See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/26258241

Differentiation between Malignant and Benign Gastric Ulcers: CT Virtual

Gastroscopy versus Optical Gastroendoscopy1

Article  in  Radiology · July 2009

DOI: 10.1148/radiol.2522081249 · Source: PubMed

CITATIONS READS
25 7,472

8 authors, including:

Yu-Ting Kuo Chien-Hung Lee

Chi-Mei Medical Center Kaohsiung Medical University
136 PUBLICATIONS   2,066 CITATIONS    284 PUBLICATIONS   5,350 CITATIONS   

SEE PROFILE SEE PROFILE

Tsyh-Jyi Hsieh Wan-Ting Huang

Kaohsiung Medical University Taipei Medical University Hospital
60 PUBLICATIONS   1,000 CITATIONS    65 PUBLICATIONS   933 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The association between regular use of aspirin and the prevalence of prostate cancer: Results from the National Health Interview Survey View project

sugar-sweetened beverages(SSB) intake, obesity, and inflammation View project

All content following this page was uploaded by Yu-Ting Kuo on 30 January 2014.

The user has requested enhancement of the downloaded file.

 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.
䡲 GASTROINTESTINAL IMAGING
Differentiation between
Malignant and Benign Gastric
Ulcers: CT Virtual Gastroscopy versus
Optical Gastroendoscopy1
ORIGINAL RESEARCH
Chiao-Yun Chen, MD
Yu-Ting Kuo, MD, PhD
Chien-Hung Lee, PhD
Tsyh-Jyi Hsieh, MD
Chang-Ming Jan, MD
Twei-Shiun Jaw, MD, MMS
Wan-Ting Huang, MD
Fang-Jung Yu, MD
1
From the Department of Medical Imaging (C.Y.C., Y.T.K.,
T.J.H., T.S.J.), Division of Gastroenterology, Department
of Internal Medicine (C.M.J., F.J.Y.), and Department of
Pathology (W.T.H.), Kaohsiung Medical University Hospital,
No. 100 Tz-You First Road, Kaohsiung City, 807 Taiwan;
Departments of Radiology (C.Y.C., Y.T.K., T.J.H., T.S.J.)
and Internal Medicine (C.M.J.), and Department of Public
Health, College of Health Science (C.H.L.), Kaohsiung Med-
ical University, Kaohsiung, Taiwan. From the 2006 RSNA
Annual Meeting. Received July 31, 2008; revision requested
September 2; revision received December 15; accepted
January 13, 2009; final version accepted March 3. Address
correspondence to F.J.Y. (e-mail:yufj@kmu.edu.tw ).
姝 RSNA, 2009
410
Purpose: To retrospectively compare computed tomographic virtual
gastroscopy (VG) and conventional optical gastroendos-
copy for the differentiation of malignant and benign gastric
ulcers.
Materials and The institutional review board approved this study and
Methods: confirmed that informed consent was not required. Gas-
tric ulcers in 115 patients (mean age, 64.7 years; range,
31– 86 years; 61 men, 54 women) were evaluated by using
endoscopy and VG. Ulcer shape, base, and margin and
periulcer folds were evaluated by two independent review-
ers. Malignant gastric ulcers were identified by irregular,
angulated, or geographic shape; uneven base; irregular or
asymmetric edges; and disrupted or moth-eaten appear-
ance of periulcer folds near the crater edge and/or clubbed
or fused folds. Benign gastric ulcers were identified by
smooth and regular shapes, even bases, clearly demar-
cated and regular edges, and folds that tapered and con-
verged toward the ulcer. The performance of VG and
endoscopy for the diagnosis of benign and malignant gas-
tric ulcers was evaluated by using histopathologic results
as the reference standard. The McNemar test was used to
compare VG and endoscopic data. A P value less than .05
was considered to indicate a significant difference.
Results: At histopathologic examination, 39 gastric ulcers were
benign, while 76 were malignant. VG and endoscopy had
sensitivities of 92.1% (70 of 76) and 88.2% (67 of 76),
respectively, for overall diagnosis of malignant gastric ul-
cers, and specificities of 91.9% (34 of 37) and 89.5% (34 of
38), respectively, for overall diagnosis of malignant gastric
ulcers. Endoscopy was more sensitive in depicting malig-
nancy according to ulcer base (85.5% [65 of 76] vs 68.4%
[52 of 76]) (P ⫽ .034), and VG was more specific in
depicting malignancy according to ulcer margin (78.4%
[29 of 37] vs 63.2% [24 of 38]) (P ⫽ .034).
Conclusion: VG and endoscopy were almost equally useful in distin-
guishing between malignant and benign gastric ulcers.
娀 RSNA, 2009
Supplemental material: http://radiology.rsnajnls.org/cgi
/content/full/2522081249/DC1
radiology.rsnajnls.org ▪ Radiology: Volume 252: Number 2—August 2009
GASTROINTESTINAL IMAGING: Virtual Gastroscopy and Endoscopy of Gastric Ulcers
G
astric cancer is one of the most
common causes of death from
cancer worldwide (1,2). It is the
second-most common cause of death
due to cancer in Asians from China,
Japan, and Korea (3). While it is one of
the least common cancers in North
America, it is the eighth leading cause of
cancer death in the United States. In
2006, there were reported to be more
than 22 280 new gastric cancer cases,
resulting in about 11 430 deaths in the
United States (4).
Early diagnosis of this malignancy is
crucial because patient outcome is de-
pendent on detecting it in its initial
stages. In Taiwan, for example, overall
5-year survival rates for this cancer
have been reported to be 97.9% for
stage IA disease, 92.9% for stage IB dis-
ease, 63.2% for stage II disease, 50.1%
for stage IIIA disease, 28.7% for stage
IIIB disease, and 9.3% for stage IV dis-
ease (5).
Patients with gastric ulcers have
been found to be at greater risk for de-
veloping gastric cancer (6,7). Because
malignancy is best treated in its early
stages, it is usually recommended that
all instances of gastric ulcers be fol-
lowed up with conventional endoscopy
and histopathologic studies until they
have healed to ensure that they are not,
Advances in Knowledge
䡲 The sensitivity and specificity for
the overall diagnosis at virtual
gastroscopy (VG) were 92.1%
and 91.9%, respectively, and
88.2% and 89.5%, respectively,
for the overall diagnosis at endos-
copy in the detection of malignant
gastric ulcers.
䡲 Endoscopy was more sensitive for
detection of malignancy based on
ulcer base criteria (P ⫽ .034),
and VG was more specific for de-
tection of malignancy based on
ulcer margin criteria (P ⫽ .034);
VG was about as accurate as en-
doscopy in identifying malignant
ulcer shapes, was more accurate
in finding the ulcer margin and
perifold, but was less accurate in
depicting ulcer base.
Radiology: Volume 252: Number 2—August 2009 ▪ radiology.rsnajnls.org
in fact, malignant (8,9). However, be-
cause endoscopy is an expensive and
invasive procedure, many patients,
even those at higher risk levels, might
refuse to undergo such essential endo-
scopic examinations.
While investigators (10–14) have
found computed tomographic (CT)
colonography to be a potent imaging
tool for the detection and screening of
colon cancer, CT has not been widely
used for the stomach (15,16). Fortu-
nately, gastric CT imaging by means of
three-dimensional virtual gastroscopy
(VG) and air distention of the stomach
now makes better detection of subtle
mucosal changes possible, and it can be
used to provide images that are almost
as detailed as those produced by using
endoscopy (15,16).
We therefore conducted this retro-
spective study to compare the perfor-
mance of CT VG with that of conven-
tional optical gastroendoscopy in regard
to the differentiation of malignant and
benign gastric ulcers.
Materials and Methods
Patients
The protocol for this retrospective
study was approved by the institutional
review board of our hospital, Kaohsiung
Medical University Hospital, which also
determined that informed consent could
be waived for this undertaking.
After performing a computerized
search of the medical records in our
hospital’s radiology files, one author
(C.Y.C., with 10 years of experience in
Implication for Patient Care
䡲 While endoscopy and VG are al-
most equally capable of depicting
malignancy by using the criteria
that we studied, decreased dis-
comfort of VG examination may
make it a preferred means of dis-
tinguishing benign from malignant
gastric ulcers in the future, partic-
ularly for patients who have con-
traindications to or who are un-
able to undergo conventional
endoscopy.
Chen et al
abdominal CT and 5 years of experience
in three-dimensional endoluminal ap-
proaches and interpretation) identified
135 patients with gastric ulcers who had
undergone endoscopic examination and
biopsy, as well as VG, at our institution
between May 2003 and August 2007. All
of the patients in our study were Tai-
wanese. Twenty patients were excluded
because they were lost during follow-
up, so no final diagnosis was made. The
remaining 115 patients, with either ma-
lignant (n ⫽ 76) or benign (n ⫽ 39)
gastric ulcers, were included in the
study. The clinical indications that re-
sulted from CT examinations in the 76
patients with malignant ulcers were that
patients were suspected of having ma-
lignant gastric ulcers at endoscopy (n ⫽
76). The indications for the 39 patients
with benign ulcers were as follows: (a)
patients were suspected of having ma-
lignant gastric ulcers at endoscopy (n ⫽
7); (b) intractable gastric ulcers did not
heal after 8 weeks of treatment with
proton-pump inhibitors (n ⫽ 8); (c) ul-
cer size of more than 3 cm in diameter
at endoscopy (n ⫽ 8); (d) gastric ulcers
in the gastric body where it was difficult
to exclude the possibility of malignancy
at endoscopy (n ⫽ 8); (e) previous bi-
opsy results showed cell dysplasia or
atypia (n ⫽ 5); and (f) gastric bleeding
(n ⫽ 3) (Fig 1).
The mean age of the 115 patients
was 64.7 years (age range, 31– 86
years). These patients included 61 men
Published online before print
10.1148/radiol.2522081249
Radiology 2009; 252:410 – 417
Abbreviation:
VG ⫽ virtual gastroscopy
Author contributions:
Guarantors of integrity of entire study, C.Y.C., C.M.J.,
T.S.J., F.J.Y.; study concepts/study design or data acqui-
sition or data analysis/interpretation, all authors; manu-
script drafting or manuscript revision for important intel-
lectual content, all authors; manuscript final version ap-
proval, all authors; literature research, C.Y.C., C.M.J.,
T.S.J., W.T.H., F.J.Y.; clinical studies, C.Y.C., Y.T.K.,
T.J.H., C.M.J., T.S.J., W.T.H., F.J.Y.; statistical analysis,
C.Y.C., C.H.L., F.J.Y.; and manuscript editing, C.Y.C.,
Y.T.K., T.J.H., C.M.J., F.J.Y.
Authors stated no financial relationship to disclose.
411
GASTROINTESTINAL IMAGING: Virtual Gastroscopy and Endoscopy of Gastric Ulcers Chen et al

(mean age, 65.9 years; range, 38 – 84 detector CT scanner (Brilliance 190P; tained from the diaphragmatic domes to 2
years) and 54 women (mean age, 63.4 Philips Medical Systems, Cleveland, Ohio) cm below the lower margin of the air-
years; range, 31– 86 years). The age dis- (26 patients). All patients had fasted for at distended gastric body. The 16-section mul-
tribution was comparable between the least 8 hours and had received 6 g of gas- tidetector CT scanner parameters were set
sexes (P ⫽ .266, t test). The elapsed producing crystals with 10 mL water orally at 1.25-mm collimation, 1.375 pitch, 27.5-
time between CT and endoscopy was to enable distention of the stomach before mm/sec table speed, 1.25-mm reconstruc-
within 7 days (mean, 3 days). Of these the procedures were performed. All of the tion thickness, 0.625-mm reconstruction
patients, 112 underwent endoscopy patients were placed in the supine position interval, 250–300 mAs, and 120 kVp. The
prior to CT VG, and three patients with with their right side elevated at approxi- 64-section multidetector CT parameters
gastric bleeding underwent CT VG prior mately 30°. To ensure adequate gastric dis- were set at 0.625-mm collimation, 1.142
to endoscopy. All of the diagnoses were tention, a scanogram was obtained. An ad- pitch, 0.75-second rotation time, 1.0-mm
confirmed with histopathologic exami- ditional 3 g of gas-producing crystals were reconstruction thickness, 0.5-mm recon-
nation results, and all of the patients given to patients determined to have insuf- struction interval, 250–300 mAs, and 120
with a diagnosis of a benign ulcer were ficient air distention. CT scans were ob- kVp.
followed up for more than 6 months.
Figure 1
Endoscopy
Endoscopy was performed with the pa-
tient in the left lateral position after ad-
ministration of oropharyngeal anesthesia
(Xylocaine; AstraZeneca, Sweden). En-
doscopy, as well as directed biopsy, was
performed by two board-certified experi-
enced gastroenterologists (F.J.Y. and
C.M.J., with 10 and 20 years of experi-
ence, respectively). We analyzed the dif-
ference in age and sex distribution for the
patients examined. Patterns for the two
characteristics were found to be compa-
rable (P ⫽ .495 for sex, ␹2 test; P ⫽ .257
for age, t test). End-viewing fiberoptic
panendoscopes (GIF-XQ240; Olympus, Figure 1: Flowchart of study enrollment. ENDO ⫽ endoscopy, FP ⫽ false positive, TP ⫽ true positive.
Tokyo, Japan) were used in all patients.
Six specific chosen directed-biopsy speci-
mens from the edge and base of the ulcer
were collected from each patient. Endo-
Figure 2
scopic evaluations were made (F.J.Y. and
C.M.J.). During analysis of the endo-
scopic features, cases of malignant and
benign ulcers were randomly intermixed.
Each gastroenterologist reviewed the en-
doscopic images and made an indepen-
dent evaluation; they then met with each
other, went over every case, and came to
a consensus evaluation.

CT VG Studies
All of the patients underwent multidetec-
tor CT examination following the routine
procedures designed for patients sus-
pected of having abnormal gastric lesions
in our department. CT examinations
were performed by using either a 16- Figure 2: Images of malignant gastric ulcer in 65-year-old man. (a) VG image shows en face view of ulcer
section multidetector CT scanner (Light- and (b) endoscopic image shows minimal oblique view of ulcer at the gastric part of the body with uneven
ulcer base, irregular ulcer shape, irregular ulcer margin, and associated gastric folds with rugae interruption
Speed H16; GE Healthcare, Milwaukee,
(arrows).
Wis) (89 patients) or a 64-section multi-

412 radiology.rsnajnls.org ▪ Radiology: Volume 252: Number 2—August 2009

GASTROINTESTINAL IMAGING: Virtual Gastroscopy and Endoscopy of Gastric Ulcers Chen et al

Reconstructed images from data are
routinely stored in our picture archiving
and communication system before three-
dimensional image processing is per-
formed. The VG images were indepen-
dently created and interpreted in the
three-dimensional workstation (AW 4.1;
GE Healthcare) by two authors (T.S.J.
and Y.T.K., with 10 and 12 years of ab-
dominal CT experience, respectively, as
well as 4 years of experience each in the
three-dimensional endoluminal ap-
proach). During analysis of the VG fea-
tures, the malignant and benign ulcer
cases were randomly intermixed. The
two authors analyzed the VG images
Figure 3
Figure 3: Images of benign gastric ulcer in 37-year-old woman. (a) VG image and (b) endoscopic image show
en face view of ulcer at the gastric angle with even ulcer base, regular triangular ulcer shape, regular ulcer margin,
and associated regular gastric folds terminating at the ulcer margin (arrows).
knowing that the images had been ob-
tained in patients with gastric ulcers but
not knowing the results of endoscopic and
pathologic examinations. Each of the two
gastrointestinal radiologists created and
reviewed the VG features and made a
record independently and then met with
each other to go over every case and
come to a consensus evaluation.
Image Evaluation
The locations of the gastric ulcers were
recorded on both VG and endoscopic
images. The benign gastric ulcers were
differentiated from the malignant gas-
tric ulcers, and this process was based
on radiologic and endoscopic criteria
established for the recognition of each
(17,18). Our diagnoses were made by
examining the available morphologic
features of ulcer shape, base, and mar-
gins and the relation of the ulcer to the
surrounding folds. Malignant gastric ul-
cers were identified by shapes that were
irregular, angulated, or geographic;
bases that were uneven; edges that
were irregular or asymmetric; and folds
that appeared to be disrupted or “moth-
eaten” near the crater edge and/or were
clubbed or fused (Fig E1, http:
//radiology.rsnajnls.org/cgi/content/full
/2522081249/DC1; Fig 2). Benign gastric
ulcers were identified by shapes that
were smooth and regular, bases that
were even, edges that were clearly de-
marcated and regular, and periulcer folds
that tapered and converged toward the
ulcer (Fig E2, http://radiology.rsnajnls
.org/cgi/content/full/2522081249/DC1;
Fig 3).
A standardized form was completed
independently by the two radiologists
who interpreted the VG images and by
the two gastroenterologists who inter-
preted the endoscopic images. On this
form, the evaluators noted whether the
shapes, bases, margins, or surrounding
folds on each image had primarily be-
nign or malignant features. In addition,
the two gastroenterologists evaluating
the endoscopic images also noted the
presence or absence of any periulcer
color change (redness, discoloration, or
spotty bleeding) (19). Radiologists and
gastroenterologists initially made inde-

Table 1

Diagnosis of Benign and Malignant Gastric Ulcers according to Criteria at VG and Endoscopy
VG Endoscopy
Periulcer Periulcer
Ulcer Ulcer Ulcer Fold Overall Ulcer Ulcer Ulcer Fold Color Overall
Shape* Base* Margin* Change† Diagnosis* Shape* Base* Margin* Change† Change Diagnosis
Parameter B M B M B M B M B M B M B M B M B M B M B M

No. of benign ulcers (n ⫽ 39) 28 9 30 7 29 8 15 3 34 3 31 7 33 5 24 14 12 5 20 19 34 4

No. of malignant ulcers (n ⫽ 76) 6 70 24 52 2 74 2 42 6 70 8 68 11 65 2 74 5 47 5 71 9 67

Note.—B ⫽ Benign criteria of images, M ⫽ malignant criteria of images.

* Information regarding VG criteria for two patients with benign ulcers was not evaluated, and information regarding endoscopic criteria for one patient with a benign ulcer was not evaluated.
†
No fold change was determined at VG examination for 21 patients with benign ulcers and 32 patients with malignant ulcers, and no fold change was determined at endoscopic examination for
22 patients with benign ulcers and 24 patients with malignant ulcers.

Radiology: Volume 252: Number 2—August 2009 ▪ radiology.rsnajnls.org 413

GASTROINTESTINAL IMAGING: Virtual Gastroscopy and Endoscopy of Gastric Ulcers Chen et al

pendent records, after which they came mucosa layers with varying degrees of
to a consensus diagnosis at VG and en- inflammation or granulomatous tissue

No fold change was determined at VG examination for 21 patients with benign ulcers and 32 patients with malignant ulcers, and no fold change was determined at endoscopic examination for 22 patients with benign ulcers and 24 patients with malignant ulcers.
88.2 (80.8, 95.5)
89.5 (79.5, 99.5)
Overall Diagnosis
doscopy, respectively. infiltration in the lamina propria.
VG results showed that 98.3%
Overall Diagnosis

88.6
(113 of 115) of patients in this study

…
To perform an overall diagnosis by us- had gastric ulcers. Two patients had

93.4 (87.8, 99.1)

51.3 (35.2, 67.3)
ing the individual VG and individual en- flat ulcers, which endoscopy, not VG,

Color Change
doscopic findings for the malignancy of could depict. While endoscopy could

0.847
gastric ulcers, we calculated the ⌽ cor- depict all of the ulcers, it could not be

79.1
relation coefficient for each individual used to differentiate between benign
criterion and the malignancy. These two and malignant ulcers in one patient

Periulcer Fold Change†

sets of coefficients were used as the rel- because the lesion was found at the

90.4 (77.4, 96.2)

70.6 (55.3, 97.6)

* Information regarding VG criteria for two patients with benign ulcers was not evaluated, and information regarding endoscopic criteria for one patient with a benign ulcer was not evaluated.
ative weight for developing a diagnostic distal antrum, and a complete view
score. The receiver operating character- was impossible.

0.913
Endoscopy

85.5
istic curve method was employed to find As can be seen in Tables 1 and 2,
the cutoff value for the optimal diagnostic the interobserver agreement in the

Diagnostic Accuracy, Sensitivity, and Specificity of Criteria of VG and Endoscopy Associated with Malignant Gastric Ulcers

97.4 (91.6, 100.0)

63.2 (44.6, 76.4)
accuracy (ie, the highest sum of the values differentiation of malignant and be-

Ulcer Margin*
for sensitivity and specificity). nign gastric ulcers by using VG or en-

0.900
doscopy showed excellent reproduc-

86.0
Statistical Analysis ibility (␬ ⫽ 0.879 – 0.937 for VG and
By using histopathologic findings as a 0.847– 0.919 for endoscopy). For the

85.5 (75.9, 92.6)

86.8 (65.7, 92.2)

Note.—See Table 1 for numbers used to calculate sensitivity, specificity, and accuracy percentages. Data in parentheses are 95% confidence intervals.
generally accepted standard of evalua- diagnosis of malignant gastric ulcers,

Ulcer Base*

0.870
tion, we compared the sensitivities, VG had a sensitivity and specificity of

86.0
specificities, and accuracies of the diag- 92.1% and 75.7%, respectively, based
noses based on the image characteris- on ulcer shape, 68.4% and 81.1%

89.5 (80.8, 95.5)

81.6 (65.7, 92.2)
tics on VG and endoscopic images. The based on ulcer base, 97.4% and 78.4%

Ulcer Shape*
sensitivities and specificities were re- based on ulcer margin, and 95.5% and

0.919
86.8
ported with corresponding 95% confi- 83.3% based on periulcer fold change
dence intervals. Interobserver variability, and had an average accuracy of more

91.9 (82.9, 100.0)

Overall Diagnosis*

92.1 (85.9, 98.3)

for benign and malignant ulcer evaluation than 72.6%. For the diagnosis of ma-
resulting from VG and endoscopic exam- lignant gastric ulcers, endoscopy had a
inations, was assessed by calculating the ␬ sensitivity and specificity of 89.5% and 92.0
statistic. ␬ Statistics were interpreted by 81.6%, respectively, based on ulcer …

using the following scale: fair agreement shape, 85.5% and 86.8% based on ul-
Periulcer Fold Change†

95.5 (85.8, 100.0)

83.3 (65.3, 100.0)

was indicated by 0.21– 0.40; moderate
agreement, 0.41– 0.60; good agreement,
0.61– 0.80; and excellent agreement,
cer base, 97.4% and 63.2% based on
ulcer margin, and 90.4% and 70.6%
based on periulcer fold change.
0.929
91.9

0.81–1.00 (20). Consistency between VG For the overall diagnosis performed

and endoscopic results was evaluated by by using the individual VG and individ-
97.4 (89.6, 99.8)
78.4 (62.5, 90.2)

using the McNemar test. A P value less ual endoscopic findings for the malig-
Ulcer Margin*
VG

than .05 was considered to indicate a sig- nancy of gastric ulcers, the correlation
P values less than .05 indicated a significant difference for ␬ values.
0.933
91.2

nificant difference. All statistical opera- coefficients were 0.69, 0.47, 0.80, and
tions were performed by using software 0.80 for ulcer shape, ulcer base, ulcer
68.4 (56.4, 77.9)
81.1 (65.7, 92.2)

(Stata, SE version 9.1 for Windows, margin, and periulcer fold change, re-
Ulcer Base*

2005; Stata, College Station, Tex). spectively, for VG findings and 0.71,
0.879

0.70, 0.68, and 0.6, respectively, for

72.6

endoscopic findings. The correlation-

Results weighted overall scores were calcu-
92.1 (85.9, 98.3)
75.7 (56.3, 85.8)

lated as follows: (0.69 䡠 shape) ⫹

Ulcer Shape*

Of the 115 patients observed in this

study, 39 had histopathologically (0.47 䡠 base) ⫹ (0.80 䡠 margin) ⫹
0.937
86.7

proved benign gastric ulcers, and 76 (0.80 䡠 fold change) for VG diagnosis
had malignant gastric ulcers (24 at and (0.71 䡠 shape) ⫹ (0.70 䡠 base) ⫹
Specificity (%)
Sensitivity (%)

Accuracy (%)

stage T1, 29 at stage T2, and 23 at (0.68 䡠 margin) ⫹ (0.60 䡠 fold change)
␬ Value‡
Parameter
Table 2

stage T3). Histopathologic study re- for endoscopic diagnosis, with a score
sults revealed that all of the benign of 1 for positive and a score of 0 for
†

ulcers had eroded or defective gastric negative (or no fold change) findings.

414 radiology.rsnajnls.org ▪ Radiology: Volume 252: Number 2—August 2009

GASTROINTESTINAL IMAGING: Virtual Gastroscopy and Endoscopy of Gastric Ulcers Chen et al

The cutoff values for VG findings were
2.29 and 1.49, respectively, for patients
with and those without the periulcer fold
change and 2.09 and 0.70, respectively,
for endoscopic findings. Patients with
overall scores that were larger or equal to
those of the corresponding cutoff values
were determined to have a malignancy.
The sensitivity and specificity for the
overall diagnosis at VG were 92.1% and
91.9%, respectively, and 88.2% and
89.5% at endoscopy. In addition, endos-
copy showed 93.4% sensitivity and
51.3% specificity for periulcer color
Figure 4
Figure 4: Images of malignant gastric ulcer in 65-year-old woman, with atypical benign morphologic fea-
ture of the ulcer base on VG image. (a) VG image shows en face view of ulcer at the low gastric body with even
ulcer base (arrow). (b) Endoscopic image shows oblique view of ulcer and an uneven ulcer base (arrow).
change and had an average accuracy of
more than 79.1% (Table 2).
As shown in Table 3, VG and endos-
copy had significantly different diagnostic
results in the evaluation of ulcer base and
ulcer margin (both P ⫽ .034, McNemar
test). Compared with VG, endoscopy was
found to more sensitively depict malig-
nancy based on the unevenness of the
ulcer base (Fig 4), while VG had better
specificity based on the irregularity of the
ulcer margin (Fig 5).
With both VG and endoscopic crite-
ria, two benign gastric ulcers were mis-
diagnosed as being malignant (Fig 6). In
addition, one malignant gastric ulcer
was misdiagnosed as benign at VG but
was correctly diagnosed at endoscopy
(Fig E3, http://radiology.rsnajnls.org
/cgi/content/full/2522081249/DC1).
Discussion
In this study, we compared the perfor-
mance of VG and endoscopy in the dif-
ferentiation between malignant and be-
nign gastric ulcers and found the diag-
nostic results at VG to be comparable
with those at endoscopy.
VG is a new noninvasive technique
capable of creating high-quality detailed
three-dimensional images of subtle mu-
cosal changes in the stomach. Its image
quality is diagnostically comparable
with that provided at optical endoscopy
(15,21). VG has a wider viewing field
than endoscopy and is better for evalu-
ating high lesser curvature sites and du-
odenal bulbs than conventional optical
endoscopy (19). Because views ob-
tained with a virtual camera can be ad-
justed without substantial limitations,
and because they can be re-formed ret-
rospectively to clear up any blind spots,
VG produces better overall views of the
whole ulcer and better views of the ul-
cer margin than endoscopy (P ⫽ .034).
Because its view comes from several an-
gles, it offers a more precise measure-
ment of the abnormality. Even more im-
portant, VG is technically less invasive
than endoscopy.

Table 3

Diagnostic Comparison of Individual and Overall Criteria for Imaging Findings at VG and Endoscopy for Patients with Gastric Ulcers
Ulcer Shape at Ulcer Base at Ulcer Margin at Periulcer Fold Change Overall Diagnosis at
Endoscopy Endoscopy Endoscopy at Endoscopy* Endoscopy
Parameter Benign Malignant Benign Malignant Benign Malignant Benign Malignant Benign Malignant

VG†
Benign 29 5 34 19 23 7 11 5 31 8
Malignant 8 70 8 51 1 81 2 40 10 63
␹2 For McNemar test‡ 0.69 (.405) 4.48 (.034) 4.50 (.034) 1.29 (.257) 0.22 (.637)

Note.—Unless otherwise indicated, data are numbers of ulcers.

* Only 58 patient pairs were available for the comparison of periulcer fold change between VG and endoscopy.
†
Patients with benign or malignant ulcers (n ⫽ 112) were compared with the same individual criteria as for those undergoing endoscopy. Data from three patients were not used: Two patients
had flat ulcers, which could not be detected at VG, and another patient had an ulcer at the distal antrum, and this could not be used to differentiate a benign from a malignant ulcer because a
complete view was impossible.
‡
Data in parentheses are P values.

Radiology: Volume 252: Number 2—August 2009 ▪ radiology.rsnajnls.org 415

GASTROINTESTINAL IMAGING: Virtual Gastroscopy and Endoscopy of Gastric Ulcers Chen et al

Figure 5

Figure 5: Images of benign gastric ulcer in 59-year-old man, with atypical malignant morphologic features of ulcer margin and periulcer fold change on endoscopic
image. VG images show the (a) profile view and (b) en face view of ulcer at gastric antrum with even ulcer base, regular oval ulcer shape, regular ulcer margin, and associ-
ated regular gastric folds terminating at the ulcer margin (arrows). (c) Endoscopic image shows oblique view of ulcer and asymmetric edema of ulcer margin (arrowhead)
and associated gastric folds with bulbous rugae enlargement (arrow). P ⫽ pylorus.

VG, however, has several disadvan-

tages as well. First, VG does not depict Figure 6
color change, which makes it difficult to
detect flat lesions or to evaluate subtle
changes at the ulcer base. In our study,
two patients had flat ulcers that were
undetectable at VG. Endoscopy was
also found to produce a better view of
the ulcer base (P ⫽ .034), partially be-
cause of its depiction of the color
change. Second, the use of recon-
structed algorithms is still far from
yielding perfect “natural” images. En-
doscopy, which has a pixel size of
around 0.03 mm/pixel, has better spa-
tial resolution than VG (pixel size
around 0.59 – 0.82 mm/pixel), which
makes it more difficult to detect subtle
changes in small lesions at VG. In our Figure 6: Images of acute benign gastric ulcer in 52-year-old man exhibit malignant morphologic features.
study, use of VG led to the misdiagnosis (a) VG image shows en face view of ulcer and (b) endoscopic image shows minimal oblique view of ulcer at
of one malignant gastric ulcer as benign, gastric body with uneven ulcer base, irregular ulcer shape, and irregular ulcer margin. Rugae interruption
while endoscopy clearly depicted the (arrow) and bulbous enlargement (arrowheads) are evident on a but not on b; this is because of the limited
characteristics needed to make an accu- field of view.
rate diagnosis. Additionally, when using
VG, there is a substantial dose of ioniz-
ing radiation, and biopsies cannot be
collected for histopathologic evaluation. most malignant gastric ulcers from be- Although VG may be as good as en-
Thus, optical endoscopy must still be nign ones, acute benign ulcers with se- doscopy in the depiction of a malignancy,
performed for confirmation. In fact, in vere periulcer edema might mimic ma- it must be performed by well-trained ra-
many instances, endoscopy is per- lignant ulcers. In our study, two acute diologists who can process a relatively
formed as a concomitant. benign ulcers were misdiagnosed as ma- large number of CT images and who are
We found that although using endo- lignant at both VG and endoscopy in familiar with the endoscopic appearance
scopic criteria can help differentiate evaluation of all criteria. of gastric ulcers beforehand. While CT is

416 radiology.rsnajnls.org ▪ Radiology: Volume 252: Number 2—August 2009

GASTROINTESTINAL IMAGING: Virtual Gastroscopy and Endoscopy of Gastric Ulcers
not currently performed as a routine ex-
amination for the detection of gastric ul-
cers, multidetector CT, with continued
advancements in CT scanners and com-
puter technology, may play an increas-
ingly important role in the detection of
malignant gastric ulcers. Whether this
will translate into improved overall sur-
vival rates for patients with gastric can-
cer, or to cost savings, is still unclear.
Randomized, double-blinded prospective
studies with larger patient groups are re-
quired to validate the role of VG in the
diagnosis of gastric cancer.
This study had some limitations.
There may have been some selection bias
caused by limiting the enrollment of pa-
tients to those with gastric ulcers who
underwent CT. This may have led to an
overestimation in the ability of VG to de-
pict gastric ulcers. However, because the
aim of this study was to compare the abil-
ity of the two modalities to help differen-
tiate malignant from benign gastric ul-
cers, rather than investigating the detect-
ability of gastric ulcers with VG, this
overestimation may not substantially
have changed the results of this study. In
addition, there was great variation in the
elapsed time between the endoscopic and
VG examinations. Some benign active ul-
cers might lose edema after intensive treat-
ment. In our study, one benign ulcer with
active bleeding showed less edema at VG
than at endoscopy after 7 days of treatment
with a proton-pump inhibitor.
In conclusion, this study found that
VG and endoscopy are almost equally
useful in distinguishing between malig-
nant and benign gastric ulcers. The rel-
ative reduction in discomfort of the non-
invasive VG examination may make it a
Radiology: Volume 252: Number 2—August 2009 ▪ radiology.rsnajnls.org
good potential alternative means for
evaluating ulcers. Further prospective
randomized studies comparing VG and
optical endoscopy are required to inves-
tigate the overall efficacy and cost-effec-
tiveness of VG.
References
1. Parkin DM, Bray F, Ferlay J, Pisani P. Global
cancer statistics, 2002. CA Cancer J Clin
2005;55:74 –108.
2. Crew KD, Neugut AI. Epidemiology of gas-
tric cancer. World J Gastroenterol 2006;12:
354 –362.
3. Leung WK, Wu MS, Kakugawa Y, et al.
Screening for gastric cancer in Asia: current
evidence and practice. Lancet Oncol 2008;9:
279 –287.
4. Jemal A, Siegel R, Ward E, et al. Cancer
statistics, 2006. CA Cancer J Clin 2006;56:
106 –130.
5. Wu CW, Lui WY. Current management of
gastric cancer in Taiwan. Asia J Surg 2001;
24:253–257.
6. Hansson LE, Nyren O, Hsing AW, et al. The
risk of stomach cancer in patients with gas-
tric or duodenal ulcer disease. N Engl J Med
1996;335:242–249.
7. Andrade MH, Silva E, Mares-Guia M. A plau-
sible identification of the secondary binding
site in trypsin and trypsinogen. Braz J Med
Biol Res 1990;23:1223–1231.
8. Mountford RA, Brown P, Salmon PR,
Alvarenga C, Neumann CS, Read AE. Gas-
tric cancer detection in gastric ulcer disease.
Gut 1980;21:9 –17.
9. Moayyedi P, Axon AT. Endoscopy and gas-
tric ulcers. Endoscopy 1995;27:689 – 693.
10. Yee J, Akerkar GA, Hung RK, Steinauer-
Gebauer AM, Wall SD, McQuaid KR. Colo-
rectal neoplasia: performance characteris-
tics of CT colonography for detection in 300
patients. Radiology 2001;219:685– 692.
Chen et al
11. Sosna J, Morrin MM, Kruskal JB, Lavin PT,
Rosen MP, Raptopoulos V. CT colonography
of colorectal polyps: a metaanalysis. AJR
Am J Roentgenol 2003;181:1593–1598.
12. Pickhardt PJ, Choi JR, Hwang I, et al. Com-
puted tomographic virtual colonoscopy to screen
for colorectal neoplasia in asymptomatic adults.
N Engl J Med 2003;349:2191–2200.
13. Kim DH, Pickhardt PJ, Taylor AJ, et al. CT
colonography versus colonoscopy for the de-
tection of advanced neoplasia. N Engl J Med
2007;357:1403–1412.
14. Neri E, Giusti P, Battolla L, et al. Colorectal
cancer: role of CT colonography in preoper-
ative evaluation after incomplete colonos-
copy. Radiology 2002;223:615– 619.
15. Inamoto K, Kouzai K, Ueeda T, Marukawa
T. CT virtual endoscopy of the stomach:
comparison study with gastric fiberscopy.
Abdom Imaging 2005;30:473– 479.
16. Ogata I, Komohara Y, Yamashita Y, Mitsu-
zaki K, Takahashi M, Ogawa M. CT evalua-
tion of gastric lesions with three-dimensional
display and interactive virtual endoscopy:
comparison with conventional barium study
and endoscopy. AJR Am J Roentgenol 1999;
172:1263–1270.
17. Gabrielsson N. Benign and malignant gastric
ulcers: evaluation of differential diagnostics
in roentgen examination and endoscopy. En-
doscopy 1972;4:73–78.
18. Maruyama M, Baba Y. Gastric carcinoma.
Radiol Clin North Am 1994;32:1233–1252.
19. Honmyo U, Misumi A, Murakami A, et al.
Mechanisms producing color change in flat
early gastric cancers. Endoscopy 1997;29:
366 –371.
20. Landis JR, Koch GG. The measurement of
observer agreement for categorical data. Bi-
ometrics 1977;33:159 –174.
21. Kim JH, Eun HW, Goo DE, Shim CS, Auh
YH. Imaging of various gastric lesions with
2D MPR and CT gastrography performed
with multidetector CT. RadioGraphics 2006;
26:1101–1116; discussion 1117–1118.
417