Life,
Journal Year:
2023,
Volume and Issue:
13(10), P. 1950 - 1950
Published: Sept. 23, 2023
In
diagnostic
imaging,
distinguishing
chromophobe
renal
cell
carcinomas
(chRCCs)
from
oncocytomas
(ROs)
is
challenging,
since
they
both
present
similar
radiological
characteristics.
Radiomics
has
the
potential
to
help
in
differentiation
between
chRCCs
and
ROs
by
extracting
quantitative
imaging.
This
a
preliminary
study
of
role
radiomic
features
using
machine
learning
models.
this
retrospective
work,
38
subjects
were
involved:
19
diagnosed
with
ROs.
The
CT
nephrographic
contrast
phase
was
selected
each
case.
Three-dimensional
segmentations
lesions
performed
extracted.
To
assess
reliability
features,
intraclass
correlation
coefficient
calculated
three
radiologists
different
degrees
expertise.
selection
based
on
criteria
excellent
(ICC),
high
correlation,
statistical
significance.
Three
models
elaborated:
support
vector
(SVM),
random
forest
(RF),
logistic
regression
(LR).
From
105
extracted
41
presented
an
ICC
6
not
highly
correlated
other.
Only
two
showed
significant
differences
according
histological
type
developed
them.
LR
better
model,
particular,
83%
precision.
