Journal of the National Cancer Center,
Год журнала:
2024,
Номер
5(2), С. 113 - 131
Опубликована: Дек. 27, 2024
Upper
gastrointestinal
cancers,
mainly
comprising
esophageal
and
gastric
are
among
the
most
prevalent
cancers
worldwide.
There
many
new
cases
of
upper
annually,
survival
rate
tends
to
be
low.
Therefore,
timely
screening,
precise
diagnosis,
appropriate
treatment
strategies,
effective
prognosis
crucial
for
patients
with
cancers.
In
recent
years,
an
increasing
number
studies
suggest
that
artificial
intelligence
(AI)
technology
can
effectively
address
clinical
tasks
related
These
focus
on
four
aspects:
treatment,
prognosis.
this
review,
we
application
AI
in
Firstly,
basic
pipelines
radiomics
deep
learning
medical
image
analysis
were
introduced.
Furthermore,
separately
reviewed
aforementioned
aspects
both
Finally,
current
limitations
challenges
faced
field
summarized,
explorations
conducted
selection
algorithms
various
scenarios,
popularization
early
applications
AI,
large
multimodal
models.
Bacterial
vaginosis
(BV)
is
an
abnormal
gynecological
condition
caused
by
the
overgrowth
of
specific
bacteria
in
vagina.
This
study
aims
to
develop
a
novel
method
for
BV
detection
integrating
surface-enhanced
Raman
scattering
(SERS)
with
machine
learning
(ML)
algorithms.
Vaginal
fluid
samples
were
classified
as
positive
or
negative
using
BVBlue
Test
and
clinical
microscopy,
followed
SERS
spectral
acquisition
construct
data
set.
Preliminary
analysis
revealed
notable
disparities
characteristic
peak
features.
Multiple
ML
models
constructed
optimized,
convolutional
neural
network
(CNN)
model
achieving
highest
prediction
accuracy
at
99%.
Gradient-weighted
class
activation
mapping
(Grad-CAM)
was
used
highlight
important
regions
images
prediction.
Moreover,
CNN
blindly
tested
on
spectra
vaginal
collected
from
40
participants
unknown
infection
status,
90.75%
compared
results
combined
microscopy.
technique
simple,
cheap,
rapid
accurately
diagnosing
bacterial
vaginosis,
potentially
complementing
current
diagnostic
methods
laboratories.
Advanced Intelligent Systems,
Год журнала:
2024,
Номер
unknown
Опубликована: Окт. 28, 2024
In
this
study,
it
is
aimed
to
establish
a
novel
method
based
on
deep‐learning‐guided
surface‐enhanced
Raman
spectroscopy
(SERS)
technique
achieve
rapid
and
accurate
classification
of
vaginal
cleanliness
levels.
We
proposed
variational
autoencoder
(VAE)
approach
enhance
spectral
quality,
coupled
with
deep
learning
algorithm
long
short‐term
memory
(LSTM)
neural
network
analyze
SERS
spectra
produced
by
secretions.
The
performance
various
machine
(ML)
algorithms
assessed
using
multiple
evaluation
metrics.
Finally,
the
reliability
optimal
model
tested
blind
test
data
(
N
=
10/group
for
each
level).
quality
fingerprints
four
types
secretions
significantly
improved
after
VAE
decoding
reconstruction.
signal‐to‐noise
ratio
generated
increased
from
original
2.58–11.13.
Among
all
algorithms,
VAE–LSTM
demonstrates
best
prediction
ability
time
efficiency.
Additionally,
datasets
yielded
an
overall
accuracy
85%.
concluded
that
holds
significant
potential
in
rapidly
distinguishing
between
different
levels
through
human
secretion
samples.
This
contributes
efficient
diagnosis
clinical
settings.
Journal of the National Cancer Center,
Год журнала:
2024,
Номер
5(2), С. 113 - 131
Опубликована: Дек. 27, 2024
Upper
gastrointestinal
cancers,
mainly
comprising
esophageal
and
gastric
are
among
the
most
prevalent
cancers
worldwide.
There
many
new
cases
of
upper
annually,
survival
rate
tends
to
be
low.
Therefore,
timely
screening,
precise
diagnosis,
appropriate
treatment
strategies,
effective
prognosis
crucial
for
patients
with
cancers.
In
recent
years,
an
increasing
number
studies
suggest
that
artificial
intelligence
(AI)
technology
can
effectively
address
clinical
tasks
related
These
focus
on
four
aspects:
treatment,
prognosis.
this
review,
we
application
AI
in
Firstly,
basic
pipelines
radiomics
deep
learning
medical
image
analysis
were
introduced.
Furthermore,
separately
reviewed
aforementioned
aspects
both
Finally,
current
limitations
challenges
faced
field
summarized,
explorations
conducted
selection
algorithms
various
scenarios,
popularization
early
applications
AI,
large
multimodal
models.
