Insights into Imaging,
Год журнала:
2024,
Номер
15(1)
Опубликована: Авг. 1, 2024
Abstract
Objectives
To
explore
the
predictive
performance
of
tumor
and
multiple
peritumoral
regions
on
dynamic
contrast-enhanced
magnetic
resonance
imaging
(MRI),
to
identify
optimal
interest
for
developing
a
preoperative
model
grade
microvascular
invasion
(MVI).
Methods
A
total
147
patients
who
were
surgically
diagnosed
with
hepatocellular
carcinoma,
had
maximum
diameter
≤
5
cm
recruited
subsequently
divided
into
training
set
(
n
=
117)
testing
30)
based
date
surgery.
We
utilized
pre-trained
AlexNet
extract
deep
learning
features
from
seven
different
transverse
cross-section
tumors
in
various
MRI
sequence
images.
Subsequently,
an
extreme
gradient
boosting
(XGBoost)
classifier
was
employed
construct
MVI
prediction
model,
evaluation
area
under
curve
(AUC).
Results
The
XGBoost
trained
data
20-mm
region
showed
superior
AUC
compared
alone.
values
consistently
increased
when
utilizing
5-mm,
10-mm,
regions.
Combining
arterial
delayed-phase
yielded
highest
performance,
micro-
macro-average
AUCs
0.78
0.74,
respectively.
Integration
clinical
further
improved
0.83
0.80.
Conclusion
Compared
those
region,
provide
more
important
information
predicting
MVI.
resulted
relatively
ideal
accurate
within
which
can
be
predicted.
Clinical
relevance
statement
holds
significance
than
grade.
Deep
indirectly
predict
by
extracting
directly
capturing
region.
Key
Points
investigated
regions,
as
well
their
fusion.
predominantly
occurs
predictor
20
mm
is
reasonable
accurately
three-grade
Graphical
European Journal of Radiology,
Год журнала:
2020,
Номер
132, С. 109312 - 109312
Опубликована: Сен. 28, 2020
PurposeTo
investigate
whether
Liver
Imaging
Reporting
and
Data
System
(LI-RADS)
imaging
features
at
preoperative
gadoxetic
acid-enhanced
MRI
can
predict
microvascular
invasion
(MVI)
histologic
grade
of
hepatocellular
carcinoma
(HCC)
to
evaluate
their
associations
with
recurrence
after
curative
resection
single
HCC.Materials
methodsFrom
July
2015
September
2018,
111
consecutive
patients
pathologically
confirmed
HCC
who
underwent
acid–enhanced
within
1
month
before
surgery
were
included
in
this
retrospective
study.
Significant
findings
clinical
parameters
for
predicting
MVI,
high-grade
HCCs
postoperative
identified
by
logistic
regression
model
Cox
proportional
hazards
model.ResultsTwenty-six
(23.4
%)
had
MVI
36
(32.4
HCCs,
whereas
44
95
(46.3
experienced
recurrence.
Tumor
size
>
5
cm
(OR
=
9.852;
p
<
0.001)
absence
nodule-in-nodule
architecture
8.302;
independent
predictors
MVI.
Enhancing
capsule
4.396;
0.004)
corona
enhancement
3.765;
0.021)
HCCs.
Blood
products
mass
(HR
2.275;
0.009),
4.332;
0.001),
serum
AFP
level
400
ng/mL
2.071;
0.023)
recurrence.ConclusionLI-RADS
be
used
as
potential
biomarkers
aggressive
pathologic
HCC.
The
identification
prognostic
LI-RADS
may
facilitate
the
selection
surgical
candidates
optimize
management
patients.
Frontiers in Oncology,
Год журнала:
2021,
Номер
11
Опубликована: Окт. 7, 2021
To
develop
and
validate
an
MR
radiomics-based
nomogram
to
predict
the
presence
of
MVI
in
patients
with
solitary
HCC
further
evaluate
performance
predictors
for
subgroups
(HCC
≤
3
cm
>
cm).Between
May
2015
October
2020,
201
were
analysed.
Radiomic
features
extracted
from
precontrast
T1WI,
arterial
phase,
portal
venous
delayed
phase
hepatobiliary
images
regions
intratumoral,
peritumoral
their
combining
areas.
The
mRMR
LASSO
algorithms
used
select
radiomic
related
MVI.
Clinicoradiological
factors
selected
by
using
backward
stepwise
regression
AIC.
A
was
developed
incorporating
clinicoradiological
radiomics
signature.
In
addition,
separately
evaluated
cm).Histopathological
examinations
confirmed
111
(55.22%).
signature
showed
a
favourable
discriminatory
ability
training
set
(AUC,
0.896)
validation
0.788).
enhancement,
tumour
growth
type
good
discrimination
0.932)
sets
0.917)
achieved
well-fitted
calibration
curves.
Subgroup
analysis
that
predictor
cohort
enhancement
cohort;
varied
between
cohort.
improved
noticeably
both
0.953)
cohorts
0.993)
compared
original
set.The
preoperative
integrating
risk
predictive
efficiency
predicting
HCC.
cm).
prediction
subgroups.
Frontiers in Oncology,
Год журнала:
2021,
Номер
11
Опубликована: Март 16, 2021
To
investigate
microvascular
invasion
(MVI)
of
HCC
through
a
noninvasive
multi-disciplinary
team
(MDT)-like
radiomics
fusion
model
on
dynamic
contrast
enhanced
(DCE)
computed
tomography
(CT).
This
retrospective
study
included
111
patients
with
pathologically
proven
hepatocellular
carcinoma,
which
comprised
57
MVI-positive
and
54
MVI-negative
patients.
Target
volume
interest
(VOI)
was
delineated
four
DCE
CT
phases.
The
tumor
core
(V
tc
)
seven
peripheral
regions
pt
,
varying
distances
2,
4,
6,
8,
10,
12,
14
mm
to
margin)
were
obtained.
Radiomics
features
extracted
from
different
combinations
phase(s)
VOI(s)
cross-validated
by
150
classification
models.
best
phase
VOI
(or
combinations)
determined.
top
predictive
models
ranked
screened
cross-validation
the
training/validation
set.
fusion,
procedure
analogous
multidisciplinary
consultation,
performed
top-3
generate
final
model,
validated
an
independent
testing
Image
V
+V
pt(12mm)
in
portal
venous
(PVP)
showed
dominant
performances.
PVP
one
gray
level
size
zone
matrix
(GLSZM)-based
feature
first-order
based
features.
Model
outperformed
single
MVI
prediction.
weighted
method
achieved
performance
AUC
0.81,
accuracy
78.3%,
sensitivity
81.8%,
specificity
75%
are
most
reliable
indicative
MVI.
MDT-like
is
promising
tool
accurate
reproducible
results
status
prediction
HCC.
