificity, positive predictive value and negative predictive value were calculated for both techniques.
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532 Table 1 Distribution of pathology. Malignant disease Pancreatic adenocarcinoma Neuroendocrine tumor Ampulloma Metastasis from melanoma GIST in duodenum Benign disease Chronic pancreatitis Adenopathy Cyst Benign alteration Focal tuberculosis Tumor localization Head Body Tail
2529
diagnose malignancy). The results of the interobserver grading and corresponding weighted kappa’s are given in Fig. 1a (MRI) and b (PET/CT). N = 10 N=4 N=1 N=1 N=1
4. Discussion
N=5 N=1 N=1 N=2 N=1 N=5 N=4 N=6
Grades 1 and 2 were accepted as negative for the presence of malignancy. Grades 4 and 5 were accepted as positive for the presence of malignancy. Regarding Grade 3, it was considered malignant since the main concern of the authors was the ability of excluding malignancy by both techniques. The inter-observer agreement was assessed with the Cohen’s kappa statistics K for the agreement about the diagnosis of malignancy. Additionally, the linear weighted Cohen’s kappa statistics was performed to assess the agreement between scores. The standard error on kappa (SE) was reported as well. The kappa statistic was categorized as follows: 0–0.2, slight agreement; 0.21–0.4, fair agreement; 0.41–0.6, moderate agreement; 0.61–0.8, substantial agreement; 0.81–1, almost perfect agreement [14]. Calculations were performed using the software Medcalc (MedCalc Software, Mariakerke, Belgium). 3. Results All 27 patients were fully evaluated by the MR observers. Given that PET/CT data archived for 3 patients could not be properly retrieved in order to obtain adequate images reconstructions, only 24 patients were fully evaluated by nuclear medicine observers. Among the 17 patients with malignancies, there were 10 adenocarcinomas, 4 neuroendocrine tumors (NET), 1 ampulloma, 1 melanoma metastasis and 1 gastrointestinal stromal tumor (GIST) of the duodenum. Benign findings included chronic pancreatitis in 5 cases, other benign pancreatic lesions (2 patients), focal tuberculosis (1 patient), cystic lesion (1 patient) and adenopathy in the context of HIV (1 patient). Histopathologic results were confirmed in all patients (Table 1). It should be noted that three of the patients with proved adenocarcinoma had background clinical history of chronic pancreatitis. The most common location of pancreatic tumors was the tail of pancreas (n = 6). Other locations included the head (n = 5) and body (n = 4). The size of pancreatic tumors ranged from 0.7 to 3 cm. Open surgery was performed in 6 patients, laparoscopy was performed in 2 patients and biopsy/cytology of the lesions under endoscopic guidance was obtained in 19 patients. The results about diagnosis of malignancy are summarized in Table 2 (Grades 3–5 to
In this study, MR imaging was the most accurate, sensitive and specific method for diagnosing benign and malignant lesions. Depending on the observer, the values of sensitivity and specificity of MRI varied between 88–94% and 50–80% respectively. Overall accuracy was calculated at 74% and 89%. A prospective evaluation of patients with suspected solid pancreatic tumors evaluated with a combination of non-enhanced and contrast-enhanced MRI, MRCP, and contrast-enhanced MR angiography reported a sensitivity of 95% and a specificity of 82% for MRI in the diagnosis of malignancy with an overall accuracy of 90% [2]. The accuracy achieved in our study by the most experienced observer was similar to accuracy of the aforementioned study. Within the group of malignant lesions, three lesions were classified as benign by the MR observers. Two of these patients had background clinical history of chronic pancreatitis. As previously described, the presence of chronic pancreatitis represents an extremely difficult challenge in the differential diagnosis of pancreatic cancer. In the same way, the majority of false-positive MR studies corresponded to patients with chronic pancreatitis. Because pancreatic tumors with equivocal imaging findings are potentially resectable, the observers are more likely to make the diagnosis of cancer in the patients with focal pancreatitis. This tendency explains the relatively high number in false-positive diagnoses of cancer in this study. For both readers, the values of sensitivity and specificity of PET/CT were 73% and 56% respectively. The sensitivity and specificity achieved by PET/CT was lower than the sensibility (89%) and specificity (64–74%) reported on recent literature by Kauhanen et al. [15] and Schick et al. [16]. One possible explanation can be the fact that in those studies the observers were not blinded to clinical information or to imaging findings in previous examinations. It can also be explained by the five-visual point score system used in this study, which can be difficult to apply to PET/CT examinations. There were 4 false-negatives for both readers at PET/CT. Among them, 3 were common to the two readers. These results might be explained with the appearance and small size of the tumors in our population. Given the diffuse pancreatic uptake seen, some tumors were erroneously interpreted at PET/CT as inflammatory activity (Fig. 2). A reduced sensitivity for small pancreatic cancers has also been described for the standard radiotracer 18 F-FDG, presumably because of partial-volume effects [17]. In this study, the pancreatic cancers ranged in size from 0.7 to 3 cm. Additionally, one of the patients had a duodenal GIST treated with imatinib that showed no uptake on PET/CT. The difference in metabolic activity after even a single dose of imatinib can be shown on 18 F-FDG PET scans and is often dramatic [18]. However, it should be noted that unlike the MR results, all the patients with chronic pancreatitis complicated with pancreatic cancer were classified as malignant by the most experienced
Table 2 PET/CT and MRI diagnosis of malignancy (Grades 3–5 to diagnose malignancy). K is the Cohen’s kappa coefficient for diagnosing malignancy. SE – standard error. Accuracy (%) MR Observer 1 Observer 2 PET-CT Observer 1 Observer 2
Sensitivity (%)
Specificity (%)
Positive predictive value (%)
Negative predictive value (%)
K (SE)
74 89
88 (15/17) 94 (16/17)
50 (5/10) 80 (8/10)
75 (15/20) 89 (16/18)
71 (5/7) 89 (8/9)
0.47 (0.19)
67 67
73 (11/15) 73 (11/15)
56 (5/9) 56 (5/9)
73 (11/15) 73 (11/15)
56 (5/9) 56 (5/9)
0.64 (0.16)
2530
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532
Fig. 1. (a) Interobserver grading variability at MRI. The weighted kappa was 0.47 (standard error 0.13). (b) Interobserver grading variability at PET/CT. The weighted kappa was 0.53 (standard error 0.15).
Fig. 2. 62-year-old man with pancreatic adenocarcinoma. T2-weighted MR image shows a mass (arrow) in the head of the pancreas (a). Peripancreatic adenopathies (arrow) are also seen (b). Corresponding axial high-b value (b = 1000 s/mm2 ) diffusion-weighted image (c). Coronal MR cholangiopancreatography image demonstrates the doubleduct sign (d). Positron emission tomography axial and coronal images show intense and diffuse FDG uptake in pancreatic head (e and f). This lesion was classified as malignant by MR readers and considered indetermined by both PET/CT readers.
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532
2531
Fig. 3. 55-year-old woman with chronic pancreatitis complicated of pancreatic adenocarcinoma. A periampullary mass (arrow) is shown in axial T2-weighted and coronal MRCP images (a and b). MRCP also demonstrates dilation of the pancreatic duct with stigmas of chronic pancreatitis. The common bile duct is not dilated (b). PET axial and coronal images show a focal FDG uptake in pancreatic head region (c and d). MR readers classified the imaging findings as benign and the most experienced PET/CT reader considered this lesion malignant.
PET/CT observer (Fig. 3). These results are consistent with previous studies, which have shown that in this group of patients PET-scans are usually positive [19]. Regarding the 5 cases with chronic pancreatitis, one the PET/CT observers classified 3 lesions as benign and one lesion as malignant with score 3; the other observer classified 3 lesions as benign and one lesion as malignant with score 5. One of the cases was not assessed due to technical problems. Therefore, the majority of false-positive cases registered at PET/CT were not related with patients with chronic pancreatitis. Falsepositive PET-scans have been noted in the presence of acute and chronic inflammatory reactions, including granulomatous disease, osteomyelitis and abdominal abscesses [5]. Three of the patients with false-positive PET/CT studies in this series had peri-pancreatic adenopathies (one patient with HIV, one patient with focal tuberculosis and one patient with resected benign disease of the common bile duct). All readers (MR and PET/CT studies) misinterpreted focal tuberculosis as malignant. In fact, the radiologic findings of focal tuberculosis are not specific and usually mimic pancreatic malignancy. The interobserver agreement calculated with kappa statistics for both MR and PET-CT readers was moderate (weighted kappa of 0.47 at MRI and 0.53 at PET/CT). This aspect can be due to the great variability of cases included in our series and to the different
experience degrees between observers. Fig. 1a and b shows a higher agreement at MRI in the attribution of scores 4 and 5, however score 3 is less reproducible. This explains the best kappa value obtained for diagnosis of malignancy based on scores 3–5 at PET/CT. This study was limited by its retrospective design and by the low number of patients. A small sample size is due to the relatively low number of patients with histopathologic confirmation who had undergone MRI and PET/CT within a short interval of time. A case selection bias must also be considered. At our institution, the majority of patients submitted to both techniques are those where the differential diagnosis between pancreatic cancer and chronic pancreatic is difficult. Apparent diffusion coefficient (ADC) values were not used to assess the probability of malignancy because of the reported overlap between malignant and benign lesions [10]. In the same way, no specific SUV threshold was used to differentiate malignant from benign lesions because a previous study has shown that malignant and inflammatory lesions present with similar SUV values and are better differentiated based on the FDG uptake distribution [20]. The results from this study indicate that MR is able to depict and characterize pancreatic lesions, especially if performed by experienced readers. The major drawbacks were noted in patients with chronic pancreatitis and in patients with chronic pancreatitis complicated of pancreatic cancer. Once our results suggest a better
2532
S. Belião et al. / European Journal of Radiology 81 (2012) 2527–2532
performance of PET/CT in this specific group of patients, the combination of both MR and PET/CT may yield the highest overall value for the diagnosis of pancreatic lesions. In fact, negative predictive value for malignancy seems higher at MRI while PET-CT could help in identifying chronic pancreatitis. In our study, information provided by DW-imaging was especially helpful for the detection of the lesions and less helpful for their characterization. Like conventional MR imaging, it is difficult to differentiate pancreatic adenocarcinoma from mass-forming pancreatitis at DW-imaging [10]. Furthermore, DW images can be fused with conventional MR images, like the fusion images obtained with PET/CT scanners, to achieve better anatomic resolution. In our study, MR readers did not have access to this kind of fusion images, but some studies have reported their advantages [6]. 5. Conclusion The results of this study show that MRI has higher performance than PET/CT for the characterization of pancreatic lesions. Nevertheless, both techniques presented diagnostic limitations when compared with previous studies. Further studies in larger clinical settings that compare diagnostic performance of each technique with the accuracy of fully integrated PET/MRI systems are needed to support our findings. References [1] Fletcher J, Djulbegovic B, Soares H, et al. Recommendations on the use of 18 FFDG PET in oncology. J Nucl Med 2008;49(3):480–508. [2] Hanninen EH, Amthauer H, Hosten N, et al. Prospective evaluation of pancreatic tumors: accuracy of MR imaging with MR cholangiopancreatography and MR angiography. Radiology 2002;224:34–41. [3] 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;10:85–8. [4] Tajima Y, Kuroki T, Tsutsumi R, Isomoto I, Uetani M, Kanematsu T. Pancreatic carcinoma coexisting with chronic pancreatitis versus tumor-forming pancreatitis: diagnostic utility of the time-signal intensity curve from dynamic contrast-enhanced MR imaging. World J Gastroenterol 2007;13(6): 858–65.
[5] Rose DM, Delbeke D, Beauchamp MDRD, et al. 18 Fluorodeoxyglucose-positron emission tomography in the management of patients with suspected pancreatic cancer. Ann Surg 1999;229(5):729. [6] Ichikawa T, Erturk SM, Motosugi U, et al. High-b value diffusion-weighted MRI for detecting pancreatic adenocarcinoma: preliminary results. AJR 2007;188:409–14. [7] Kwee T, Takahara T, Ochiai R, et al. Complementary roles of whole-body diffusion-weighted MRI and 18 F-FDG PET: the state of the art and potential applications. J Nucl Med 2010;51:1549–58. [8] Sahani DV, Zarine KS, Catalano OA, Boland GW, Brugge WR. Radiology of pancreatic adenocarcinoma: current status of imaging. J Gastroenterol Hepatol 2008;23:23–33. [9] Kalra M, Maher M, Boland G, Saini S, Fischman A. Correlation of positron emission tomography and CT in evaluating pancreatic tumors: technical and clinical implications. AJR 2003;181:387–93. [10] Wang Y, Miller F, Chen Z, et al. Diffusion-weighted MR imaging of solid and cystic lesions of the pancreas. Radiographics 2011;31:E47–65. [11] Kalra M, Maher M, Mueller P, et al. State of the art imaging of pancreatic neoplasms. Br J Radiol 2003;76:857–65. [12] Balci NV, Semelka RC. Radiologic diagnosis and staging of pancreatic ductal adenocarcinoma. Eur J Radiol 2001;38:105–12. [13] Bipat S, Phoa S, Delden O, et al. Ultrasonography, computed tomography and magnetic resonance imaging for diagnosing and determining resectability of pancreatic adenocarcinoma, a meta-analysis. J Comput Assist Tomogr 2005;29:438–45. [14] Svanholm H, Starklint H, Gundersen HJ, Fabricius J, Barlebo H, Olsen S. Reproducibility of histomorphologic diagnoses with special reference to the kappa statistic. AP-MIS 1989;97:689–98. [15] Kauhanen S, Komar G, Seppänen M, et al. A Prospective diagnostic accuracy study of 18 F-fluorodeoxyglucose positron emissiontomography/computed tomography multidetector rowcomputed tomography, and magnetic resonance imaging in primary diagnosis and staging of pancreatic cancer. Ann Surg 2009;250:957–63. [16] Schick V, Franzius C, Beyna T, et al. Diagnostic impact of 18 F-FDG PET-CT evaluating solid pancreatic lesions versus endosonography, endoscopic retrograde cholangio-pancreatography with intraductal ultrasonography and abdominal ultrasound. Eur J Nucl Med Mol Imaging 2008;35:1775–85. [17] Herrmann K, Eckel F, Schmidt S, et al. In vivo characterization of proliferation for discriminating cancer from pancreatic pseudotumors. J Nucl Med 2008;49(9):1437–44. [18] Holdsworth CH, Badawi RD, Manola JB, et al. CT and PET: early prognostic indicators of response to imatinib mesylate in patients with gastrointestinal stromal tumor. AJR 2007;189:W324–30. [19] Kouwen M, Jansen J, Goor H, Castro S, Oyen W, Drenth J. FDG-PET is able to detect pancreatic carcinoma in chronic pancreatitis. Eur J Nucl Med Mol Imaging 2005;32(4):399–404. [20] Lee TY, Kim MH, Park do H, et al. Utility of 18 F-FDG PET/CT for differentiation of autoimmune pancreatitis with atypical pancreatic imaging findings from pancreatic cancer. Am J Roentgenol 2009;193:343–8.