Ultrasonic Imaging,
Год журнала:
2022,
Номер
44(1), С. 25 - 38
Опубликована: Янв. 1, 2022
U-Net
based
algorithms,
due
to
their
complex
computations,
include
limitations
when
they
are
used
in
clinical
devices.
In
this
paper,
we
addressed
problem
through
a
novel
architecture
that
called
fast
and
accurate
for
medical
image
segmentation
task.
The
proposed
model
contains
four
tuned
2D-convolutional,
2D-transposed
convolutional,
batch
normalization
layers
as
its
main
layers.
There
blocks
the
encoder-decoder
path.
results
of
our
were
evaluated
using
prepared
dataset
head
circumference
abdominal
tasks,
public
(HC18-Grand
challenge
dataset)
fetal
measurement.
network
significantly
improved
processing
time
comparison
with
U-Net,
dilated
R2U-Net,
attention
MFP
U-Net.
It
took
0.47
seconds
segmenting
image.
addition,
over
model,
Dice
Jaccard
coefficients
97.62%
95.43%
segmentation,
95.07%,
91.99%
segmentation.
Moreover,
have
obtained
97.45%
95.00%
HC18-Grand
dataset.
Based
on
results,
concluded
fine-tuned
simple
well-structured
devices
can
outperform
models.
Phenomics,
Год журнала:
2023,
Номер
3(3), С. 285 - 299
Опубликована: Янв. 5, 2023
Abstract
The
rapid
development
of
such
research
field
as
multi-omics
and
artificial
intelligence
(AI)
has
made
it
possible
to
acquire
analyze
the
multi-dimensional
big
data
human
phenomes.
Increasing
evidence
indicated
that
phenomics
can
provide
a
revolutionary
strategy
approach
for
discovering
new
risk
factors,
diagnostic
biomarkers
precision
therapies
diseases,
which
holds
profound
advantages
over
conventional
approaches
realizing
medicine:
first,
patients'
phenomes
remarkably
richer
information
than
genomes;
second,
phenomic
studies
on
diseases
may
expose
correlations
among
cross-scale
parameters
well
mechanisms
underlying
correlations;
third,
phenomics-based
are
data-driven
studies,
significantly
enhance
possibility
efficiency
generating
novel
discoveries.
However,
still
in
early
developmental
stage,
facing
multiple
major
challenges
tasks:
there
is
significant
deficiency
analytical
modeling
analyzing
phenomes;
crucial
establish
universal
standards
acquirement
management
patients;
methods
devices
patients
under
clinical
settings
should
be
developed;
fourth,
significance
regulatory
ethical
guidelines
diseases;
fifth,
important
develop
effective
international
cooperation.
It
expected
would
profoundly
comprehensively
our
capacity
prevention,
diagnosis
treatment
diseases.
Diagnostics,
Год журнала:
2025,
Номер
15(6), С. 689 - 689
Опубликована: Март 11, 2025
The
widespread
use
of
medical
imaging
techniques
such
as
X-rays
and
computed
tomography
(CT)
has
raised
significant
concerns
regarding
ionizing
radiation
exposure,
particularly
among
vulnerable
populations
requiring
frequent
imaging.
Achieving
a
balance
between
high-quality
diagnostic
minimizing
exposure
remains
fundamental
challenge
in
radiology.
Artificial
intelligence
(AI)
emerged
transformative
solution,
enabling
low-dose
protocols
that
enhance
image
quality
while
significantly
reducing
doses.
This
review
explores
the
role
AI-assisted
imaging,
CT,
X-ray,
magnetic
resonance
(MRI),
highlighting
advancements
deep
learning
models,
convolutional
neural
networks
(CNNs),
other
AI-based
approaches.
These
technologies
have
demonstrated
substantial
improvements
noise
reduction,
artifact
removal,
real-time
optimization
parameters,
thereby
enhancing
accuracy
mitigating
risks.
Additionally,
AI
contributed
to
improved
radiology
workflow
efficiency
cost
reduction
by
need
for
repeat
scans.
also
discusses
emerging
directions
AI-driven
including
hybrid
systems
integrate
post-processing
with
data
acquisition,
personalized
tailored
patient
characteristics,
expansion
applications
fluoroscopy
positron
emission
(PET).
However,
challenges
model
generalizability,
regulatory
constraints,
ethical
considerations,
computational
requirements
must
be
addressed
facilitate
broader
clinical
adoption.
potential
revolutionize
safety,
optimizing
quality,
improving
healthcare
efficiency,
paving
way
more
advanced
sustainable
future
Computers in Biology and Medicine,
Год журнала:
2021,
Номер
136, С. 104752 - 104752
Опубликована: Авг. 8, 2021
The
aim
of
this
study
was
to
identify
the
most
important
features
and
assess
their
discriminative
power
in
classification
subtypes
NSCLC.This
involved
354
pathologically
proven
NSCLC
patients
including
134
squamous
cell
carcinoma
(SCC),
110
large
(LCC),
62
not
other
specified
(NOS),
48
adenocarcinoma
(ADC).
In
total,
1433
radiomics
were
extracted
from
3D
volumes
interest
drawn
on
malignant
lesion
identified
CT
images.
Wrapper
algorithm
multivariate
adaptive
regression
splines
implemented
relevant/discriminative
features.
A
multivariable
multinomial
logistic
employed
with
1000
bootstrapping
samples
based
selected
classify
four
main
NSCLC.The
results
revealed
that
texture
features,
specifically
gray
level
size
zone
matrix
(GLSZM),
significant
indicators
subtypes.
optimized
classifier
achieved
an
average
precision,
recall,
F1-score,
accuracy
0.710,
0.703,
0.706,
0.865,
respectively,
by
wrapper
algorithm.Our
approach
demonstrated
impressive
potential
for
histological
NSCLC,
It
is
anticipated
could
be
useful
treatment
planning
precision
medicine.
European Journal of Nuclear Medicine and Molecular Imaging,
Год журнала:
2022,
Номер
50(4), С. 980 - 995
Опубликована: Дек. 5, 2022
Quantitative
SPECT-CT
is
a
modality
of
growing
importance
with
initial
developments
in
post
radionuclide
therapy
dosimetry,
and
more
recent
expansion
into
bone,
cardiac
brain
imaging
together
the
concept
theranostics
generally.
The
aim
this
document
to
provide
guidelines
for
nuclear
medicine
departments
setting
up
developing
their
quantitative
service
guidance
on
protocols,
harmonisation
clinical
use
cases.These
practice
were
written
by
members
European
Association
Nuclear
Medicine
Physics,
Dosimetry,
Oncology
Bone
committees
representing
current
major
stakeholders
SPECT-CT.
have
also
been
reviewed
approved
all
EANM
endorsed
Medicine.The
present
will
help
practitioners,
scientists
researchers
perform
high-quality
framework
continuing
development
as
an
established
modality.
Journal of Digital Imaging,
Год журнала:
2021,
Номер
34(5), С. 1086 - 1098
Опубликована: Авг. 11, 2021
The
aim
of
this
work
is
to
investigate
the
applicability
radiomic
features
alone
and
in
combination
with
clinical
information
for
prediction
renal
cell
carcinoma
(RCC)
patients'
overall
survival
after
partial
or
radical
nephrectomy.
Clinical
studies
210
RCC
patients
from
Cancer
Imaging
Archive
(TCIA)
who
underwent
either
nephrectomy
were
included
study.
Regions
interest
(ROIs)
manually
defined
on
CT
images.
A
total
225
extracted
analyzed
along
59
features.
An
elastic
net
penalized
Cox
regression
was
used
feature
selection.
Accelerated
failure
time
(AFT)
shared
frailty
model
determine
effects
selected
time.
Eleven
twelve
based
their
non-zero
coefficients.
Tumor
grade,
tumor
malignancy,
pathology
t-stage
most
significant
predictors
(OS)
among
(p
<
0.002,
0.02,
0.018,
respectively).
OS
flatness,
area
density,
median
0.05,
Along
important
features,
such
as
heterogeneity
imaging
biomarkers
are
significantly
correlated
patients.
Cancers,
Год журнала:
2023,
Номер
15(3), С. 708 - 708
Опубликована: Янв. 24, 2023
The
incidence
of
thyroid
nodules
diagnosed
is
increasing
every
year,
leading
to
a
greater
risk
unnecessary
procedures
being
performed
or
wrong
diagnoses
made.
In
our
paper,
we
present
the
latest
knowledge
on
use
artificial
intelligence
in
diagnosing
and
classifying
nodules.
We
particularly
focus
usefulness
ultrasonography
for
diagnosis
characterization
pathology,
as
these
are
two
most
developed
fields.
search
innovations,
reviewed
only
publications
specific
types
published
from
2018
2022.
analyzed
930
papers
total,
which
selected
33
that
were
relevant
topic
work.
conclusion,
there
great
scope
future
nodule
classification
diagnosis.
addition
typical
uses
cancer
differentiation,
identified
several
other
novel
applications
during
review.
Neural Regeneration Research,
Год журнала:
2023,
Номер
18(10), С. 2134 - 2134
Опубликована: Янв. 1, 2023
The
scientists
are
dedicated
to
studying
the
detection
of
Alzheimer's
disease
onset
find
a
cure,
or
at
very
least,
medication
that
can
slow
progression
disease.
This
article
explores
effectiveness
longitudinal
data
analysis,
artificial
intelligence,
and
machine
learning
approaches
based
on
magnetic
resonance
imaging
positron
emission
tomography
neuroimaging
modalities
for
estimation
onset.
significance
feature
extraction
in
highly
complex
data,
identification
vulnerable
brain
regions,
determination
threshold
values
plaques,
tangles,
neurodegeneration
these
regions
will
extensively
be
evaluated.
Developing
automated
methods
improve
aforementioned
research
areas
would
enable
specialists
determine
link
between
biomarkers
more
accurate
