The
rapid
proliferation
of
plant
diseases
poses
a
grave
threat
to
global
food
security
and
agricultural
productivity.
To
effectively
address
these
challenges
ensure
sustainable
practices,
the
timely
accurate
identification
becomes
paramount.
Over
recent
years,
deep
learning
techniques
namely
Convolutional
Neural
Networks
(CNNs),
have
emerged
as
pivotal
tool
with
potential
revolutionize
disease
by
performing
effective
feature
extraction.
This
study
focuses
on
development
CNN
model
for
automated
identification.
dataset
implementation
has
been
collected
from
Kaggel,
which
contains
32
varieties
leaf
including
normal
leaves.
proposed
13
different
convolutional,
4
max
pooling,
1
flattening
dense
layer
performance
implemented
in
four
scenarios
applying
complete
5,
10,
15,
20
epoch
values.
results
depict
that
shown
highest
accuracy
98.70%
at
while
97.87%,
95.92%,
87.09%
accuracies
resulted
values
respectively.
effectiveness
this
also
compared
existing
work
achieving
accuracy.
Diagnostics,
Год журнала:
2024,
Номер
14(4), С. 390 - 390
Опубликована: Фев. 11, 2024
In
the
domain
of
AI-driven
healthcare,
deep
learning
models
have
markedly
advanced
pneumonia
diagnosis
through
X-ray
image
analysis,
thus
indicating
a
significant
stride
in
efficacy
medical
decision
systems.
This
paper
presents
novel
approach
utilizing
convolutional
neural
network
that
effectively
amalgamates
strengths
EfficientNetB0
and
DenseNet121,
it
is
enhanced
by
suite
attention
mechanisms
for
refined
classification.
Leveraging
pre-trained
models,
our
employs
multi-head,
self-attention
modules
meticulous
feature
extraction
from
images.
The
model’s
integration
processing
efficiency
are
further
augmented
channel-attention-based
fusion
strategy,
one
complemented
residual
block
an
attention-augmented
enhancement
dynamic
pooling
strategy.
Our
used
dataset,
which
comprises
comprehensive
collection
chest
images,
represents
both
healthy
individuals
those
affected
pneumonia,
serves
as
foundation
this
research.
study
delves
into
algorithms,
architectural
details,
operational
intricacies
proposed
model.
empirical
outcomes
model
noteworthy,
with
exceptional
performance
marked
accuracy
95.19%,
precision
98.38%,
recall
93.84%,
F1
score
96.06%,
specificity
97.43%,
AUC
0.9564
on
test
dataset.
These
results
not
only
affirm
high
diagnostic
accuracy,
but
also
highlight
its
promising
potential
real-world
clinical
deployment.
IEEE Access,
Год журнала:
2024,
Номер
12, С. 34691 - 34707
Опубликована: Янв. 1, 2024
Pneumonia
is
a
potentially
life-threatening
infectious
disease
that
typically
diagnosed
through
physical
examinations
and
diagnostic
imaging
techniques
such
as
chest
X-rays,
ultrasounds,
or
lung
biopsies.
Accurate
diagnosis
crucial
wrong
diagnosis,
inadequate
treatment
lack
of
can
cause
serious
consequences
for
patients
may
become
fatal.
The
advancements
in
deep
learning
have
significantly
contributed
to
aiding
medical
experts
diagnosing
pneumonia
by
assisting
their
decision-making
process.
By
leveraging
models,
healthcare
professionals
enhance
accuracy
make
informed
decisions
suspected
having
pneumonia.
In
this
study,
six
models
including
CNN,
InceptionResNetV2,
Xception,
VGG16,
ResNet50,
Efficient-NetV2L
are
implemented
evaluated.
study
also
incorporates
the
Adam
optimizer,
which
effectively
adjusts
epoch
all
models.
trained
on
dataset
5856
X-ray
images
show
87.78%,
88.94%,
90.7%,
91.66%,
87.98%,
94.02%
ResNet50
EfficientNetV2L,
respectively.
Notably,
EfficientNetV2L
demonstrates
highest
proves
its
robustness
detection.
These
findings
highlight
potential
accurately
detecting
predicting
based
images,
providing
valuable
support
clinical
improving
patient
treatment.
Symmetry,
Год журнала:
2025,
Номер
17(3), С. 469 - 469
Опубликована: Март 20, 2025
Artificial
intelligence
(AI)
is
playing
a
dominant
role
in
advancing
heart
failure
detection
and
diagnosis,
significantly
furthering
personalized
healthcare.
This
review
synthesizes
AI-driven
innovations
by
examining
methodologies,
applications,
outcomes.
We
investigate
the
integration
of
machine
learning
algorithms,
diverse
datasets
including
electronic
health
records
(EHRs),
medical
records,
imaging
data,
clinical
notes,
deep
models,
neural
networks
to
enhance
diagnostic
accuracy.
Key
advancements
include
prediction
models
that
leverage
real-time
data
from
wearable
devices
alongside
state-of-the-art
AI
systems
trained
on
patient
hospitals
clinics.
Notably,
recent
studies
have
reported
accuracies
ranging
86.7%
as
high
99.9%,
with
sensitivity
specificity
values
often
exceeding
97%,
underscoring
potential
these
improve
early
decision-making
substantially.
Our
further
explores
impact
symmetry
asymmetry
model
design,
highlighting
symmetric
architectures
like
U-Net
offer
computational
efficiency
structured
feature
extraction.
In
contrast,
asymmetric
rare
conditions
subtle
patterns.
Incorporating
(DL)
methods
anomaly
disease
progression
modeling
reinforces
their
positive
accuracy
Furthermore,
this
identifies
challenges
current
such
quality,
algorithmic
transparency,
bias,
evaluation
metrics,
while
outlining
future
research
directions,
integrating
generative
hybrid
architectures,
explainable
techniques
optimize
practice.
IEEE Access,
Год журнала:
2024,
Номер
12, С. 57613 - 57632
Опубликована: Янв. 1, 2024
Deep
learning
models
have
demonstrated
promising
results
in
the
early
and
accurate
diagnosis
of
osteonecrosis
femoral
head
(ONFH),
enabling
detection
informed
surgical
decision-making.
The
objective
this
review
is
to
summarize
applications
deep
on
medical
images
ONFH.
English
papers
were
searched
from
CINAHL
via
EBSCOhost,
Embase,
IEEE
Xplore®
Digital
Library,
PubMed,
Scopus,
Web
Science.
Sixteen
studies
(n
=
16)
eligible
for
data
synthesis.
Among
these,
five
5)
focusing
radiographs,
ten
10)
magnetic
resonance
imaging,
one
study
1)
computed
tomographic
images.
these
included
identifying
ONFH
normal
or
other
hip
pathologies,
classifying
severity,
segmenting,
detecting
necrotic
regions,
predicting
signs
symptoms
ONFH,
potential
after
fracture
fixation.
Generally,
good
excellent
classification
performance
discriminatory
power;
generally
comparable
that
experienced
physicians
superior
less
physicians.
However,
external
validity
only
moderate,
as
evidenced
by
testing
set
might
be
attributed
relatively
small
size
used
during
model
training.
we
observed
a
shift
CNN-based
U-Net
based
(i.e.,
with
encoder-decoder
architecture).
In
addition
streamlining
segmentation,
detection,
procedures,
future
will
explore
multimodal
attention,
self-supervised
learning,
explainable
models,
augmentation
through
generative
models.
A
fungal
infection
in
humans
is
a
pathological
state
resulting
from
the
infiltration
and
proliferation
of
fungi
within
body.
Microorganisms
known
as
are
present
air,
water,
soil,
plants.
The
can
cause
skin
to
become
red
inflamed
causing
bad
oral
genital
effects
article
presents
deep
learning
technique
for
identifying
infections
using
MobileNetV3,
which
compact
resilient
convolutional
neural
network
(CNN).
model
trained
on
wide
variety
datasets,
demonstrating
its
efficiency
mobility
real-time
detection
portable
devices.
categorize
identify
various
across
different
conditions
capabilities.
findings
result
an
excellent
accuracy
speed
infections,
indicating
potential
rapid
accessible
healthcare,
agriculture,
environmental
monitoring.
work
investigates
effectiveness
MobileNetV3
named
DeepFungusDet
broad
dataset
containing
infections.
This
has
been
implemented
at
numbers
epochs
highest
identification
93.14%
epoch
13
loss
0.4494,
promise
recognizing
tool
provides
option
via
mobile
devices,
paving
way
future
research
use
crucial
field
fungus
identification.
represent
major
step
forward
provide
prospects
developing
practical
diagnostic
tools
healthcare
industry
related
fields.