Cancers,
Год журнала:
2025,
Номер
17(5), С. 768 - 768
Опубликована: Фев. 24, 2025
Radiomics
has
seen
substantial
growth
in
medical
imaging;
however,
its
potential
optical
coherence
tomography
(OCT)
not
been
widely
explored.
We
systematically
evaluate
the
repeatability
and
reproducibility
of
handcrafted
radiomics
features
(HRFs)
from
OCT
scans
benign
nevi
examine
impact
bin
width
(BW)
selection
on
HRF
stability.
The
effect
using
stable
a
classification
model
was
also
assessed.
In
this
prospective
study,
20
volunteers
underwent
test-retest
imaging
40
nevi,
resulting
80
scans.
HRFs
extracted
manually
delineated
regions
interest
(ROIs)
were
assessed
concordance
correlation
coefficients
(CCCs)
across
BWs
ranging
5
to
50.
A
unique
set
identified
at
each
BW
after
removing
highly
correlated
eliminate
redundancy.
These
robust
incorporated
into
multiclass
classifier
trained
distinguish
basal
cell
carcinoma
(BCC),
Bowen's
disease.
Six
all
BWs,
with
25
emerging
as
optimal
choice,
balancing
ability
capture
meaningful
textural
details.
Additionally,
intermediate
(20-25)
yielded
53
reproducible
features.
six
achieved
90%
accuracy
AUCs
0.96
0.94
for
BCC
disease,
respectively,
compared
76%
0.86
0.80
conventional
feature
approach.
This
study
highlights
critical
role
enhancing
stability
provides
methodological
framework
optimizing
preprocessing
radiomics.
By
demonstrating
integration
diagnostic
models,
we
establish
promising
tool
aid
non-invasive
diagnosis
dermatology.
European Radiology Experimental,
Год журнала:
2024,
Номер
8(1)
Опубликована: Май 14, 2024
Overall
quality
of
radiomics
research
has
been
reported
as
low
in
literature,
which
constitutes
a
major
challenge
to
improve.
Consistent,
transparent,
and
accurate
reporting
is
critical,
can
be
accomplished
with
systematic
use
guidelines.
The
CheckList
for
EvaluAtion
Radiomics
(CLEAR)
was
previously
developed
assist
authors
their
radiomic
reviewers
evaluation.
To
take
full
advantage
CLEAR,
further
explanation
elaboration
each
item,
well
literature
examples,
may
useful.
main
goal
this
work,
Explanation
Elaboration
Examples
CLEAR
(CLEAR-E3),
improve
CLEAR's
usability
dissemination.
In
international
collaborative
effort,
members
the
European
Society
Medical
Imaging
Informatics-Radiomics
Auditing
Group
searched
identify
representative
examples
item.
At
least
two
demonstrating
optimal
reporting,
were
presented
All
selected
from
open-access
articles,
allowing
users
easily
consult
corresponding
full-text
articles.
addition
these,
item's
expanded
elaborated.
For
easier
access,
resulting
document
available
at
https://radiomic.github.io/CLEAR-E3/
.
As
complementary
effort
we
anticipate
that
initiative
will
greater
ease
transparency,
editors
reviewing
manuscripts.Relevance
statement
Along
original
checklist,
CLEAR-E3
expected
provide
more
in-depth
understanding
items,
concrete
evaluating
research.Key
points•
aims
research,
manuscripts.•
Based
on
positive
by
EuSoMII
Group,
item
elaborated
CLEAR-E3.•
accessed
European Radiology,
Год журнала:
2025,
Номер
unknown
Опубликована: Янв. 9, 2025
Abstract
Objectives
To
investigate
how
studies
determine
the
sample
size
when
developing
radiomics
prediction
models
for
binary
outcomes,
and
whether
meets
estimates
obtained
by
using
established
criteria.
Methods
We
identified
that
were
published
from
01
January
2023
to
31
December
in
seven
leading
peer-reviewed
radiological
journals.
reviewed
justification
methods,
actual
used.
calculated
compared
used
three
criteria
proposed
Riley
et
al.
investigated
which
characteristics
factors
associated
with
sufficient
Results
included
116
studies.
Eleven
out
of
one
hundred
sixteen
justified
size,
6/11
performed
a
priori
calculation.
The
median
(first
third
quartile,
Q1,
Q3)
total
is
223
(130,
463),
those
training
are
150
(90,
288).
(Q1,
difference
between
minimum
according
−100
(−216,
183),
differences
more
restrictive
approach
based
on
−268
(−427,
−157).
presence
external
testing
specialty
topic
size.
Conclusion
Radiomics
often
designed
without
justification,
whose
may
be
too
small
avoid
overfitting.
Sample
encouraged
model.
Key
Points
Question
critical
help
minimize
overfitting
model,
but
overlooked
underpowered
research
.
Findings
Few
justified,
calculated,
or
reported
their
most
them
did
not
meet
recent
formal
Clinical
relevance
justification.
Consequently,
many
It
should
justify,
perform,
report
considerations
Magnetic Resonance in Medical Sciences,
Год журнала:
2023,
Номер
22(4), С. 401 - 414
Опубликована: Янв. 1, 2023
Due
primarily
to
the
excellent
soft
tissue
contrast
depictions
provided
by
MRI,
widespread
application
of
head
and
neck
MRI
in
clinical
practice
serves
assess
various
diseases.
Artificial
intelligence
(AI)-based
methodologies,
particularly
deep
learning
analyses
using
convolutional
neural
networks,
have
recently
gained
global
recognition
been
extensively
investigated
research
for
their
applicability
across
a
range
categories
within
medical
imaging,
including
MRI.
Analytical
approaches
AI
shown
potential
addressing
limitations
associated
with
In
this
review,
we
focus
on
technical
advancements
deep-learning-based
methodologies
utility
field
encompassing
aspects
such
as
image
acquisition
reconstruction,
lesion
segmentation,
disease
classification
diagnosis,
prognostic
prediction
patients
presenting
We
then
discuss
current
offer
insights
regarding
future
challenges
field.