International Journal of Imaging Systems and Technology,
Journal Year:
2025,
Volume and Issue:
35(2)
Published: March 1, 2025
ABSTRACT
Addressing
the
challenges
posed
by
colorectal
polyp
variability
and
imaging
inconsistencies
in
endoscopic
images,
we
propose
multiscale
feature
fusion
booster
network
(MFFB‐Net),
a
novel
deep
learning
(DL)
framework
for
semantic
segmentation
of
polyps
to
aid
early
cancer
detection.
Unlike
prior
models,
such
as
pyramid
vision
transformer‐based
cascaded
attention
decoder
(PVT‐CASCADE)
parallel
reverse
(PraNet),
MFFB‐Net
enhances
accuracy
efficiency
through
unique
extraction
both
encoder
stages,
coupled
with
module
refining
fine‐grained
details
bottleneck
efficient
compression.
The
leverages
multipath
skip
connections,
capturing
local
global
contextual
information,
is
rigorously
evaluated
on
seven
benchmark
datasets,
including
Kvasir,
CVC‐ClinicDB,
CVC‐ColonDB,
ETIS,
CVC‐300,
BKAI‐IGH,
EndoCV2020.
achieves
state‐of‐the‐art
(SOTA)
performance,
Dice
scores
94.38%,
91.92%,
91.21%,
80.34%,
82.67%,
76.92%,
74.29%
EndoCV2020,
respectively,
outperforming
existing
models
computational
efficiency.
real‐time
processing
speeds
26
FPS
only
1.41
million
parameters,
making
it
well
suited
real‐world
clinical
applications.
results
underscore
robustness
MFFB‐Net,
demonstrating
its
potential
deployment
computer‐aided
diagnosis
systems
setting
new
automated
segmentation.
Deleted Journal,
Journal Year:
2024,
Volume and Issue:
21(4), P. 617 - 630
Published: April 12, 2024
Abstract
Recently,
Meta
AI
Research
approaches
a
general,
promptable
segment
anything
model
(SAM)
pre-trained
on
an
unprecedentedly
large
segmentation
dataset
(SA-1B).
Without
doubt,
the
emergence
of
SAM
will
yield
significant
benefits
for
wide
array
practical
image
applications.
In
this
study,
we
conduct
series
intriguing
investigations
into
performance
across
various
applications,
particularly
in
fields
natural
images,
agriculture,
manufacturing,
remote
sensing
and
healthcare.
We
analyze
discuss
limitations
SAM,
while
also
presenting
outlook
its
future
development
tasks.
By
doing
so,
aim
to
give
comprehensive
understanding
SAM’s
This
work
is
expected
provide
insights
that
facilitate
research
activities
toward
generic
segmentation.
Source
code
publicly
available
at
https://github.com/LiuTingWed/SAM-Not-Perfect
.
Deleted Journal,
Journal Year:
2024,
Volume and Issue:
37(2), P. 851 - 863
Published: Jan. 12, 2024
Accurate
and
early
detection
of
precursor
adenomatous
polyps
their
removal
at
the
stage
can
significantly
decrease
mortality
rate
occurrence
disease
since
most
colorectal
cancer
evolve
from
polyps.
However,
accurate
segmentation
by
doctors
are
difficult
mainly
these
factors:
(i)
quality
screening
with
colonoscopy
depends
on
imaging
experience
doctors;
(ii)
visual
inspection
is
time-consuming,
burdensome,
tiring;
(iii)
prolonged
inspections
lead
to
being
missed
even
when
physician
experienced.
To
overcome
problems,
computer-aided
methods
have
been
proposed.
they
some
disadvantages
or
limitations.
Therefore,
in
this
work,
a
new
architecture
based
residual
transformer
layers
has
designed
used
for
polyp
segmentation.
In
proposed
segmentation,
both
high-level
semantic
features
low-level
spatial
utilized.
Also,
novel
hybrid
loss
function
The
focal
Tversky
loss,
binary
cross-entropy,
Jaccard
index
reduces
image-wise
pixel-wise
differences
as
well
improves
regional
consistencies.
Experimental
works
indicated
effectiveness
approach
terms
dice
similarity
(0.9048),
recall
(0.9041),
precision
(0.9057),
F2
score
(0.8993).
Comparisons
state-of-the-art
shown
its
better
performance.
Medical Imaging 2018: Computer-Aided Diagnosis,
Journal Year:
2024,
Volume and Issue:
unknown
Published: April 2, 2024
Automatic
segmentation
of
colon
polyps
can
significantly
reduce
the
misdiagnosis
cancer
and
improve
physician
annotation
efficiency.
While
many
methods
have
been
proposed
for
polyp
segmentation,
training
large-scale
networks
with
limited
colonoscopy
data
remains
a
challenge.
Recently,
Segment
Anything
Model
(SAM)
has
recently
gained
much
attention
in
both
natural
image
medical
segmentation.
SAM
demonstrates
superior
performance
several
vision
benchmarks
shows
great
potential
In
this
study,
we
propose
Poly-SAM,
finetuned
model
compare
its
to
state-of-the-art
models.
We
also
two
transfer
learning
strategies
without
finetuning
encoders.
Evaluated
on
five
public
datasets,
our
Polyp-SAM
achieves
datasets
impressive
three
dice
scores
all
above
88%.
This
study
adapting
tasks.
2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM),
Journal Year:
2023,
Volume and Issue:
unknown
Published: Dec. 5, 2023
Medical
image
segmentation
is
a
critical
component
in
variety
of
clinical
applications,
facilitating
accurate
diagnosis
and
treatment
planning.
The
Segment
Anything
Model
(SAM),
deep
learning
architecture,
has
emerged
as
promising
solution
to
the
challenges
inherent
medical
segmentation.
SAM's
superior
zero-shot
capability
allows
it
generalize
effectively,
even
absence
task-specific
samples.
This
unique
characteristic
broadens
its
application
potential
across
various
modalities.
paper
provides
an
in-depth
review
SAM,
focusing
on
discusses
advantages
over
traditional
methods,
emphasizing
accuracy,
efficiency,
automation
that
models
offer.
also
highlights
applications
SAM
imaging
modalities,
demonstrating
versatility
adaptability.
A
taxonomy
approaches
presented,
categorizing
them
based
modality,
dimension,
organ,
dataset,
prompt,
performance.
Despite
results
remain
field
identifies
these
suggests
directions
for
future
research.
In
conclusion,
this
aims
provide
comprehensive
understanding
revolutionize
analysis
contribute
advancements
healthcare.
