Note: This copy is for your personal non-commercial use only.
To order presentation-ready
copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.
Intraductal Papillary Mucinous
Neoplasms of the Pancreas:
Evaluation of Malignant Potential and
Surgical Resectability by Using MR
Imaging with MR Cholangiography1
Seong Ho Kim, MD
Jeong Min Lee, MD
Eun Sun Lee, MD
Jee Hyun Baek, MD
Jung Hoon Kim, MD
Joon Koo Han, MD
Byung Ihn Choi, MD
1
 From the Department of Radiology (S.H.K., J.M.L., J.H.B.,
J.H.K., J.K.H., B.I.C.) and Institute of Radiation Medicine
(J.M.L., J.H.K., J.K.H., B.I.C.), Seoul National University
College of Medicine, Seoul National University Hospital, 101
Daehak-ro, Jongno-gu, Seoul 110-744, Republic of Korea;
and Department of Radiology, Research Institute and Hos-
pital of National Cancer Center, Goyang, Republic of Korea
(E.S.L.). Received December 29, 2013; revision requested
February 3, 2014; revision received April 13; accepted June
16; final version accepted August 12. Address correspon-
dence to J.M.L. (e-mail: jmsh@snu.ac.kr).
q
 RSNA, 2014
Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org
Original Research  n  Gastrointestinal
Imaging
Purpose: To evaluate the diagnostic performance of magnetic reso-
nance (MR) imaging with MR cholangiopancreatography
(MRCP) in determining the malignant potential and sur-
gical resectability of pancreas intraductal papillary mucin-
ous neoplasms (IPMNs).
Materials and Institutional review board approval was obtained, and the
Methods: requirement for informed consent was waived. Ninety-
eight patients with pathologically proved pancreas IPMNs
who underwent MR imaging with MRCP comprised the
study population. MR images were analyzed for findings
suggestive of high-risk stigmata or worrisome features,
as proposed by the international consensus guidelines
2012. Interobserver agreement between two experienced
observers (observers 1 and 2) and one inexperienced ob-
server (observer 3) was assessed. Diagnostic performance
of MR imaging in the evaluation of the malignant potential
and surgical resectability of IPMNs was analyzed in these
three observers by using receiver operating curve analysis.
Results: MR imaging with MRCP showed sensitivity of 83%
(35/42), 79% (33/42), and 90% (38/42); specificity of
80% (41/51), 51% (26/51), and 24% (12/51); and accu-
racy of 82% (76/93), 63% (59/93), and 54% (50/93) for
observers 1, 2, and 3, respectively, in the evaluation of the
malignant potential of pancreas IPMNs when at least one
worrisome feature was present. Interobserver agreement
in the detection of intramural nodules (k = 0.349–0.574),
enhanced solid components (k = 0.318–0.574), and mea-
surement of main pancreatic duct diameter (intraclass
correlation coefficient = 0.9477) was fair to high. The
respective sensitivity, specificity, and accuracy in deter-
mination of surgical resectability were 95% (81/85), 99%
(84/85), and 88% (75/85); 69% (9/13), 69% (9/13), and
54% (7/13); and 92% (90/98), 95% (93/98), and 84%
(82/98) for observers 1, 2, and 3.
Conclusion: MR imaging with MRCP is a useful modality in the evalua-
tion of the malignant potential and resectability of IPMNs,
with high sensitivity and moderate specificity in the ex-
perienced radiologists but relatively low specificity in the
inexperienced radiology trainee.
q
 RSNA, 2014
Online supplemental material is available for this article.
723
GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al
I
ntraductal papillary mucinous neo-
plasm (IPMN) of the pancreas is de-
fined as a tumor growing in the main
duct or branch duct of the pancreas,
Advances in Knowledge
nn MR imaging with MR cholangio-
pancreatography (MRCP) showed
sensitivity of 83% (35/42), 79%
(33/42), and 90% (38/42); speci-
ficity of 80% (41/51), 51%
(26/51), and 24% (12/51); and
accuracy of 82% (76/93), 63%
(59/93), and 54% (50/93) in two
experienced observers and one
inexperienced observer, respec-
tively, who were evaluating the
malignant potential of pancreatic
intraductal papillary mucinous
neoplasms (IPMNs) when at
least one worrisome feature was
present.
nn Interobserver agreement in the
detection of intramural nodules
(k = 0.349–0.574) and enhanced
solid components infiltrating the
parenchyma (k = 0.318–0.574)
and for the measurement of
main pancreatic duct diameter
(intraclass correlation coefficient
= 0.9477) and pancreatic cyst
size (intraclass correlation coeffi-
cient = 0.6982) was fair to high,
whereas interobserver agreement
in the detection of enhanced cyst
walls (k = 0.066–0.211) was rela-
tively limited among the two
experienced observers and the
one inexperienced observer.
nn Agreement for detection of
abrupt caliber changes in the
main pancreatic duct (k = 0.294–
0.612) was variable between the
three observers.
nn Overall accuracy of MR imaging
with MRCP in determining sur-
gical resectability was 92%
(90/98), 95% (93/98), and 84%
(82/98) in two experienced ob-
servers and one inexperienced
observer, respectively, while
showing a tendency toward un-
derestimation of vascular
encasement.
724 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015
with differentiated papillary features
and production of atypical mucin, as
well as segmental or diffuse dilation of
the main pancreatic duct (MPD), cys-
tic dilation of the secondary branches,
or both (1). It is clinically subcatego-
rized into three types: main-duct IPMN,
branch-duct IPMN, and mixed IPMN,
and the malignant potential is known
to be higher in main-duct IPMNs and
mixed IPMNs (2,3). In recent years,
the incidence and number of surgical
resections for pancreatic IPMNs has
increased significantly (4,5). However,
this increase has come without a cor-
responding increase in IPMN-related or
overall pancreatic cancer–related mor-
tality. Thus, it is likely to have resulted
from an increase in diagnostic scru-
tiny, particularly the expansive use of
high-resolution cross-sectional imaging
techniques, such as multidetector row
computed tomography (CT) or magnetic
resonance (MR) imaging, rather than
from a greater incidence of patients with
IPMNs (4). According to the 2010 World
Health Organization classification, pan-
creatic IPMNs can be subcategorized
into IPMNs with low- or moderate-grade
dysplasia, which are considered benign,
and IPMNs with high-grade dysplasia
or those associated with invasive can-
cers, which are considered malignant
(3,6–8). Previous studies have described
the imaging criteria in the differentia-
tion of malignant IPMNs from benign
IPMNs (7,9–18), and partly on the ba-
sis of those studies, the international
consensus guidelines 2012 for the man-
agement of IPMNs and mucinous cystic
neoplasms of the pancreas have been
issued (7,9–18). The new international
consensus guidelines recommend mul-
tidetector row CT or MR imaging with
Implication for Patient Care
nn MR imaging with MRCP is a
useful modality in the evaluation
of malignant potential and re-
sectability of IPMNs, with high
sensitivity and moderate speci-
ficity in the experienced radiolo-
gists, albeit with relatively low
specificity in the inexperienced
radiology trainee.
MR cholangiopancreatography (CP) in
the evaluation of pancreatic cysts larger
than 1 cm to check for high-risk stig-
mata or worrisome features. In cysts
that show high-risk stigmata, surgical
management is recommended. In cysts
that show worrisome features or that
are larger than 3 cm without worrisome
features, endoscopic ultrasonography
(US) is recommended (7). If no worri-
some features are present, no further
diagnostic work-up is recommended, al-
though surveillance is still required.
Even though CT is still a mainstay
in the evaluation of the surgical re-
sectability of pancreatic IPMNs with
a strong preference among pancreatic
surgeons, MR imaging with MRCP has
come into wide acceptance in recent
days (7,14,19–27).
Thus, the aim of our study was to
evaluate the diagnostic performance
of MR imaging with MRCP in deter-
mining the malignant potential and
surgical resectability of pancreatic
IPMNs by using pathologic and surgi-
cal analyses as reference standards.
In addition, we attempted to evaluate
the interobserver agreement in apply-
ing the diagnostic criteria proposed
by the international consensus guide-
lines 2012 for the management of ma-
lignant IPMNs by using MR imaging
with MRCP in a larger sample size.
Published online before print
10.1148/radiol.14132960  Content code:
Radiology 2015; 274:723–733
Abbreviations:
CI = confidence interval
ICC = intraclass correlation coefficient
IPMN = intraductal papillary mucinous neoplasm
MPD = main pancreatic duct
MRCP = MR cholangiopancreatography
3D = three-dimensional
Author contributions:
Guarantors of integrity of entire study, S.H.K., J.M.L.;
study concepts/study design or data acquisition or data
analysis/interpretation, all authors; manuscript drafting or
manuscript revision for important intellectual content, all
authors; approval of final version of submitted manuscript,
all authors; literature research, S.H.K., J.M.L., E.S.L., J.H.K.;
clinical studies, S.H.K., J.M.L.; statistical analysis, S.H.K.;
and manuscript editing, all authors
Conflicts of interest are listed at the end of this article.
GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al

Materials and Methods Figure 1

Patients
Institutional review board approval was
obtained for this retrospective study,
and the informed consent requirement
was waived. Between May 2009 and
February 2013, 616 patients underwent
MR examinations for suspected IPMNs
of the pancreas either at our institution
or at an outside hospital. Among them,
123 patients underwent either surgery
or US-guided biopsy at our institution
after diagnosis of pancreatic IPMNs
with a suspicion of at least borderline
malignancy. We included (a) patients
with pancreatic IPMNs who had under-
gone curative or palliative treatment at
our hospital, (b) patients who had un-
dergone a preoperative or preprocedure
three-dimensional (3D) MRCP examina-
tion and a contrast material–enhanced
3D dynamic MR imaging examination
within 3 months prior to surgery or bi-
opsy, and (c) patients who had a diag-
nosis of pancreatic IPMN at pathologic
examination. A total of 493 patients who
did not undergo surgery or biopsy were
excluded from the study. Thereafter, 25
patients were further excluded because Figure 1:  Flow diagram of enrolled patients. ∗Only patients who underwent
of the absence of contrast-enhanced surgical resection were enrolled in the evaluation of presumed malignant poten-
multiphasic MR imaging (n = 11) or sub- tial. ∗∗Five patients were included in addition to the 93 patients who underwent
diagnostic quality MR images or MRCP surgical resection to evaluate surgical resectability at MR imaging with MRCP.
images due to severe motion artifacts
from limited breath-holding capability
or recent interventional procedures (n had undergone only biopsy and who had MR Technique
= 14). Finally, our retrospective analysis been excluded from evaluation of tumor Because of the retrospective nature of
included 98 patients (mean age, 67.8 staging, were included. However, for this study, various MR imagers were
years; age range, 40–85 years), 93 of evaluation of the diagnostic performance used. MR imaging was performed with
whom had undergone surgical resection of MR imaging in determining the malig- either a 1.5-T MR unit (Signa Excite
and five of whom had received only non- nant potential of pancreatic IPMNs, only HDXT, GE Medical Systems, Milwau-
surgical palliative treatment because of 93 patients who had undergone surgical kee, Wis [n = 18]; Achieva, Philips
the advanced stage of malignant IPMNs resection were included, as the pathol- Healthcare, Best, the Netherlands [n =
at preoperative MR examination, as de- ogy reports of the five patients who un- 2]) with an eight-channel phased-array
termined by a consensus of two expe- derwent only biopsy were not sufficiently torso coil or a 3-T MR unit (Magne-
rienced abdominal radiologists (K.J.H., adequate to determine the malignancy tom Verio or Trio, Siemens Medical
L.J.M.) with more than 20 years of of pancreatic IPMNs (Fig 1). Among Solutions, Erlangen, Germany [n =
experience, but had pathologically con- the 93 patients who underwent surgical 74]; Ingenia or Achieva, Philips Health-
firmed pancreatic IPMNs (Fig 1). resection for pancreatic IPMNs, 51 had care [n = 4]) with a 32- or 12-channel
Our study used two different sam- IPMNs with low- or moderate-grade dys- phased-array torso coil. All MR im-
ple sizes for the respective evaluation of plasia, 17 had IPMNs with high-grade ages were obtained in the axial plane
the surgical resectability of pancreatic dysplasia, and 25 had IPMNs associated with a rectangular field of view of 380
IPMNs and the malignant potential of with an invasive cancer. Mean interval 3 380 mm that was adjusted for each
IPMNs. For resectability assessment, all between MR examinations and surgery patient. Baseline MR imaging examina-
98 patients, including five patients who was 25.8 days (range, 1–90 days). tions included breath-hold transverse

Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 725

GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al

T2-weighted imaging with a single-

shot fast spin-echo sequence or a half-

320 3 224
Fourier acquisition single-shot turbo

300–350
Imager 2

4.6/2.2
Dynamic Image
spin-echo sequence with or without

4.8
12
1
fat saturation, transverse T1-weighted
imaging with in-phase and opposed-

384 3 278
300–380
Imager 1
phase spoiled 3D gradient-echo se-

3.4/1.2
quences, and an unenhanced fat-sup-

11
3
1
pressed 3D gradient-echo sequence

320 3 320
(VIBE, Siemens Medical Solutions).

3750/820

256–320
Imager 2
Diffusion-weighted images were ob-

3D MRCP
tained with both 1.5- and 3-T MR units.

90
2
1
The 3D MRCP data were then trans-

384 3 366
2320/815
ferred to a dedicated 3D workstation

Imager 1

0.9
(Advanced Workstation, GE Medical

380
69
130
Systems) and reconstructed by using
a maximum intensity projection algo-

320 3 256
4000/1202
Thick-Slab Single-Section

Imager 2
rithm to obtain oblique images rotat-

260
1
90
50
ing about the z-axis in 10° increments.
Dynamic MR images were obtained be- MRCP

320 3 256
130 or 180
fore and after administration of 1.0-M
2500/909

50 or 60
Imager 1
gadobutrol (7.5 mL of Gadovist; Bayer

240
256
Healthcare, Berlin, Germany) at a
dose of 0.1 mmol per kilogram of body

144 3 144
3275/81.9

weight and an injection rate of 2 mL/

Imager 2
Diffusion-weighted

Note.—Imager 1 was a Magnetom Verio unit (Siemens Medical Solutions). Imager 2 was a Signa Excite HDXT unit (GE Medical Systems).
sec. Images were acquired before con-
380
1
90
7
Imaging

trast material injection (precontrast)

and 8 seconds (arterial phase), 60 sec- 256 3 205
Imager 1

4500/52

onds (portal venous phase), and 2, 3,

1
180
7
380

and 5 minutes after contrast material

arrival at the distal thoracic aorta. The
320 3 192

timing of arterial phase imaging was de-

300–350
T2-weighted Imaging
Imager 2

905/159

termined by using the MR fluoroscopic

90
7
1

technique, which allowed real-time

visualization of the heart and aorta
384 3 307
300–380

during repetitive measurements at the

Imager 1

800/93

same coronal position as the T1-weight-

7
256
130

ed gradient-echo sequence. Postcon-

trast subtracted images were obtained
6.6/2.1, 6.6/4.4*

in all patients who underwent MR ex-

320 3 224

* Data are for opposed-phase and in-phase sequences, respectively.

300–350

aminations at our institution. All pa-

T1-weighted Imaging
Imager 2

rameters for the MR sequences of the

4.8
15
1

two imagers are summarized in Table 1.

The acquisition of 3D gradient-echo
4.0/1.3, 4.0/2.3*

data in each phase was finished during

320 3 285
300–380

one breath hold at the end of expira-

Imager 1

tion (time range, 16–22 seconds; mean

1
9
3

time, 18 seconds) (Table 1).

msec/echo time msec

Image Analysis
Flip angle (degrees)
MR Parameters

Field of view (mm)

Evaluation of the malignant potential of

Echo train length

Thickness (mm)
Repetition time

pancreas IPMNs.—Morphologic features

of pancreatic IPMNs and parenchymal
Parameter
Table 1

Matrix

changes on MR images were indepen-

dently evaluated by two experienced at-
tending radiologists with a subspecialty

726 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015

GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al

Table 2
List of Criteria in the International Consensus Guideline 2012 and Scoring of Malignant Potential in Pancreatic IPMNs
Classification Findings Score

High-risk stigmata Enhanced solid components infiltrating pancreas parenchyma 5, presence of at least one finding of high-risk stigmata
  and MPD size 10 mm
Worrisome features MPD size of 5–9 mm, presence of intramural nodule, acute 4, presence of two or more worrisome features; 3, presence of only one
 caliber change in the MPD, enhanced cyst walls, cystic mass  worrisome feature; 2, no worrisome feature with an unclear duct type;
size  30 mm, lymphadenopathy 1, no worrisome feature and a definite branch duct type

in abdominal imaging (L.E.S., B.J.H.;
7 and 10 years of experience, respec-
tively) and one inexperienced radiology
trainee without any subspecialty (H.E.J.,
4 years of experience in radiology) on
a picture archiving and communication
system workstation monitor (m-view;
Marotech, Seoul, Korea). They were
blinded to the patients’ histologic diag-
nosis, as well as to any clinical or labo-
ratory information; however, they were
aware that the study population had
pancreatic IPMNs. They were requested
to evaluate findings of high-risk stigmata
and worrisome features on MRCP im-
ages (Table 2). Thereafter, to assess
interobserver agreement among two
experienced observers and one inexpe-
rienced observer, imaging findings were
evaluated by using the criteria in Appen-
dix E1 (online). In addition, consensus
for individual MR findings was obtained
to determine the most common individ-
ual findings that may suggest malignant
pancreatic IPMNs.
The two experienced radiologists
and one inexperienced radiology trainee
were then asked to determine the ma-
lignant potential and surgical resect-
ability of pancreatic IPMNs. The pre-
sumed malignant potential of pancreatic
IPMNs was evaluated with a five-point
scale. First, they scored their diagnostic
confidence in differentiating malignant
IPMNs (IPMN with high-grade dysplasia
and IPMN with an invasive cancer) from
benign IPMNs (low-grade or moderate
dysplasia) as follows: a score of 1 indi-
cated a definitely benign IPMN; a score
of 2, a probably benign IPMN; a score
of 3, a possibly malignant IPMN; a score
of 4, a probably malignant IPMN; and a
score of 5, a definitely malignant IPMN.
MR criteria for presuming the malignant
potential of pancreatic IPMNs were
based on those recommended by the
international consensus guidelines 2012
(7). The method of scoring is provided
in Table 2. Interobserver agreement in
determining the malignant potential of
pancreatic IPMNs was also obtained.
Evaluation of the surgical resect-
ability of pancreatic IPMNs.—Surgical
resectability of pancreatic IPMNs was
evaluated by using the criteria used in
previous studies of pancreatic adeno-
carcinoma or IPMN, including vascular
encasement, regional or distant lymph
node metastases, and metastases to
other solid organs (7,15,28–32). Unre-
sectable disease was defined as clearly
identifiable distant metastases; overt
lymphadenopathy at portocaval, retro-
caval, paraaortic, and mesentery areas
with the shortest diameter greater than
10 mm; tumor encasement (.180° of
circumferential involvement) of arteries,
such as the superior mesenteric artery,
celiac trunk, common hepatic artery, or
origin or long segment of the portal vein
(.3 cm in length) or superior mesen-
teric vein; or complete occlusion of the
portal vein (15,28). Thereafter, the two
experienced readers and one inexpe-
rienced reader rated the possibility of
resectability on a five-point scale: 1 for
definitely unresectable, 2 for probably
unresectable, 3 for possibly unresect-
able or undetermined, 4 for probably re-
sectable, and 5 for definitely resectable.
The observers were aware that a rating
of 1–3 indicated an IPMN was unresect-
able and that a rating of 4–5 indicated
an IPMN was resectable.
Surgical Resection
One of two experienced pancreatobili-
ary surgeons (K.S.H. and J.J.Y., both
with more than 20 years of experience)
performed a number of different oper-
ations in 93 (95%) of 98 patients for
pancreatic IPMNs according to the lo-
cation of the tumor. These procedures
included subtotal pancreatectomy (n =
26), pancreas head resection (n = 1),
total pancreatectomy (n = 6), Whipple
operation (n = 8), pylorus-preserving
pancreaticoduodenectomy (n = 51), and
choledochojejunostomy with excisional
biopsy (n = 1) (Appendix E2, Table E1
[online]).
Histopathologic Analysis
All resected IPMNs were reviewed by
an experienced pathologist (L.K.B.)
with more than 10 years of experience
in the evaluation of hepatopancreato-
biliary diseases. The pathologist de-
scribed the following findings in the
pathologic report: type of IPMN, loca-
tion of tumor, size of tumor, presence
of mural nodule, diameter of MPD,
peripancreatic infiltration, vascular
involvement, lymph node metastasis,
perineural invasion, and involvement
of adjacent solid organs. R0, R1, and
R2 were recorded on either pathology
or surgery reports with the following
definitions: R0, no evidence of resid-
ual tumor at the resection margin both
microscopically and macroscopically;
R1, direct tumor growth (high-grade
dysplasia or invasive carcinoma) up
to or within 1 mm of the margin; and
R2, macroscopic residual tumor tis-
sue at the resection margin (33). In
this study, R1 and R2 resection cases
were classified as unresectable cases,
and R0 resection cases were thought
to be resectable. The incidence of un-
resectable cases is listed in Appendix
E3 (online).

Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 727

GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al

Table 3
Frequency of MR Findings in Benign versus Malignant IPMNs with Statistical Significance and Their Interobserver Agreement on the
Basis of the International Consensus Guidelines 2012
Weighted k† or ICC Value
Benign IPMNs Malignant IPMNs
MR Finding (n = 51) (n = 42) P Value* Observers 1 and 2 Observers 2 and 3 Observers 1 and 3

Type of IPMN
  Main duct type 2 (4) 4 (10) … … …
  Branch duct type 37 (73) 13 (31) … … …
  Mixed type 12 (24) 25 (60) ,.001 k = 0.394 k = 0.421 k = 0.594
High-risk stigmata
  Enhanced solid component infiltrating 1 (2) 14 (33) ,.001 k = 0.318 k = 0.398 k = 0.574
 parenchyma
  MPD diameter 10 mm 3 (6) 15 (36) ,.001 … … …
Worrisome features
  Intramural nodule 4 (8) 21 (50) ,.001 k = 0.574 k = 0.449 k = 0.349
  Abrupt caliber change in the MPD 4 (8) 12 (29) .012 k = 0.398 k = 0.294 k = 0.612
  Enhanced cyst wall 0 (0) 4 (10) .038 k = 0.195 k = 0.211 k = 0.066
  MPD diameter 5–9 mm 18 (35) 17 (40) ,.001 … … …
  Cystic mass 3 cm 32 (63) 29 (69) NS … … …
Measurement of MPD or tumor‡
  Mean diameter of MPD (mm) 4.4 6 2.8 8.9 6 5.9 ,.001 ICC = 0.9477
  Mean size of cystic mass (mm) 32.5 6 11.7 37.0 6 16.2 NS ICC = 0.6982

Note.—Unless otherwise indicated, data are for all three observers in consensus. Data in parentheses are percentages. NS = not significant
* P , .05 indicates a significant difference.
†
Weighted k value of less than 0.20 indicates slight agreement; k value of 0.20–0.39, fair agreement; k value of 0.40–0.59, moderate agreement; k value of 0.60–0.79, substantial agreement; and
k value of 0.80 or greater, outstanding agreement. ICC value of 0.75 or higher indicates high agreement.
‡
Data are mean 6 standard deviation.

Statistical Analysis
Preoperative or preprocedural MR
evaluation of imaging findings, ma-
lignant potential, and overall tumor
resectability were compared with the
surgical findings and the pathology
reports. Statistical analyses were per-
formed with commercially available
software (MedCalc for Windows, ver-
sion 12.7.0; MedCalc Software, Mar-
iakerke, Belgium). The prevalence
of each stigmata or worrisome MR
finding in benign pancreatic IPMNs
and malignant pancreatic IPMNs was
analyzed by using the Fisher exact
test. Interobserver agreement was
evaluated with the weighted k sta-
tistic for noncontinuous scales and
with the intraclass correlation coef-
ficient (ICC) for continuous scales.
Weighted k values of less than 0.20
indicated slight agreement; weighted
k values of 0.20–0.39, fair agreement;
weighted k values of 0.40–0.59, mod-
erate agreement; weighted k values
of 0.60–0.79, substantial agreement,
and weighted k values of more than
0.80, outstanding agreement. ICCs of
more than 0.75 were considered to be
in high agreement. Diagnostic perfor-
mance of MR imaging with MRCP in
the determination of malignant poten-
tial and its resectability was analyzed
by using receiver operating character-
istic curve analysis. Sensitivity, spec-
ificity, accuracy, and interobserver
agreement were determined, and P ,
.05 was considered to indicate a signif-
icant difference.
Results
Interpreting Individual MR Imaging
Findings and Their Interobserver Variation
Among the MR imaging findings of high-
risk stigmata and worrisome features,
the prevalence of all imaging findings
except cyst size 3 cm in diameter or
larger (mean size of the cystic mass) and
MPD diameter of 5–9 mm were higher
in patients with malignant IPMNs than
in those with benign IPMNs (Table 3).
In addition, intramural nodule was the
most common finding among the imag-
ing criteria observed in malignant pan-
creatic IPMNs, while enhanced solid
component was the second most com-
mon finding according to the consensus
reading of the three reviewers (Table
3). Most of the high-risk stigmata and
worrisome features proposed by the in-
ternational consensus guidelines 2012
were shown to have fair to moderate or
high interobserver agreement, except
for the enhanced cyst wall (Table 3).
The three observers showed high inter-
observer agreement (ICC = 0.9477) for
measurement of MPD diameter and fair
to moderate interobserver agreement
for measurement of pancreatic cyst size
(ICC = 0.6982), classification of IPMN
type (k = 0.394–0.594), detection of
enhanced solid components infiltrat-
ing the parenchyma (k = 0.318–0.574),

728 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015

GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al

Figure 2

Figure 2:  Intraductal papillary mucinous neoplasm associated with an invasive carcinoma in an 85-year-old woman who underwent a 3-T MR examination shows
both a high-risk stigmata and a worrisome feature of MPD 10 mm or larger in diameter and intramural nodules (score 5). (a) Axial T2-weighted half-Fourier acqui-
sition single-shot turbo spin-echo MR image shows marked dilatation of the MPD larger than 10 mm in diameter and multiple intraluminal nodules showing lower
signal intensity than that of the pancreatic duct fluid (arrow). (b) A 3D MRCP image obtained with T2-weighted half-Fourier acquisition single-shot turbo spin-echo
sequence shows marked dilatation of the MPD with multiple filling defects (arrows), which is suggestive of multiple mural nodules. (c) Delayed contrast-enhanced 3D
axial gradient-echo MR image shows enhancement of multiple mural nodules (arrows) filling in the dilated MPD. Both reviewers determined the mass did not abut any
major vessel. The mass was confirmed as an invasive carcinoma without vascular encasement.

Figure 3

Figure 3:  Images in a 70-year-old man who was confirmed to have pancreatic IPMN of high-grade dysplasia at 3-T MR imaging with three worrisome features
of a cystic mass at least 3 cm in diameter, an enhanced cystic wall, and an MPD diameter of 5–9 mm on MR images (score 4). (a) Axial T2-weighted half-Fourier
acquisition single-shot turbo spin-echo MR image shows a lobulated cystic mass (arrow) larger than 3 cm in diameter in the pancreas uncinate process. (b) Delayed
contrast-enhanced 3D axial gradient-echo MR image shows a cystic mass with thickened enhanced cyst walls (arrow) at the uncinate process abutting and com-
pressing the major portal vein but without vascular encasement. (c) A 3D MRCP image shows an irregularly shaped cystic mass (arrow) at the uncinate process with
associated diffuse upstream pancreatic duct dilatation (5–9 mm in diameter). Both reviewers interpreted the mass as separate from major vessels and assigned a
diagnosis of resectable malignant IPMN (score 4). The mass was confirmed as a pancreatic IPMN with high-grade dysplasia negative for vascular encasement.

and identification of intramural nodules
(k = 0.349–0.574). However, the three
observers showed slight interobserver
agreement in the detection of an en-
hanced cyst wall (k = 0.066–0.211).
Diagnostic Performance of MR Imaging
with MRCP in Detecting Malignant Potential
When the presence of at least one
worrisome feature or any high-risk
stigmata (score 3) was regarded as
indicating the malignant potential of
IPMNs or specifically high-grade dys-
plasia and invasive carcinoma, the
sensitivity of MR imaging with MRCP
was 83% (35/42), 79% (33/42),
and 90% (38/42) for observers 1, 2,
and 3, respectively; specificity was
80% (41/51), 51% (26/51), and 22%
(12/51), respectively; and accuracy
was 82% (76/93), 63% (59/93), and
54% (50/93), respectively (Figs 2, 3);
Fig E1 [online]). The area under the
receiver operating characteristic curve
of MR imaging with MRCP was 0.843
(95% confidence interval [CI]: 0.753,
0.910) for observer 1, 0.778 (95% CI:
0.680, 0.857) for observer 2, and 0.823
(95% CI: 0.730, 0.894) for observer 3 in
determining the malignant potential of

Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 729

GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al
Table 4
Diagnostic Performance of MR Imaging with MRCP in the Evaluation of the Malignant
Potential of Pancreas IPMNs
Score and Statistic Observer 1 (%)
Score 5 only
 Sensitivity 33 (14/42)
 Specificity 98 (50/51)
 Accuracy 69 (64/93)
Score 4
 Sensitivity 62 (26/42)
 Specificity 96 (49/51)
 Accuracy 81 (75/93)
Score 3
 Sensitivity 83 (35/42)
 Specificity 80 (41/51)
 Accuracy 82 (76/93)
Note.—Numbers in parentheses are raw data. Worrisome features and high-risk stigmata are based on the international
consensus guidelines 2012.
pancreatic IPMNs. However, when only
the presence of any high-risk stigmata
regardless of worrisome features (score
5) was regarded as a criterion indicat-
ing the malignant potential of pancre-
atic IPMNs, the sensitivity of MR imag-
ing with MRCP was 33% (14/42), 26%
(11/42), and 57% (24/42) for observers
1, 2, and 3, respectively; specificity was
98% (50/51), 100% (51/51), and 98%
(50/51), respectively; and accuracy was
69% (64/93), 67% (62/93), and 80%
(74/93), respectively (Table 4). The
two experienced observers showed
substantial interobserver agreement
(k = 0.652), while agreement between
the experienced observers and the in-
experienced observer was moderate
(k = 0.460 and 0.475) in scoring the
malignant potential when the suggested
criteria of the international consensus
guidelines 2012 were applied.
Diagnostic Performance of MR Imaging
with MRCP in Evaluating Surgical
Resectability
The sensitivity of MR imaging with
MRCP in determining surgical resect-
ability was 95% (81/85), 99% (84/85),
and 88% (75/85), respectively, for ob-
servers 1, 2, and 3 (Table 5). However,
specificity was 69% (nine of 13) for ob-
servers 1 and 2 but 54% (seven of 13)
forn observer 3, showing a tendency
730 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015
Observer 2 (%) Observer 3 (%)
26 (11/42) 57 (24/42)
100 (51/51) 98 (50/51)
67 (62/93) 80 (74/93)
62 (26/42) 83 (35/42)
96 (49/51) 65 (33/51)
81 (75/93) 73 (68/93)
79 (33/42) 90 (38/42)
51 (26/51) 24 (12/51)
63 (59/93) 54 (50/93)
toward underestimation of vascular en-
casement (Fig 4). Overall accuracy of
observers 1, 2, and 3 was 92% (90/98),
95% (93/98), and 84% (82/98), respec-
tively. The area under the receiver op-
erating characteristic curve was 0.859
(95% CI: 0.728, 0.990) in observer 1,
0.831 (95% CI: 0.688, 0.974) in ob-
server 2, and 0.751 (95% CI: 0.654,
0.833) in observer 3. The two expe-
rienced readers showed outstanding
interobserver agreement in the evalua-
tion of surgical resectability (k = 0.870),
while the inexperienced reader and the
experienced readers showed substan-
tial interobserver agreement in the
evaluation of surgical resectability (k =
0.706 and 0.739).
Both experienced observers cor-
rectly evaluated unresectability in nine
(69%) of 13 unresectable cases, while
both experienced observers underesti-
mated four (31%) of 13 truly unresect-
able cases. Of these, three were under-
estimation of portal vein invasion, and
one was underestimation of superior
mesenteric artery invasion. Conversely,
observer 1 correctly evaluated resect-
ability in 81 (95%) of 85 truly resect-
able cases, and observer 2 correctly
evaluated resectability in 84 (99%) of
85 truly resectable cases. The inexpe-
rienced observer correctly evaluated
unresectability in seven (54%) of 13
unresectable cases, underestimated six
(46%) of 13 truly unresectable cases,
and correctly evaluated resectability in
75 (88%) of 85 truly resectable cases.
The positive predictive value of MR im-
aging with MRCP for observers 1, 2,
and 3 in determining whether IPMNs
were unresectable was 69% (nine of
13), 90% (nine of 10), and 39% (seven
of 18), respectively (Table 5).
Discussion
The recently introduced international
consensus guidelines 2012 for the
management of IPMNs and mucin-
ous cystic neoplasms of the pancreas
gave us useful recommendations for
the management of IPMNs. However,
questions remained as to the diagnos-
tic performance of MR imaging with
MRCP in determining the malignant
potential and surgical resectability of
pancreas IPMNs with the new criteria.
Our study results showed that MR im-
aging with MRCP showed acceptable
diagnostic performance in the pre-
diction of the malignant potential of
IPMNs when using the image criteria
of the international consensus guide-
lines 2012. Specifically, we observed
that the diagnostic accuracy of MR
imaging with MRCP ranged from 63%
(59/93) to 82% (76/93) in experienced
observers and was 54% (50/93) in an
inexperienced observer when we con-
sidered the presence of at least one
worrisome feature or any high-risk
stigmata (score 3) as criteria for di-
agnosis of the malignant potential of
IPMNs, with high sensitivity. Under
these combined criteria, for observers
1, 2, and 3, respectively, sensitivity was
83% (35/42), 79% (33/42), and 90%
(38/42); specificity was 80% (41/51),
51% (26/51), and 22% (12/51); and ac-
curacy was 82% (76/93), 63% (59/93),
and 54% (50/93). The low specificity
of experienced observer 2 and inexpe-
rienced observer 3 and the substantial
gap between the three observers may
be explained by the low interobserver
reliability in the evaluation of enhanced
cyst walls and abrupt caliber changes
in the MPD. The enhanced cyst wall is
difficult to discriminate on MR images,
GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al

Table 5
Surgical Resectability According to MR Imaging with MRCP Results in 98 Patients with IPMNs
R0 Resection* (n = 85) R1† or R2‡ Resection or Not Resected (n = 13)
MR Result Observer 1 Observer 2 Observer 3 Observer 1 Observer 2 Observer 3

Resectable 81 (95) 84 (99) 75 (88) 4 (31) 4 (31) 6 (46)

Not resectable 4 (5) 1 (1) 10 (12) 9 (69) 9 (69) 7 (54)

* R0 resection = no evidence of residual tumor at the resection margin both microscopically or macroscopically.
†
R1 resection = direct tumor growth (high-grade dysplasia or invasive carcinoma) up to or within 1 mm of the margin.
‡
R2 resection = macroscopic residual tumor tissue at the resection margin.

Figure 4

Figure 4:  Images in a 72-year-old woman with invasive cancer who underwent a 3-T MR examination and had an incomplete resection (R2 resection). (a) A 3D
MRCP image shows a multiseptated cystic mass in the pancreas head with upstream pancreatic duct dilatation and segmental narrowing (arrow) of the distal
common bile duct, suggesting possible common bile duct invasion. (b) Axial contrast-enhanced T1-weighted MR image shows the ill-defined enhancing solid portion
(arrow) of the mass. (c) Coronal contrast-enhanced T1-weighted MR image shows segmental narrowing of the distal common bile duct with enhancing wall thicken-
ing (arrow), suggesting distal common bile duct invasion by the tumor. The observers interpreted this case as a resectable invasive cancer (score 5); however, during
surgery, the mass was deemed to be not resectable owing to severe invasion into the main portal vein and the superior mesenteric vein.

even with the help of postcontrast
subtraction sequences. Moreover, the
evaluation of MPD strictures is rela-
tively subjective, particularly when as-
sociated with pancreatitis; we did not
count findings as strictures in such a
situation. It was even more difficult
for the inexperienced observer to dis-
criminate such findings, and he had a
tendency to overrate imaging findings
of worrisome features or high-risk stig-
mata, probably owing to a lack of expe-
rience and a sense of self-preservation.
This might have resulted in higher sen-
sitivity and lower specificity in the inex-
perienced observer. Indeed, Barron et
al (34) reported that it may not be pos-
sible to classify main duct-type IPMNs
appropriately solely on the basis of im-
aging studies. Our results showed fair
to moderate interobserver reliability
in the evaluation of duct-type IPMNs,
which also may contribute to the low
specificity in diagnosing the malignant
potential of pancreatic IPMNs. Further
refinement for enhanced cyst walls and
abrupt caliber changes of the MPD may
be required to improve specificity and
the gap between different observers.
Furthermore, we observed that spec-
ificity increased markedly (.96%)
when we applied the criteria of a score
of more than 4. In addition, among
the findings of worrisome features and
high-risk stigmata, intramural nod-
ule was the most commonly detected
finding in malignant pancreatic IPMNs,
and an enhancing solid component was
the second most common finding when
considering only the proposed findings
for high-risk stigmata and worrisome
features.
Interobserver agreement for worri-
some features or high-risk stigmata was
fair to high. Interobserver agreement
was especially high in the measurement
of MPD diameter. The two experienced
observers and the one inexperienced
observer exhibited fair to moderate
interobserver agreement in the mea-
surement of cystic mass size, detection
of enhanced solid components infiltrat-
ing the parenchyma, and the presence
of abrupt caliber change in the MPD,
whereas they showed only slight inter-
observer agreement (k = 0.066–0.211)
in the detection of the presence of en-
hanced cyst walls. In addition, moder-
ate interobserver agreement was ob-
served in the detection of the presence
of mural nodules, which was in good
agreement with the results of a previ-
ous study by Manfredi et al (14). The

Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 731

GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al
poor interobserver agreement observed
in the detection of enhanced cyst walls
in our study could be explained by the
fact that IPMNs are composed of di-
lated pancreatic ducts; thus, it may have
been difficult to differentiate enhanced
cyst walls from enhanced pancreatic
duct walls. To continue to improve in-
terobserver agreement, further clarifi-
cation and refinement of the criteria in
the international consensus guidelines
for multidetector row CT and MR im-
aging suggesting malignancy may be
necessary.
In our study, the sensitivity of the
two experienced observers was quite
high (95% [81/85] and 99% [84/85] for
observers 1 and 2, respectively) in the
evaluation of the surgical resectability of
IPMN lesions, whereas specificity was
moderate (69% [nine of 13] for both
reviewers) and overall accuracy (92%
[90/98] and 95% [93/98] for observers
1 and 2, respectively) was satisfactory.
Sensitivity (88%, [75/85]), specificity
(54% [seven of 13]), and accuracy (84%
[82/98]) were slightly lower in the inex-
perienced observer. These results are
considerably better than those reported
by Vullierme et al (15), who reported
that the overall accuracy of helical CT
in determining the surgical resectability
of malignant IPMNs with invasive cancer
was 74% but that the positive predictive
value of CT in determining unresect-
ability was only 17%, mainly owing to
overestimation of arterial invasion. Our
study showed that the positive predic-
tive value of MR imaging with MRCP
for experienced observers 1 and 2 in
determining whether IPMNs were unre-
sectable was 69% (nine of 13) and 90%
(nine of 10), respectively. One impor-
tant difference between our study and a
previous study (15), which showed that
CT usually resulted in overestimation of
tumor extent and misdiagnosis of unre-
sectability even in resectable cases, was
that our results showed that MR imaging
with MRCP tended to result in underes-
timation of tumor extent and misdiagno-
sis of unresectable cases as resectable,
even though this modality was accurate
in the detection of resectable lesions.
Our study had several limitations.
First, as the evaluation of cases was
732 radiology.rsna.org  n Radiology: Volume 274: Number 3—March 2015
retrospective, there was unavoidable
selection bias and verification bias.
However, as MR imaging with MRCP
is one of the standard preoperative di-
agnostic modalities used in the evalua-
tion of IPMNs at our institution, selec-
tion bias might not have been serious.
In addition, verification bias may have
been unavoidable, as we used a strict
reference standard to assess resectabil-
ity. Second, MR images were obtained
from multiple institutions, resulting in
use of a variety of MR machines, MR
sequences, and contrast materials.
These variations might have interfered
with the interpretation of images dur-
ing the retrospective image analyses.
Third, to decrease selection bias in the
evaluation of surgical resectability, we
enrolled patients with lesions deemed
to be unresectable by two experienced
abdominal imaging radiologists in con-
sensus. However, there is a possibility
that the good results obtained in our
study may be related to the inclusion of
relatively advanced cases in our study
population. Fourth, as multidetector
row CT is the primary modality used to
evaluate pancreativ IPMNs at our insti-
tution, many macroscopically unresect-
able cases were excluded on the basis
of CT findings, without the need for an
MR examination. Thus, only patients
with difficult cases of unresectable
IPMN that were equivocal in the deter-
mination of tumor resectability at CT
and that required further MR examina-
tion were enrolled in our study. Further
studies involving a direct comparison
between multidetector row CT and MR
imaging with MR CP in determining the
surgical resectability of IPMNs with in-
vasive cancer are warranted to better
address this limitation.
In conclusion, MR imaging with MR
CP has high sensitivity in the diagnosis
of malignant IPMNs when the presence
of at least one worrisome feature or any
high-risk stigmata (score 3) is used as
a criterion for malignant IPMN, with
high interobserver agreement. In addi-
tion, MR imaging with MRCP showed
high sensitivity and diagnostic accuracy
in the determination of surgical resect-
ability, albeit with a relatively high rate of
underestimating surgical unresectability.
Acknowledgments: We thank Eui Jin Hwang,
MD, for his assistance in reviewing MR imaging
findings. We also thank Chris Woo, BA, for his
English editorial assistance.
Disclosures of Conflicts of Interest: S.H.K.
disclosed no relevant relationships. J.M.L. Ac-
tivities related to the present article: disclosed
no relevant relationships. Activities not related
to the present article: received grants from
Bayer Healthcare, GE Healthcare, CMS, Acuzen,
and Guerbe Starmed; is on the board of Bayer
Healthcare; gave lectures for Bayer Healthcare;
is a consultant for Siemens Healthcare. Other re-
lationships: disclosed no relevant relationships.
E.S.L. disclosed no relevant relationships. J.H.B.
disclosed no relevant relationships. J.H.K. dis-
closed no relevant relationships. J.K.H. dis-
closed no relevant relationships. B.I.C. Activities
related to the present article: disclosed no rel-
evant relationships. Activities not related to the
present article: received a grant from Samsung
Electronics. Other relationships: disclosed no
relevant relationships.
References
1. Lim JH, Lee G, Oh YL. Radiologic spec-
trum of intraductal papillary mucinous
tumor of the pancreas. RadioGraphics
2001;21(2):323–337; discussion 337–340.
2. Augustin T, Vandermeer TJ. Intraductal
papillary mucinous neoplasm: a clinico-
pathologic review. Surg Clin North Am
2010;90(2):377–398.
3. Shi C, Hruban RH. Intraductal papillary
mucinous neoplasm. Hum Pathol 2012;
43(1):1–16.
4. Klibansky DA, Reid-Lombardo KM, Gordon
SR, Gardner TB. The clinical relevance of
the increasing incidence of intraductal papil-
lary mucinous neoplasm. Clin Gastroenterol
Hepatol 2012;10(5):555–558.
5. de Jong K, Bruno MJ, Fockens P. Epidemi-
ology, diagnosis, and management of cystic
lesions of the pancreas. Gastroenterol Res
Pract 2012;2012:147465.
6. Hruban RH, Takaori K, Klimstra DS, et al.
An illustrated consensus on the classification
of pancreatic intraepithelial neoplasia and
intraductal papillary mucinous neoplasms.
Am J Surg Pathol 2004;28(8):977–987.
7. Tanaka M, Fernández-del Castillo C, Ad-
say V, et al. International consensus guide-
lines 2012 for the management of IPMN
and MCN of the pancreas. Pancreatology
2012;12(3):183–197.
8. Bassi C, Sarr MG, Lillemoe KD, Reber HA.
Natural history of intraductal papillary mu-
cinous neoplasms (IPMN): current evidence
and implications for management. J Gastro-
intest Surg 2008;12(4):645–650.
GASTROINTESTINAL IMAGING: Intraductal Papillary Mucinous Neoplasms of the Pancreas Kim et al
9. Kimura W, Sasahira N, Yoshikawa T, Muto
T, Makuuchi M. Duct-ectatic type of mucin
producing tumor of the pancreas: new con-
cept of pancreatic neoplasia. Hepatogastro-
enterology 1996;43(9):692–709.
10. Ogawa H, Itoh S, Ikeda M, Suzuki K, Naga-
nawa S. Intraductal papillary mucinous neo-
plasm of the pancreas: assessment of the
likelihood of invasiveness with multisection
CT. Radiology 2008;248(3):876–886.
11. Kawamoto S, Horton KM, Lawler LP,
Hruban RH, Fishman EK. Intraductal pap-
illary mucinous neoplasm of the pancreas:
can benign lesions be differentiated from
malignant lesions with multidetector CT?
RadioGraphics 2005;25(6):1451–1468; dis-
cussion 1468–1470.
12. Taouli B, Vilgrain V, Vullierme MP, et al. In-
traductal papillary mucinous tumors of the
pancreas: helical CT with histopathologic
correlation. Radiology 2000;217(3):757–764.
13. Sugiyama M, Atomi Y. Intraductal papillary
mucinous tumors of the pancreas: imaging
studies and treatment strategies. Ann Surg
1998;228(5):685–691.
14. Manfredi R, Graziani R, Motton M, et al.
Main pancreatic duct intraductal papillary
mucinous neoplasms: accuracy of MR im-
aging in differentiation between benign and
malignant tumors compared with histo-
pathologic analysis. Radiology 2009;253(1):
106–115.
15. Vullierme MP, Giraud-Cohen M, Hammel P,
et al. Malignant intraductal papillary mucin-
ous neoplasm of the pancreas: in situ versus
invasive carcinoma surgical resectability.
Radiology 2007;245(2):483–490.
16. Uchiyama S, Chijiiwa K, Hiyoshi M, et al.
Mucin-producing bile duct tumor of the cau-
date lobe protruding into the common he-
patic duct. J Gastrointest Surg 2007;11(11):
1570–1572.
17. Sugiyama M, Izumisato Y, Abe N, Masaki
T, Mori T, Atomi Y. Predictive factors for
malignancy in intraductal papillary-muci-
Radiology: Volume 274: Number 3—March 2015  n  radiology.rsna.org 733
nous tumours of the pancreas. Br J Surg
2003;90(10):1244–1249.
18. Sugiyama M, Atomi Y, Kuroda A. Two types
of mucin-producing cystic tumors of the
pancreas: diagnosis and treatment. Surgery
1997;122(3):617–625.
19. Waters JA, Schmidt CM, Pinchot JW, et
al. CT vs MRCP: optimal classification of
IPMN type and extent. J Gastrointest Surg
2008;12(1):101–109.
20. Song SJ, Lee JM, Kim YJ, et al. Differenti-
ation of intraductal papillary mucinous neo-
plasms from other pancreatic cystic masses:
comparison of multirow-detector CT and
MR imaging using ROC analysis. J Magn Re-
son Imaging 2007;26(1):86–93.
21. Sahani DV, Kadavigere R, Blake M, Fernan-
dez-Del Castillo C, Lauwers GY, Hahn PF.
Intraductal papillary mucinous neoplasm
of pancreas: multi-detector row CT with
2D curved reformations—correlation with
MRCP. Radiology 2006;238(2):560–569.
22. Pedrosa I, Boparai D. Imaging consider-
ations in intraductal papillary mucinous
neoplasms of the pancreas. World J Gastro-
intest Surg 2010;2(10):324–330.
23. Macari M, Eubig J, Robinson E, et al.
Frequency of intraductal papillary mucin-
ous neoplasm in patients with and without
pancreas cancer. Pancreatology 2010;10(6):
734–741.
24. Kim JH, Hong SS, Kim YJ, Kim JK, Eun
HW. Intraductal papillary mucinous neo-
plasm of the pancreas: differentiate from
chronic pancreatits by MR imaging. Eur J
Radiol 2012;81(4):671–676.
25. Kang KM, Lee JM, Shin CI, et al. Added
value of diffusion-weighted imaging to MR
cholangiopancreatography with unenhanced
MR imaging for predicting malignancy or in-
vasiveness of intraductal papillary mucinous
neoplasm of the pancreas. J Magn Reson
Imaging 2013;38(3):555–563.
26. Gourgiotis S, Ridolfini MP, Germanos S. In-
traductal papillary mucinous neoplasms of
the pancreas. Eur J Surg Oncol 2007;33(6):
678–684.
27. Berland LL, Silverman SG, Gore RM, et
al. Managing incidental findings on abdom-
inal CT: white paper of the ACR inciden-
tal findings committee. J Am Coll Radiol
2010;7(10):754–773.
28. Pezzilli R, Serra C, Calculli L, Ferroni F, Iam-
marino MT, Casadei R. Three-dimensional
contrast-enhanced ultrasonography of in-
traductal papillary mucinous neoplasms of
the pancreas: a comparison with magnetic
resonance imaging. Pancreas 2013;42(7):
1164–1168.
29. Raman SP, Kawamoto S, Blackford A, et al.
Histopathologic findings of multifocal pancre-
atic intraductal papillary mucinous neoplasms
on CT. AJR Am J Roentgenol 2013;200(3):
563–569.
30. Diehl SJ, Lehmann KJ, Sadick M, Lach-
mann R, Georgi M. Pancreatic cancer: value
of dual-phase helical CT in assessing resect-
ability. Radiology 1998;206(2):373–378.
31. Hough TJ, Raptopoulos V, Siewert B, Mat-
thews JB. Teardrop superior mesenteric vein:
CT sign for unresectable carcinoma of the
pancreas. AJR Am J Roentgenol 1999;173(6):
1509–1512.
32. O’Malley ME, Boland GW, Wood BJ, Fer-
nandez-del Castillo C, Warshaw AL, Muel-
ler PR. Adenocarcinoma of the head of
the pancreas: determination of surgical
unresectability with thin-section pancreat-
ic-phase helical CT. AJR Am J Roentgenol
1999;173(6):1513–1518.
33. Verbeke CS. Resection margins in pancre-
atic cancer. Surg Clin North Am 2013;93(3):
647–662.
34. Barron MR, Roch AM, Waters JA, et al.
Does preoperative cross-sectional imaging
accurately predict main duct involvement in
intraductal papillary mucinous neoplasm? J
Gastrointest Surg 2014;18(3):447–455; dis-
cussion 5455–5456.