Bioengineering,
Journal Year:
2025,
Volume and Issue:
12(6), P. 608 - 608
Published: June 3, 2025
The
rapid
advancement
of
prompt-based
models
in
natural
language
processing
and
image
generation
has
revolutionized
the
field
segmentation.
introduction
Segment
Anything
Model
(SAM)
further
invigorated
this
domain
with
its
unprecedented
versatility.
However,
applicability
to
medical
segmentation
remains
uncertain
due
significant
disparities
between
images,
which
demand
careful
consideration.
This
study
comprehensively
analyzes
recent
efforts
adapt
SAM
for
segmentation,
including
empirical
benchmarking
methodological
refinements
aimed
at
bridging
gap
SAM’s
capabilities
unique
challenges
imaging.
Furthermore,
we
explore
future
directions
field.
While
direct
application
complex,
multimodal,
multi-target
datasets
may
not
yet
yield
optimal
results,
insights
from
these
provide
crucial
guidance
developing
foundational
tailored
intricacies
analysis.
Despite
existing
challenges,
holds
considerable
potential
demonstrate
advantages
robust
near
future.
PLoS ONE,
Journal Year:
2025,
Volume and Issue:
20(4), P. e0321096 - e0321096
Published: April 14, 2025
In
nephrology
research,
multi-glomerular
segmentation
in
immunofluorescence
images
plays
a
crucial
role
the
early
detection
and
diagnosis
of
chronic
kidney
disease.
However,
obtaining
accurate
segmentations
often
requires
labor-intensive
annotations
existing
methods
are
hampered
by
low
recall
rates
limited
accuracy.
Recently,
general
Segment
Anything
Model
(SAM)
has
demonstrated
promising
performance
several
tasks.
this
paper,
SAM-based
model
(GlomSAM)
is
introduced
to
employ
SAM
pathology
domain.
The
fusion
encoder
strategy
utilizing
advantages
both
convolution
networks
transformer
structures
with
prompts
conducted
facilitate
SAM’s
transfer
learning
applications
pathological
analysis.
Moreover,
rough
mask
generator
employed
generate
preliminary
glomerular
masks,
enabling
automated
input
prompting
improving
final
results.
Extensive
comparative
experiments
ablation
studies
show
its
state-of-the-art
surpassing
other
relevant
research.
We
hope
report
will
provide
insights
advance
field
promote
more
interesting
work
future.
PLoS ONE,
Journal Year:
2025,
Volume and Issue:
20(5), P. e0322751 - e0322751
Published: May 27, 2025
Purpose:
Detection
of
crucial
components
is
a
fundamental
problem
in
surgical
scene
understanding.
Limited
by
the
huge
cost
spatial
annotation,
current
studies
mainly
focus
on
recognition
three
elements
⟨
instrument,
verb,
target
id="M2">⟩
,
while
detection
id="M3">⟨
id="M4">⟩
remains
highly
challenging.
Some
efforts
have
been
made
to
detect
components,
yet
their
limitations
include:
(1)
performance
depends
amount
manual
annotations;
(2)
No
previous
study
has
investigated
targets.
Methods:
We
introduce
weakly
supervised
method
for
detecting
key
novelly
combining
triplet
model
and
foundation
Segment
Anything
Model
(SAM).
First,
setting
appropriate
prompts,
we
used
SAM
generate
candidate
regions
components.
Then,
preliminarily
localize
extracting
positive
activation
areas
class
maps
from
model.
However,
using
instrument’s
as
position
attention
guide
leads
positional
deviations
target’s
resulting
activation.
To
tackle
this
issue,
propose
RDV-AGC
introducing
an
Attention
Guide
Correction
(AGC)
module.
This
module
adjusts
guidance
according
forward
direction.
Finally,
match
initial
localization
instruments
targets
with
generated
SAM,
achieving
precise
scene.
Results:
Through
ablation
comparisons
similar
works,
our
achieved
remarkable
without
requiring
any
annotations.
Conclusion:
introduced
novel
integrating
visual
Bioengineering,
Journal Year:
2025,
Volume and Issue:
12(6), P. 608 - 608
Published: June 3, 2025
The
rapid
advancement
of
prompt-based
models
in
natural
language
processing
and
image
generation
has
revolutionized
the
field
segmentation.
introduction
Segment
Anything
Model
(SAM)
further
invigorated
this
domain
with
its
unprecedented
versatility.
However,
applicability
to
medical
segmentation
remains
uncertain
due
significant
disparities
between
images,
which
demand
careful
consideration.
This
study
comprehensively
analyzes
recent
efforts
adapt
SAM
for
segmentation,
including
empirical
benchmarking
methodological
refinements
aimed
at
bridging
gap
SAM’s
capabilities
unique
challenges
imaging.
Furthermore,
we
explore
future
directions
field.
While
direct
application
complex,
multimodal,
multi-target
datasets
may
not
yet
yield
optimal
results,
insights
from
these
provide
crucial
guidance
developing
foundational
tailored
intricacies
analysis.
Despite
existing
challenges,
holds
considerable
potential
demonstrate
advantages
robust
near
future.