<s>
Real-time	O
magnetic	O
resonance	O
imaging	O
(	O
RT-MRI	O
)	O
refers	O
to	O
the	O
continuous	O
monitoring	O
(	O
"	O
filming	O
"	O
)	O
of	O
moving	O
objects	O
in	O
real	O
time	O
.	O
</s>
<s>
Because	O
MRI	B-Algorithm
is	O
based	O
on	O
time-consuming	O
scanning	O
of	O
k-space	B-Algorithm
,	O
real-time	B-Algorithm
MRI	I-Algorithm
was	O
possible	O
only	O
with	O
low	O
image	O
quality	O
or	O
low	O
temporal	O
resolution	O
.	O
</s>
<s>
Using	O
an	O
iterative	B-Algorithm
reconstruction	I-Algorithm
algorithm	O
these	O
limitations	O
have	O
recently	O
been	O
removed	O
:	O
a	O
new	O
method	O
for	O
real-time	B-Algorithm
MRI	I-Algorithm
achieves	O
a	O
temporal	O
resolution	O
of	O
20	O
to	O
30	O
milliseconds	O
for	O
images	O
with	O
an	O
in-plane	O
resolution	O
of	O
1.5	O
to	O
2.0mm	O
.	O
</s>
<s>
Real-time	B-Algorithm
MRI	I-Algorithm
promises	O
to	O
add	O
important	O
information	O
about	O
diseases	O
of	O
the	O
joints	O
and	O
the	O
heart	O
.	O
</s>
<s>
1977/1978	O
-	O
Raymond	O
Damadian	O
built	O
the	O
first	O
MRI	B-Algorithm
scanner	O
and	O
achieved	O
the	O
first	O
MRI	B-Algorithm
scan	O
of	O
a	O
healthy	O
human	O
body	O
(	O
1977	O
)	O
with	O
the	O
intent	O
of	O
diagnosing	O
cancer	O
.	O
</s>
<s>
Additionally	O
,	O
Peter	O
Mansfield	O
develops	O
the	O
echo-planar	O
technique	O
,	O
producing	O
images	O
in	O
seconds	O
and	O
becoming	O
the	O
basis	O
for	O
fast	O
MRIs	B-Algorithm
.	O
</s>
<s>
In	O
general	O
,	O
real	O
time	O
MRI	B-Algorithm
relies	O
on	O
gradient	O
echo	O
sequences	O
,	O
efficient	O
k-space	B-Algorithm
sampling	O
,	O
and	O
fast	O
reconstruction	O
methods	O
to	O
speed	O
up	O
the	O
image	O
acquisition	O
process	O
.	O
</s>
<s>
Efficient	O
k-space	B-Algorithm
sampling	O
also	O
decreases	O
data	O
collection	O
time	O
.	O
</s>
<s>
Rectilinear	O
scanning	O
has	O
become	O
the	O
standard	O
k-space	B-Algorithm
sampling	O
method	O
for	O
MRI	B-Algorithm
.	O
</s>
<s>
However	O
,	O
the	O
process	O
takes	O
a	O
relatively	O
long	O
time	O
as	O
it	O
samples	O
the	O
entire	O
k-space	B-Algorithm
equally	O
.	O
</s>
<s>
However	O
,	O
it	O
is	O
important	O
to	O
note	O
that	O
the	O
EPI	O
method	O
is	O
still	O
a	O
Cartesian	O
sampling	O
method	O
,	O
like	O
the	O
rectilinear	O
scan	O
,	O
equally	O
sampling	O
the	O
entire	O
k-space	B-Algorithm
.	O
</s>
<s>
Spiral	O
sampling	O
,	O
like	O
EPI	O
,	O
only	O
requires	O
a	O
single	O
RF	O
pulse	O
to	O
sample	O
the	O
entire	O
k-space	B-Algorithm
.	O
</s>
<s>
Radial	O
and	O
spiral	O
sampling	O
are	O
also	O
used	O
as	O
methods	O
to	O
efficiently	O
sample	O
the	O
k-space	B-Algorithm
,	O
with	O
spiral	O
also	O
only	O
requiring	O
a	O
single	O
RF	O
pulse	O
to	O
sample	O
the	O
k-space	B-Algorithm
.	O
</s>
<s>
Both	O
radial	O
and	O
spiral	O
sampling	O
are	O
more	O
efficient	O
than	O
the	O
Cartesian	O
methods	O
because	O
they	O
oversample	O
low	O
frequencies	O
,	O
which	O
allows	O
for	O
general	O
motion	O
capture	O
and	O
better	O
real-time	O
image	B-Algorithm
reconstruction	I-Algorithm
.	O
</s>
<s>
Thus	O
,	O
radial	O
or	O
spiral	O
sampling	O
of	O
the	O
k-space	B-Algorithm
are	O
now	O
the	O
preferred	O
methods	O
for	O
real-time	B-Algorithm
MRI	I-Algorithm
reconstruction	O
.	O
</s>
<s>
While	O
early	O
applications	O
were	O
based	O
on	O
echo	O
planar	O
imaging	O
,	O
which	O
found	O
an	O
important	O
application	O
in	O
real-time	O
functional	B-Algorithm
MRI	I-Algorithm
(	O
rt-fMRI	O
)	O
,	O
recent	O
progress	O
is	O
based	O
on	O
iterative	B-Algorithm
reconstruction	I-Algorithm
and	O
FLASH	B-Algorithm
MRI	I-Algorithm
.	O
</s>
<s>
The	O
real-time	O
imaging	O
method	O
proposed	O
by	O
Uecker	O
and	O
colleagues	O
combines	O
radial	O
FLASH	B-Algorithm
MRI	I-Algorithm
,	O
which	O
offers	O
rapid	O
and	O
continuous	O
data	O
acquisition	O
,	O
motion	O
robustness	O
,	O
and	O
tolerance	O
to	O
undersampling	O
,	O
with	O
an	O
iterative	B-Algorithm
image	I-Algorithm
reconstruction	I-Algorithm
method	O
based	O
on	O
the	O
formulation	O
of	O
image	B-Algorithm
reconstruction	I-Algorithm
as	O
a	O
nonlinear	O
inverse	O
problem	O
.	O
</s>
<s>
parallel	O
MRI	B-Algorithm
)	O
and	O
exploiting	O
the	O
redundancy	B-General_Concept
in	O
the	O
time	O
series	O
of	O
images	O
with	O
the	O
use	O
of	O
regularization	O
and	O
filtering	O
,	O
this	O
approach	O
enhances	O
the	O
possible	O
degree	O
of	O
data	O
undersampling	O
by	O
one	O
order	O
of	O
magnitude	O
,	O
so	O
that	O
high-quality	O
images	O
may	O
be	O
obtained	O
out	O
of	O
as	O
little	O
as	O
5	O
to	O
10%	O
of	O
the	O
data	O
required	O
for	O
a	O
normal	O
image	B-Algorithm
reconstruction	I-Algorithm
.	O
</s>
<s>
Another	O
GRE	O
sequence	O
commonly	O
used	O
in	O
RT-MRI	O
is	O
balanced	O
steady	B-Algorithm
state	I-Algorithm
free	I-Algorithm
precession	I-Algorithm
(	O
bSSFP	B-Algorithm
)	O
,	O
as	O
mentioned	O
above	O
with	O
balanced	O
gradients	O
.	O
</s>
<s>
Steady	B-Algorithm
state	I-Algorithm
free	I-Algorithm
precession	I-Algorithm
involves	O
a	O
repetition	O
time	O
(	O
TR	O
)	O
that	O
is	O
shorter	O
than	O
T2	O
.	O
</s>
<s>
The	O
short	O
TR	O
also	O
makes	O
bSSFP	B-Algorithm
ideal	O
for	O
RT-MRI	O
.	O
</s>
<s>
The	O
equation	O
for	O
peak	O
MR	O
signal	O
in	O
bSSFP	B-Algorithm
is	O
given	O
as	O
:	O
</s>
<s>
Due	O
to	O
this	O
strong	O
fluid/tissue	O
contrast	O
,	O
RT-MRI	O
with	O
bSSFP	B-Algorithm
lends	O
itself	O
to	O
cardiac	O
imaging	O
and	O
visualizing	O
blood	O
flow	O
.	O
</s>
<s>
Certain	O
image	B-Algorithm
reconstruction	I-Algorithm
algorithms	O
used	O
alongside	O
parallel	O
imaging	O
address	O
the	O
potential	O
issues	O
that	O
can	O
arise	O
from	O
undersampling	O
the	O
k-space	B-Algorithm
.	O
</s>
<s>
SENSitivity	O
Encoding	O
(	O
SENSE	O
)	O
is	O
a	O
method	O
that	O
reconstructs	O
the	O
partial	O
k-space	B-Algorithm
data	O
from	O
each	O
coil	O
and	O
combines	O
the	O
partial	O
images	O
into	O
the	O
final	O
scan	O
in	O
the	O
spatial	O
domain	O
.	O
</s>
<s>
During	O
the	O
rest	O
of	O
imaging	O
,	O
the	O
k-space	B-Algorithm
is	O
undersampled	O
to	O
skip	O
every	O
other	O
line	O
,	O
resulting	O
in	O
a	O
½	O
FOV	O
.	O
</s>
<s>
As	O
a	O
two-point	O
example	O
,	O
pixels	O
on	O
the	O
original	O
aliased	B-Error_Name
images	O
can	O
be	O
“	O
unfolded	O
”	O
through	O
the	O
following	O
equations	O
to	O
give	O
the	O
final	O
scan	O
:	O
</s>
<s>
and	O
denote	O
the	O
image	O
signal	O
for	O
the	O
aliased	B-Error_Name
image	O
.	O
</s>
<s>
GRAPPA	O
fills	O
in	O
the	O
undersampled	O
k-space	B-Algorithm
data	O
in	O
the	O
k-space	B-Algorithm
domain	O
before	O
reconstructing	O
the	O
final	O
image	O
.	O
</s>
<s>
Lines	O
through	O
the	O
center	O
of	O
the	O
k-space	B-Algorithm
are	O
fully	O
sampled	O
,	O
typically	O
alongside	O
the	O
actual	O
image	O
,	O
to	O
give	O
the	O
autocalibration	O
signal	O
(	O
ACS	O
)	O
region	O
.	O
</s>
<s>
Then	O
,	O
the	O
filled-in	O
k-space	B-Algorithm
data	O
undergoes	O
the	O
inverse	O
Fourier	O
transform	O
to	O
construct	O
the	O
partial	O
,	O
non-aliased	O
images	O
.	O
</s>
<s>
If	O
the	O
k-space	B-Algorithm
data	O
is	O
non-Cartesian	O
,	O
reconstruction	O
is	O
computationally	O
more	O
difficult	O
,	O
since	O
the	O
fast	O
Fourier	O
transform	O
(	O
FFT	O
)	O
requires	O
Cartesian	O
values	O
.	O
</s>
<s>
Typically	O
,	O
k-space	B-Algorithm
data	O
must	O
be	O
resampled	O
into	O
Cartesian	O
coordinates	O
before	O
applying	O
the	O
FFT	O
.	O
</s>
<s>
Lastly	O
,	O
within	O
parallel	O
image	B-Algorithm
reconstruction	I-Algorithm
there	O
is	O
another	O
factor	O
to	O
consider	O
,	O
which	O
is	O
the	O
signal	O
to	O
noise	O
ratio	O
(	O
SNR	O
)	O
.	O
</s>
<s>
Although	O
applications	O
of	O
real-time	B-Algorithm
MRI	I-Algorithm
cover	O
a	O
broad	O
spectrum	O
ranging	O
from	O
non-medical	O
studies	O
of	O
turbulent	O
flow	O
to	O
the	O
noninvasive	O
monitoring	O
of	O
interventional	O
(	O
surgical	O
)	O
procedures	O
,	O
the	O
most	O
important	O
application	O
making	O
use	O
of	O
the	O
new	O
capabilities	O
is	O
cardiovascular	O
imaging	O
.	O
</s>
<s>
Previous	O
cardiac	B-Algorithm
MR	I-Algorithm
(	O
CMR	O
)	O
used	O
cine	O
techniques	O
to	O
capture	O
the	O
periodic	O
motion	O
of	O
the	O
heart	O
.	O
</s>
<s>
However	O
,	O
this	O
is	O
not	O
feasible	O
for	O
patients	O
with	O
arrhythmia	B-Application
,	O
where	O
the	O
cardiac	O
cycle	O
is	O
unpredictable	O
.	O
</s>
<s>
With	O
the	O
new	O
method	O
it	O
is	O
possible	O
to	O
obtain	O
movies	O
of	O
the	O
beating	O
heart	O
in	O
real	O
time	O
with	O
up	O
to	O
50	O
frames	O
per	O
second	O
during	O
free	O
breathing	O
and	O
without	O
the	O
need	O
for	O
a	O
synchronization	O
to	O
the	O
electrocardiogram	B-Application
.	O
</s>
<s>
al	O
demonstrated	O
that	O
RT-MRI	O
produced	O
higher	O
quality	O
images	O
with	O
a	O
higher	O
SNR	O
than	O
cine	O
CMR	O
with	O
a	O
bSSFP	B-Algorithm
sequence	O
and	O
radial	O
k-space	B-Algorithm
sampling	O
.	O
</s>
<s>
RT-MRI	O
also	O
removes	O
the	O
need	O
for	O
breath-holding	O
while	O
imaging	O
,	O
leading	O
to	O
a	O
more	O
comfortable	O
experience	O
for	O
the	O
patient	O
as	O
well	O
.	O
</s>
<s>
Apart	O
from	O
cardiac	B-Algorithm
MRI	I-Algorithm
other	O
real-time	O
applications	O
deal	O
with	O
functional	O
studies	O
of	O
joint	O
kinetics	O
(	O
e.g.	O
,	O
temporomandibular	O
joint	O
,	O
knee	O
and	O
the	O
wrist	O
)	O
or	O
address	O
the	O
coordinated	O
dynamics	O
of	O
the	O
articulators	O
such	O
as	O
lips	O
,	O
tongue	O
,	O
soft	O
palate	O
and	O
vocal	O
folds	O
during	O
speaking	O
(	O
articulatory	O
phonetics	O
)	O
or	O
swallowing	O
.	O
</s>
<s>
Researchers	O
at	O
the	O
NYU	O
Grossman	O
School	O
of	O
Medicine	O
developed	O
a	O
RT-MRI	O
glove	O
for	O
imaging	O
movement	O
of	O
the	O
hand	O
.	O
</s>
<s>
The	O
glove	O
uses	O
high	O
impedance	O
coils	O
to	O
prevent	O
the	O
generation	O
of	O
eddy	O
currents	O
from	O
rapidly	O
changing	O
magnetic	O
fields	O
and	O
bSSFP	B-Algorithm
for	O
rapid	O
imaging	O
times	O
.	O
</s>
<s>
Applications	O
in	O
interventional	B-Algorithm
MRI	I-Algorithm
,	O
which	O
refers	O
to	O
the	O
monitoring	O
of	O
minimally	O
invasive	O
surgical	O
procedures	O
,	O
are	O
possible	O
by	O
interactively	O
changing	O
parameters	O
such	O
as	O
image	O
position	O
and	O
orientation	O
.	O
</s>
<s>
It	O
requires	O
an	O
in-room	O
display	O
for	O
the	O
physician	O
to	O
use	O
during	O
the	O
procedure	O
as	O
well	O
as	O
the	O
use	O
of	O
MRI-safe	O
surgical	O
tools	O
.	O
</s>
<s>
By	O
using	O
bSSFP	B-Algorithm
and	O
parallel	O
imaging	O
with	O
multiple	O
coils	O
,	O
frame	O
rates	O
of	O
5-10	O
frames	O
per	O
second	O
have	O
been	O
accomplished	O
,	O
allowing	O
for	O
the	O
visualization	O
of	O
cardiac	O
procedures	O
.	O
</s>
<s>
Image	B-Algorithm
reconstruction	I-Algorithm
in	O
RT-MRI	O
benefits	O
from	O
machine	O
learning	O
(	O
ML	O
)	O
or	O
deep	B-Algorithm
learning	I-Algorithm
(	O
DL	O
)	O
.	O
</s>
<s>
A	O
nonlinear	O
kernel	B-Algorithm
,	O
or	O
mapping	O
function	O
,	O
can	O
be	O
developed	O
from	O
the	O
ACS	O
to	O
fill	O
in	O
k-space	B-Algorithm
data	O
and	O
generate	O
the	O
final	O
image	O
.	O
</s>
<s>
This	O
process	O
as	O
a	O
whole	O
significantly	O
accelerates	O
the	O
MRI	B-Algorithm
process	O
.	O
</s>
<s>
In	O
deep	B-Algorithm
learning	I-Algorithm
,	O
with	O
a	O
convolutional	B-Architecture
neural	I-Architecture
network	I-Architecture
,	O
the	O
mapping	O
function	O
can	O
be	O
specified	O
by	O
the	O
network	O
.	O
</s>
<s>
High-performance	O
,	O
low-field	O
MRI	B-Algorithm
scanners	O
are	O
also	O
an	O
area	O
of	O
development	O
.	O
</s>
<s>
These	O
scanners	O
operate	O
at	O
relatively	O
low	O
magnetic	O
field	O
strengths	O
,	O
such	O
as	O
0.35	O
T	O
or	O
0.55	O
T	O
.	O
Many	O
RT-MRI	O
acquisition	O
sequences	O
,	O
such	O
as	O
bSSFP	B-Algorithm
,	O
experience	O
significant	O
off-resonance	O
effects	O
.	O
</s>
<s>
This	O
allows	O
for	O
longer	O
TRs	O
,	O
which	O
then	O
opens	O
the	O
door	O
for	O
a	O
wider	O
range	O
of	O
k-space	B-Algorithm
sampling	O
methods	O
and	O
sequence	O
designs	O
.	O
</s>
<s>
Finally	O
,	O
lower	O
strength	O
MRI	B-Algorithm
scanners	O
will	O
reduce	O
the	O
dangers	O
associated	O
with	O
heating	O
of	O
metallic	O
implants	O
and	O
decrease	O
the	O
cost	O
of	O
MRI	B-Algorithm
.	O
</s>
