diagnostics

Article
Improved Image Analysis for Measuring Gastric Ulcer Index in
Animal Models and Clinical Diagnostic Data
Chi-Yeol Yoo 1 , Hyeong-U Son 1 , Sung-kook Kim 2 , Si-Oh Kim 3 and Sang-Han Lee 1,2, *

1 Department of Food Science and Biotechnology, Graduate School of Kyungpook National University,
Daegu 41566, Korea; [email protected] (C.-Y.Y.); [email protected] (H.-U.S.)
2 Department of Gastroenterology & Hepatology, Kyungpook National University Hospital,
Daegu 41944, Korea; [email protected]
3 Department of Anesthesiology, Kyungpook National University Chilgok Hospital, Daegu 41404, Korea;
[email protected]
* Correspondence: [email protected]

Abstract: Gastric ulcers are one of the most common gastrointestinal diseases. In this study, as an
attempt to reduce the minimal error in clinical observations during the diagnosis of gastric ulcers,
the applicability of improved ImageJ analysis (IA) was investigated by comparing the results of
animal experiments and clinical data. As a result, IA exhibited a significantly improved potential
for determining the ulcer index (UI) of clinical data sheets compared to those rated directly by
conventional clinical observation (CCO). This indicated that IA enhanced the reproducibility of the
measurement of gastric UI using a Bland–Altman plot, resulting in a reduced deviation of each UI
value. In addition, it was confirmed that errors in gastric UI decisions can be reduced by adjusting
RGB values in diagnostic clinical data (i.e., adjusting to 100 is relatively better than adjusting to
50 or 200). Together, these results suggest that the new enhanced IA could be compatible with
novel applications for measuring and evaluating gastric ulcers in clinical settings, meaning that the
Citation: Yoo, C.-Y.; Son, H.-U.; Kim, developed method could be used not only as an auxiliary tool for CCO, but also as a pipeline for
S.-k.; Kim, S.-O.; Lee, S.-H. Improved ulcer diagnosis.
Image Analysis for Measuring
Gastric Ulcer Index in Animal Keywords: gastric ulcer; image analysis; ImageJ; conventional clinical observation; ulcer index
Models and Clinical Diagnostic Data.
Diagnostics 2022, 12, 1233. https://
doi.org/10.3390/diagnostics12051233

Academic Editors: Anastasios 1. Introduction

Koulaouzidis and Consolato M. Sergi Gastric ulcers have been established as major gastrointestinal tract diseases which
affect 10% of the world’s population to varying degrees, have a high incidence rate, induce
Received: 25 April 2022
Accepted: 11 May 2022
serious complications, and are caused by various factors. Gastric ulcers can be caused
Published: 14 May 2022
by Helicobacter pylori infection [1], stress, smoking, excessive use of nonsteroidal anti-
inflammatory drugs, and excessive alcohol consumption [2]. The most common causal
Publisher’s Note: MDPI stays neutral factor is alcohol, which can induce gastric tissue lesions, such as gastritis, gastric ulcers, and
with regard to jurisdictional claims in
even gastric cancer. In vivo studies using ethanol induction, indomethacin administration,
published maps and institutional affil-
and water immersion-restraint (WIRE) stress-induced models have established effective
iations.
methods to determine the protective effects against gastric lesions [3]. To confirm the
expression of the MUC5 and MMPs genes, which protect the gastric mucosa, in vivo
experiments were conducted using a mouse model [4,5]. The characteristics of gastric ulcers
Copyright: © 2022 by the authors.
are well-exposed in the damaged feature of the stomach. Therefore, the state of the stomach
Licensee MDPI, Basel, Switzerland. is commonly evaluated by an endoscope. In previous studies using a gastric ulcer mouse
This article is an open access article model, a data-standardized data-scoring technique was used to determine the gastric ulcer
distributed under the terms and index (UI). The gastric UI is well-known for showing standardized scoring criteria, however,
conditions of the Creative Commons it has problems. First, the patterns of ulcer damage caused by artificial means are not the
Attribution (CC BY) license (https:// same. Several types of animal models of gastric ulcers, such as ethanol-, indomethacin-, or
creativecommons.org/licenses/by/ stress-induced models (including the WIRE model), have been established [6]. The clinical
4.0/). diagnosis revealed that the shapes of the damaged areas are totally different. Second,

Diagnostics 2022, 12, 1233. https://doi.org/10.3390/diagnostics12051233 https://www.mdpi.com/journal/diagnostics

Diagnostics 2022, 12, 1233 2 of 15

gastric ulcer scoring is only two-dimensional (gastric ulcer surface). Even if the damaged
areas are the same, their depths are not. It is difficult to determine by existing scoring
which is more damaged: slight erythema with an extensive area or severe erythema with a
small area. In various in vivo experiments, some parts of the indicator cannot be evaluated
by a quantitative and automatic instrument, such as a spectrophotometer. For example,
although the skin or organ erythema is generally calculated by a caliper [7], this method is
possibly affected by the subjectivity of the researcher, leading to low reproducibility and
objectivity. To overcome this, analyses with specific software programs, such as ImageJ (a
freeware offered from the NIH homepage), photoshop, and MATLAB, are suggested as
an alternative method. These computer programs have various benefits with their high
accuracy and broad application for biological analysis [8]. Nonetheless, photography has
to be supplemented by constantly taking photos with optimized factors, such as ISO value
and controlled light source (usually daylight 50~60), because the photo data are not only
scientific but also affected by individual differences [9].
However, the ImageJ analysis has been applied for various agricultural, biochemical,
and medical purposes. The use of the ImageJ analysis to detect the “woody breast” condi-
tion in broiler carcasses has allowed processors to sort broiler carcasses according to woody
breast severity among commercial in-line vision grading systems [10]. Furthermore, the Im-
ageJ method is experience-independent, i.e., it can be executed without expert participation
in criminological, anthropology, and human anatomy [11]. According to a previous study,
by calculating all histological phenotypic parameters—such as mean seminiferous tubule
diameter, mean seminiferous lumen diameter, epithelial height, tubular area, and luminal
area—by comparing treated and untreated groups using ImageJ, it could be concluded that
astaxanthin has beneficial effects [12].
For these reasons, we modified the image analysis to compare the gastric lesions with
two types of induced models and clinical diagnostic data. Furthermore, we found that the
current method is more efficient, sensitive, and reproducible for calculating the current
ulcer index.

2. Materials and Methods

2.1. Mice and Care
The specific-pathogen-free 7-week-old male BALB/C mice were purchased at 20~22 g
of weight from Samtaco Korea (Osan, Korea). The mice were treated under conditions
of 12 h of light per day, 22 ± 1 ◦ C temperature, and 50% ± 5% humidity. The mice were
allowed to acclimate in the laboratory environment for 1 week prior to the experiment. Ani-
mal care was conducted according to the guidelines of the Kyungpook National University
Animal Care Committee (approval No. KNU-2014-0069).

2.2. Gastric Ulcer Mouse Models

Ethanol was used as an inducer of acute gastric ulcers. The mice were fasted for 24 h
before the experiment. All treatment groups were orally administered with 200 µL of
80% ethanol per approximately 25 g of mouse weight (concentration: 8 mL/kg), and the
negative control (no treatment (NT) group) group was administrated with saline 30 min
before the positive control treatment, as described previously [13–18]. The mice were
separated into four groups of seven mice each, as follows: NT (no induction of ethanol
and saline treatment), saline (ethanol induction and saline treatment), omeprazole (ethanol
induction and omeprazole treatment at 3 mg/kg), and gastric sample (ethanol induction
and aqueous fraction of Opuntia humifusa treatment at 100 mg/kg) groups. For the positive
control, 3 mg/kg of omeprazole was orally administrated [13]. Over time, the mice in the
treatment groups had drunken symptoms, but no mice died. After 30 min of treatment,
euthanasia was performed using carbon dioxide, and the stomach tissues of mice were
immediately extracted. The stomach was cut along its greater curvature, and saline was
applied to the inner surface area. Then, erythema on the ethanol-induced group (control)
was checked, and pictures were taken. For the water immersion-restraint stress (WIRE)
Diagnostics 2022, 12, 1233 3 of 15

model, a previously described protocol with a slight modification was applied to 5 mice
per group [3].

2.3. General Analysis of Gastric Ulcer

To compare the general and modified analysis methods, they were first observed and
evaluated using a criterion from previous papers [3,14–17]. These methods are the same as
that of the panel with professional knowledge who observed the inner gastric tissues of the
mice using a caliper and scored them based on the criteria (Figures 1 and 2). The criteria
for gastric ulceration were: 1 point was assigned for small, round, hemorrhagic corrosions,
2 points were assigned when the length of hemorrhagic corrosions was <1 mm, 3 points
were assigned when the length was 1–2 mm, 4 points were assigned when the length was
2–3 mm, and 5 points were assigned when the length was >4 mm. Then, the scores were
added and doubled when the width of the corrosions was >1 mm. A panel of eight trained
researchers participated to determine the UI, and its general measurement based on the
length of the stripe erythema taken by a researcher was considered to be not a quantitative
measurement because of subjectivity (Figure 2A). We designated this analysis method to
conventional clinical observation (CCO).

2.4. Image Analysis of Gastric Ulcer

To take consistent images, the light source and digital raw file were controlled
(Figure 2B). To avoid the effect of sunlight and interior lighting, the photos were taken in a
dark room, and a strobe (Hyundai Fomax strobe light E400, Seoul, Korea), a Nikon digital
camera D5100 (Nikon Co., Tokyo, Japan), and AF-S Micro NIKKOR 40 mm f/2.8 G lens
(Nikon Co., Tokyo, Japan) with a Nikon ML-L3 wireless remote control were used. Initially,
a color image scale (QP card 203, QP card AB, Helsingborg, Sweden) was photographed
after the picture target was placed in a mini studio with a diffuser and background fab-
ric cover [18]. The color image scale is a useful tool to control the color profile for the
color character reproducibility and color restoration [19]. Using the same conditions as
in the photographed color image scale data, the gastric organs were photographed using
the methodology stated in Section 2.2. Gastric ulcer mouse models were placed in the
mini studio and photographed. The photographs of the gastric organs were saved as
raw files, and their color was adjusted using QP calibration v1.21 software (QP card AB,
Helsingborg, Sweden).

2.5. Comparison of Optical Measurement and Image Analysis

To evaluate the accuracy of the ImageJ analysis program on visible inflammatory
damages, the spectrophotometer and digital camera data information were compared.
Since gastric organ damages generally revealed the erythema through gastric mucosal
blood flow [20], this value was compared with real blood concentrations. The blood to use a
standard erythema was collected from mouse venipuncture. The blood was serially diluted
with saline into 50%, 25%, 12.5%, 6.25%, 3.125%, and 0% concentrations in a 6-well plate. A
spectrophotometer was used for various assays because of its many advantages, such as
high precision, high sensitivity, minimal sample requirements, nonconducting feature, and
dynamic range [21]. Therefore, the optical density based on a spectrophotometer at 520 nm
was accurate and was used as a standard.
Diagnostics 2022, 12, x FOR PEER REVIEW
Diagnostics 2022, 12, 1233
4 of 15
4 of 15

Figure
Figure1.1.Application
Applicationof of shielding and3D
shielding and 3Dpicture
picture onon gastric
gastric ulcer
ulcer image image analysis.
analysis. (A) The(A)digital
The digital
camera
camerawas shielded
was shieldedwith
withaacellophane
cellophanefilm
film (green, blue, and
(green, blue, and red),
red),and
andthe
thecolor
colormarker
markerpicture
picture was
taken.
was Each
taken.color
Eachwascolorhidden with the
was hidden withsame colorcolor
the same shielding afterafter
shielding photographing.
photographing. (B)(B)
TheThe
ethanol-
induced gastric organ
ethanol-induced in aorgan
gastric mousein amodel
mousewas model photographed with red
was photographed withcellophane film film
red cellophane andand
showed a
customized photographphotograph
showed a customized file, whichfile,
waswhich
only wasdetermined with a red
only determined withcolor.
a red(C) The(C)
color. color
Themarker
color was
photographed by a molecular
marker was photographed by aimager.
molecularRed, green,
imager. Red,and blue
green, andwithout whitewhite
blue without indicate the the
indicate height of
color density.
height of color(D) As the
density. (D)color
As thedensity increases,
color density the the
increases, height becomes
height becomeshigher in the
higher in the3D
3Dpicture.
picture. (E)
The
(E)ethanol-induced
The ethanol-inducederythema
erythema damage
damagein inaa mouse
mouse isisdisplayed
displayed as as
a 3Da 3D image
image ofgastric
of the the gastric
organ.organ.
iagnostics 2022, 12, x FOR
Diagnostics 2022,PEER REVIEW
12, 1233 5 of 15
5 of 15

Figure 2.Figure
Application of ImageJ
2. Application analysisanalysis
of ImageJ on gastric ulcers ulcers
on gastric by comparing the conventional
by comparing clinical
the conventional clinical
observation (CCO) method.
observation (A,B) General
(CCO) method. (A,B) description of CCO analysis
General description (A) and (A)
of CCO analysis ImageJ
and analysis (B)
ImageJ analysis
during determination of the gastric ulcer index. (C–E) Photographs of ethanol-induced gastric tissue
(C) are shown to score the gastric ulcer index by CCO (D) or by ImageJ analysis (E). (F) Classical
photographs of gastric organ in an ethanol-induced gastric ulcer model. In the model, the damaged
zone was classified as slight or severe on the surface of the inner gastric organ by the ImageJ analysis
Diagnostics 2022, 12, 1233 6 of 15

(B) during determination of the gastric ulcer index. (C–E) Photographs of ethanol-induced gas-
tric tissue (C) are shown to score the gastric ulcer index by CCO (D) or by ImageJ analysis (E).
(F) Classical photographs of gastric organ in an ethanol-induced gastric ulcer model. In the model,
the damaged zone was classified as slight or severe on the surface of the inner gastric organ by
the ImageJ analysis (G). (H) Histogram of the counted pixel of erythema from the photograph. The
pixels were divided according to the red values in the photograph. (I) Determination of gastric ulcer
index by ImageJ analysis. EtOH, ethanol; Ome, Omeprazole. # p < 0.05 and * p < 0.01, versus NT
(not treated).

2.6. Application into Multidirectional Evaluation via Image Processing

To calculate with minimum background of color information, cellophane films were
used (Figure 2A,B). This process corrected the background for green, red, and blue. The
digital camera lens was covered with green, blue, or red cellophane films, and pictures
were taken under the same condition (Figure 2B).

2.7. Image Analysis into Clinical Diagnostic Data

The clinical features were provided from Kyungpook National University Hospital,
Korea. Pictures of randomly selected gastric ulcers were used. Their photographs were
taken using an Olympus HD260 CCD (charge-coupled device) (Olympus Co., Tokyo, Japan).
In total, 140 photographs were analyzed by the ImageJ software. They were assessed by
the panel with the same protocol as the mouse model (KNU-2019-0162).

2.8. Statistical Analysis

All statistical analyses were performed using SPSS version 20.0 (SPSS Institute, Chicago,
IL, USA). The data are presented as means ± standard deviations (SD). Statistical analyses
were performed using one-way analysis of variance and Tukey’s test for post hoc compari-
son. p-values < 0.05 were considered statistically significant. To calculate the reproducibility
of the analysis methods, a Bland–Altman plot was created [22].

3. Results
3.1. Application into Multidirectional Evaluation via Image Processing
Some information was excluded from the data analysis as it interrupted the measuring
factor in some cases. To analyze the tissues/photograph, large amounts of data were first
taken; then, specific information, such as the RGB value, pixel count, or brightness, was
selected from the image files. In the analysis of erythema, a low-red value was possibly an
interrupting factor as basic organ tissues with no erythematous region also had a red value
and can be miscalculated when this region is included in the data. Therefore, covering
the low-red value can be a challenge. To cover the RGB information, a cellophane film
was used due to its low cost and easy use. The pictures were displayed, blinding each
color marker by a green, blue, and red cellophane film in front of the camera lens. The
results showed that each blinded color was not exhibited (Figure 1A). Thus, the mouse
gastric ulcer tissues with the red cellophane film were adapted, and it was found that the
low-red value pixel was unified with the background (Figure 1B). The photograph data
of the erythematous lesions were calculated using the ImageJ software, showing that the
erythematous lesions had a higher value than the background with the cellophane film.

3.2. Measurement of Gastric Ulcer with an Improved Method Using ImageJ Analysis
After euthanasia and opening the inner area of the stomach, severe spots and linear-
shaped erythema were observed in the ethanol only group (Figure 2F, second image).
However, in the omeprazole-treated group, the gastric lesions in the positive control group
were significantly recovered (Figure 2F, third image). To quantitatively determine the
gastric ulceration, ImageJ analysis was used to quantify the ulcer index using ImageJ
Diagnostics 2022, 12, 1233 7 of 15

software. First, the photographs were processed, and the damaged zone was divided into
white or black (Figure 2G, second and third images). The bisection of the damage degree
was adjusted based on the medical doctors’ suggestions or according to experimental
differentiation. Thus, the separation of the damaged zone in the photograph (Figure 2G)
was easier and more useful than that of the CCO (Figure 2F). The results showed that the
ethanol-untreated control group (Figure 2F, first image) rarely indicated the erythematous
zone, whereas the ethanol-treated group (Figure 2F, second image) substantially revealed it;
however, that of the omeprazole-treated group (Figure 2F, third image) decreased. Second,
the pixel number of erythematous zones in the photographs was counted (Figure 2G).
Since erythema expresses a red color, only the red value was set as a standard and divided
on the x-axis [23]; then, the number of red pixels was counted on the y-axis (Figure 2H).
Similar to the photograph data (Figure 2F,G), the ethanol-untreated control group had a low
pixel number that indicated erythema, whereas the ethanol-treated group had the highest
(Figure 2H). Third, erythematous parts were counted, and gastric UI was determined
relative to the ethanol-untreated control group, adjusting it to 100% (Figure 1).

3.3. ImageJ Analysis Enhanced the Reproducibility of the Gastric UI Measurement

The eye observation analysis (designated as CCO) technique can possibly make a
statistical error during gastric UI observation. Thus, it was compared with the ImageJ
analysis of gastric UI. Eight trained professionals hand-operated the two methods. Since
the ethanol-induced control group was the most damaged and had a high UI, only the
ethanol-treated group was compared with all mouse groups. Figure 3A shows the UI of
five mice, whereas SD signifies the deviation of the eight professionals. The SD value of the
ImageJ analysis was more acceptable than that of the CCO analysis. The Bland–Altman plot,
which is a useful statistical analysis used to determine the reproducibility of the experiment,
demonstrated that the eye observation had a significantly lower reproducibility than the
ImageJ analysis, which also clearly classified the mouse entities (Figure 3C,D). Therefore,
it was assumed that the ImageJ analysis was also reasonable for other methods. WIRE
was tested in stress-induced mouse models to determine whether the reproducibility of
the eye observation and ImageJ analysis had the same pattern in the ethanol model. Since
the symptoms of gastric ulcers in the WIRE model, such as the erythematous color, were
different than the ethanol model, the technical protocol for using the ImageJ analysis
was modified. The ImageJ analysis of the reproducibility of the WIRE model was more
important than the CCO of the ethanol model (Figure 3B,E,F).

3.4. Determination of the Erythema Value by Image Analysis Using a Blood Standard
Since gastric ulcers were revealed with inflammation, severe redness, and mucosal
blood flow, their degree was affected by erythematic values, which can be explained by
the red color values [24]. To establish a real erythematic value, mouse blood was added
to a 6-well plate (Figure 4A) and was measured at 520 nm using a spectrophotometer
(Figure 4B). Using a spectrophotometer was a patently elaborate analysis; therefore, the
method presented here has the advantage of quantifying the degree and evaluating the
distribution of erythema (pigmentation), with minimal interference from pigmentation
(erythema). Unlike the color coordinates, such as L*a*b* in the Commission Internationale
de l’Eclairage color space, EI and MI are not the indicators in evaluating ‘color’ [25]. The
erythema standard was calculated using a spectrophotometer and ImageJ (Figure 4C).
Comparing the spectrophotometer and ImageJ analysis of the erythema standard, both
methods were fairly precise, with an R2 of 0.9951 and 0.9905, respectively (Figure 4B,C).
To divide them based on damage degrees, the formula of the erythema standard based
on ImageJ analysis (y = 0.4432x − 0.0282) was applied to the gastric tissue photographs
and classified as 0–80% of blood-concentrated erythematic values in minimum values.
This figure shows that the severe part of erythema, which was also swollen, was balanced
at ~60%.
R REVIEW 8 of 15
Diagnostics 2022, 12, 1233 8 of 15

Figure 3. Comparison of gastric ulcer scores observed by CCO and IA. (A,B) The SD bar indicates the
Figure 3. Comparison of gastric ulcer scores observed by CCO and IA. (A,B) The SD bar indicates
differences among the eight professionals for individual mouse ulcers. To compare the reproducibility
the differences among theaccording
eight professionals fordifferent
to the same object and individual
panels, a mouse ulcers.
Bland–Altman Tocreated.
plot was compare theshows
The y-axis repro-
ducibility according to thethesame object
percentage and
value different
difference panels,
between a Bland–Altman
the professionals plot was
and all professionals’ created.
standard The
deviation.
The x-axis shows the gastric ulcer index of CCO (C,E) and image analysis
y-axis shows the percentage value difference between the professionals and all professionals’ stand- (D,F). The lesser deviation
(D,E) of IA means a more reliable gastric UI than that of CCO was retrieved (C,E). CCO, conventional
ard deviation. The x-axis shows the gastric ulcer index of CCO (C,E) and image analysis (D,F). The
clinical observation; IA, ImageJ analysis; WIRS, water immersion-restraint stress.
lesser deviation (D,E) of IA means a more reliable gastric UI than that of CCO was retrieved (C,E).
CCO, conventional clinical observation; IA, ImageJ analysis; WIRS, water immersion-restraint
stress.

3.4. Determination of the Erythema Value by Image Analysis Using a Blood Standard
 ImageJ analysis (y = 0.4432x − 0.0282) was applied to the gastric tissue photographs and
classified as 0–80% of blood-concentrated erythematic values in minimum values. This
figure shows that the severe part of erythema, which was also swollen, was balanced
Diagnostics 2022, 12, 1233 9 of 15
at
~60%.

Figure 4. ImageJ analysis using a phased erythema standard. (A) Photograph of diluted blood on
Figure 4. ImageJ analysis
6-wellusing a phased
plate. The blood was erythema
collected fromstandard.
a mouse, and(A) Photograph
the blood of were
concentrations diluted blood on 6-
0%, 3.125%,
well plate. The blood6.25%,
was 12.5%,
collected from
25%, and 50%. aThe
mouse,
erythemaand the was
standard blood concentrations
calculated were (B)
by spectrophotometer 0%, and3.125%,
6.25%, 12.5%, 25%, and 50%.
ImageJ The(C).
analysis erythema
The erythemastandard was
range on the calculated
photograph by spectrophotometer
was adjusted as a blood. (B) and
ImageJ analysis (C). The erythemainrange
3.5. Application Clinical on the photograph
Diagnosis was adjusted
by Measuring Gastric UI as a blood.
Clinical gastric features were obtained from a 140-image dataset. The clinical diagnosis
was determined by selected panels at Kyungpook National University, Korea, and gastric
3.5. Application in Clinical Diagnosis by Measuring Gastric UI
ulcer photographs were classified as alcohol-induced, stress-induced, and drug side effect.
Clinical gastricSince
features
clinicalwere obtained
photographing wasfrom a 140-image
not available to controldataset. Thewith
the condition clinical diagno-
an exact
light and distance, each one’s optional condition was followed on the ImageJ software
sis was determined(Figures
by selected panels at Kyungpook National University, Korea, and gas-
5 and 6). Gastric ulcers were assessed and randomly scored using the ImageJ
tric ulcer photographs were
analysis classified
program as alcohol-induced,
for approximately 140 patients (Figure stress-induced, and drug
5A–E). The results exhibited a side
effect. Since clinicalhighly similar pattern to that obtained in a previous ImageJ analysis [26], an AI-based
photographing was not available to control the condition with an ex-
deep learning program, although the model did not match that obtained by gastric ulcer
act light and distance, each The
specialists. one’s
modeloptional
diagnosiscondition
gastric ulcers was
at the followed
same level ason the ImageJ
residents software
and provides
primary treatment,
(Figures 5 and 6). Gastric ulcers suchwere as anti-inflammatory
assessed and drugs. randomlyFurthermore,
scoredthe ImageJ
usinganalysis
the ImageJ
was more effective in diagnosing gastric ulcers when used by an internal medicine doctor
analysis program for approximately
alone. 140that
These findings indicate patients
the ImageJ(Figure 5A–E).
analysis exhibits The results
excellent exhibited a
results compared
highly similar pattern toscored
to those that directly
obtained by thein a previous
panels that measureImageJ
gastric UI.analysis [26], an AI-based

deep learning program, although the model did not match that obtained by gastric ulcer
specialists. The model diagnosis gastric ulcers at the same level as residents and provides
primary treatment, such as anti-inflammatory drugs. Furthermore, the ImageJ analysis
was more effective in diagnosing gastric ulcers when used by an internal medicine doctor
FOR PEER REVIEW 10 of 15
Diagnostics 2022, 12, 1233 10 of 15

Figure 5. Measurement of gastric

Figure ulcers from
5. Measurement clinical
of gastric specimens/images.
ulcers A 140-image
from clinical specimens/images. dataset ofdataset
A 140-image
gastric ulcers was analyzed
of gastric ulcers was analyzed by the ImageJ software, and a reader study was using
by the ImageJ software, and a reader study was performed the using
performed
140 images (A). Several
the types of gastric
140 images symptoms
(A). Several types ofwere
gastricobserved
symptomsand wereclassified
observed andintoclassified
severe into
ery-severe
thematous area with erythematous
red spot area area(B),
withslight erythema
red spot (C), erythema
area (B), slight H. pylori(C),
infection
H. pylori(D), and(D),
infection defense
and defense
against ulcer with gastric juice
against (E)with
ulcer as the white
gastric area.
juice (F)the
(E) as Gastric
white ulcer measurements
area. (F) in clinical spec-
Gastric ulcer measurements in clinical
imens were analyzed specimens
with ImageJ wereby
analyzed with RGB
adjusting ImageJvalues
by adjusting RGB 0values
between between 0 and 225.
and 225.
Diagnostics 2022, 12, x FOR PEER REVIEW 11 of 15
Diagnostics 2022, 12, 1233 11 of 15

Figure
Figure 6. Measurement
6. Measurement of gastric
of gastric ulcers
ulcers in clinical
in clinical photographs
photographs underunder different
different light.
light. The The brightness
brightness
of of
thethe images
images should
should bebe optimized
optimized enough
enough totoidentify
identifythethesymptoms
symptomsofofgastricgastriculcer/gastritis.
ulcer/gastritis. In In
some
some cases, dark
cases, darkerythema
erythemaisismoremoreclearly
clearly distinguished
distinguished from the thenormal
normalrange
rangeasasititisisbrightened
brightened (A),
(A),
butbut light
light erythema
erythema is difficult
is difficult to distinguish
to distinguish fromfrom
normalnormal
if it is ifbrightened
it is brightened
beyondbeyond
a certaina range
certain(B).
range (B). However,
However, when thewhenphoto’sthe photo’s
light is notlight is notitsufficient,
sufficient, is difficultittoisdiscriminate
difficult to the
discriminate
data valuethe data
regardless
value regardless
of light or darkof light or (A,B).
erythema dark erythema
The gastric(A,B).
ulcerThe gastric ulcerinmeasurements
measurements clinical specimensin clinical speci-
were analyzed
mens were analyzed with ImageJ (C). The photographs obtained from the first observation were set
with ImageJ (C). The photographs obtained from the first observation were set as the original data,
as the original data, and based on their brightness value, they were classified as high 50, medium
and based on their brightness value, they were classified as high 50, medium 100, and low 200
100, and low 200 by setting them as the white balance. The gastric ulcer measurements in clinical
by setting
specimens them
were as the white
analyzed balance. light
with different The gastric ulcer
intensities measurements
using ImageJ. Theiny-axisclinical specimens
shows were
the value
analyzed with different light intensities using ImageJ. The y-axis shows
of RGB. RGB color has to be maintained between 0 and 255. The x-axis shows the high (50), medium the value of RGB. RGB color
has to be maintained between 0 and
(100), and minimum (200) brightness contrast. 255. The x-axis shows the high (50), medium (100), and minimum
(200) brightness contrast.
3.6. Classification of Gastric Ulcer Lesions by Image Analysis
3.6. Classification of Gastric Ulcer Lesions by Image Analysis
The degree of gastric ulcers was classified by internal medicine specialists and their
The degree of gastric ulcers was classified by internal medicine specialists and their
values were evaluated through ImageJ analysis. The RGB value of the gastric ulcer pic-
values were evaluated through ImageJ analysis. The RGB value of the gastric ulcer pictures
tures was judged as normal, with an average of 247.964. In contrast, the values of mild
was judged as normal, with an average of 247.964. In contrast, the values of mild and
and severe gastric ulcer pictures judged were 233.184 and 206.857, respectively (Figure
severe gastric ulcer pictures judged were 233.184 and 206.857, respectively (Figure 5F).
5F). Therefore, the ImageJ analysis value measured according to the diagnosis of gastric
Therefore, the ImageJ analysis value measured according to the diagnosis of gastric ulcers
ulcers was clearly distinguished. However, it was difficult to accurately define their range
was clearly distinguished. However, it was difficult to accurately define their range with
with RGB values compared to animal experiments, because the light and camera in the
RGB values compared to animal experiments, because the light and camera in the patient’s
patient’s photographs cannot be consistently controlled (Figure 6A,B). In addition, the re-
photographs cannot be consistently controlled (Figure 6A,B). In addition, the results of
sults of human
human gastricgastric
endoscopyendoscopy
were more were more complicated
complicated than those than those of When
of animals. animals.the When
amount
theofamount
light wasof light
high,was high, iteasier
it became became to easier
observe to erythema;
observe erythema;
however,however, if the bright-
if the brightness of the
ness of the
picture picture
itself itself wasafter
was adjusted adjusted
takingafter taking
it, errors it, errors
occurred occurred
more than inmore than in value
the medium the
medium value (Figure 6C). This phenomenon was due to the difference
(Figure 6C). This phenomenon was due to the difference in redness in the normal region. in redness in the
normal region. In contrast, it was confirmed that when the brightness
In contrast, it was confirmed that when the brightness was lowered, the erythematous area was lowered, the
erythematous
was broadened areaaswas broadened
a whole (Figure as 6C,ahigh).
wholeThe (Figure 6C, high).
data values wereThe data values
erroneously were
recognized
erroneously
as dark areasrecognized
instead ofaserythema
dark areas instead
because theseof erythema because
did not receive eventhese
lightdid not receive
during shooting.
even light during shooting. In conclusion, the light intensity should be
In conclusion, the light intensity should be adjusted to a level where normal/erythematous adjusted to a level
where normal/erythematous areas can be distinguished. To obtain
areas can be distinguished. To obtain a consistent value, it was necessary to implement a consistent value, it
was necessary to implement a constant illuminance. Therefore, in this
a constant illuminance. Therefore, in this clinical result, the error in the analysis was clinical result, the
error in theofanalysis
because was because
the unoptimized of the unoptimized
brightness on some original brightness on some
photographs, andoriginal photo-
the original data
graphs, and theaccurate
can produce originalvalues
data can produce
if the accurateerrors
data without values if the data
received without
constant lighterrors re-
in clinical
ceived constant We
observation. lightsuggest
in clinical
thatobservation. We suggest
making a diagnosis that
with making
ImageJ a diagnosis
analysis with Im-
was impossible
ageJ analysis
before was impossible
supplementation, and before
we alsosupplementation,
suggest that thisand only wehelps
also in
suggest thatsetting.
a clinical this only
helps in a clinical setting.
4. Discussion
4. Discussion
The alleviatory effects of natural and food ingredients against gastric ulcers have been
well-documented. Of these, our recent investigation showed that a polymer fraction of
Diagnostics 2022, 12, 1233 12 of 15

Aloe vera improved the ethanol-induced gastric ulcer based on UI [15]. The gastric UI was
determined using the conventional method for scoring and was compared to the normal
group. The alleviated group was easily calculable, whereas the severe groups were not
calculable due to ambiguous erythematous areas. It is speculated that the calculation time
required for CCO is lengthy, and the reproducibility between the observation panels is
inconsistent. Therefore, an improved method to determine gastric ulcers was developed
using the ImageJ analysis. It was found that the phenotypic ulcer symptoms were unlike
those observed when using WIRE- or ethanol-induced mouse models. Red-colored erythe-
matous lesions with consistent lines were observed in ethanol-induced gastric ulcers only,
whereas WIRE-induced gastric ulcer symptoms included randomly formed black foreign
spots on the inner surface of the mouse stomach [6]. The indomethacin-induced gastric
ulcer model also exhibited symptoms similar to those of the ethanol-induced gastric ulcer
model according to previous reports [27,28].
Therefore, we investigated whether the ImageJ analysis could be applied to various
gastric ulcer-induced experiments. The ethanol model (Figure 3C,D) was relatively effective
in inducing gastric ulcers, whereas the WIRE model was not (Figure 3E,F). Optimizing
the light brightness and contrast of images derived from the WIRE model proved difficult
because erythema appeared deeply during stress induction. The result for the ethanol
model was close to the a* value on CIE-Lab, whereas the result for the WIRE model
was close to the L value, which was influenced by the background color. Moreover, the
gene expression in various induction models was different. The indomethacin-treated
model showed upregulated EGF, VEGF, myeloperoxidase, and malondialdehyde, and
downregulated GSH, GST, SOD, and GPx [29,30], whereas the ethanol-induced model
upregulated MMP-9, MMP-2, iNOS, and eNOS [31,32]; moreover, the WIRE-induced
model upregulated HSP-70, COX-2, PPAR-γ, NF-κB, ghrelin, and IL-1β, but downregulated
PGE2 [33,34]. Therefore, the genetic factors of the gastric ulcer induction models seem to
differ, although the regulatory factors are similarly related to inflammation, antioxidant
potential, and growth factors. Thereafter, it was predicted that these various aspects of gene
expression should also be reflected in the phenotypic stomach symptoms. Furthermore, the
gastric ulcer measurements in the clinical specimens were analyzed with ImageJ (Figure 5C).
The photo brightness of the human gastric ulcer image was set to 100, which was designated
as a medium measure. Additionally, relatively high and low values of 50 and 200 were
indicated, respectively, as shown in Figure 6. The gastric ulcer measurements of clinical
specimens were analyzed with ImageJ at various luminosities. Errors due to false detection
of erythema lesions were found in values derived by manipulating the brightness and
contrast of the images.
Thus far, ImageJ analysis has been applied to various biochemical data. One method
to measure the amplified PCR bands or protein spots observed by antibodies after elec-
trophoresis is taking the quantitative measurements of band strength [35]. In addition, cell
counting using an analysis image is simple in mammalian cell cultures [36,37]. Further-
more, because immunohistochemistry (IHC) data are difficult to analyze visually, ImageJ
can be used to quantitatively calculate the area of stained cells [38]. The advantages of
the ImageJ analysis method against a spectrophotometric technique with image analysis
were described earlier. When the graph was shown as a linear equation, the coefficient of
determination (R2 ) value validated the data. Therefore, image analysis could be adapted
to various diagnoses and predictions of disease symptoms. To accomplish this, a paraffin-
embedded tissue section could be observed by hematoxylin and eosin staining or IHC, and
computer-aided image analysis would be useful to determine phenotypic characteristics.
One related study previously showed that IHC results can be derived from visual scores via
the ImageJ analysis [39]. In breast cancer prognosis, the image analysis of the histopatho-
logical features of paraffin-embedded breast cancer specimens had a fairly significant effect
on the 8-year disease-free survival of 230 patients with breast cancer [40]. Among the
730 features retrieved by the latter study’s clinical results, 4 factors, including the features
of tumor nests, tumor nest cell nuclei, and tumor nest cell density, were newly identified
Diagnostics 2022, 12, 1233 13 of 15

as prognostic factors. Meanwhile, bone analysis is usually measured by computed tomog-

raphy and X-ray microtomography, which are expensive and inflexible. In contrast, the
ImageJ image analysis is relatively inexpensive and easy to conduct when analyzing the
trabecular bones, whole bones, and osteocyte lacunae [41].
Since the image analysis of gastric ulcers is only suitable for phenotypic data or routine
observation, even though it can be applied to IHC, additional analysis from a mechanism
study by Western blot or RT-PCR is required. The ImageJ analysis was verified as more
rapid, reproducible, and unbiased than CCO, which was affected by the panel’s subjectivity
when determining UI. However, for an accurate determination of the image analysis, certain
conditions related to the camera background, such as illumination and chromaticity, must
be controlled because these can affect the information in the photograph [42]. Specifically,
if a large amount of light is present, it becomes easier to observe erythema, but if the
brightness of the picture itself is adjusted post-capture (as in this experiment), errors occur
more frequently (Figure 6C, minimum). This phenomenon was due to the difference in
redness in the normal region. However, it was confirmed that the area recognized as
erythema was broadened as a whole when the brightness darkened (Figure 6C, high). Thus,
the data values were erroneously recognized as dark areas due to insufficient light during
shooting, as with erythema. In conclusion, the intensity of light should be adjusted to
a level where normal/erythematous areas can be distinguished. To obtain a consistent
value, it is necessary to implement a constant illumination. With such results, it is easier to
distinguish whether errors occurred in the analysis because the brightness of some of the
original photographs was not optimized. The original data can include accurate values if
error-free data, wherein a constant light was received in clinical observation, are secured.
We assumed that it would be difficult to accurately diagnose gastric ulcers without
trained experience in clinical practice because the degree and cause of ulceration differ
among patients. Accordingly, an auxiliary method that complements the CCO method
was found, which could be assessed by clinical experience and by effectively analyzing the
photos/images from radiology department computers currently used by clinicians. The
reliability of this method was confirmed using the ImageJ software (Figures 1 and 2), and
it was confirmed that IA can reduce the deviation of gastric ulcer scores compared with
CCO in an animal experiment (Figures 2 and 3). Thus, it might be safe to assume that the
deviation of the RGB values in the IA according to the progression of gastric ulcers among
the 140 clinical samples could be reduced. To the best of our knowledge, this is the first
study to provide scientific evidence that an image analysis can be complementarily applied
to the clinical diagnosis of gastric ulcers.
If the limitations of the current method are controlled and improved, the image
analysis has the potential to become a beneficial tool for the phenotypic determination
of other disease targets in vivo and as a subsidiary for clinical ulcer diagnosis. Although
previous studies have demonstrated that image analysis can identify gastric lesions with
better diagnostic accuracy than physicians’ clinical observations, the clinical applicability of
the method remains uncertain. Further studies are required to demonstrate clinical efficacy
in real-world settings to ensure that the improved methodology can be used to determine
the intensity of gastric ulcers.

5. Patents
Int. Cl. A61B 5/00 (2006. 01) 10-2013-0097146 (KR. B1).

Author Contributions: Conceptualization, C.-Y.Y., H.-U.S. and S.-H.L.; methodology, C.-Y.Y. and
H.-U.S.; software, C.-Y.Y. and H.-U.S.; validation, C.-Y.Y., H.-U.S., S.-k.K., S.-O.K. and S.-H.L.; formal
analysis, C.-Y.Y., H.-U.S., S.-k.K., S.-O.K. and S.-H.L.; investigation, C.-Y.Y., S.-k.K. and S.-H.L.; re-
sources, S.-k.K.; data curation, C.-Y.Y., H.-U.S., S.-k.K. and S.-H.L.; writing—original draft preparation,
C.-Y.Y., H.-U.S. and S.-H.L.; writing—review and editing, C.-Y.Y., H.-U.S., S.-k.K., S.-O.K. and S.-H.L.;
supervision, S.-H.L. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Diagnostics 2022, 12, 1233 14 of 15

Institutional Review Board Statement: The study was conducted according to the guidelines of the
Declaration of Helsinki, and approved by the Institutional Review Board of Kyungpook National
University (approval number: KNU-2019-0162, approved on 1 November 2019).
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are openly available.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Boyacioglu, M.; Kum, C.; Sekkin, S.; Yalinkilinc, H.S.; Avci, H.; Epikmen, E.T.; Karademir, U. The effects of lycopene on DNA
damage and oxidative stress on indomethcin-induced gastric ulcer in rats. Clin. Nutr. 2015, 35, 428–435. [CrossRef] [PubMed]
2. Dore, M.P.; Fanciulli, G.; Tomasi, P.A.; Realdi, G.; Delitala, G.; Graham, D.Y.; Malaty, H.M. Gastrointestinal symptoms and
Helicobacter pylori infection in school-age children residing in Porto Torres, Sardinia, Italy. Helicobacter 2012, 17, 369–373.
[CrossRef] [PubMed]
3. Park, C.H.; Son, H.U.; Son, M.; Lee, S.-H. Protective effect of Acer mono Max. sap on water immersion restraint stress-induced
gastric ulceration. Exp. Ther. Med. 2011, 2, 843–848. [CrossRef] [PubMed]
4. Longman, R.J.; Douthwaite, J.; Sylvester, P.A.; Poulsom, R.; Corfield, A.P.; Thomas, M.G.; Wright, N.A. Coordinated localisation of
mucins and trefoil peptides in the ulcer associated cell lineage and the gastrointestinal mucosa. Gut 2000, 47, 792–800. [CrossRef]
[PubMed]
5. Jansson, E.A.; Petersson, J.; Reinders, C.; Sobko, T.; Björne, H.; Phillipson, M.; Weitzberg, E.; Holm, L.; Lundberg, J.O. Protection
from nonsteroidal anti-inflammatory drug (NSAID)-induced gastric ulcers by dietary nitrate. Free Radic. Biol. Med. 2007, 42,
510–518. [CrossRef] [PubMed]
6. Al Moutaery, A.R. Effect of centrophenoxine on water-immersion restraint stress-and chemically-induced gastric ulcers in rats.
Res. Commun. Mol. Pathol. Pharmacol. 2002, 113, 39–56.
7. Ning, J.W.; Lin, G.B.; Ji, F.; Xu, J.; Sharify, N. Preventive effects of geranylgeranylacetone on rat ethanol-induced gastritis. World J.
Gastroenterol. 2012, 18, 2262–2269. [CrossRef]
8. Collins, T.J. ImageJ for microscopy. Biotechniques 2007, 43, 25–30. [CrossRef]
9. Liepinsh, E.; Kuka, J.; Dambrova, M. Troubleshooting digital macro photography for image acquisition and the analysis of
biological samples. J. Pharmacol. Toxicol. Methods 2013, 67, 98–106. [CrossRef]
10. Caldas-Cueva, J.P.; Mauromoustakos, A.; Sun, X.; Owens, C.M. Use of image analysis to identify woody breast characteristics in
8-week-old broiler carcasses. Poult. Sci. 2021, 100, 100890. [CrossRef]
11. Bonicelli, A.; Zioupos, P.; Arnold, E.; Rogers, K.D.; Xhemali, B.; Kranioti, E.F. Age related changes of rib cortical bone matrix and
the application to forensic age-at-death estimation. Sci. Rep. 2021, 11, 2086. [CrossRef] [PubMed]
12. Bašković, M.; Bojanac, A.K.; Sinčić, N.; Perić, M.H.; Krsnik, D.; Ježek, D. The effect of astaxanthin on testicular torsion-detorsion
injury in rats—Detailed morphometric evaluation of histological sections. J. Pedatr. Urol. 2021, 27, S1477. [CrossRef] [PubMed]
13. Sangiovanni, E.; Vrhovsek, U.; Rossoni, G.; Colombo, E.; Brunelli, C.; Brembati, L.; Trivulzio, S.; Gasperotti, M.; Mattivi, F.; Bosisio,
E.; et al. Ellagitannins from Rubus berries for the control of gastric inflammation: In vitro and in vivo studies. PLoS ONE 2013,
8, e71762. [CrossRef] [PubMed]
14. Khattab, M.M.; Gad, M.Z.; Abdallah, D. Protective role of nitric oxide in indomethacin-induced gastric ulceration by a mechanism
independent of gastric acid secretion. Pharmacol. Res. 2001, 43, 463–467. [CrossRef] [PubMed]
15. Xing, J.; Yang, B.; Dong, Y.; Wang, B.; Wang, J.; Kallio, H.P. Effects of sea buckthorn (Hippophae rhamnoides L.) seed and pulp oils
on experimental models of gastric ulcer in rats. Fitoterapia 2002, 73, 644–650. [CrossRef]
16. Park, C.H.; Nam, D.Y.; Son, H.U.; Lee, S.-R.; Lee, H.-J.; Heo, J.-C.; Cha, T.-Y.; Baek, J.-H.; Lee, S.-H. Polymer fraction of Aloe vera
exhibits a protective activity on ethanol-induced gastric lesions. Int. J. Mol. Med. 2011, 27, 511–518.
17. Park, C.H.; Son, H.U.; Yoo, C.Y.; Lee, S.H. Low molecular-weight gel fraction of Aloe vera exhibits gastroprotection by inducing
matrix metalloproteinase-9 inhibitory activity in alcohol-induced acute gastric lesion tissues. Pharm. Biol. 2017, 55, 2110–2115.
[CrossRef]
18. Seo, S.H.; Kim, J.H.; Kim, J.W.; Kye, Y.C.; Ahn, H.H. Better understanding of digital photography for skin color measurement:
With a special emphasis on light characteristics. Skin Res. Technol. 2011, 17, 20–25. [CrossRef]
19. Kibune, S.; Kuriyama, S. Color restoration with visual feature code. In International Workshop on Advanced Image Technology; SPIE:
Bellingham, WA, USA, 2011.
20. Kawano, S.; Tsuji, S. Role of mucosal blood flow: A conceptional review in gastric mucosal injury and protection. J. Gastroenterol.
Hepatol. 2000, 15, 1–6. [CrossRef]
21. Tran, C.D.; Simianu, V. Multiwavelength thermal lens spectrophotometer based on an acousto-optic tunable filter. Anal. Chem.
1992, 64, 1419–1425. [CrossRef]
22. Bartlett, J.W.; Frost, C. Reliability, repeatability and reproducibility: Analysis of measurement errors in continuous variables.
Ultrasound Obstet. Gynecol. 2008, 31, 466–475. [CrossRef] [PubMed]
Diagnostics 2022, 12, 1233 15 of 15

23. Wilson, C.; Gies, P.; Niven, B.; McLennan, A.; Bevin, N. The relationship between UV transmittance and color—visual description
and instrumental measurement. Textile Res. J. 2008, 78, 128–137. [CrossRef]
24. Wallace, J.L. Prostaglandins, NSAIDs, and gastric mucosal protection: Why doesn’t the stomach digest itself? Physiol. Rev. 2008,
88, 1547–1565. [CrossRef] [PubMed]
25. Yamamoto, T.; Takiwaki, H.; Arase, S.; Ohshima, H. Derivation and clinical application of special imaging by means of digital
cameras and Image J freeware for quantification of erythema and pigmentation. Skin Res. Technol. 2008, 14, 26–34. [CrossRef]
[PubMed]
26. Han, S.S.; Park, I.W.; Chang, S.E.; Lim, W.; Kim, M.S.; Park, G.H.; Chae, J.B.; Huh, C.H.; Na, J.-I. Augmented intelligence
dermatology: Deep neural networks empower medical professionals in diagnosing skin cancer and predicting treatment options
for 134 skin disorders. J. Investig. Dermatol. 2020, 140, 1753–1761. [CrossRef]
27. Bandyopadhyay, S.K.; Pakrashi, S.C.; Pakrashi, A. The role of antioxidant activity of Phyllanthus emblica fruits on prevention from
indomethacin induced gastric ulcer. J. Ethnopharmacol. 2000, 70, 171–176. [CrossRef]
28. Vahalia, M.K.; Vemula, S.K.; Chawada, M.B.; Thakur, K.S. Antiulcer activity of Amlapitta Mishran suspension in rats: A pilot study.
Ancient Sci. Life 2012, 32, 112. [CrossRef]
29. Banerjee, D.; Maity, B.; Bandivdeker, A.H.; Bandyopadhyay, S.K.; Chattopadhyay, S. Angiogenic and cell proliferating action of
the natural diarylnonanoids, malabaricone B and malabaricone C during healing of indomethacin-induced gastric ulceration.
Pharmaceut. Res. 2008, 25, 1601–1609. [CrossRef]
30. Suleyman, H.; Albayrak, A.; Bilici, M.; Cadirci, E.; Halici, Z. Different mechanisms in formation and prevention of indomethacin-
induced gastric ulcers. Inflammation 2010, 33, 224–234. [CrossRef]
31. Swarnakar, S.; Mishra, A.; Ganguly, K.; Sharma, A.V. Matrix metalloproteinase-9 activity and expression is reduced by melatonin
during prevention of ethanol-induced gastric ulcer in mice. J. Pineal Res. 2007, 43, 56–64. [CrossRef]
32. Pan, L.; Tang, Q.; Fu, Q.; Hu, B.-R.; Xiang, J.-Z.; Qian, J.-Q. Roles of nitric oxide in protective effect of berberine in ethanol-induced
gastric ulcer mice. Acta Pharmacol. Sin. 2005, 26, 1334. [CrossRef] [PubMed]
33. Konturek, P.; Sliwowski, Z.; Koziel, J.; Burnat, G.; Brzozowski, T.; Konturek, S.J. Probiotic bacteria Escherichia coli strain Nissle
1917 attenuates acute gastric lesions induced by stress. J. Physiol. Pharmacol. 2009, 60, 41–48. [PubMed]
34. İzgüt-Uysal, V.N.; Gemici, B.; Birsen, İ.; Acar, N.; Ustunel, I. The protective effect of apelin against water-immersion and restraint
stress-induced gastric damage. J. Physiol. Sci. 2014, 64, 279–289. [CrossRef] [PubMed]
35. Huleihel, L.; Ben-Yehudah, A.; Milosevic, J.; Yu, G.; Pandit, K.; Sakamoto, K.; Yousef, H.; LeJeune, M.; Coon, T.A.; Redinger,
C.J.; et al. Let-7d microRNA affects mesenchymal phenotypic properties of lung fibroblasts. Am. J. Physiol.-Lung Cell. Mol. Physiol.
2014, 306, L534–L542. [CrossRef] [PubMed]
36. Rhim, H.J.; Luo, X.; Xu, X.; Gao, D.; Zhou, T.; Li, F.; Qin, L.; Wang, P.; Xia, X.; Wong, S.T.C. A high-content screen identifies
compounds promoting the neuronal differentiation and the midbrain dopamine neuron specification of human neural progenitor
cells. Sci. Rep. 2015, 5, 16237. [CrossRef] [PubMed]
37. Ghali, W.; Vaudry, D.; Jouenne, T.; Marzouki, M.N. Lycium Europaeum fruit extract: Antiproliferative activity on A549 human
lung carcinoma cells and PC12 Rrat adrenal medulla cancer cells and assessment of its cytotoxicity on cerebellum granule cells.
Nutr. Cancer 2015, 67, 637–646. [CrossRef]
38. Owens, P.; Pickup, M.W.; Novitskiy, S.V.; Chytil, A.; Gorska, A.E.; Aakre, M.E.; West, J.; Moses, H.L. Disruption of bone
morphogenetic protein receptor 2 (BMPR2) in mammary tumors promotes metastases through cell autonomous and paracrine
mediators. Proc Natl. Acad. Sci. USA 2012, 109, 2814–2819. [CrossRef]
39. Rizzardi, A.E.; Johnson, A.T.; Vogel, R.I.; Pambuccian, S.E.; Henriksen, J.; Skubitz, A.P.; Metzger, G.J.; Schmechel, S.C. Quantitative
comparison of immunohistochemical staining measured by digital image analysis versus pathologist visual scoring. Diagn. Pathol.
2012, 7, 42. [CrossRef]
40. Chen, J.M.; Qu, A.P.; Wang, L.W.; Yuan, J.-P.; Yang, F.; Xiang, Q.-M.; Maskey, N.; Yang, G.-F.; Liu, J.; Li, Y. New breast cancer
prognostic factors identified by computer-aided image analysis of HE stained histopathology images. Sci. Rep. 2015, 5, 10690.
[CrossRef]
41. Gurcan, M.N.; Boucheron, L.E.; Can, A.; Madabhushi, A.; Rajpoot, N.M.; Yener, B. Histopathological image analysis: A review.
IEEE Rev. Biomed. Eng. 2009, 2, 147–171. [CrossRef]
42. Weng, C.-C.; Chen, H.; Fuh, C.-S. A novel automatic white balance method for digital still cameras. In Proceedings of the 2005
IEEE International Symposium on Circuits and Systems (ISCAS), Kobe, Japan, 23–26 May 2005.