Introduction
Gastric cancer is the second most common cause of cancer-associated deaths worldwide
[1]; therefore, early detection and curative treatment are the best strategies to improve
patient survival. Detection of gastric cancers ≤ 20 mm in diameter is particularly
ideal because they are curable by minimally invasive treatment modalities, including
endoscopic mucosal resection and endoscopic submucosal dissection [2]
[3]. Esophagogastroduodenoscopy (EGD) is the most sensitive method for early detection
of gastric cancers, but it is often difficult while using conventional white-light
imaging (C-WLI) endoscopy [4]. Hence, new modalities are required.
In recent years, image-enhanced endoscopy (IEE), including narrow-band imaging (NBI)
and blue-laser imaging (BLI), has been used to improve visualization of microstructures
and microvessels in gastrointestinal lesions and to increase diagnostic accuracy [5]
[6]
[7]. Previous studies have shown that conventional IEE, including NBI and BLI, was effective
for early detection of neoplastic lesions in the esophagus and colon [8]
[9]
[10]. Although conventional IEE was also expected to facilitate early detection of gastric
mucosal cancers, no studies have demonstrated its utility in this regard; this lack
of evidence may be because conventional IEE probably does not provide sufficient brightness
for effective examination of the stomach, which has a wider lumen compared with the
esophagus and colon.
Linked color imaging (LCI) (Fujifilm Co., Tokyo, Japan) is a newly developed IEE modality
that provides images with enhanced brightness and visualization of red lesions and
whitish lesions during routine endoscopy. Another merit is that LCI images resemble
color-enhanced C-WLI images; thus, eliminating the need for special training. We reported
that LCI increases visibility of flat colorectal lesions [11]. Previous studies, which were case series involving a small number of patients,
reported that LCI enhances early gastric cancer lesions [12]
[13]. However, the efficacy of this modality compared with conventional IEE remains unknown.
With the aim of examining the utility of LCI in improving detection of gastric mucosal
cancer, we conducted a retrospective study to compare three modalities, including
WLI, BLI-bright, and LCI, with regard to their ability to facilitate the visualization
of gastric mucosal cancer.
Patients and methods
Between December 2014 and July 2017, 113 consecutive early gastric cancers were examined
by EGD and were histologically proven mucosal cancers that measured ≤ 20 mm in diameter
at Chiba Cancer Center, Chiba, Japan. One expert (Y. K.) took all the endoscopic images.
Thirteen cases examined using a NBI system (Olympus Medical Systems, Tokyo, Japan)
were excluded from the analysis. Thus, the current study used endoscopic images of
100 early gastric cancers, examined using a LASEREO system (FUJIFILM, Tokyo, Japan)
with an EG-L590ZW or L600ZW endoscope. For all lesions, we selected one image each
acquired by WLI, BLI-bright, and LCI from the same site or from the intermediate or
distant view. A total of 100 images were prepared for each modality.
Before the study, surface colors in WLI were evaluated by two experts (Y. K. and T. H.)
and then divided into three groups; whitish, normochromic, and reddish lesions. They
were diagnosed based on the predominant color. The endoscopic images of WLI, BLI-bright,
and LCI were randomly displayed and independently reviewed by four endoscopists, including
two experts and two trainees, in a single session without prior knowledge of the lesions.
In this study, the endoscopists were defined as experts when their experience in using
IEE was over 5 years or as trainees when their experience was less than 1 year. Images
were electronically presented without zooming; image size was predetermined and similar
to that used during routine endoscopic monitoring. The reviewers did not refer to
earlier images. They scored the lesions based on a previously reported 4-point visibility
scale [11]
[14], as follows: score 4, excellent (easily detectable); score 3, good (detectable with
careful observation); score 2, fair (hardly detectable without careful examination);
and score 1, poor (not detectable without repeated careful examination). Previous
study demonstrated that this score was related to an objective indicator such as color
difference value [15]. A representative image for each score is shown in [Fig. 1].
Fig. 1 Representative lesions with visibility scores 1 – 4 using the three modalities. WLI,
white-light imaging; BLI-bright, blue-laser imaging-bright; LCI, linked color imaging.
The current study was approved by the ethics committee of Chiba Cancer Center and
the contents were displayed on the notice board for inpatients and outpatients. The
study was carried out in accordance with the World Medical Association’s Declaration
of Helsinki, and all the patients provided informed consent for undergoing EGD.
Histopathologic diagnosis
Lesion specimens were obtained by endoscopic resection and were fixed with 10 % formalin.
Histopathological diagnosis was performed according to the Japanese Classification
of Gastric Carcinoma [16].
LCI, the novel image-enhancing mode
We used an endoscopic system employing a laser light source (LASEREO), which has two
lasers with different wavelengths; white-light laser (wavelength 450 ± 10 nm) provides
a wide spectrum of white light illumination, suitable for general observation, whereas
BLI mode laser comprises short wavelength (410 ± 10 nm) and narrow band. BLI mode
provides high contrast signals to obtain information regarding capillaries on the
mucosal surface, slight mucosal irregularities, and deep blood vessels. The intermediate
mode BLI-bright, which has a higher white light intensity ratio, provides a brighter
image. LCI used in the current study was a novel image-enhanced mode based on BLI-bright
imaging with additional processing that enhanced red color separation to depict red
and white colors more vividly. Because LCI enhanced color contrast, color difference
between the surrounding tissue and lesions, including light-colored lesions, is emphasized,
thereby facilitating visualization.
Statistical analysis
Clinical data were expressed as percentage, mean, and range. The mean (± SD) of the
visibility scores was calculated. Mean scores of all endoscopists, the expert group,
and the trainee group were analyzed and compared among the modalities using paired
t-test. The modalities were also compared in terms of various clinical characteristics,
including surface color and macroscopic type. P values < 0.05 were considered to be significant. Statistical analyses were performed
using SPSS software, version 17.0 (SPSS Inc, Chicago, IL).
Results
Patient characteristics are summarized in [Table 1]. Median patient age was 74.5 (62 – 90) years, with a preponderance of men (n = 77).
Median lesion size was 10 mm (2 – 20 mm), and surface colors were whitish in 16, normochromic
in 55, and reddish in 29 lesions. Macroscopic types were as follows: protruding in
33, flat in 14, and depressed in 53. Gastric lesions were located in the upper third
in 17 patients, in the middle third in 16 patients, and in the lower third in 67 patients.
There were 94 cases of well-differentiated tubular adenocarcinoma and six cases of
moderately differentiated tubular adenocarcinoma. Representative cases are shown in
[Fig. 2], [Fig. 3], [Fig. 4], and [Fig. 5].
Table 1
Baseline clinicopathologic characteristics.
|
No. of patients
|
100
|
|
Male/female, no.
|
77/23
|
|
Median age (range), years
|
74.5 (62 – 90)
|
|
Lesion size, median (range), mm
|
10.0 (2 – 20)
|
|
Surface color
|
|
|
16
|
|
|
55
|
|
|
29
|
|
Macroscopic type
|
|
|
33
|
|
|
14
|
|
|
53
|
|
Location of the gastric lesion
|
|
|
17
|
|
|
16
|
|
|
67
|
|
Histology
|
|
|
94
|
|
|
6
|
Tub1, well-differentiated tubular adenocarcinoma; Tub2, moderately differentiated
tubular adenocarcinoma.
Fig. 2 a White-light imaging shows a whitish lesion (arrow) in the lower third of the stomach:
mean visibility score, 3.75. b Blue-laser imaging-bright: mean visibility score, 3.50. c Linked color imaging enables enhanced visualization of the whitish lesion: mean visibility
score, 4.00.
Fig. 3 a White-light imaging shows a reddish lesion (arrow) in the lower third of the stomach:
mean visibility score, 3.50. b Blue-laser imaging-bright: mean visibility score, 3.75. c Linked color imaging enables enhanced visualization of the reddish lesion: mean visibility
score, 4.00.
Fig. 4 Endoscopic images of the same gastric mucosal cancer (case #1). a White-light imaging shows a depressed lesion (arrow) in the upper third of the stomach:
mean visibility score, 1.75. b Blue-laser imaging-bright: mean visibility score, 2.75. c Linked color imaging, which provides a high color contrast image to vividly differentiate
between the cancerous lesion and the surrounding normal mucosa: mean visibility score,
3.50.
Fig. 5 Endoscopic images of the same gastric mucosal cancer (case #2). a White-light imaging shows a flat lesion (arrow) in the lower third of the stomach:
mean visibility score, 2.25. b Blue-laser imaging-bright: mean visibility score, 2.75. c Linked color imaging, which enhances mucosal irregularity: mean visibility score,
3.00.
Mean (± SD) visibility scores of the four endoscopists were 2.54 ± 1.10 for WLI, 3.02 ± 1.07
for BLI-bright, and 3.28 ± 0.97 for LCI. The score for BLI-bright was significantly
higher than that for WLI (P < 0.001) and the score for LCI (P < 0.001) was significantly higher than that for BLI-bright mode. For the experts,
the score for BLI-bright was significantly higher than that for WLI (3.23 ± 0.95 vs.
2.51 ± 1.08; P < 0.001), but there was no difference between the scores for BLI-bright and LCI (3.23 ± 0.95
vs. 3.38 ± 0.83). For the trainees, there was no difference between the scores for
WLI (2.57 ± 1.12) and BLI-bright (2.79 ± 1.13), but the score for LCI (3.18 ± 1.08)
was significantly higher than that for WLI and BLI-bright (P < 0.001) ([Table 2]).
Table 2
Mean visibility scores of endoscopists (experts and trainees) for WLI, BLI-bright,
and LCI modalities.
|
Endoscopist
|
WLI
|
BLI-bright
|
LCI
|
WLI vs. BLI-bright, P value
|
WLI vs. LCI, P value
|
BLI-bright vs. LCI, P value
|
|
All, mean ± SD
|
2.54 ± 1.10
|
3.02 ± 1.07
|
3.28 ± 0.97
|
< 0.001
|
< 0.001
|
< 0.001
|
|
Expert, mean ± SD
|
2.51 ± 1.08
|
3.23 ± 0.95
|
3.38 ± 0.83
|
< 0.001
|
< 0.001
|
NS
|
|
Trainee, mean ± SD
|
2.57 ± 1.12
|
2.79 ± 1.13
|
3.18 ± 1.08
|
NS
|
< 0.001
|
< 0.001
|
WLI, white-light imaging; BLI, blue-laser imaging; LCI, linked color imaging; SD,
standard deviation; NS, not significant.
Distribution of scores for the three modalities is shown in [Fig. 6]. The proportion of cases with scores 3 and 4, which represented good visibility,
was 50.3 % for WLI compared with 79 % for LCI. Conversely, the proportion of cases
with scores 1 and 2, which represented poor visibility, was 49.8 % for WLI compared
with 21 % for LCI. Next, when the mean score of the four endoscopists was calculated
for each lesion, 28 cases had mean scores of 2 or lower on WLI, 15 (53.6 %) of which
had scores of 3 or higher on LCI. Among 13 cases with scores of 2 or lower on BLI-bright,
3 cases (23.1 %) showed scores of 3 or higher on LCI. Based on the assessment of four
endoscopists, two lesions had mean scores of 2 or lower on LCI, but neither of the
two showed scores of 3 or higher on WLI or BLI-bright.
Fig. 6 Distribution of visibility scores. For each modality, the total scores of all endoscopists
were counted. WLI, white-light imaging; BLI-bright, blue-laser imaging-bright; LCI,
linked color imaging.
Mean visibility scores of the four endoscopists, in terms of various clinical characteristics
including surface color and macroscopic type, are shown in [Table 3]. With regard to surface color, mean (± SD) visibility scores for WLI, BLI-bright,
and LCI in normochromic lesions were 2.15 ± 1.03, 2.78 ± 1.10, and 3.08 ± 1.04, respectively,
which were lower those for whitish lesions (3.14 ± 1.10, 3.47 ± 0.93, and 3.69 ± 0.64,
respectively) and reddish lesions (2.95 ± 0.91, 3.23 ± 0.97, and 3.42 ± 0.88, respectively).
These findings indicated that in normochromic lesions, although visibility scores
for all three modalities were lower than those in whitish and reddish lesions, the
score for LCI remained significantly higher than those for the other two modalities.
With regard to macroscopic type, mean (± SD) visibility scores in flat and depressed
lesions were 2.33 ± 1.04 for WLI, 2.84 ± 1.09 for BLI-bright, and 3.15 ± 1.02 for
LCI, which were lower than the scores in protruding lesions (2.95 ± 1.11, 3.39 ± 0.93,
and 3.53 ± 0.80, respectively). These findings showed that in flat and depressed lesions,
although visibility scores of all three modalities were lower than those in protruding
lesions, the score for LCI remained significantly higher than those for the other
two modalities.
Table 3
Mean visibility scores of endoscopists, in terms of various clinical characteristics.
|
WLI
|
BLI-bright
|
LCI
|
WLI vs. BLI-bright, P value
|
WLI vs. LCI,
P value
|
BLI-bright vs. LCI,
P value
|
|
Surface color
|
|
|
3.14 ± 1.10
|
3.47 ± 0.93
|
3.69 ± 0.64
|
NS
|
< 0.001
|
NS
|
|
|
2.15 ± 1.03
|
2.78 ± 1.10
|
3.08 ± 1.04
|
< 0.001
|
< 0.001
|
< 0.001
|
|
|
2.95 ± 0.91
|
3.23 ± 0.97
|
3.42 ± 0.88
|
< 0.001
|
< 0.001
|
NS
|
|
Macroscopic type
|
|
|
2.95 ± 1.11
|
3.39 ± 0.93
|
3.53 ± 0.80
|
< 0.001
|
< 0.001
|
NS
|
|
|
2.33 ± 1.04
|
2.84 ± 1.09
|
3.15 ± 1.02
|
< 0.001
|
< 0.001
|
< 0.001
|
WLI, white-light imaging; BLI, blue-laser imaging; LCI, linked color imaging; SD,
standard deviation; NS, not significant.
Discussion
In the current study, we proved that, compared with WLI and BLI-bright modes, LCI
mode improved visibility of gastric mucosal cancers measuring ≤ 20 mm in diameter.
Upon comparing visibility scores of experts with those of trainees, those for BLI-bright
and LCI were significantly higher than those for WLI, with no significant difference
between the scores for BLI-bright and LCI in experts; conversely, in trainees, the
score for LCI was significantly higher than the scores for WLI and BLI-bright, with
no significant difference between WLI and BLI-bright. One possible reason for these
results was that the images of conventional IEE, such as BLI-bright, had completely
different color tones compared with those of standard WLI and that some familiarization
was required. In contrast, because LCI images resembled color-enhanced WLI images,
LCI may have a steep learning curve and is easy to adopt even in facilities with limited
opportunities to diagnose early gastric cancer. Therefore, this modality could be
extremely useful in actual clinical practice, may enable IEE for observation, and
can potentially become the standard method for detection of gastric cancer.
In this study, distribution of surface colors was similar to that reported in a previous
study [17]. With regard to surface color, the score for LCI in normochromic lesions was significantly
higher than those for WLI and BLI-bright. This is because LCI enhances color separation
and depicts red lesions even redder and white lesions whiter, which contributes to
demonstrating large color differences in color contrast between a malignant lesion
and the surrounding area, even if poor tone difference is visualized in the other
two modalities. In contrast, scores for LCI in whitish and reddish lesions were not
significantly higher than those for BLI-bright. This result may be explained by the
imaging principle of BLI-bright, which involves use of light that is easily absorbed
by the hemoglobin in capillaries, thereby clearly delineating blood vessels in the
reddish lesions as brownish in color. In addition, when a whitish lesion is present,
red blood vessels in the surrounding area are also amplified. Interruption of the
underlying mucosal vascular pattern provides an indirect finding to facilitate lesion
recognition.
In our study, we found that LCI was effective in enhancing flat and depressed lesions,
which may not be visible on WLI because of their subtle appearance and the limited
contrast of the surrounding mucosa [18]
[19]. Previous case studies also reported that LCI enhanced early gastric cancer lesions
in patients with these clinical features [12]
[13]. Although BLI-bright provides brighter endoscopic views, the imaging is still criticized
for darkness when used in the stomach [20]. Therefore, BLI-bright probably does not contribute to detection of subtle changes
in early gastric cancer. In contrast, LCI used in the current study yielded very bright
images and enhanced mucosal irregularities, even in the middle to distant views, and
enabled identification of abnormal lesions. The previous study using same visibility
scores also showed that LCI improved endoscopic visibility of colorectal flat lesions
[11]. Thus, the potential benefit of LCI in detecting flat and depressed lesions may
have clinical implications, especially in screening for early gastric cancers.
Early gastric cancer is often overlooked during ordinary endoscopy because it is more
difficult to detect than most other upper gastrointestinal tract cancers [4]. The reported rate of missing superficial gastric cancers is 4.6 % to 25.8 % [21]
[22]
[23]. Failure to detect early gastric cancer during endoscopy may contribute to poor
prognosis [21]. Thus, a variety of premedication and screening techniques are available to ensure
the quality of EGD [24]
[25]. When the distribution of scores was analyzed in our study, the rate of poor visibility
of LCI mode was lower than that of WLI. With regard to the mean score of the four
endoscopists for each lesion, over half of the cases that had poor visibility scores
on WLI showed good visibility scores on LCI. We supposed that LCI was advantageous
in improving diagnosis of subtle, early gastric cancer.
Recently, some studies concerning efficacy of LCI for early gastric cancer have been
reported [26]
[27]. One report showed objective data on color change, but this study can demonstrate
only improved visibility by color deference of each modality [24]. Another report used scores for improvements, which cannot evaluate visibility of
baseline modality [25]. Thus, additional reports are required to build consensus about usefulness of LCI,
which is a newly developed IEE modality. The strong point of this study is evaluation
of each modality including WLI images, which are the basis of comparison. In addition,
this study used the visibility scale, which has been used in several reports of IEE.
Although this report used the subjective indicator, the result clearly showed a sufficient
improvement of LCI for early gastric cancer.
The current study had some limitations. In this study, still images and not videos
were reviewed, and the possibility of selection bias persists. The method of consecutive
reviews of the same image in different modalities during a single session could have
led to another bias. Because this study included only gastric mucosal cancers, further
studies are required to clarify whether LCI is superior to WLI in differentiating
between malignant and nonmalignant lesions. Furthermore, this was a single-center
retrospective study; a multicenter prospective study may be more accurate in assessing
the effectiveness of this new modality for early detection of gastric cancers. Nonetheless,
the current study did provide evidence based on a relatively large number of cases
that LCI significantly improved visibility compared with the other modalities under
the same conditions.
