in vivo 19: 983-988 (2005)
Diagnostic Accuracy of MRI for Preoperative Staging
of Pancreatic Carcinoma: Tendency for Understaging
T.A. BLEY1, M. UHL1, P. SIMON2, J. MAYERLE2, N.A. GHANEM1,
B. GEML1, U. SAUERESSIG1 and M. LANGER1
1Departmentof Diagnostic Radiology, University-Hospital, Freiburg;
2Department of Gastroenterology, Endocrinology and Nutrition, Ernst-Moritz-Arndt-Universitt, Greifswald, Germany
Abstract. Objective: To evaluate the diagnostic accuracy of
magnetic resonance imaging (MRI) for preoperative staging in
pancreatic carcinoma. Materials and Methods: MRI
investigations, including MR-angio and MR-cholangio-
pancreatography (MRCP) of 19 patients who underwent surgery
for pancreatic carcinoma were retrospectively evaluated by two
radiologists. The size, localization of the tumor and possible
infiltration of neighboring organs, as well as the presence of
enlarged lymph nodes, were determined to define a preoperative,
radiological TN stage. Lymph node metastasis was defined as
peripancreatic lymphoma greater than 10 mm. Our findings were
correlated to postoperative diagnosis. Results: The T-stage was
correctly evaluated in 52.6% of the cases (10/19). Understaging
took place in 31.6% (6/19) and overstaging in 15.8% (3/19). In
three cases of understaging, a micro-infiltration of the
peripancreatic tissue was not visible in MRI. Pathologically
enlarged lymph nodes were correctly found in 63.2% of the cases
(12/19). Overstaging took place in 21.1% of the cases (4/19) and
understaging in 15.8% (3/19). Conclusion: MRI for preoperative
staging of pancreatic carcinoma showed a tendency to understage
tumor size in this study population. Especially in cases of small
tumor size, micro-infiltration of peripancreatic tissue or the
common bile duct may not be detected by MRI. Concerning
N-stage, the 95% confidence interval reveals a distribution of
over- and understaged.
Pancreatic cancer is a disease with poor prognosis, being the
fourth most common cause of cancer-related deaths in
Western countries (1). Ninety percent of all pancreatic
cancers are histologically ductal adenocarcinomas that are
Correspondence to: Thorsten A. Bley, MD, Department of
Diagnostic Radiology, University Hospital Freiburg, Hugstetter
Strae 55, 79106 Freiburg, Germany. Tel: +49 761 270 3802, Fax:
+49 761 270 3842, e-mail: bley@mrs1.ukl.uni-freiburg.de
Key Words: Pancreatic carcinoma, diagnostic preoperative staging,
MRI, CT.
0258-851X/2005 $2.00+.40
associated with a poor median survival of less than 6 months
in patients with unresectable tumors (2). The main reason
for the poor prognosis of pancreatic cancer is its tendency
to form early metastases before clinical symptoms arise and
before the tumor is detectable by diagnostic imaging
techniques. Therefore, detection of pancreatic cancer in an
early stage with small tumor size and absence of metastases
is needed for potential cure with surgery (3). However, in
80% of patients with pancreatic cancer, only palliative
treatment options are available (4). In order to determine
surgical respectability and, therefore, the patients
prognosis, a correct staging of the tumor including possible
infiltration of neighboring organs and the presence of
metastases is crucial. A combination of magnetic resonance
imaging, MR-angiography and three-dimensional MR-
cholangiopancreatography (MRCP) gives information about
diagnosis, tumor stage and assessment of resectability in a
single MR investigation of less than 45 min acquisition time
(5). Recently, many studies have been published which
compared the diagnostic value of CT and MRI in detecting
pancreatic carcinoma. These demonstrated the good
diagnostic sensitivity and specificity of both methods (6-8).
However, some authors claim that MRI-technology will
replace most other staging methods because of the
possibility to avoid endoscopy, vascular cannulation, allergic
reactions and X-radiation (9).
The aim of this study was to evaluate the diagnostic
accuracy of MRI for preoperative staging in histologically
proven pancreatic carcinoma. The focus was set on pancreatic
carcinoma of an early and surgically treatable stage.
Materials and Methods
Patients. The study population consisted of 19 patients (eight
female, eleven male; age range = 47-76 years, mean age = 62.8
years) with histologically proven and surgically treatable pancreatic
carcinoma. Patients with an advanced, initially non-resectable
tumor stage were not included in the study. All patients underwent
preoperative staging investigations by MRI. All patients were
preoperatively evaluated as surgically treatable and underwent
983
 in vivo 19: 983-988 (2005)
resection of the pancreatic carcinoma within 6-42 days (mean = 16
days) after the MRI examination.
Magnetic resonance imaging. MRI was performed using a 1.5T
superconducting magnetic resonance unit (Magnetom Vision or
Symphony, Siemens AG, Erlangen, Germany). A standard phased-
array coil with four elements (Siemens Medical Systems) was used.
The following sequences were acquired: Axial T2-weighted turbo-
spin-echo (TSE) in breath hold technique (TE 138, TR 2900, slice
thickness 6 mm, one acquisition, field of view (FOV) 215x350,
matrix 116x256, TA 0:15 min); axial T1-weighted spoiled gradient
echo 2D (TE 4.1, TR 94, slice thickness 6 mm, one acquisition,
FOV 215x350, matrix 116x256, TA 0:15 min) before and after
application of gadopentetate dimeglumine intravenously (i.v.) in a
dose of 0.1 mmol/kg (either Magnevist, Schering or Omniscan,
Nycomed); coronal T2-weighted TSE (TE 125, TR 4275, slice
thickness 6 mm, one acquisition, FOV 215x350, matrix 116x256, TA
0:15 min); coronal T1-weighted TSE (TE 4.1, TR 86, slice thickness
6 mm, one acquisition, FOV 215x350, matrix 116x256, TA 0:15 min)
before and after application of i.v. contrast agent.
MRCP images were acquired with half-Fourier single-shot TSE
sequences (flip angle of 180, matrix size of 256 x 240, acquisition
time of 2 seconds per section) (10). The total examination time
(patient-in-room-time) was approximately 35-40 minutes.
Image analysis. Two radiologists, with thorough experience in
diagnostic radiology (mean 7.5 years), retrospectively evaluated the
MRI scans in a consensus methodology. Both readers had no
previous knowledge of the cases, and neither knew the results of
other studies such as ERCP and angiography. Histological findings
were used as gold standard for an intra-individual comparison with
the radiological findings concerning tumor size and occurrence of
lymph node metastases.
According to the TNM Classification of the International
Union Against Cancer (UICC), T- (tumor) and N-stages (lymph
node) were determined. Therefore, the localization and size of
the tumor were evaluated as well as the presence of infiltration
of neighboring organs. A tumor of less than 2 cm restricted to
the pancreas revealed stage T1, a tumor greater than 2 cm
restricted to the pancreas revealed stage T2. If the tumor directly
infiltrated the duodenum, the common bile duct or
peripancreatic fatty tissue, a stage T3 was found. Infiltration of
the stomach, colon and/or mesenteric vessels led to stage T4
(Figure 1). Tumor size was measured using axial or coronal
T1-w imaging.
In order to determine the N-stage, peripancreatic lymph nodes
of 10 mm or greater were evaluated as pathologically enlarged. If
no regional lymph node was enlarged stage N0 was diagnosed. A
single enlarged regional lymph node revealed stage N1a, multiple
enlarged lymph nodes led to stage N1b.
Integrity and possible congestion of the pancreatic duct and
common bile duct were evaluated using MRCP imaging.
Results
Histological findings. According to the pathological findings
84.2% of the pancreatic tumors were adenocarcinomas
(16/19), 10.5% were adenosquamous carcinomas (2/19), and
in 1 case (5,3%) a papilla carcinoma was diagnosed; 15.8%
984
Table I. Histological findings concerning tumor entity and tumor grading.
Cases
Adenocarcinoma 16/19 84.2%
Adenosquamous carcinoma 2/19 10.5%
Papilla carcinoma 1/19 5.3%
Highly-differentiated G1 3/19 15.8%
Intermediately-differentiated G2 11/19 57.9%
Poorly-differentiated G3 5/19 26.3
of the tumors (3/19) were highly-differentiated, resulting in
G1 grading; 57.9% (11/19) were intermediately-differentiated
G2 and 26.3% (5/19) were poorly-differentiated leading to
G3 grading (Table I).
In 89.5% (17/19) of the cases, the tumor was located in
the head of the pancreas, while in 10.5% (2/19) it was
located in the body. No pancreatic mass occurred in the tail
of the pancreas in the study population.
Radiological findings. The average tumor size was 3.4x2.9 cm
(range from 0.5 cm to 5.4 cm maximal diameter) according
to the T1-w axial or coronal MR images. In 84.2% of the
cases (16/19) the pancreatic edges were diffuse. Necrosis
was seen in 26.3% (5/19). In the majority (15/19, 78.9%) the
tumor signal was hypointense in T1-w images. Only in
21.1% (4/19) of the cases was the tumor signal isointense to
pancreatic tissue. In T2-w images the tumor signal was in
almost the same amount of cases hypointense as it was
isointense to pancreatic tissue (10/19, 52.6% hypointense;
9/19, 47.4% isointense), respectively.
In 84.2% (16/19) of the patients, the tumor tissue showed
a moderate to strong enhancement of the contrast medium
Gd-DTPA. The same number of tumors (16/19) showed a
heterogenous arterial perfusion of the pancreas and an
absent typical lobulation of the organ.
According to the UICC TNM staging, T- and N-staging
of the investigated tumors were evaluated as shown in Table
IIa. Concerning the staging results of each patient
individually, we found a tendency for understaging as shown
in Table IIb,c. Overstaging took place in 3/19, understaging
in 6/19 patients. Among those understaged, 3 were severely
understaged: MRI revealed stage T1, but histology proved
stage T3 (11).
Discussion
The poor prognosis of pancreatic cancer may be improved
by earlier detection of the disease. The aim of this study was
to evaluate the accuracy of MRI for preoperative staging of
pancreatic carcinoma with a focus on surgically resectable
carcinomas.
 Bley et al: MRI for Preoperative Staging of Pancreatic Carcinoma

Figure 1. Contrast-enhanced axial (right) and coronal (left) T1-weighed fat-saturated MRI shows pancreatic mass of more than 2 cm (arrows), tumor
invasion in the peripancreatic fatty tissue (arrowhead) and the congested common bile duct (bold arrow) are clearly visible (UICC T4-stage). Histology
revealed a pT4-Stage.

Figure 2. Contrast-enhanced axial (top row) and coronal (bottom row) T1-weighed fat-saturated MRI shows pancreatic mass of less than 2 cm (arrows),
peripancreatic tumor invasion is not detectable (radiological IUCC T1-stage). Histology revealed micro-focal infiltration of peripancreatic fatty tissue
leading to a pT3-Stage. Please note the congested pancreatic duct (arrowhead).

985
 in vivo 19: 983-988 (2005)

Table IIa. Comparison of radiological and pathological findings fatty tissue was not detectable in MRI. These tumors were
concerning T-stage and N-stage in the total of 19 cases. staged as T1 in MRI, but histology revealed pT3 (Figure 2).
Nishiharu et al. prospectively evaluated MRI examinations
MRI Histology
of patients suspected of having pancreatic cancer. As a
T-stage T1 4/19 pT1 0/19 secondary study objective, they evaluated the detection of
T2 1/19 pT2 3/19 peripancreatic tumor invasion. Concerning this matter,
T3 12/19 pT3 14/19 their two observers had a sensitivity of 75% and 87% and
T4 2/19 pT4 2/19
specificity of 63% and 64%, respectively (12). Since our
N-stage N0 9/19 pN0 9/19 study was a retrospective evaluation, we did not calculate
N1a 0/19 pN1a 1/19 sensitivity and specificity. The 3 severely understaged cases,
N1b 10/19 pN1b 9/19 however, indicate that we encountered the same problems
leading to low specificity: in cases of small tumor size, MRI
may not reveal any peripancreatic tissue alterations at all. If
Table IIb. Comparison of radiological and pathological findings irregular signal intensity of the peripancreatic tissue is
concerning T-stage case by case. T-stage was correctly evaluated in 10/19 found, it may be difficult or even impossible to differentiate
cases. Overstaging took place in 3/19 cases, understaging in 6/19 cases. peripancreatic tumor infiltration from desmoplastic
Please note 3 cases of severe understaging marked with bold print: MRI reaction. Furthermore, chemical shift artefacts between
showed T1 stage, histology revealed pT3.
pancreas parenchyma and peripancreatic fatty tissue as well
T-stage pT1 pT2 pT3 pT4 as motion-, pulsation- and susceptibility artefacts may
impair the image quality of MRI.
T1 - 1 3 - No case of undetectable micro-infiltration of the common
T2 - 1 - - bile duct was encountered in our study population.
T3 - 1 9 2
Nevertheless, this may be another reason for falsely staging
T4 - - 2 -
a pT3 tumor as T1-stage in MRI. More severe understaging
would take place if micro-infiltration of the portal vein or
the superior mesenteric vessels was overseen. Infiltration of
Table IIc. Comparison of radiological and pathological findings the major vessels would lead to a T4 stage. If surgery is
concerning N-stage case by case. N-stage was correctly evaluated in 12/19
performed by experienced surgeons, infiltration of the
cases. Almost equal variance to overstaging (4/19 cases) and understaging
(3/19 cases). portal or superior mesenteric vein is not a criterion for non-
resectability (13), nevertheless, it appears to mean a poorer
N-stage pN0 pN1a pN1b prognosis for the patients outcome (14). Therefore, the
question of infiltration of the major mesenteric vessels and
N0 6 1 2
the portal vein is most crucial for the surgeon.
N1a - 1 -
N1b 3 1 5
References
1 Greenlee RT, Murray T, Golden S and Wingo PA: Cancer
statistics. CA Cancer J Clin 50: 7-33, 2000.
2 Real FX: The cell biology of pancreatic cancer: an overview. In:
The results show a tendency for understaging concerning Neoptolemos JP, Lemoine NR (eds.). Pancreatic Cancer,
T-stage. In 3 cases, severe radiological understaging took Molecular and Clinical Advances. Oxford: Blackwell Science,
place compared to the histological gold standard: in all 3 pp. 3-19, 1996.
3 Warshaw AL and Fernandez-Del Castillo C: Pancreatic
cases a pT3-stage carcinoma was determined to be a
carcinoma. N Engl J Med 326: 455-465, 1992.
T1-stage tumor by MRI findings. A closer look at the TNM 4 Sternheim ET, Voigt J, Kaspar W and Dippold WG: Das
staging of the IUCC criteria may explain this severe Pankreaskarzinom. Internist 41: 848-859, 2000.
discrepancy: a carcinoma less than 2 cm within the 5 Catalano C, Pavone P, Laghi A, Panebianco V, Scipioni A,
pancreatic head is considered stage T1. If infiltration of the Fanelli F, Brillo R and Passariello R: Pancreatic
common bile duct or peripancreatic tissue is found, the adenocarcinoma: combination of MR imaging, MR angiography
tumor is determined as T3 stage, regardless of size. Tiny and MR cholangiopancreatography for the diagnosis and
assessment of resectability. Eur Radiol 8: 428-434, 1998.
micro-infiltration may not be detectable in MRI and may
6 Megibow AJ, Zhou XH, Rotterdam H, Francis IR, Zerhouni
not result in changes of the MRCP appearance of the EA, Balfe DM, Weinreb JC, Aisen A, Kuhlmann J, Heiken JP
common bile duct. In such cases understaging would occur. et al: Pancreatic adenocarcinoma: CT versus MR imaging in the
In 3 cases of our study population, the tumor was less evaluation of resectability: report of the radiology diagnostic
than 2 cm, and a micro-infiltration of the peripancreatic oncology group. Radiology 195: 327-332, 1995.

986
 Bley et al: MRI for Preoperative Staging of Pancreatic Carcinoma
7 Semelka RC, Kelekis NL, Molina PL, Sharp TJ and Calvo B:
Pancreatic masses with inconclusive findings on spiral CT: is
there a role for MRI? J Magn Reson Imaging 6(4): 585-588,
1996.
8 Ichikawa T, Haradome H, Hachiya J, Nitatori T, Ohtomo K,
Kinoshita T and Araki T: Pancreatic ductal adenocarcinoma:
preoperative assessment with helical CT versus dynamic MR
imaging. Radiology 202: 655-662, 1997.
9 Trede M, Rumstadt B, Wendl K, Gaa J, Tesdal K, Lehmann
KJ, Meier-Willersen HJ, Pescatore P and Schmoll J: Ultrafast
magnetic resonance imaging improves the staging of pancreatic
tumors. Ann Surg 226: 393-405, 1997.
10 Laubenberger J, Bchert M, Schneider B, Blum U, Hennig J
and Langer M: Breathhold MR-cholangio-pancreatography
(MRCP): a new method for examination of the bile and
pancreatic ducts. Magn Res Med 33: 18-23, 1995.
11 Bley TA, Uhl M, Geml B, Ghanem NA, Saueressig U, Springer
OS and Langer M: Diagnostic of MRI for preoperative staging
of pancreatic carcinoma: tendency of understaging. ECR
Abstract B-0039, 2003.
12 Nishiharu T, Yamashita Y, Abe Y, Mitsuzaki K, Tsuchigame T,
Nakayama Y and Takahashi M: Local extension of pancreatic
carcinoma: assessment with thin-section helical CT versus with
breath-hold fast MR imaging ROC analysis. Radiology 212:
445-452, 1999.
13 Baer HU, Wagner M, Sadowski C and Buchler MW: Pancreatic
carcinoma: surgical therapy. Ther Umsch 53(5): 394-400, 1996.
14 Wren SM, Ralls PW, Stain SC, Kasiraman A, Carpenter CL
and Parekh D: Assessment of resectability of pancreatic head
and periampullary tumors by color flow Doppler sonography.
Arch Surg 131(8): 812-817, discussion 817-818, 1996.
Received June 9, 2005
Accepted July 7, 2005
987