Advances in Multimedia,
Journal Year:
2024,
Volume and Issue:
2024(1)
Published: Jan. 1, 2024
Gastrointestinal
(GI)
diseases
are
a
significant
global
health
issue,
causing
millions
of
deaths
annually.
This
study
presents
novel
method
for
classifying
GI
using
endoscopy
videos.
The
proposed
involves
three
major
phases:
image
processing,
feature
extraction,
and
classification.
processing
phase
uses
wavelet
transform
segmentation
an
adaptive
median
filter
denoising.
Feature
extraction
is
conducted
concatenated
recurrent
vision
transformer
(RVT)
with
two
inputs.
classification
employs
ensemble
four
classifiers:
support
vector
machines,
Bayesian
network,
random
forest,
logistic
regression.
system
was
trained
tested
on
the
Hyper–Kvasir
dataset,
largest
publicly
available
tract
achieving
accuracy
99.13%
area
under
curve
0.9954.
These
results
demonstrate
improvement
in
performance
disease
compared
to
traditional
methods.
highlights
potential
combining
RVTs
standard
machine
learning
techniques
enhance
automated
diagnosis
diseases.
Further
validation
larger
datasets
different
medical
environments
recommended
confirm
these
findings.
Computers, materials & continua/Computers, materials & continua (Print),
Journal Year:
2024,
Volume and Issue:
78(3), P. 3377 - 3390
Published: Jan. 1, 2024
Pulmonary
nodules
are
small,
round,
or
oval-shaped
growths
on
the
lungs.
They
can
be
benign
(noncancerous)
malignant
(cancerous).
The
size
of
a
nodule
range
from
few
millimeters
to
centimeters
in
diameter.
Nodules
may
found
during
chest
X-ray
other
imaging
test
for
an
unrelated
health
problem.
In
proposed
methodology
pulmonary
classified
into
three
stages.
Firstly,
2D
histogram
thresholding
technique
is
used
identify
volume
segmentation.
An
ant
colony
optimization
algorithm
determine
optimal
threshold
value.
Secondly,
geometrical
features
such
as
lines,
arcs,
extended
and
ellipses
detect
oval
shapes.
Thirdly,
Histogram
Oriented
Surface
Normal
Vector
(HOSNV)
feature
descriptors
different
sizes
shapes
by
using
scaled
rotation-invariant
texture
description.
Smart
classification
was
performed
with
XGBoost
classifier.
results
tested
validated
Lung
Image
Consortium
Database
(LICD).
method
has
sensitivity
98.49%
sized
3–30
mm.
Advances in Multimedia,
Journal Year:
2024,
Volume and Issue:
2024(1)
Published: Jan. 1, 2024
Gastrointestinal
(GI)
diseases
are
a
significant
global
health
issue,
causing
millions
of
deaths
annually.
This
study
presents
novel
method
for
classifying
GI
using
endoscopy
videos.
The
proposed
involves
three
major
phases:
image
processing,
feature
extraction,
and
classification.
processing
phase
uses
wavelet
transform
segmentation
an
adaptive
median
filter
denoising.
Feature
extraction
is
conducted
concatenated
recurrent
vision
transformer
(RVT)
with
two
inputs.
classification
employs
ensemble
four
classifiers:
support
vector
machines,
Bayesian
network,
random
forest,
logistic
regression.
system
was
trained
tested
on
the
Hyper–Kvasir
dataset,
largest
publicly
available
tract
achieving
accuracy
99.13%
area
under
curve
0.9954.
These
results
demonstrate
improvement
in
performance
disease
compared
to
traditional
methods.
highlights
potential
combining
RVTs
standard
machine
learning
techniques
enhance
automated
diagnosis
diseases.
Further
validation
larger
datasets
different
medical
environments
recommended
confirm
these
findings.
