Traitement du signal,
Год журнала:
2024,
Номер
41(06), С. 2909 - 2922
Опубликована: Дек. 31, 2024
A
tumor
develops
when
brain
cells
exhibit
abnormal
growth
patterns
within
various
body
locations,
characterized
by
irregular
boundaries
and
shapes.Typically,
these
tumors
rapid
proliferation,
increasing
at
a
rate
of
approximately
1.6%
per
day.This
cell
can
lead
to
invisible
illnesses
alterations
in
psychological
behavioral
functions,
contributing
rising
trend
adult
mortality
rates
worldwide.Therefore,
Brain
must
be
detected
early.Failure
do
so
may
cause
deadly,
incurable
condition.Effective
therapy
improves
survival
if
early.Magnetic
Resonance
Imaging
(MRI)
is
essential
for
finding
classifying
tumors.The
manual
nature
diagnosis
classification
makes
it
prone
errors,
necessitating
the
development
automated
processes
improved
accuracy.In
light
considerations,
we
have
devised
with
fully
way
use
MR
images
find
classify
tumors.Our
approach
encompasses
three
key
phases:
pre-processing,
segmentation,
classification.To
detect
brain,
utilized
MRI,
employing
deep
transfer
transformed
VGG19
model.Notably,
our
research
demonstrates
superior
using
other
pre-trained
Convolutional
Neural
Network
(CNN)
models
such
as
AlexNet
VGG-16.The
learning
model
yielded
accuracy
achieving
98.65%
(dataset
1)
99.18%
2)
different
datasets.
Bioengineering,
Год журнала:
2025,
Номер
12(3), С. 212 - 212
Опубликована: Фев. 20, 2025
Breast
cancer
(BC)
remains
a
leading
cause
of
cancer-related
mortality
among
women
worldwide,
necessitating
advancements
in
diagnostic
methodologies
to
improve
early
detection
and
treatment
outcomes.
This
study
proposes
novel
twin-stream
approach
for
histopathological
image
classification,
utilizing
both
histopathologically
inherited
vision-based
features
enhance
precision.
The
first
stream
utilizes
Virchow2,
deep
learning
model
designed
extract
high-level
features,
while
the
second
employs
Nomic,
transformer
model,
capture
spatial
contextual
information.
fusion
these
streams
ensures
comprehensive
feature
representation,
enabling
achieve
state-of-the-art
performance
on
BACH
dataset.
Experimental
results
demonstrate
superiority
approach,
with
mean
accuracy
98.60%
specificity
99.07%,
significantly
outperforming
single-stream
methods
related
studies.
Statistical
analyses,
including
paired
t-tests,
ANOVA,
correlation
studies,
confirm
robustness
reliability
model.
proposed
not
only
improves
but
also
offers
scalable
efficient
solution
clinical
applications,
addressing
challenges
resource
constraints
increasing
demands.
2022 IEEE International Conference on Image Processing (ICIP),
Год журнала:
2023,
Номер
unknown, С. 2910 - 2914
Опубликована: Сен. 11, 2023
Prostate
cancer
(PCa)
is
a
widespread
type
of
that
leads
to
numerous
fatalities
and
high
financial
cost.
The
chance
survival
for
PCa
patients
increases
when
the
disease
detected
at
an
early
stage.
This
study
discusses
development
non-invasive
computer-aided
diagnosis
(CAD)
system
utilizes
intravoxel
incoherent
motion
(IVIM)
parameters
detect
diagnose
prostate
cancer.
focuses
on
IVIM,
which
can
separate
diffusion
water
molecules
in
capillaries
from
molecular
outside
vessels,
its
diagnostic
efficacy
central
peripheral
zones
proposes
two-step
segmentation
approach
tumor
detection,
starting
with
precise
localization
gland
using
robust
level-sets
technique
then
Attention
U-Net
extract
tumor-containing
region
interest
(ROI)
segmented
image.
evaluates
performance
CAD
system,
best
classifier
IVIM
differentiation,
value
compared
ADC.
results
this
contribute
methods
detection
diagnosis.
(CZ
+
PZ)
utilized
extra
trees
(ETC)
were
implemented
without
principal
component
analysis
(PCA)
standardization
scaling
achieved
metrics.
They
produced
accuracy
84.62%,
balanced
82.58%,
precision
80%,
specificity
67.86%,
sensitivity
97.30%,
F1-score
87.12%,
IoU
78.26%,
ROC
83.88%,
weighted
sum
metric
(WSM)
82.79%.
2022 IEEE 19th International Symposium on Biomedical Imaging (ISBI),
Год журнала:
2023,
Номер
unknown, С. 1 - 4
Опубликована: Апрель 18, 2023
In
the
field
of
pediatric
oncology,
Wilms'
tumor
is
a
common
occurrence
and
known
for
its
high
rate
recurrence.
The
study's
purpose
was
to
create
computer-based
prediction
system
response
preoperative
chemotherapy.
developed
based
on
contrast-enhanced
CT
scans
using
six
methods.
Firstly,
images
were
delineated,
followed
by
characterization
tumor's
form
3D
histogram
oriented
gradients.
Shape
features
then
extracted
spherical
harmonics,
sphericity,
elongation.
tumors'
functionality
also
demonstrated
determining
intensity
changes
in
contrast
phases.
Feature
fusion
applied
features,
responsive/non-responsive
results
found
classifier
support
vector
machine.
an
accuracy
96.83%
total,
detecting
97.83%
sensitivity
accurately
identifying
94.12%
specificity.
Additionally,
imaging
markers
used
predict
early
