Introduction
Mucosal healing (MH) represents an important therapeutic goal for patients with ulcerative
colitis (UC), as MH is a predictor of clinical remission, reduced colectomy rates,
and an improved quality of life [1]
[2]
[3]. In recent clinical trials, endoscopic MH was defined as a Mayo endoscopic score
(MES) of 0 or 1 with conventional white-light imaging (WLI) [4]
[5]
[6]. However, it was reported that patients with a MES of 1 had a higher risk of relapse
than those with a score of 0, and the concept of MH should be limited to patients
with a MES of 0. MH is usually diagnosed and confirmed based on endoscopic observation
using WLI [7].
It has recently been reported that histological intestinal inflammation is a valuable
therapeutic target [8]
[9]
[10]
[11]
[12]. In the diagnosis of histological activity in UC, multiple mucosal biopsies are
performed, despite the invasiveness of the procedure, because it has been reported
that endoscopic and histological activity sometimes differ in patients with inflammatory
bowel disease [13]. Thus, an endoscopic method is needed which can predict histological findings precisely
with minimal invasion.
Image-enhanced endoscopy (IEE) facilitates the detection and characterization of colorectal
neoplasms [14]. Recently, linked-color imaging (LCI), a color enhancement function of the LASEREO
system, was developed as a new IEE system. This system enhances red and white areas
making red areas appear redder, and white areas appear brighter. Thus, it is a useful
tool for recognizing color differences in the mucosa, and facilitates the detection
and recognition of colorectal neoplasms [15]
[16]
[17]
[18]
[19].
However, there are few reports on the utility of LCI in the assessment of histological
intestinal inflammation in UC patients. If the mucosal inflammation of patients with
UC could be evaluated by endoscopic observation using LCI, without tissue biopsy,
this would reduce the invasiveness of diagnostic examinations in patients with UC.
The aim of the present study was to evaluate the relationship between endoscopic MH
and histological activity in UC patients as assessed using LCI.
Materials and methods
Patients’ characteristics
In total, 21 UC patients (8 female and 13 male; median age at diagnosis, 39.1 years)
who underwent colonoscopy between August and December 2016 at Kagoshima University
Hospital were enrolled in the present study. The background information on all 21
UC patients who were enrolled in the present study is summarized in [Table 1]. All patients were diagnosed with UC using established endoscopic, radiological,
histological, and clinical criteria. Patients with severe ulcerative colitis, infectious
colitis, including Clostridium difficile infection, a history of total colectomy, or who had colorectal cancer, were excluded
from the present study.
Table 1
Clinical characteristics of the study patients.
|
Total number of patients
|
21
|
|
Sex, M/F
|
13/8
|
|
Age, years
|
47.4 ± 15.4
|
|
Disease duration, years
|
10.5 ± 7.9
|
|
Extent of UC
|
|
|
13
|
|
|
6
|
|
|
2
|
|
Severity of UC
|
|
|
13
|
|
|
8
|
|
|
0
|
|
Clinical course
|
|
|
14
|
|
|
7
|
|
|
0
|
|
Treatment
|
|
|
19
|
|
|
2
|
|
|
1
|
|
|
4
|
|
Biological therapy
|
5
|
UC, ulcerative colitis.
After routine bowel preparation with Moviprep that contained polyethylene glycol 3350,
sodium sulfate, sodium chloride, potassium chloride, sodium ascorbate, and ascorbic
acid, colonoscopy was performed by two experienced endoscopists. A total of 73 biopsied
regions were analyzed in the present study ([Table 1]).
Each study participant underwent colonoscopy for the evaluation of treatment response,
which included the documentation of endoscopic activity and MH, or as a surveillance
procedure. Mayo endoscopic scores (MESs) of 0 or 1 were defined as inactive disease
and MESs of 2 or 3 as active disease [12]. The present study was approved by the Kagoshima University Hospital Institutional
Review Board, and was performed in accordance with the Declaration of Helsinki. Written
informed consent was obtained from all of the patients who participated in the study.
Study design
Conventional colonoscopy was performed with an EC-L600ZW endoscope with the LASEREO
system that consists of VP-4450HD processor and LL-4450 light source (FUJIFILM Co.,
Tokyo, Japan), which is capable of producing light suitable for WLI and LCI.
Each study participant underwent colonoscopy. The endoscopic images with WLI and LCI
were captured from the same position in a row before the biopsy. Then the biopsy was
performed targeting the reddest site in the observed area while capturing the images
during the biopsy for reference of the biopsied point. In each case, the MES was evaluated
under WLI observation by two experienced endoscopists.
The endoscopic images obtained by WLI and LCI before the biopsy were analyzed using
Photoshop CC (Adobe Systems Inc., San Jose, California, USA). To evaluate the color
value, the region of interest (ROI) was determined not only including the biopsied
point but also avoiding any adverse objects such as halation, bleeding, or obvious
submucosal blood vessels with a size of 50 pixels square on each image ([Fig. 1]). The color value of the ROI was defined as the average of each pixel in the ROI
using the International Commission on Illumination 1976 (L*, a* and b*) color space.
The CIE 1976 (L*, a* and b*) color space (CIELAB), which is designed to approximate
human perception [20], is a three-dimensional model composed of a black – white axis (L*), a red – green
axis (a*), and a yellow – blue axis (b*). L* defines brightness, a* defines the red – green
component, and b* defines the yellow – blue component.
Fig. 1 Representative case of ulcerative colitis (UC) with a Mayo endoscopic score of 1
(MES 1). The selection of sample points for the measurement of color value. Left panel:
the square green area, which is the region of interest, shows a conventional image
obtained by white-light imaging (WLI). Right panel: the same region was also observed
by linked-color imaging (LCI).
The color difference (ΔE) between the average color of intestinal histological activity and of inactivity
was calculated according to the following procedure by CIELAB: first, the averages
of the L*, a* and b* color values of each lesion with histological mucosal activity,
defined as L1, a1, b1, and inactivity, defined as L2, a2, b2, respectively, on WLI
and LCI were calculated, then the ΔE between the lesion with histological mucosal activity and inactivity was represented
using the following formula:
ΔE = {(L1 – L2)2 + (a1 – a2)2 + (b1 – b2)2}1 /2 [21]
ΔE was expressed according to the evaluation criterion of the National Bureau of Standards
(NBS) units of color difference ([Table 2]). ΔE was converted to NBS units using the following formula: NBS units = ΔE × 0.92 [22].
Table 2
Evaluation criteria of color difference, based on the National Bureau of Standards
(NBS) unit.
|
NBS unit
|
Evaluation criterion
|
|
0 – 0.5
|
Trace
|
|
0.5 – 1.5
|
Slight
|
|
1.5 – 3.0
|
Noticeable
|
|
3.0 – 6.0
|
Appreciable
|
|
6.0 – 12.0
|
Much
|
|
12.0 –
|
Very much
|
Histological assessment
The histological examinations of all biopsy specimens were performed by an experienced
pathologist who was blinded to the patient’s endoscopic activity. The histological
assessment was based on the Geboes index [23]. The scale included six grades: 0, structural change only; 1, chronic inflammation;
2, neutrophils in the lamina propria; 3, neutrophils in the epithelium; 4, crypt destruction;
and 5, erosion or ulcers. A grade of 3 indicated the presence of neutrophils in the
epithelium, which was representative of acute inflammation and a predictor of relapse
[24]. MH was defined by a Geboes score ≤ 2.
Statistical analyses
The results were analyzed using the Mann – Whitney U test, or Wilcoxon signed-rank test, or Jonckheere – Terpstra trend test, as appropriate.
Correlation coefficients were calculated by a Spearman’s rank correlation analysis.
The discriminatory power of each putative marker was described using the receiver
operating characteristics area under the curve (ROC-AUC). Cutoff values were obtained
from the ROC-AUC analysis. All statistical analyses were conducted using the SPSS
software program (version 15, SPSS Inc., Chicago, Illinois, United States). P values < 0.05 were considered to indicate statistical significance.
Results
Distribution of endoscopic and pathological activity
The distribution of endoscopic activity was as follows: MES 0 in 33 ROIs, MES 1 in
18 ROIs and MES 2 in 22 ROIs at the lesion. No cases of MES 3 were included, as patients
with severe UC were excluded from this study. The distribution of histological activity
was as follows: Geboes 1 in 7 ROIs, Geboes 2 in 35 ROIs, Geboes 3 in 16 ROIs, Geboes
4 in 9 ROIs, and Geboes 5 in 6 ROIs. The relationships between the L*, a* and b* color
values on WLI or LCI and the Geboes score were investigated. The WLI-L, WLI-a, WLI-b
and LCI-L, LCI-a, LCI-b values were indicated as L*, a* and b* color values on WLI
and LCI, respectively. The Jonckheere – Terpstra trend test showed that the color
values of LCI-a, LCI-b, and WLI-b tended to be significantly higher when the Geboes
score was also high (P = 0.003, 0.03, and 0.01, respectively) ([Fig. 2]). No significant differences were observed between the Geboes score and the WLI-L,
WLI-a, or LCI-L color values. Furthermore, only the LCI-a color value was significantly
correlated with the Geboes score using Spearman’s rank correlation analysis (r = 0.36, P < 0.01). No significant differences were observed between the Geboes score and the
WLI-L, WLI-a, WLI-b, LCI-L, or LCI-b color values using Spearman’s rank correlation
analysis.
Fig. 2 Distribution of the L*, a* and b* color values obtained by white-light imaging (WLI)
and linked-color imaging (LCI) for each pathological activity (Geboes score). The
color values of LCI-a, LCI-b, and WLI-b tended to be significantly higher, when the
Geboes score was also high. Each symbol represents one patient. (Jonckheere-Terpstra
trend test)
Comparison of the L*, a* and b* color values of WLI or LCI in patients with UC
The average WLI-L, WLI-a, WLI-b, LCI-L, LCI-a, and LCI-b color values in UC patients
with mucosal activity were 52.4, 40.6, 40.9, 57.5, 29.1, and 23.4, respectively. The
average WLI-L, WLI-a, WLI-b, LCI-L, LCI-a, and LCI-b color values in UC patients with
MH were 52.5, 38.7, 38.2, 56.9, 23.0, and 18.8, respectively. The differences in the
L*, a* and b* color values on WLI or LCI according to mucosal inflammation were analyzed.
The average LCI-a and LCI-b values of patients with high mucosal activity were statistically
significantly higher than those of patients with MH, while there was no significant
difference in the other color values between patients with high mucosal activity and
those with MH ([Fig. 3]).
Fig. 3 Comparison of the L*, a* and b* color values obtained by white-light imaging (WLI)
and linked-color imaging (LCI) in patients with histologically inactive ulcerative
colitis (UC) versus those with histologically active UC. The L*, a* and b* color values
of UC patients with histologically active disease were higher than those of patients
with histologically inactive disease. In particular, the median LCI-a and LCI-b levels
in patients with histologically active UC were significantly higher than those in
patients with histologically inactive UC. Each symbol represents one patient. (Mann-Whitney
U test)
The color difference (ΔE) between the average color of intestinal histological activity and of inactivity
was 3.1 on WLI and 7.1 on LCI, which was a statistically significant difference (P < 0.05). The assessment of the color difference-based NBS unit indicated that it
was “much” with LCI, whereas it was “appreciable” with WLI ([Table 2]).
Diagnostic utility of the L*, a* and b* color values on WLI or LCI and the Geboes
score of intestinal histological activity
The WLI-a, WLI-b, LCI-a, and LCI-b color values were investigated to evaluate their
utility in distinguishing between histological mucosal activity and inactivity. The
ROC-AUCs for WLI-a and WLI-b did not differ to a statistically significant extent
(P = 0.27 and P = 0.06, respectively). In contrast, the ROC-AUCs for LCI-a and LCI-b were 0.67 and
0.66, respectively (P = 0.01 and P = 0.02) ([Fig. 4]). The ROC-AUC analysis revealed that an LCI-a of 23.2 and an LCI-b of 19.9 were
the optimum cutoff values for discriminating between the histologically active mucosa
and histological MH (sensitivity, 74.2 %; specificity, 57.1 %; positive predictive
value, 56.1 %; negative predictive value, 75 %; accuracy, 64.4 %; LCI-a and LCI-b
had exactly the same values). In addition, the ROC-AUC analysis revealed that an WLI-a
of 38.8 and an WLI-b of 38.5 were the optimum cutoff values for discriminating between
the histologically active mucosa and histological MH. The sensitivity of WLI-a and
WLI-b were 64.5 % and 71.0 %; specificity, 57.1 % and 54.8 %; positive predictive
value, 52.6 % and 53.7 %; negative predictive value, 68.6 % and 71.9 %; accuracy,
60.2 % and 61.7 %, respectively ([Table 3]).
Fig. 4 The receiver operating characteristic (ROC) curves of WLI-a, WLI-b, LCI-a, and LCI-b
for the diagnosis of mucosal healing of ulcerative colitis (UC). The ROC curves for
each white-light imaging (WLI) and linked-color imaging (LCI) color value were obtained
by plotting sensitivity versus 1 – specificity. The ROC areas under the curve for
WLI-a, WLI-b, LCI-a, and LCI-b were 0.58, 0.63, 0.67, and 0.66, respectively (P = 0.3, 0.06, 0.01, and 0.02, respectively).
Table 3
Receiver operating characteristics analysis for the diagnosis of pathological mucosal
healing in ulcerative colitis.
|
ROC-AUC
|
Sensitivity, %
|
Specificity, %
|
PPV, %
|
NPV, %
|
Accuracy, %
|
P value
|
|
WLI-a
|
0.58
|
64.5
|
57.1
|
52.6
|
68.6
|
60.2
|
0.3
|
|
WLI-b
|
0.63
|
71.0
|
54.8
|
53.7
|
71.9
|
61.7
|
0.06
|
|
LCI-a
|
0.67
|
74.2
|
57.1
|
56.1
|
75
|
64.4
|
0.01
|
|
LCI-b
|
0.66
|
74.2
|
57.1
|
56.1
|
75
|
64.4
|
0.02
|
WLI, white-light image; LCI, linked-color image; ROC-AUC, receiver operating characteristic
area under the curve; PPV, positive predictive value; NPV, negative predictive value.
Discussion
This is the first study to investigate the correlation between the color values detected
by the L*, a* and b* color values on WLI and LCI and histological assessment based
on the Geboes index in patients with UC. The findings of endoscopic observations by
LCI-a and LCI-b showed a significant difference between mucosal inflammation and non-inflammation,
and LCI-a was significantly correlated with the histological mucosal inflammation
score. The color difference between inflammation and non-inflammation was more than
twice the average difference for LCI compared with WLI. Therefore, LCI would be more
useful than WLI for inflammatory diagnosis of UC.
The endoscopic color values observed by LCI were correlated with microscopic/histological
mucosal activity. We evaluated the color difference between pathological inflammation
and remission. Since microscopic/histological mucosal activity is reported to be a
predictor of relapse in UC patients [8]
[9]
[10]
[11]
[12], the endoscopic color difference on LCI may be used to diagnose mucosal activity
in addition to a biopsy.
Recent clinical trials of endoscopic criteria have used the Mayo endoscopic score
(MES), which is easy to apply in clinical trials and clinical practice. The Mayo endoscopic
scores are defined as follows: MES 0, normal mucosa or inactive disease; MES 1, mild
activity (erythema, decreased vascular pattern, mild friability); MES 2, moderate
activity (marked erythema, lack of vascular pattern, friability, erosions); MES 3,
severe activity (spontaneous bleeding, large ulcerations). Conventional endoscopy
with WLI is normally performed to evaluate intestinal activity in UC. However, the
findings of conventional endoscopy using WLI do not always correlate with the histological
findings [25]
[26]. Indeed, in this study, 28 % of patients with MES 1 showed histological activity
(data not shown) ([Fig. 4]). Thus, the MES, as determined by WLI, was insufficient for the diagnosis of histological
activity, especially with regard to the diagnosis of mild inflammatory activity [12]. LCI observation can emphasize redness (i. e., reddish areas indicate erythema),
the color difference presumably increasing between active and nonactive lesions, and
it is easi to recognize lesions showing erythema. We therefore focused on the relationship
between the colors observed by endoscopy using LCI and tissue inflammation.
Uchiyama et al. recently reported that LCI could be used both to evaluate colonic
mucosal inflammation and predict the outcome in UC patients. They classified endoscopic
findings on LCI based on redness with or without visible vessels and reported that
the endoscopic LCI classification was correlated with the LCI index [27]. We also investigated the color values of the images using CIELAB, which is designed
to approximate human perception [20]. It was reported that the color of mucosa on WLI and LCI consists of two dimensions
between a positive a* value, indicated in red, and a positive b* value, indicated
in yellow [21]. The observation of LCI-a and LCI-b emphasized the redness and yellowness of the
lesion and highlighted the areas of inflammation in this study. The color difference,
including LCI-a and LCI-b, under LCI observation was significantly higher than that
under WLI. We therefore demonstrated that LCI observation was useful for detecting
and visualizing mucosal inflammation.
Previous studies have reported the utility of endoscopic observation using LCI in
screening for gastrointestinal tumors, as this approach increases the discriminatory
power of the red color tone [22]
[28]
[29]
[30]. In the present study, LCI was also able to improve the visibility of inflammation
in the red region of the colonic mucosa in patients with UC and provided brighter
images, allowing for both the lesion and the background to be visualized more clearly
than with other IEE systems such as narrow-band imaging or blue laser imaging, in
which the background appears dark. For these reasons, LCI is suitable for use in screening
for mucosal inflammation.
Many histological scores for assessing disease activity in UC have been described
since the 1950 s, although none have been fully validated. The Geboes score is commonly
used as a histological index of disease activity in UC. This index assesses the following
features: architectural change, lamina propria neutrophils and eosinophils, neutrophils
in epithelium, crypt destruction, and erosion or ulceration. Given that a previous
study reported this score to be the best validated of all available scores [31], we evaluated the mucosal inflammation of biopsied tissue using the Geboes score.
We showed that the endoscopic color value on LCI was correlated with microscopic/histological
mucosal activity; as such, LCI would be useful not only for the detection of inflammation
but also for the characterization of mucosal inflammation. However, the color value
of the ROI was only calculated after colonoscopy for clinical application. It would
be more useful if the system could immediately show the color values of the ROI to
predict mucosal inflammation during colonoscopy. Furthermore, evaluation of the ROI
was only performed for a single point; another tool able to analyze the color value
of a large area is therefore needed. In addition, since the accuracy of the LCI-a
diagnosis of mucosal inflammation at the ROI only showed mild correlation with the
pathological diagnosis, the LCI diagnosis cannot completely replace pathological diagnosis
at present. Therefore, a new classification system using LCI is needed as an alternative
to existing classification systems using conventional WLI.
Only the color value examination findings by WLI or LCI showed a low diagnosis rate.
We have found that, even in patients without UC, the color value of the colonic mucosa-based
endoscopic findings showed individual variations regardless of whether mucosal inflammation
was present or not. As a result, individual differences in mucosal color was one of
the factors giving increased color values. Therefore, there are limitations when attempting
to diagnose mucosal inflammation based on the absolute color values alone. However,
since a correlation was observed between the color value of LCI-a and the degree of
pathological inflammation, the color value is therefore considered to be an indicator
for comprehensively distinguishing mucosal inflammation. The color value by LCI was
significantly correlated with mucosal inflammation compared to WLI. It would also
be useful if such a system based on the combination of color values by LCI and other
parameters, such as any visible vessels, or ulceration lesions similar to the Mayo
classification, could diagnose mucosal activity with high accuracy. It therefore appears
to be promising to compare the findings of LCI observations with those of conventional
WLI observations.
The present study has several limitations. Although this was a prospective study,
it was performed at a single center as a pilot study, and the population was relatively
small. It will be difficult to perform evaluations based on the L*, a* and b* color
values in cases with ulcerative lesions that may be coated with purulent mucus or
have inflammatory polyps. In addition, the L*, a* and b* color values changed slightly
depending on the observation distance. The color value of the ROI could not be obtained
quickly during endoscopic examination; thus, a system must be established in the future
to obtain the color value of the ROI. Finally, because the ROI was a tiny spot, the
color value was affected by the color of small structures, such as capillaries at
the mucosal surface and partially hidden vessels in the submucosal layer; we should
therefore investigate the method to decide the suitable ROI which is able to estimate
mucosal inflammation most efficiently.
In conclusion, we investigated the relationship between histological activity and
color value using LCI and WLI. The LCI-a and LCI-b color values were useful parameters
for diagnosing histological mucosal activity, and LCI observation was useful for the
visualization and evaluation of mucosal inflammation in UC.
