Physics in Medicine and Biology,
Journal Year:
2023,
Volume and Issue:
68(19), P. 195026 - 195026
Published: Aug. 11, 2023
Abstract
Objective.
Automatic
segmentation
of
fundus
vessels
has
the
potential
to
enhance
judgment
ability
intelligent
disease
diagnosis
systems.
Even
though
various
methods
have
been
proposed,
it
is
still
a
demanding
task
accurately
segment
vessels.
The
purpose
our
study
develop
robust
and
effective
method
in
human
color
retinal
images.
Approach.
We
present
novel
multi-level
spatial-temporal
attentional
information
deep
fusion
network
for
vessels,
called
MSAFNet,
which
enhances
performance
robustness.
Our
utilizes
encoding
module
obtain
Self-Attention
capture
feature
correlations
different
levels
network.
Based
on
encoder
decoder
structure,
we
combine
these
features
get
final
results.
Main
Through
abundant
experiments
four
public
datasets,
achieves
preferable
compared
with
other
SOTA
vessel
methods.
Accuracy
Area
Under
Curve
achieve
highest
scores
96.96%,
96.57%,
96.48%
98.78%,
98.54%,
98.27%
DRIVE,
CHASE_DB1,
HRF
datasets.
Specificity
score
98.58%
99.08%
DRIVE
STARE
Significance.
experimental
results
demonstrate
that
strong
learning
representation
capabilities
can
detect
blood
thereby
serving
as
tool
assisting
diagnosis.
Healthcare Analytics,
Journal Year:
2023,
Volume and Issue:
4, P. 100261 - 100261
Published: Sept. 22, 2023
Retinal
fundus
images
play
a
crucial
role
in
the
early
detection
of
eye
problems,
aiding
timely
diagnosis
and
treatment
to
prevent
vision
loss
or
blindness.
With
advancements
technology,
Convolutional
Neural
Network
(CNN)
algorithms
have
emerged
as
effective
tools
for
recognition,
delineation,
classification
tasks.
This
study
proposes
comprehensive
review
CNN
used
retinal
image
segmentation
classification.
Our
follows
systematic
approach,
exploring
diverse
repositories
identify
studies
employing
segment
classify
images.
Utilizing
CNNs
can
enhance
precision
outcomes
alleviate
sole
dependence
on
human
experts.
approach
enables
more
accurate
results,
reducing
burden
A
total
sixty-two
are
included
our
review,
analyzing
aspects
such
database
usage
advantages
disadvantages
methods
employed.
The
provides
valuable
insights,
limitations,
observations,
future
directions
field.
Despite
certain
findings
indicate
that
consistently
achieve
high
accuracies.
examination
sheds
light
potential
analysis.
Applied Sciences,
Journal Year:
2023,
Volume and Issue:
13(8), P. 5111 - 5111
Published: April 19, 2023
Diabetes
is
a
global
problem
which
impacts
people
of
all
ages.
Diabetic
retinopathy
(DR)
main
ailment
the
eyes
resulting
from
diabetes
can
result
in
loss
eyesight
if
not
detected
and
treated
on
time.
The
current
process
detecting
DR
its
progress
involves
manual
examination
by
experts,
time-consuming.
Extracting
retinal
vasculature,
segmentation
optic
disc
(OD)/fovea
play
significant
part
DR.
Detecting
lesions
like
microaneurysms
(MA),
hemorrhages
(HM),
exudates
(EX),
helps
to
establish
stage
Recently
with
advancement
artificial
intelligence
(AI),
deep
learning(DL),
division
AI,
widely
being
used
related
studies.
Our
study
surveys
latest
literature
“DR
lesion
detection
fundus
images
using
DL”.
