GAN-SkipNet: A Solution for Data Imbalance in Cardiac Arrhythmia Detection Using Electrocardiogram Signals from a Benchmark Dataset
Mathematics,
Год журнала:
2024,
Номер
12(17), С. 2693 - 2693
Опубликована: Авг. 29, 2024
Electrocardiography
(ECG)
plays
a
pivotal
role
in
monitoring
cardiac
health,
yet
the
manual
analysis
of
ECG
signals
is
challenging
due
to
complex
task
identifying
and
categorizing
various
waveforms
morphologies
within
data.
Additionally,
datasets
often
suffer
from
significant
class
imbalance
issue,
which
can
lead
inaccuracies
detecting
minority
samples.
To
address
these
challenges
enhance
effectiveness
efficiency
arrhythmia
detection
imbalanced
datasets,
this
study
proposes
novel
approach.
This
research
leverages
MIT-BIH
dataset,
encompassing
total
109,446
beats
distributed
across
five
classes
following
Association
for
Advancement
Medical
Instrumentation
(AAMI)
standard.
Given
dataset’s
inherent
imbalance,
1D
generative
adversarial
network
(GAN)
model
introduced,
incorporating
Bi-LSTM
synthetically
generate
two
signal
classes,
represent
mere
0.73%
fusion
(F)
2.54%
supraventricular
(S)
The
generated
are
rigorously
evaluated
similarity
real
data
using
three
key
metrics:
mean
squared
error
(MSE),
structural
index
(SSIM),
Pearson
correlation
coefficient
(r).
In
addition
addressing
work
presents
deep
learning
models
tailored
classification:
SkipCNN
(a
convolutional
neural
with
skip
connections),
SkipCNN+LSTM,
SkipCNN+LSTM+Attention
mechanisms.
further
accuracy,
test
dataset
assessed
an
ensemble
model,
consistently
outperforms
individual
models.
performance
evaluation
employs
standard
metrics
such
as
precision,
recall,
F1-score,
along
their
average,
macro
weighted
average
counterparts.
Notably,
SkipCNN+LSTM
emerges
most
promising,
achieving
remarkable
F1-scores
99.3%,
were
elevated
impressive
99.60%
through
techniques.
Consequently,
innovative
combination
balancing
techniques,
GAN-SkipNet
not
only
resolves
posed
by
but
also
provides
robust
reliable
solution
detection.
stands
poised
clinical
applications,
offering
potential
be
deployed
hospitals
real-time
detection,
thereby
benefiting
patients
healthcare
practitioners
alike.
Язык: Английский
Next-Generation Diagnostics: The Impact of Synthetic Data Generation on the Detection of Breast Cancer from Ultrasound Imaging
Mathematics,
Год журнала:
2024,
Номер
12(18), С. 2808 - 2808
Опубликована: Сен. 11, 2024
Breast
cancer
is
one
of
the
most
lethal
and
widespread
diseases
affecting
women
worldwide.
As
a
result,
it
necessary
to
diagnose
breast
accurately
efficiently
utilizing
cost-effective
widely
used
methods.
In
this
research,
we
demonstrated
that
synthetically
created
high-quality
ultrasound
data
outperformed
conventional
augmentation
strategies
for
diagnosing
using
deep
learning.
We
trained
deep-learning
model
EfficientNet-B7
architecture
large
dataset
3186
images
acquired
from
multiple
publicly
available
sources,
as
well
10,000
generated
generative
adversarial
networks
(StyleGAN3).
The
was
five-fold
cross-validation
techniques
validated
four
metrics:
accuracy,
recall,
precision,
F1
score
measure.
results
showed
integrating
produced
into
training
set
increased
classification
accuracy
88.72%
92.01%
based
on
score,
demonstrating
power
models
expand
improve
quality
datasets
in
medical-imaging
applications.
This
larger
comprising
synthetic
significantly
improved
its
performance
by
more
than
3%
over
genuine
with
common
augmentation.
Various
procedures
were
also
investigated
set’s
diversity
representativeness.
research
emphasizes
relevance
modern
artificial
intelligence
machine-learning
technologies
medical
imaging
providing
an
effective
strategy
categorizing
images,
which
may
lead
diagnostic
optimal
treatment
options.
proposed
are
highly
promising
have
strong
potential
future
clinical
application
diagnosis
cancer.
Язык: Английский
Advanced Segmentation of Gastrointestinal (GI) Cancer Disease Using a Novel U-MaskNet Model
Life,
Год журнала:
2024,
Номер
14(11), С. 1488 - 1488
Опубликована: Ноя. 15, 2024
The
purpose
of
this
research
is
to
contribute
the
development
approaches
for
classification
and
segmentation
various
gastrointestinal
(GI)
cancer
diseases,
such
as
dyed
lifted
polyps,
resection
margins,
esophagitis,
normal
cecum,
pylorus,
Z
line,
ulcerative
colitis.
This
relevant
essential
because
current
challenges
related
absence
efficient
diagnostic
tools
early
diagnostics
GI
cancers,
which
are
fundamental
improving
diagnosis
these
common
diseases.
To
address
above
challenges,
we
propose
a
new
hybrid
model,
U-MaskNet,
combination
U-Net
Mask
R-CNN
models.
Here,
utilized
pixel-wise
instance
segmentation,
together
forming
solution
classifying
segmenting
cancer.
Kvasir
dataset,
includes
8000
endoscopic
images
validate
proposed
methodology.
experimental
results
clearly
demonstrated
that
novel
model
provided
superior
compared
other
well-known
models,
DeepLabv3+,
FCN,
DeepMask,
well
improved
performance
state-of-the-art
(SOTA)
including
LeNet-5,
AlexNet,
VGG-16,
ResNet-50,
Inception
Network.
quantitative
analysis
revealed
our
outperformed
achieving
precision
98.85%,
recall
98.49%,
F1
score
98.68%.
Additionally,
achieved
Dice
coefficient
94.35%
IoU
89.31%.
Consequently,
developed
increased
accuracy
reliability
in
detecting
cancer,
it
was
proven
can
potentially
be
used
process
and,
consequently,
patient
care
clinical
environment.
work
highlights
benefits
integrating
opening
way
further
medical
image
segmentation.
Язык: Английский
The Improved Network Intrusion Detection Techniques Using the Feature Engineering Approach with Boosting Classifiers
Mathematics,
Год журнала:
2024,
Номер
12(24), С. 3909 - 3909
Опубликована: Дек. 11, 2024
In
the
domain
of
cybersecurity,
cyber
threats
targeting
network
devices
are
very
crucial.
Because
exponential
growth
wireless
devices,
such
as
smartphones
and
portable
risks
becoming
increasingly
frequent
common
with
emergence
new
types
threats.
This
makes
automatic
accurate
detection
network-based
intrusion
essential.
this
work,
we
propose
a
system
utilizing
comprehensive
feature
engineering
approach
combined
boosting
machine-learning
(ML)
models.
A
TCP/IP-based
dataset
25,192
data
samples
from
different
protocols
has
been
utilized
in
our
work.
To
improve
dataset,
used
preprocessing
methods
label
encoding,
correlation
analysis,
custom
iterative
encoding.
model’s
accuracy
for
prediction,
then
unique
methodology
that
included
novel
scaling
random
forest-based
selection
techniques.
We
three
conventional
models
(NB,
LR,
SVC)
four
classifiers
(CatBoostGBM,
LightGBM,
HistGradientBoosting,
XGBoost)
classification.
The
10-fold
cross-validation
were
employed
to
train
each
model.
After
an
assessment
using
numerous
metrics,
best-performing
model
emerged
XGBoost.
With
mean
metric
values
99.54
±
0.0007
accuracy,
99.53
0.0013
precision,
0.001
recall,
F1-score
0.0014,
XGBoost
produced
best
performance
overall.
Additionally,
showed
ROC
curve
evaluating
model,
which
demonstrated
all
obtained
perfect
AUC
value
one.
Our
suggested
methodologies
show
effectiveness
detecting
intrusions,
setting
stage
be
real
time.
method
provides
strong
defensive
measure
against
malicious
intrusions
into
infrastructures
while
keep
varying.
Язык: Английский