Gastrointestinal Imaging • Original Research

Kim et al.
CT Differentiation of Causes of Isolated MPD Dilatation

Gastrointestinal Imaging
Original Research

Isolated Main Pancreatic

Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved

Duct Dilatation: CT
Differentiation Between
Benign and Malignant Causes
Se Woo Kim1 OBJECTIVE. The purpose of this study is to retrospectively evaluate the differential CT
Se Hyung Kim1,2 features of isolated benign and malignant main pancreatic duct (MPD) dilatation and to inves-
Dong Ho Lee1,2 tigate whether the diagnostic performance of radiologists can be improved with knowledge of
Sang Min Lee1 these differential CT features.
Yeon Soo Kim1 MATERIALS AND METHODS. Forty-one patients who had isolated MPD dilatation
without any visible mass on CT from January 2000 to October 2016 were retrospectively en-
Jin Young Jang 3
rolled in the study. Two radiologists reviewed CT images in consensus for the location, shape
Joon Koo Han1,2,4 (smooth vs abrupt), length of transition, dilated pancreatic duct (PD) diameter, presence of
Kim SW, Kim SH, Lee DH, et al. duct penetrating sign, parenchymal atrophy, attenuation difference, associated pancreatitis,
calcification, PD or common bile duct (CBD) enhancement, and perilesional cyst. The chi-
square test, Fisher exact test, and t test were used to find the differential CT features of be-
nign and malignant MPD dilatation. Two successive review sessions for differentiation be-
tween the two disease entities were then independently performed by three other reviewers
with differing expertise, with the use of a 5-point confidence scale. The first session provided
no information for differentiation; however, reviewers were aware of the results of univariate
analyses in the second session. The diagnostic performance of the radiologists was evaluated
using a pairwise comparison of ROC curves.
Keywords: benign stricture, isoattenuating pancreatic RESULTS. A total of 19 benign and 22 malignant MPD dilatations were identified. In pa-
cancer, isolated main pancreatic duct dilatation, tients with benign MPD dilatation, transition areas were frequently located in the head (57.9%
pancreatic ductal adenocarcinoma
[11/19] vs 13.6% [3/22], p = 0.003) and showed significantly shorter (< 6.1 mm) (78.9% [15/19]
DOI:10.2214/AJR.17.17963 vs 9.1% [2/22], p < 0.0001) and smooth transition (89.5% [17/19] vs 9.1% [2/22], p < 0.0001).
Duct penetrating sign was exclusively observed in patients with benign MPD dilatation (73.7%
Received January 16, 2017; accepted after revision [14/19] vs 0% [0/22], p < 0.0001). In contrast, malignant MPD dilatation frequently was accom-
March 20, 2017.
panied by attenuation difference (63.6% [14/22] vs 10.5% [2/19], p = 0.001) and associated PD
Supported by the Basic Science Research Program or CBD enhancement (36.4% [8/22] vs 0% [0/19], p = 0.003). The AUC values of three review-
through the National Research Foundation of Korea ers significantly increased from 0.653, 0.587, and 0.884 to 0.864, 0.964, and 0.908, respectively,
funded by the Ministry of Science, ICT and Future with knowledge of significant CT features (p = 0.013, p < 0.0001, and p = 0.701, respectively).
Planning (2016R1A2B4007762). CONCLUSION. Distal, long (≥ 6.1 mm), and abrupt transition, the absence of duct pen-
1 etrating sign, and the presence of attenuation difference and PD or CBD enhancement were
Department of Radiology, Seoul National University
Hospital, 101 Daehangno, Jongno-gu, Seoul, 03080, highly suggestive CT findings for differentiation of malignant from benign MPD dilatation.
Korea. Address correspondence to S. H. Kim The diagnostic performance of radiologists with regard to differentiation was significantly
(shkim7071@gmail.com). improved with knowledge of these highly suggestive CT criteria.
2
Seoul National University College of Medicine,
Seoul, Korea.
ancreatic cancer is the fourth borderline resectable at the time of diagno-

P leading cause of death from can- sis [1, 2].

3
Department of Surgery, Seoul National University
Hospital, Seoul, Korea.
cer among both males and fe- CT generally plays a primary role in the
4
Institute of Radiation Medicine, Seoul National males in the United States [1]. diagnosis of pancreatic disease [3–5]. The
University Medical Research Center, Seoul, Korea. Although surgical resection is the only cura- introduction of MDCT as well as dynamic
tive treatment option for this disease, the multiphase imaging after rapid contrast ad-
AJR 2017; 209:1046–1055
5-year survival rate of patients with pancre- ministration has substantially improved the
0361–803X/17/2095–1046 atic cancer in any stage is dismal (approxi- diagnostic performance of CT for the detec-
mately 7%) because only 20% of patients tion and characterization of pancreatic le-
© American Roentgen Ray Society present with disease that is resectable or sions [5, 6]. However, several challenging

1046 AJR:209, November 2017

 CT Differentiation of Causes of Isolated MPD Dilatation

issues remain regarding the evaluation of ab-
normalities of the pancreas using CT.
Pancreatic ductal adenocarcinoma (PDAC)
typically manifests as a hypoattenuating mass
or nodule with upstream main pancreatic duct
(MPD) dilatation on contrast-enhanced dy-
Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved
cluded. A total of 19 patients with benign MPD
dilatation and 22 patients with malignant MPD
dilatation were ultimately enrolled.
MPD dilatation was defined by an MPD with a
maximal diameter greater than 3 mm or with dila-
tation more than 2 mm larger than its visible nar-
hydrate antigen [CA] 19-9 levels), the diagnostic
method used (biopsy, surgery, or clinical and imag-
ing follow-up), and the type of surgery performed.
The CEA level was evaluated in 29 patients (10
with benign and 19 with malignant MPD dilata-
tion), and the CA 19-9 level was checked in 29 pa-

namic CT [5], thereby facilitating identifica-
tion of the cause of duct dilatation. However,
it is not rare in clinical practice to encounter
an isoattenuating pancreatic adenocarcinoma
in which the tumor attenuation on CT is indis-
tinguishable from the attenuation of the adja-
cent pancreatic parenchyma, thereby manifest-
ing only as an isolated MPD dilatation [7–9].
In this situation, differentiation from benign
stricture or benign MPD dilatation may be-
come difficult. CT features of PDAC, includ-
ing isoattenuating pancreatic cancer, have
been investigated [7]; however, to our knowl-
edge, CT differentiation between benign and
malignant MPD dilatation has not yet been
studied. Therefore, the present study retrospec-
tively evaluates differential CT features of iso-
lated benign versus malignant MPD dilatation.
We also investigate whether the diagnostic
performance of radiologists can be improved
with knowledge of differential CT features.
Materials and Methods
The institutional review board at Seoul Nation-
al University Hospital approved this retrospective
study and waived the requirement for informed
consent from patients.
Study Population
Between January 2000 and March 2016, a list
of patients who had the term “pancreatic duct dil-
atation,” “p duct dilatation,” or “MPD dilatation”
appear in their CT report was collected through a
text search function in our PACS. A total of 11,170
patients were identified. The following patients
were then excluded: 9095 patients for whom the
phrase “no or without pancreatic duct dilatation,”
“no or without p duct dilatation,” or “no or without
MPD dilatation” appeared in their CT report; 1119
patients with a visible pancreatic mass at the tran-
sition site; and 289 patients with diffuse pancreat-
ic duct (PD) dilatation caused by senescent change
(n = 270) or an ampulla of Vater or duodenal prob-
lem (n = 19). Senescent change was considered to
occur when both the PD and the common bile duct
(CBD) were dilated without any discernable mass
in the ampulla of Vater or duodenum in patients
older than 60 years as well as when the dilatation
did not progress for at least 1 year. Patients with
intraductal papillary mucinous neoplasm (n =
207), chronic pancreatitis (n = 224), PD stone (n =
6), or a history of surgery (n = 189) were also ex-
rowest portion [10]. The benign MPD dilatation tients (13 with benign and 16 with malignant MPD
group consisted of 14 men and five women with a dilatation). Histologic confirmation was obtained
mean age of 66.6 years (range, 43–83 years). The for 17 of 21 patients with malignant MPD dilata-
malignant MPD dilatation group consisted of 12 tion. For the remaining 23 patients (four with ma-
men and 10 women with a mean age of 60.9 years lignant and 19 with benign MPD dilatation), clin-
(range, 38–76 years) (Table 1). For patients with ical or radiologic follow-up was used for disease
benign MPD dilatation, a diagnosis was made on confirmation (Table 1). For histologic confirmation
the basis of clinical findings when MPD dilatation in the malignant MPD dilatation group, endoscop-
was stable for 1 year or more of imaging follow- ic ultrasound-guided biopsy was used for 23.5%
up. We selected 1 year as the period for assess- of patients (4/17), whereas surgery was performed
ing benign dilatation because the median surviv- for 76.5% (13/17). Various surgical methods were
al of patients with untreated advanced pancreatic used, including pylorus-preserving pancreaticodu-
cancer is approximately 3.5 months and because, odenectomy for seven patients, distal pancreatec-
even with good treatment, survival increases only tomy for two, subtotal pancreatectomy for two, to-
to 8 months [11]. The mean (± SD) follow-up was tal pancreatectomy for one, and open laparotomy
60.9  ± 45.7 months (range, 12.6–198.6 months). with biopsy for one (Table 1). The mean (± SD) fol-
Malignant MPD dilatation was confirmed by his- low-up for both groups was also reviewed.
topathologic analysis in 17 patients and by clinical
follow-up in five patients. CT Acquisition
The retrospective design of the present study ne-
Clinical and Histologic Features cessitated the use of various CT scanners. Most pa-
One author reviewed the electronic medical re- tients (92.7% [38/41]) with benign (89.5% [17/19])
cords of all patients to determine laboratory find- and malignant (95.5% [21/22]) MPD dilatation un-
ings (carcinoembryonic antigen [CEA] and carbo- derwent MDCT performed using scanners with
TABLE 1: Summary of Clinical and Histologic Findings for Patients
With Benign (n = 19) and Malignant (n = 22) Main Pancreatic
Duct (MPD) Dilatation
Benign MPD Malignant MPD
Finding Dilatation Group Dilatation Group p
Age (y), mean ± SD 66.6 ± 9.9 60.9 ± 11.2 0.082
Sex 0.205
Male 14 12
Female 5 10
CEA level (ng/mL), mean ± SD 1.8 ± 0.8 10.8 ± 25.6 0.701
Carbohydrate antigen level 19-9 (U/mL), mean ± SD 21.6 ± 26.3 527.4 ± 1715.5 0.249
Confirmation method < 0.0001a
Histologic analysis 0 17
Follow-upb 19 5
Histologic confirmation
Endoscopic ultrasound-guided biopsy 4
Surgeryc 13
Follow-up duration (mo), mean ± SD 60.9 ± 45.7 30.3 ± 33.6 0.004a
Note—Except where otherwise indicated, data are number of patients. CEA = carcinoembryonic antigen.
aStatistically significant.
bPatients were monitored using clinical and radiologic methods.
cVarious surgical methods were used, including pylorus-preserving pancreaticoduodenectomy for seven
patients, distal pancreatectomy for two, subtotal pancreatectomy for two, total pancreatectomy for one, and
open laparotomy with biopsy for one.

AJR:209, November 2017 1047


2–320 detector rows. The remaining three pa-
tients (7.3%), two with benign and one with ma-
lignant MPD dilatation, underwent CT performed
using a single-detector CT scanner. The acquisi-
tion parameters used for MDCT were as follows:
tube voltage, 100–120 kVp; tube current, 150–250
Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved
mAs; slice thickness, 1.3–5 mm; reconstruction in-
terval, 0.6–5 mm; pitch, 0.9–1; and rotation time,
0.5–1 second. For single-detector CT, the acquisi-
tion parameters were as follows: tube voltage, 120
kVp; tube current, 150–200 mAs; pitch, 1; rotation
time, 0.9–1 second; slice thickness, 5–10 mm; and
reconstruction interval, 5–8 mm.
Iodinated contrast agent (350 or 370 mg I/mL)
was injected using an automatic power injector at
a rate of 3–5 mL/s and at a dose of 1.5–1.6 mL/kg
for 30 seconds. Saline chase was performed at
the same rate for 10 seconds. Dynamic contrast-
enhanced scans were performed for 15 patients
(78.9% [15/19]) with benign MPD dilatation (ear-
ly arterial, late arterial, and portal phases [n = 6];
late arterial, portal, and delayed phases [n = 2];
late arterial and portal phases [n = 6]; and por-
tal and delayed phases [n = 1]) and for 15 patients
(68.2% [15/22]) with malignant MPD dilatation
(early arterial, late arterial, and portal phases
[n = 7]; late arterial, portal, and delayed phases
[n = 3]; and late arterial and portal phases [n =
5]). The remaining 11 patients with benign (n =
4) and malignant (n = 7) MPD dilatation had CT
scans acquired in the portal phase only. By use
of the bolus tracking method, early and late ar-
terial phase scans were started 5–9 seconds and
18–23 seconds, respectively, after the attenuation
threshold (100 HU) was reached in the descend-
ing thoracic aorta. For portal and delayed phase
scans, a fixed delay of 60–75 seconds and 3 min-
utes, respectively, was used.
Image Analysis
CT images were analyzed in three differ-
ent reading sessions: one for consensus reading
to find significant CT features and two for inde-
pendent reviewing with and without information
on significant CT features. First, two radiologists
(one attending physician with 20 years of experi-
ence and one resident with 1 year of experience)
who were blinded to the specific diagnosis and
clinical information interpreted CT images in
consensus to determine significant CT findings
differentiating between benign and malignant
MPD dilatation. The images were presented to re-
viewers in random sequence and were reviewed on
a PACS using stack mode.
For the consensus reading session, the follow-
ing CT findings were analyzed: location of duct
transition (head, neck, body, and tail), shape of
duct transition (smoothly tapered versus abrupt),
1048 AJR:209, November 2017
Kim et al.
length of duct transition (expressed in millime-
ters) (Fig. 1), maximum diameter of dilated MPD
(expressed in millimeters), presence of parenchy-
mal atrophy, attenuation (isoattenuation, hypoat-
tenuation, or hyperattenuation) of the upstream
pancreatic parenchyma relative to normal paren-
chyma, associated pancreatitis, intraductal or pa-
renchymal calcification, intra- or extrapancreatic
cyst adjacent to duct caliber transition point, PD
or CBD enhancement at the transition point, pres-
ence of duct penetrating sign, associated CBD dil-
atation, and soft-tissue cuffing around the celiac
axis or superior mesenteric artery. For parenchy-
mal atrophy, if the width of upstream parenchyma
was smaller than that of the normal pancreas, we
determined that there was parenchymal atrophy.
In addition, if the attenuation of the upstream pa-
renchyma was different from that of the normal
pancreas, we considered the presence of a differ-
ence in attenuation. Associated pancreatitis was
considered present when peripancreatic infiltra-
tion and fluid collection were shown. PD or CBD
enhancement was considered present when ductal
wall enhancement was greater than that of adja-
cent pancreatic parenchyma in the portal venous
phase. The duct penetrating sign was defined as a
thin intact PD that was visible traversing the tran-
sition zone (Figs. 2–4). CBD dilatation was con-
sidered present when the maximum CBD diam-
eter was 8 mm or larger.
After the consensus reading, three other radi-
ologists with differing expertise (one with 1 year
of experience in abdominal imaging, one with 5
years of experience, and one with 10 years of ex-
perience) were recruited for the second and third
sessions. In the second session, they independent-
ly interpreted CT images for differentiation be-
tween benign and malignant MPD dilatation with-
out knowledge of the differential CT features.
Two weeks after the second session, they were in-
formed of significant CT findings derived from the
consensus reading session and were asked to inter-
pret the CT images again. Through these two seri-
al readings, we were able to assess whether the di-
agnostic performance of the radiologists improved
with knowledge of these significant CT findings.
The radiologists were asked to determine wheth-
er the lesion was benign or malignant MPD dila-
tation on the basis of a 5-point confidence scor-
ing system, with a score of 1 denoting definitely
benign MPD dilatation; 2, probably benign MPD
dilatation; 3, possibly malignant MPD dilatation;
4, probably malignant MPD dilatation; and 5, def-
initely malignant MPD dilatation. Lesions with
scores of 3–5 were considered to denote malignant
MPD dilatation. To minimize recall bias, two sep-
arate interpretation sessions were scheduled with
a 2-week interval and randomly presented images.
Statistical Analysis
The differences in CT features between benign
and malignant MPD dilatation were evaluated us-
ing univariate statistical tests (the chi-square test
or Fisher exact test [for categoric variables] and
the independent t test [for continuous variables]).
To find the most discriminating CT features, mul-
tivariate binary logistic regression analysis was
performed using significant CT variables ob-
tained from univariate analysis. To solve the mul-
ticollinearity issue, a stepwise variable selection
method was used for binary logistic regression
analysis. ROC analysis was also used to find the
optimized cutoff point of the length of transition
site to differentiate malignant from benign MPD
dilatation. Two-tailed p < 0.05 was considered to
indicate statistical significance, and a 95% CI was
reported for each variable.
The individual performance of the three inde-
pendent reviewers for differentiation between be-
nign and malignant MPD was evaluated using ROC
analysis and the McNemar test. A pairwise com-
parison of ROC curves or McNemar test was done
between the two successive independent review
sessions (the second and third sessions) to assess
the improvement in radiologists’ diagnostic perfor-
mance once they had knowledge of significant CT
findings. A value of p < 0.05 was considered to de-
note a statistically significant difference. Statistical
analysis was performed using statistical software
(SPSS, version 22.0, SPSS; and MedCalc for Win-
dows, version 8.0.0.1, MedCalc Software).
Results
Clinical and Histologic Findings
Table 1 presents the clinical and histo-
logic findings. Age (p = 0.082) and sex (p =
0.205) were not significantly different be-
tween the two groups. In the malignant MPD
dilatation group, the CEA level (p = 0.701)
and CA19-9 level (p = 0.249) tended to be
higher than in the benign MPD dilatation
group; however, the differences were not sta-
tistically significant. The mean follow-up for
the benign group (60.9  ± 45.7 months) was
significantly longer than that for the malig-
nant group (30.3 ± 33.6 months) (p = 0.004).
Analysis of CT Findings
Table 2 summarizes the analyzed CT find-
ings of benign and malignant MPD dilata-
tion. In terms of the transition site, it was fre-
quently located at the head (57.9% [11/19]) in
the benign group, whereas it was located any-
where throughout the entire pancreas in the
malignant group (p = 0.003). In benign MPD
dilatation, the transition site usually showed
smoothly tapered narrowing (89.5% [17/19]),
 CT Differentiation of Causes of Isolated MPD Dilatation

TABLE 2: Results of Univariate Statistical Analysis Comparing Benign and whereas in the malignant MPD dilatation
Malignant Main Pancreatic Duct (MPD) Dilatation group, it showed abrupt narrowing (90.9%
Benign MPD Malignant MPD
[20/22]) (p < 0.0001). The mean length of the
Finding Dilatation Group Dilatation Group pa transition site was significantly longer in the
malignant group (13.8 ± 7.7 mm) than the be-
Age (y), mean ± SD 66.6 ± 9.9 60.9 ± 11.2 0.093
nign group (4.2 ± 4.1 mm). For the length of
Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved

Sex, no. of patients 0.205 the transition site, the optimized cutoff value
Male 14 12 to differentiate malignant from benign MPD
Female 5 10 dilatation was 6.1 mm (AUC value, 0.884;
sensitivity, 90.9% [20/22]; specificity, 78.9%
Maximum diameter of dilated PD (mm), mean ± SD 7.2 ± 3.0 6.6 ± 2.8 0.498
[15/19]; p  < 0.0001). PD or CBD enhance-
Transition site ment at the transition site was found exclu-
Location 0.003a sively in the malignant MPD dilatation group
Head 11 (57.9) 3 (13.6) (0% [0/19] vs 36.4% [8/22], p = 0.003).
Regarding upstream pancreatic paren-
Nonhead 8 (42.1) 19 (86.4)
chyma, a difference in attenuation was more
Shape < 0.0001a frequently found in the malignant MPD dil-
Smooth 17 (89.5) 2 (9.1) atation group (63.6% [14/22]) than in the be-
Abrupt 2 (10.5) 20 (90.9) nign MPD dilatation group (10.5% [2/17])
(p = 0.001). Of the 14 patients with malig-
Length (mm), mean ± SD 4.2 ± 4.1 13.8 ± 7.7 < 0.0001a
nant MPD dilatation who showed an attenu-
PD or common bile duct enhancement 0.003a ation difference, the upstream pancreatic pa-
Absent 19 (100) 14 (63.6) renchyma appeared to have lower attenuation
Present 0 (0) 8 (36.4) than normal downstream pancreatic paren-
chyma in 12 patients and higher attenuation
Perilesional cyst 0.058
in two patients. In the two patients with be-
Absent 17 (89.5) 14 (63.6) nign MPD dilatation who showed an attenu-
Present 2 (10.5) 8 (36.4) ation difference, the upstream pancreatic pa-
Upstream pancreatic parenchyma renchyma appeared to have lower attenuation
Parenchymal atrophy 0.138
than the normal downstream pancreatic pa-
renchyma in one patient and higher attenua-
Absent 16 (84.2) 14 (63.6) tion in the other patient.
Present 3 (15.8) 8 (36.4) Duct penetrating sign was observed exclu-
Attenuation difference 0.001a sively in patients with benign MPD dilatation
Absent 17 (89.5) 8 (36.4)
(73.7% [14/19] vs 0% [0/22]), with statistical
significance noted (p  < 0.0001). Similarly,
Present 2 (10.5) 14 (63.6)
soft-tissue cuffing around the celiac axis or
Associated pancreatitis 0.537 superior mesenteric artery was exclusive-
Absent 19 (100) 21 (95.5) ly found in the malignant MPD dilatation
Present 0 (0) 1 (4.5) group (18.2%, 4/22); however, statistical sig-
nificance was not achieved (p = 0.072). Fig-
Common bile duct dilatation 0.400
ures 2–6 show representative examples.
Absent 14 (73.7) 18 (81.8) Multivariate analysis using a binary logis-
Present 5 (26.3) 4 (18.2) tic regression test revealed that abrupt nar-
Duct penetrating sign < 0.0001a rowing at the transition site was the only
significant CT variable for the differentiation
Absent 5 (26.3) 22 (100)
of malignant MPD dilatation from benign
Present 14 (73.7) 0 (0) MPD dilatation (odds ratio [OR], 85.0; 95%
Intraductal or parenchymal calcification 0.556 CI, 10.8–669.5) (p < 0.0001).
Absent 18 (94.7 20 (90.9)
Diagnostic Performance of Radiologists
Present 1 (5.3) 2 (9.1)
for Differentiation
Soft-tissue cuffing around celiac axis and SMA 0.072 Table 3 presents the individual perfor-
Absent 19 (100) 18 (81.8) mance of the three radiologists for differen-
Present 0 (0) 4 (18.2) tiation between benign and malignant MPD
dilatation during the two successive inde-
Note—Except where otherwise indicated, data are number (%) of patients. PD = pancreatic duct, SMA =
superior mesenteric artery. pendent review sessions. When information
aStatistically significant. on significant CT findings was provided, ac-

AJR:209, November 2017 1049

 Kim et al.
TABLE 3: Results for Differentiation Without and With Knowledge of
Significant CT Features
Second Review Third Review
Finding and Reviewer Session Session pa
Diagnostic accuracy, % (no. of patients/total)
Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved
Reviewer 1 58.5 (24/41) 78.0 (32/41) 0.021b
Reviewer 2 65.9 (27/41) 82.9 (34/41) 0.180
Reviewer 3 68.3 (28/41) 87.8 (36/41) 0.013b
AUC value (95% CI)
Reviewer 1 0.653 (0.489–0.795) 0.864 (0.720–0.951) 0.013b
Reviewer 2 0.587 (0.423–0.739) 0.964 (0.854–0.998) < 0.0001b
Reviewer 3 0.884 (0.745–0.963) 0.908 (0.776–0.976) 0.701
Note—The second session provided no information for the differentiation; however, the third session provided
the three reviewers with information on the significant CT findings for differential diagnosis.
aObtained using a McNemar test or pairwise comparison of ROC curves.
bStatistically significant.
curacy in the diagnosis of benign and malig- tients who have CT features favoring malig-
nant MPD dilatation increased for all three nant MPD dilatation.
reviewers. The improvement was statisti- Our results regarding CT features fre-
cally significant for reviewer 1 (p = 0.021) quently found in the malignant MPD dilata-
and reviewer 3 (p = 0.013). Regardless of tion group are similar to those in previously
the expertise of the reviewers, the AUC val- published literature that reported ancillary
ues for all three reviewers also increased CT features associated with isoattenuat-
from 0.653, 0.587, and 0.884 for reviewers ing PDAC [7]. According to Kim et al. [7],
1, 2, and 3, respectively, to 0.864, 0.964, and abrupt interruption of the PD, upstream pan-
0.908, respectively (p = 0.013 for reviewer creatic parenchymal atrophy, mild peripan-
1, who had 1 year of experience; p < 0.0001 creatic infiltration, perilesional retention
for reviewer 2, who had 5 years of experi- cysts, and soft-tissue cuffing around the celi-
ence; and p = 0.701 for reviewer 3, who had ac artery are frequently found in isoattenuat-
10 years of experience). ing PDAC on CT, with a reported frequency
of 3.3–86.7%.
Discussion Among several valuable CT findings, we
The results of the present study show that found that abrupt PD narrowing at the transi-
several CT findings (i.e., nonhead location, tion area was the only statistically significant
long transition of the PD [≥ 6.1 mm], abrupt feature for malignant MPD dilatation on
transition of the PD, absence of duct penetrat- multivariate logistic regression analysis (OR,
ing sign, PD or CBD enhancement at the tran- 85.0; 95% CI, 10.8–669.5). This result can
sition area, and presence of attenuation dif- be expected because abrupt narrowing, in
ference in upstream pancreatic parenchyma) contrast to smooth U- or V-shaped narrow-
were statistically significant predictors of ma- ing, is a known imaging feature of malignant
lignant MPD dilatation in the differentiation CBD or PD obstruction [12–15]. As with CT
of benign and malignant MPD dilatation. In or dynamic contrast-enhanced MRI, this dif-
addition, the diagnostic performance of ra- ferential point between benign and malig-
diologists, as analyzed by the ROC method, nant CBD or PD obstruction is commonly
was improved with knowledge of significant used in cholangiopancreaticography, such as
CT criteria. If the results of the present study ERCP or MRCP [12, 13, 15]. In a recent ar-
are applied for differentiating malignant from ticle, smooth U- or V-shaped narrowing (a
benign MPD dilatation using CT, it may help PD icicle or ice pick sign) at the transition
radiologists obtain an accurate diagnosis that site is reported as a characteristic feature of
will help avoid futile surgery or other invasive autoimmune pancreatitis differentiating it
procedures, such as endoscopic ultrasound- from PDAC [12]. Furthermore, presence of
guided biopsy, for patients with asymptomat- a duct penetrating sign at the transition site,
ic benign MPD dilatation. Conversely, metic- the visible nonobstructed PD penetrating the
ulous and careful follow-up with aggressive mass, was exclusively observed in the benign
management should be recommended for pa- group (73.7% [14/19]). This result is similar
1050 AJR:209, November 2017
to results indicated in previously published
reports [12, 16]. According to those articles,
the presence of a duct penetrating sign is a
characteristic feature of inflammatory pan-
creatic lesion, including autoimmune pancre-
atitis compared with PDAC [12, 16]. The ma-
lignant mass is too small or isoattenuating to
be visible on CT; however, it may cause com-
plete obstruction of the PD and may result in
absence of a duct penetrating sign.
The head is the most common transition
site in benign PD dilatation; however, transi-
tion (tumor) in malignant PD dilatation can
occur anywhere throughout the entire pancre-
as. More than half of patients with benign PD
dilatation (57.9% [11/19]) had a transition site
in the head. When we traced down the dilated
PD on contrast-enhanced CT in the MPD dil-
atation benign group, we found that the tran-
sition zone was usually located at the area of
union between the main Wirsung duct and ac-
cessory Santorini duct. The true cause of this
locational prevalence is unknown; however,
we hypothesized that the area of embryologic
fusion between two ducts might be vulnerable
for benign PD narrowing. Indeed, insignificant
narrowing of the caliber at the site of fusion
for the dorsal and ventral ducts is considered
a common anatomic variant of the pancre-
as [17]. Further study using MRCP or ERCP
might prove the validity of our hypothesis.
The presence of an attenuation difference
at upstream pancreatic parenchyma was also
a significant CT feature of malignant MPD
dilatation. There was peripancreatic infiltra-
tion or fluid collection, which are CT features
of associated acute pancreatitis, in only one
patient with malignant MPD dilatation; how-
ever, frequent presence of low (n = 12) or high
(n = 2) attenuation in upstream pancreatic pa-
renchyma in the malignant MPD dilatation
group might be caused by occult obstruct-
ing pancreatitis. Statistical significance was
not achieved; however, recurrent occult pan-
creatitis at upstream pancreatic parenchyma
might also be responsible for a frequent asso-
ciation of upstream parenchymal atrophy in
the malignant MPD dilatation group (36.4%
[8/22] vs 15.8% [3/19], p = 0.138).
Contrary to our expectation, CBD dil-
atation (≥ 8 mm) can occur in both the be-
nign (26.3% [5/19]) and malignant (18.2%
[4/22]) dilatation groups without a statisti-
cally significant difference (p = 0.400). This
finding is also contrary to findings from a
previous report by Edge et al. [9]. According
to Edge and colleagues, double duct dilata-
tion (both PD and CBD dilatation) can oc-
 CT Differentiation of Causes of Isolated MPD Dilatation
cur more frequently in the malignant MPD
dilatation group, with statistical significance
(p  < 0.05). In the present study, the mean
age of patients in the benign MPD dilatation
group was 66.6 years, and old age might have
been responsible for accompanying CBD
Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved
dilatation resulting from senescent change.
Therefore, we should be cautious that the
presence of double duct dilatation does not
always indicate a malignant cause in patients
with MPD dilatation.
We initially hypothesized that the diagnos-
tic performance of a novice reader might im-
prove with knowledge of significant CT cri-
teria and that the performance of an expert
reader might not improve. However, in the
present study, diagnostic performance for the
differentiation improved in the third review
session, regardless of the expertise of the ra-
diologist. This result indicates that, unless in-
formation on significant CT features is given,
differentiation between benign and malig-
nant MPD dilatation is difficult regardless
of expertise. We hope that our results can be
widely accepted for use in clinical practice;
consequently, futile surgery or other unneces-
sary invasive procedures, such as endoscopic
ultrasound-guided biopsy, can be omitted for
patients with benign MPD dilatation.
The present study has several limitations.
First, only a small number of patients were
enrolled in the study because of the rarity of
isoattenuating pancreatic cancer and the use
of strict inclusion criteria. Second, the ret-
rospective study design required the use of
various CT scanners with various acquisition
protocols, such as scan parameters and CT
phases. However, we believe that this may be
insignificant because most CT scans (92.7%
[38/41]) were acquired with MDCT using a
slice thickness and reconstruction interval
not thicker than 5 mm (except for two pa-
tients). Nonetheless, a future prospective val-
idation study with a large study population
and uniform CT protocol is strongly warrant-
ed. Third, the same dataset was used for both
consensus and independent review sessions.
This might also create some bias toward im-
provement of the radiologists’ diagnostic per-
formance in the independent review session.
Therefore, a follow-up study with a different
AJR:209, November 2017 1051
CT dataset is strongly recommended to gen-
eralize the study results. Furthermore, for a
consensus reading session, the experience of
two radiologists (one who had 20 years of ex-
perience and one who was a 1st-year resident)
was quite different. Because the opinion of
the novice observer might be outweighed by
that of the experienced investigator, the lat-
ter dominant radiologist could essentially
make all decisions. Fourth, our study ana-
lyzed CT images only; however, the results
might be different if MR or PET/CT images
were added. We still believe that our results
are noteworthy because CT is the primary
imaging modality for the diagnosis of pan-
creatic diseases. Finally, histologic confirma-
tion could not be achieved for all 19 patients
with benign MPD dilatation and five patients
with malignant MPD dilatation. However, to
circumvent this limitation, we used a rela-
tively long-term follow-up (mean follow-up,
60.9 months; range, 12.6–198.6 months) for
patients with benign MPD dilatation. Fur-
thermore, we did not analyze histopathologic
findings for CT features, such as a perilesion-
al cyst or difference in attenuation.
In conclusion, distal, long (≥ 6.1 mm), and
abrupt transition; the absence of duct pen-
etrating sign; and the presence of attenua-
tion difference and PD or CBD enhance-
ment were found to be highly suggestive CT
findings for the differentiation of malignant
from benign MPD dilatation. Consequently,
knowledge of highly suggestive CT criteria
improved the diagnostic performance of ra-
diologists with regard to differentiation.
References
1. Siegel RL, Miller KD, Jemal A. Cancer statistics,
2015. CA Cancer J Clin 2015; 65:5–29
2. Vincent A, Herman J, Schulick R, Hruban RH,
Goggins M. Pancreatic cancer. Lancet 2011;
378:607–620
3. Bashir MR, Gupta RT. MDCT evaluation of the
pancreas: nuts and bolts. Radiol Clin North Am
2012; 50:365–377
4. Shrikhande SV, Barreto SG, Goel M, Arya A.
Multimodality imaging of pancreatic ductal ade-
nocarcinoma: a review of the literature. HPB
(­Oxford) 2012; 14:658–668
5. Takeshita K, Kutomi K, Haruyama T, et al. Imag-
ing of early pancreatic cancer on multidetector
row helical computed tomography. Br J Radiol
2010; 83:823–830
6. Fukukura Y, Takumi K, Kamiyama T, Shindo T,
­Higashi R, Nakajo M. Pancreatic adenocarcinoma: a
comparison of automatic bolus tracking and empiri-
cal scan delay. Abdom Imaging 2010; 35:548–555
7. Kim JH, Park SH, Yu ES, et al. Visually isoattenu-
ating pancreatic adenocarcinoma at dynamic-en-
hanced CT: frequency, clinical and pathologic
characteristics, and diagnosis at imaging exami-
nations. Radiology 2010; 257:87–96
8. Yoon SH, Lee JM, Cho JY, et al. Small (≤ 20 mm)
pancreatic adenocarcinomas: analysis of enhance-
ment patterns and secondary signs with multiphasic
multidetector CT. Radiology 2011; 259:442–452
9. Edge MD, Hoteit M, Patel AP, Wang X,
­Baumgarten DA, Cai Q. Clinical significance of
main pancreatic duct dilation on computed to-
mography: single and double duct dilation. World
J ­Gastroenterol 2007; 13:1701–1705
10. Fukukura Y, Fujiyoshi F, Sasaki M, et al. Pancre-
atic duct: morphologic evaluation with MR chol-
angiopancreatography after secretin stimulation.
Radiology 2002; 222:674–680
11. Pancreatica website. Pancreatic cancer prognosis.
­pancreatica.org/pancreatic-cancer/pancreatic-cancer-
prognosis/. Accessed March 5, 2017
12. Kim HJ, Kim YK, Jeong WK, Choi D. Pancreatic
duct “icicle sign” on MRI for distinguishing auto-
immune pancreatitis from pancreatic ductal ade-
nocarcinoma in the proximal pancreas. Eur Radi-
ol 2015; 25:1551–1560
13. Chung YE, Kim MJ, Kim HM, et al. Differentiation
of benign and malignant ampullary obstructions on
MR imaging. Eur J Radiol 2011; 80:198–203
14. Choi SH, Han JK, Lee JM, et al. Differentiating
malignant from benign common bile duct stric-
ture with multiphasic helical CT. Radiology 2005;
236:178–183
15. Soto JA, Alvarez O, Lopera JE, Múnera F,
­Restrepo JC, Correa G. Biliary obstruction: find-
ings at MR cholangiography and cross-sectional
MR imaging. RadioGraphics 2000; 20:353–366
16. Ichikawa T, Sou H, Araki T, et al. Duct-penetrat-
ing sign at MRCP: usefulness for differentiating
inflammatory pancreatic mass from pancreatic
carcinomas. Radiology 2001; 221:107–116
17. Mortelé KJ, Rocha TC, Streeter JL, et al. Multimo-
dality imaging of pancreatic and biliary congenital
anomalies. RadioGraphics 2006; 26:715–731
(Figures start on next page)
 Kim et al.

Fig. 1—Illustration showing measurement of length of duct transition. (Drawing by Kim SH, used with permission)
Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved

A B
Fig. 2—69-year-old man with benign main pancreatic duct (MPD) dilatation.
A, Curved coronal portal phase multiplanar reconstruction (MPR) images show smooth tapered narrowing (arrow) at head of pancreas. Upstream pancreatic duct is
mildly dilated. However, no atrophy or difference in attenuation exists in upstream pancreatic parenchyma. Duct penetrating sign (arrowheads) at transition site is clearly
shown. In this patient, length of transition site is 2 mm, which is shorter than length cutoff (6.1 mm) for differentiation of malignant MPD dilatation from benign MPD
dilatation.
B, Curved MPR image obtained at arterial phase 4 years later shows finding similar to that in image in A.

Fig. 3—79-year-old man with benign main pancreatic duct (MPD) dilatation.
Curved arterial phase multiplanar reconstruction image shows smooth tapered
narrowing (arrow) at head of pancreas. Upstream pancreatic duct (PD) is markedly
dilated. Common bile duct (CBD) is also dilated, probably because of senescent
change. However, no atrophy or difference in attenuation in upstream pancreatic
parenchyma exists. Duct penetrating sign (arrowheads) at transition site is clearly
present. In this patient, length of transition site is measured as 2 mm, which is
shorter than length cutoff (6.1 mm) for differentiation of malignant MPD dilatation
from benign MPD dilatation. PD dilatation showed no change on CT images
obtained 1 year earlier (not shown).

1052 AJR:209, November 2017

 CT Differentiation of Causes of Isolated MPD Dilatation
Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved

A B
Fig. 4—63-year-old woman with benign main pancreatic duct (MPD) dilatation.
A–C, Axial contrast-enhanced arterial (A) and portal phase (B) CT images as
well as oblique coronal minimum-intensity-projection CT image (C) show smooth
tapered narrowing (arrow) of pancreatic duct (PD) at body of pancreas. Upstream
PD is mildly dilated. Note penetrating PD (arrowhead, C) at transition site. Common
bile duct (CBD, C) is also dilated, probably because of senescent change. However,
no atrophy or difference in attenuation exists in upstream pancreatic parenchyma.
In this patient, length of transition site is measured as 13 mm, which is longer than
length cutoff (6.1 mm) for differentiation of malignant MPD dilatation from benign
dilatation. PD dilation remained stable for next 4 years (not shown).

A B
Fig. 5—77-year-old man with malignant main pancreatic duct (MPD) dilatation.
A, Coronal contrast-enhanced arterial phase CT image shows abrupt cutoff (arrow) of pancreatic duct (PD) in pancreatic body and upstream PD dilatation and
parenchymal atrophy (arrowheads). Pancreatic parenchyma has subtle lower attenuation than does downstream normal pancreatic parenchyma.
B, Coronal contrast-enhanced portal phase CT image shows abrupt cutoff (arrow) of pancreatic duct (PD) pancreatic body and upstream PD dilatation and parenchymal
atrophy (arrowheads). Subtle enhancement of PD is seen at transition site (arrow). Length of transition site was measured as 8.8 mm, which was longer than length cutoff
(6.1 mm) for differentiation of malignant MPD dilatation from benign dilatation. Endoscopic ultrasound-guided biopsy revealed diagnosis of adenocarcinoma (not shown);
distal pancreatectomy subsequently was performed.
(Fig. 5 continues on next page)

AJR:209, November 2017 1053

 Kim et al.
Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved

C D
Fig. 5 (continued)—77-year-old man with malignant main pancreatic duct (MPD) dilatation.
C, Photograph of gross specimen shows abrupt narrowing of MPD (arrow) at body portion.
D, Micrograph (H and E, ×4) of histologic specimen shows cancer nodule (arrows). MPD (arrowheads) is collapsed by tumor. Final histopathologic analysis confirmed
diagnosis of 3.5-cm poorly differentiated pancreatic ductal adenocarcinoma of T3N1.

A B

Fig. 6—67-year-old woman with malignant main pancreatic duct (MPD) dilatation.
A and B, Axial contrast-enhanced arterial (A) and portal (B) CT images show
abrupt cutoff (arrow) of pancreatic duct in neck portion as well as upstream
pancreatic duct dilatation without demonstrable mass or nodule in transition site
(arrow). Upstream pancreatic parenchyma (arrowheads) has prominent lower
attenuation compared with downstream normal pancreatic parenchyma with
peripancreatic infiltration, suggesting accompanying acute pancreatitis.
C, Delayed phase CT image shows abrupt cutoff (arrow) of pancreatic duct in neck
portion as well as upstream pancreatic duct dilatation without demonstrable
mass or nodule in transition site (arrow). Upstream pancreatic parenchyma
(arrowheads) has prominent lower attenuation compared with downstream
normal pancreatic parenchyma with peripancreatic infiltration, suggesting
accompanying acute pancreatitis. Subtle enhancement of pancreatic duct is seen
at transition site (arrow). In this patient, length of transition site is measured as
10 mm, which is longer than length cutoff (6.1 mm) for differentiation of malignant
MPD dilatation from benign dilatation. Endoscopic ultrasound-guided biopsy
revealed diagnosis of adenocarcinoma (not shown); subsequently, pylorus-
preserving total pancreatectomy was performed.
C (Fig. 6 continues on next page)

1054 AJR:209, November 2017

 CT Differentiation of Causes of Isolated MPD Dilatation
Downloaded from www.ajronline.org by 139.228.164.11 on 01/31/18 from IP address 139.228.164.11. Copyright ARRS. For personal use only; all rights reserved

D E
Fig. 6 (continued)—67-year-old woman with malignant main pancreatic duct (MPD) dilatation.
D, Photograph of gross specimen shows tiny yellowish nodule (arrow) at neck portion. D = duodenum.
E, Photomicrograph (H and E, ×4) of histologic specimen shows cancer nodule (area enclosed by red dots). Final histopathologic analysis confirmed diagnosis of 0.7-cm
moderately differentiated T1N0 pancreatic ductal adenocarcinoma.

AJR:209, November 2017 1055