BMC Medical Imaging,
Год журнала:
2024,
Номер
24(1)
Опубликована: Сен. 16, 2024
Recently
emerged
SAM-Med2D
represents
a
state-of-the-art
advancement
in
medical
image
segmentation.
Through
fine-tuning
the
Large
Visual
Model,
Segment
Anything
Model
(SAM),
on
extensive
datasets,
it
has
achieved
impressive
results
cross-modal
However,
its
reliance
interactive
prompts
may
restrict
applicability
under
specific
conditions.
To
address
this
limitation,
we
introduce
SAM-AutoMed,
which
achieves
automatic
segmentation
of
images
by
replacing
original
prompt
encoder
with
an
improved
MobileNet
v3
backbone.
The
performance
multiple
datasets
surpasses
both
SAM
and
SAM-Med2D.
Current
enhancements
lack
applications
field
classification.
Therefore,
SAM-MedCls,
combines
our
designed
attention
modules
to
construct
end-to-end
classification
model.
It
performs
well
various
modalities,
even
achieving
results,
indicating
potential
become
universal
model
for
Electronics,
Год журнала:
2023,
Номер
12(4), С. 955 - 955
Опубликована: Фев. 14, 2023
The
study
of
neuroimaging
is
a
very
important
tool
in
the
diagnosis
central
nervous
system
tumors.
This
paper
presents
evaluation
seven
deep
convolutional
neural
network
(CNN)
models
for
task
brain
tumor
classification.
A
generic
CNN
model
implemented
and
six
pre-trained
are
studied.
For
this
proposal,
dataset
utilized
Msoud,
which
includes
Fighshare,
SARTAJ,
Br35H
datasets,
containing
7023
MRI
images.
magnetic
resonance
imaging
(MRI)
belongs
to
four
classes,
three
tumors,
including
Glioma,
Meningioma,
Pituitary,
one
class
healthy
brains.
trained
with
input
images
several
preprocessing
strategies
applied
paper.
evaluated
Generic
CNN,
ResNet50,
InceptionV3,
InceptionResNetV2,
Xception,
MobileNetV2,
EfficientNetB0.
In
comparison
all
models,
best
was
obtained
an
average
Accuracy
97.12%.
development
these
techniques
could
help
clinicians
specializing
early
detection
IEEE Access,
Год журнала:
2023,
Номер
11, С. 72518 - 72536
Опубликована: Янв. 1, 2023
Around
the
world,
brain
tumors
are
becoming
leading
cause
of
mortality.
The
inability
to
undertake
a
timely
tumor
diagnosis
is
primary
this
pandemic.
Brain
cancer
crucial
procedure
that
relies
on
expertise
and
experience
doctor.
Radiologists
must
use
an
automated
classification
model
find
cancers.
current
model's
accuracy
has
be
improved
get
suitable
therapies.
can
consult
various
computer-aided
diagnostic
(CAD)
models
in
literature
medical
imaging
assist
them
with
their
patients.
Previous
research
widely
used
CNN
for
detection
classification,
which
typically
require
large
datasets.
This
proposed
Caps-VGGNet
hybrid
model,
integrates
CapsNet
VGGNet
by
adding
layers
VGGNet.
presented
addresses
challenge
requiring
datasets
automatically
extracting
classifying
features.
suggested
algorithm's
effectiveness
was
assessed
using
Brats-2020
Brats-2019
dataset,
contains
high-quality
images
tumors.
Compared
other
conventional
models,
empirical
outcomes
indicate
it
exhibited
highest
level
superior
efficacy
terms
accuracy,
specificity,
sensitivity.
Specifically,
attained
0.99,
specificity
sensitivity
0.98
Brats20
dataset.
Cancers,
Год журнала:
2023,
Номер
15(6), С. 1767 - 1767
Опубликована: Март 14, 2023
Brain
tumors
and
other
nervous
system
cancers
are
among
the
top
ten
leading
fatal
diseases.
The
effective
treatment
of
brain
depends
on
their
early
detection.
This
research
work
makes
use
13
features
with
a
voting
classifier
that
combines
logistic
regression
stochastic
gradient
descent
using
extracted
by
deep
convolutional
layers
for
efficient
classification
tumorous
victims
from
normal.
From
first
second-order
tumor
features,
model
training.
Using
helps
to
increase
precision
non-tumor
patient
classification.
proposed
along
convoluted
produces
results
show
highest
accuracy
99.9%.
Compared
cutting-edge
methods,
approach
has
demonstrated
improved
accuracy.
Cancers,
Год журнала:
2025,
Номер
17(1), С. 121 - 121
Опубликована: Янв. 2, 2025
Background/Objectives:
Brain
tumor
classification
is
a
crucial
task
in
medical
diagnostics,
as
early
and
accurate
detection
can
significantly
improve
patient
outcomes.
This
study
investigates
the
effectiveness
of
pre-trained
deep
learning
models
classifying
brain
MRI
images
into
four
categories:
Glioma,
Meningioma,
Pituitary,
No
Tumor,
aiming
to
enhance
diagnostic
process
through
automation.
Methods:
A
publicly
available
Tumor
dataset
containing
7023
was
used
this
research.
The
employs
state-of-the-art
models,
including
Xception,
MobileNetV2,
InceptionV3,
ResNet50,
VGG16,
DenseNet121,
which
are
fine-tuned
using
transfer
learning,
combination
with
advanced
preprocessing
data
augmentation
techniques.
Transfer
applied
fine-tune
optimize
accuracy
while
minimizing
computational
requirements,
ensuring
efficiency
real-world
applications.
Results:
Among
tested
Xception
emerged
top
performer,
achieving
weighted
98.73%
F1
score
95.29%,
demonstrating
exceptional
generalization
capabilities.
These
proved
particularly
effective
addressing
class
imbalances
delivering
consistent
performance
across
various
evaluation
metrics,
thus
their
suitability
for
clinical
adoption.
However,
challenges
persist
improving
recall
Glioma
Meningioma
categories,
black-box
nature
requires
further
attention
interpretability
trust
settings.
Conclusions:
findings
underscore
transformative
potential
imaging,
offering
pathway
toward
more
reliable,
scalable,
efficient
tools.
Future
research
will
focus
on
expanding
diversity,
model
explainability,
validating
settings
support
widespread
adoption
AI-driven
systems
healthcare
ensure
integration
workflows.
Neural Computing and Applications,
Год журнала:
2024,
Номер
unknown
Опубликована: Ноя. 20, 2024
Abstract
Brain
tumors
are
very
dangerous
as
they
cause
death.
A
lot
of
people
die
every
year
because
brain
tumors.
Therefore,
accurate
classification
and
detection
in
the
early
stages
can
help
recovery.
Various
deep
learning
techniques
have
achieved
good
results
tumor
classification.
The
traditional
methods
training
neural
network
from
scratch
time-consuming
last
for
weeks
training.
this
work,
we
proposed
an
ensemble
approach
depending
on
transfer
that
utilizes
pre-trained
models
DenseNet121
InceptionV3
to
detect
three
forms
tumors:
meningioma,
glioma,
pituitary.
While
developing
model,
some
changes
were
made
architecture
by
replacing
their
classifiers
(fully
connected
SoftMax
layers)
with
a
new
classifier
adopt
recent
task.
In
addition,
gradient-weighted
class
activation
maps
(Grad-CAM)
explainable
model
verify
achieve
high
confidence.
suggested
was
validated
using
publicly
available
dataset
99.02%
accuracy,
98.75%
precision,
98.98%
recall,
98.86%
F1
score.
outperformed
others
detecting
classifying
MRI
data,
verifying
degree
trust.
STUDIES IN ENGINEERING AND EXACT SCIENCES,
Год журнала:
2024,
Номер
5(1), С. 19 - 35
Опубликована: Янв. 12, 2024
Brain
tumors
(BT)
are
fatal
and
debilitating
conditions
that
shorten
the
typical
lifespan
of
patients.
Patients
with
BTs
who
receive
inadequate
treatment
an
incorrect
diagnosis
have
a
lower
chance
survival.
Magnetic
resonance
imaging
(MRI)
is
often
employed
to
assess
tumor.
However,
because
massive
quantity
data
provided
by
MRI,
early
BT
detection
complex
time-consuming
procedure
in
biomedical
imaging.
As
consequence,
automated
efficient
strategy
required.
The
brain
or
malignancies
has
been
done
using
variety
conventional
machine
learning
(ML)
approaches.
manually
collected
properties,
however,
provide
main
problem
these
models.
constraints
previously
stated
addressed
fusion
deep
model
for
binary
classification
presented
this
study.
recommended
method
combines
two
different
CNN
(Efficientnetb0,
VGG-19)
models
automatically
extract
features
make
use
feature’s
Cubic
SVM
classifier
model.
Additionally,
approach
displayed
outstanding
performance
various
measures,
including
Accuracy
(99.78%),
Precision
Recall
F1-Score
on
same
Kaggle
(Br35H)
dataset.
proposed
performs
better
than
current
approaches
classifying
from
MRI
images.