Deep Learning-based Acceleration and Denoising of 0.55T MRI for Enhanced Conspicuity of Vestibular Schwannoma Post Contrast Administration
Abstract
Background
and
Purpose
Deep-learning
(DL)
based
MRI
denoising
techniques
promise
improved
image
quality
shorter
examination
times.
This
advancement
is
particularly
beneficial
for
0.55T
MRI,
where
the
inherently
lower
signal-to-noise
(SNR)
ratio
can
compromise
quality.
Sufficient
SNR
crucial
reliable
detection
of
vestibular
schwannoma
(VS).
The
objective
this
study
to
evaluate
VS
conspicuity
acquisition
time
(TA)
examinations
with
contrast
agents
using
a
DL-denoising
algorithm.
Materials
Methods
From
January
2024
October
2024,
we
retrospectively
included
30
patients
(9
women).
We
acquired
clinical
reference
protocol
cerebellopontine
angle
containing
T1w
fat-saturated
(fs)
axial
(number
signal
averages
[NSA]
4)
Spectral
Attenuated
Inversion
Recovery
(SPAIR)
coronal
(NSA
2)
sequences
after
agent
(CA)
application
without
advanced
DL-based
(w/o
DL).
reconstructed
fs
CA
sequence
SPAIR
full
mode
change
NSA
secondly
1
(DL
&
1NSA).
Each
was
rated
on
5-point
Likert
scale
(1:
insufficient,
3:
moderate,
clinically
sufficient;
5:
perfect)
for:
overall
quality;
conspicuity,
artifacts.
Secondly,
analyzed
reliability
size
measurements.
Two
radiologists
focus
head
neck
imaging
performed
reading
Wilcoxon
Signed-Rank
Test
used
non-parametric
statistical
comparison.
Results
DL
4NSA
axial/coronal
achieved
highest
IQ
(median
4.9).
(IQ)
1NSA
higher
(M:
4.0)
than
;
median
4.0
versus
3.5,
each
p
<
0.01).
Similarly,
best
4.9),
decreased
&1NSA
4.1)
but
still
sufficient
w/o
3.7,
TA
post-contrast
were
8:59
minutes
3:24
DL&
1NSA.
Conclusions
underlines
that
reduce
by
more
half
while
simultaneously
improving
Published: May 13, 2025
Language: Английский