A 24-ch Phased-Array System for Hyperpolarized Helium Gas Parallel MRI to Evaluate Lung Functions.
ABSTRACT Hyperpolarized 3He gas MRI has a serious potential for assessing pulmonary functions. Due to the fact that the non-equilibrium of the gas results in a steady depletion of the signal level over the course of the excitations, the signal-tonoise ratio (SNR) can be independent of the number of the data acquisitions under certain circumstances. This provides a unique opportunity for parallel MRI for gaining both temporal and spatial resolution without reducing SNR. We have built a 24-channel receive / 2-channel transmit phased array system for 3He parallel imaging. Our in vivo experimental results proved that the significant temporal and spatial resolution can be gained at no cost to the SNR. With 3D data acquisition, eight fold (2x4) scan time reduction can be achieved without any aliasing in images. Additionally, a rigid analysis using the low impedance preamplifier for decoupling presented evidence of strong coupling.
A 24-ch Phased-Array System for
Hyperpolarized Helium Gas Parallel MRI to
Evaluate Lung Functions
Ray F. Lee, Glyn Johnson,#Bernd Stoeckel, Cornel Stefanescu, Robert Trampel, Georgeann McGuinness
New York University Medical Center, 660 First Ave. First Floor, New York, NY 10016
# Siemens Medical Solution, Inc., New York, NY 10016
Abstract- Hyperpolarized3He gas MRI has a serious potential
for assessing pulmonary functions. Due to the fact that the
non-equilibrium of the gas results in a steady depletion of the
signal level over the course of the excitations, the signal-to-
noise ratio (SNR) can be independent of the number of the
data acquisitions under certain circumstances. This provides a
unique opportunity for parallel MRI for gaining both temporal
and spatial resolution without reducing SNR. We have built a
24-channel receive / 2-channel transmit phased array system
proved that the significant temporal and spatial resolution can
be gained at no cost to the SNR. With 3D data acquisition,
eight fold (2x4) scan time reduction can be achieved without
any aliasing in images. Additionally, a rigid analysis using the
low impedance preamplifier for decoupling presented evidence
of strong coupling.
3He parallel imaging. Our in vivo experimental results
A. SNR vs. the number of excitations
It is well known that the SNR is proportional to the
square root of the number of the excitation in the
conventional equilibrium MRI. However, this fundamental
concept is altered for the non-equilibrium hyperpolarized
gas MRI because of a steady depletion of the signal level
over the course of the excitations, in which SNR could be
increasing, decreasing, or independent when the number of
the excitation gets larger.
In the case of the gradient-recalled echo pulse
sequence, the k-space signal from the nth excitation is
cos ) 0 ()(
Here Mz(0) is the initial longitudinal magnetization, n is the
excitation index, and ? is the flip angle. T1 is spin-lattice
relaxation and TR is repetition time. Since TR<<T1, Eq.
(1) can be simplified to
?? sincos )(
Its corresponding point-spread-function (PSF) in the image
domain can be calculated by DFT; it is
Here N is the total number of excitations, x is the spatial
index. If the peak of the PSF is used to estimate the signal,
and noise is known to be proportional to N , then the
The plots of Eq. (4) are in Figure 1.
11/ ) 1
When the nuclear spin system of a noble gas is pre-
polarized by optical pumping, its polarization exceeds the
thermal equilibrium polarization by several orders of
magnitude . Hyperpolarized helium gas MRI takes
advantage of such significant signal-to-noise enhancement
for imaging human lungs, which has serious potential for
assessing pulmonary ventilation, microstructual changes,
and gas exchange .
An analytical estimation of the SNR in terms of
excitation number and flip angles is developed here. It
suggests that due to the fact that the non-equilibrium of the
gas results in a steady depletion of the signal level over the
course of the excitations, the signal-to-noise ratio (SNR) can
be independent of the number of data acquisitions at certain
flip angles. This provides a unique opportunity for parallel
MRI [3, 4] for gaining both temporal and spatial resolution
without reducing SNR.
So far only a birdcage type rigid quadrature coil or a
flexible wrap-around chest quadrature coil was used for3He
MRI, which is not capable of multiple channel data
acquisition. Here we propose a rigid coil system that
includes a 24-element phased-array for reception and a two-
large-loop for transmittance, for the purposes of either
parallel imaging [3, 4] or SNR enhancement. . A
prototype of such a system at 1.5T was built. Its decoupling
scheme was analytically and experimentally examined. The
3He parallel imaging of the in vivo lung was conducted with
almost eight-fold scan time reduction.
1) 0 (
Proceedings of the 2005 IEEE
Engineering in Medicine and Biology 27th Annual Conference
Shanghai, China, September 1-4, 2005
0-7803-8740-6/05/$20.00 ©2005 IEEE.
Figure 1A shows that when the flip angle is about 10°, the
SNR becomes independent of N when N>30, which
indicates that SNR is independent of the number of data
acquisitions at such a flip angle. This offers a unique
opportunity for using parallel MRI to gain spatial and
temporal resolution without costing SNR given that the g-
factor is not an issue. Figure 1B is a zoom-in of 20 bottom
lines in Fig. 1A, which suggests that a 90° single-shot MRI
has the highest SNR if its g-factor can be contained in
B. 2-ch transmit and 24-element receive arrays
Both the transmit and receive coils were tuned to
48.5MHz, matched to 50?, and housed in the two rigid
shells shown in Fig. 2. There are two levels inside each
shell. The receive arrays are in the inner levels and the
transmit coils are in the outer levels of the shells.
The layout of 24 receive copper loops (diameter 13cm) is
shown in Fig. 3, which was a result of compromising
between the competing goals of decoupling and parallel
imaging . In the left-right direction, the overlapping of
the nearest neighbor gets rid of the strong coupling between
adjacent loop pairs; and the weak coupling between non-
adjacent pairs can be easily removed by the low impedance
preamplifiers . However, in the superior-inferior
direction, to avoid the complication of overlapping and
parallel imaging encoding along the B0 direction, no
overlapping was applied along the inferior/superior
direction. This arrangement results in using a low
impedance preamplifier to decouple the strong coupling. Its
availability and limitations will be discussed in the next
The layout of the two transmit rectangular loops
(40cmx37cm), as well as 24 receive coils in the top and
bottom shells, are shown in Fig 4 (A) and (B).
C. Analysis of Preamplifier decoupling
The quantitative effects of using a low impedance
preamplifier to decouple the strong coupling are analyzed
by the theory in Ref. , where any two adjacent coils can
be modeled with their equivalent circuit in Fig. 5. In this
model, L is the inductance of the loop, M is the mutual
inductance between two loops, C is the tuning capacitor, R
is the shunted loading resistance, and Z2
impedance of the preamplifier.
g represents the
Based on Ref. , the impedance measured at the port 1
while the port 2 is terminated with the low impedance of the
LZ CRZ LRb
Here ? is the angular frequency.
The circuit parameters in Fig. 5 are set to L=280nH,
C=38.45pF, R=10k? based on the experiments. If there is
no coupling between two coils at M=0, each coil is tuned to
the3He resonance frequency 48.5MHz, see Fig. 6
In the case of the coil pair that is not connected with
inductive coupling of two coils increases, the resonance
frequency starts splitting and shifts away from the
resonance frequency, as illustrated in Fig. 7A.
g=10k?? open circuit): when the mutual
However, if the coils are connected with an ideal low
impedance preamplifier (Z2
occurs. In the weak coupling (M < 20% L), even the
resonance frequency of the two coils is maintained. But
g=0), no frequency splitting
when the coupling becomes strong, such non-split frequency
will shift to higher frequency, as shown in Fig. 7F. This
extra frequency shift makes tuning a strongly coupled coil
pair more complicated if the preamplifiers alone are used for
In reality, a preamplifier cannot maintain stability when
its input impedance Z2
shows that if Z2
preamplifier is sustained, as shown in Fig. 7E. Our model
further shows that if Z2
still can avoid frequency splitting in both weak and strong
coupling, but Q-factor of the coils reduces significantly,
especially in the strong coupling situation, see Fig. 7D.
coupling is strong, see Fig. 7C; and when Z2
preamplifier cannot even decouple the weak coupling, see
This analysis quantitatively suggests that in order to use a
preamplifier to decouple non-overlapped strongly coupled
coil pairs, like our current layout in superior-inferior
direction, the input impedance of the preamplifier should be
around 2?, which we did achieve in all of our preamplifiers.
A.Bench test results
A.1 Coil tuning and matching measurements
Each of the 24 loop coils was tuned to 48.5MHz, its
unloaded impedance was about 370?, and its unloaded Q
was about 405. When the coil was loaded with the chest, its
impedance was matched to 50?, and its Q was 49, see Fig.
9 (A) and (B).
g is zero. Fortunately our model
g =2?, the decoupling capability of the
g =20?; although the preamplifier
g =200?, its decoupling function vanishes when the
g =500?, the
Fig. 9 The impedance spectrum of the unloaded and the loaded coil
When the two coils are closely next to each other, but
without any overlap, the coupling causes frequency splitting
as shown in Fig. 10A. However, if the port of one coil is
shorted, as if an ideal zero impedance preamplifier is
connected, the frequency splitting disappears, the resonance
frequency shifts to 48.525MHz, see Fig. 10B. Such a
frequency shift was predicted by the analysis in the Method
section (Fig. 7F). Besides the frequency shift, its unloaded
Q was also dropped to 203, so the loading factor was
reduced to about 5: still very good performance.
D. External frequency synthesis
Currently the software in our MRI scanner (Siemens
AVANTO, Erlangen, Germany) only supports proton
frequency (63.67MHz). To make the
work with the
external frequency synthesis scheme was setup, as shown in
Fig. 8: A 10MHz clock was taken from the scanner’s
synthesis board to synchronize the external frequency
mixing system. This 0.3dB clock from the scanner was
amplified to 20dB and attenuated to 12dB and then fed in to
a frequency synthesizer (Agilent 8648A, Palo Alto, CA).
The 50MHz output of the synthesizer is fed into the
modulator and receive boards of the scanner in order to
change their local oscillation frequency so that they can
transmit and receive 48.5MHz3He signals.
1H frequency system
3He frequency coils and preamplifiers, an
Fig. 10 The impedance spectrum of a pair of coupled (A) and decoupled
coils (B). Here the decoupling is done by the ideal low impedance at the
A. 2 Preamplifier measurements
24 units of 1T proton preamplifiers (42.6MHz) were
retuned to 48.5MHz. A typical S-parameter of these
preamplifiers, measured by the network analyzer (Agilent
E5070B), is shown in Fig. 11, (A) and (B) are the
magnitude and phase of S21, (30dB ~ 32dB); (C) and (D)
are the magnitude and phase of input impedance (0.8? ~
2.4?); (E) is the magnitude of the S11 its in linear scale; (F)
is S11 has polar coordinates, which indicates the phases of
S11 are all about 180˚.
We have theoretically proved that SNR of
can be independent to the number of the data acquisitions,
which allows parallel MRI to enhance the temporal and
spatial resolution with SNR reduction. A 24-element
phased array system was built. The feasibility of applying
parallel imaging in 3He MRI to reduce scanning time was
demonstrated in the in vivo studies.
This project is partially funded by grants from Siemens
and Amersham. I’d like to thank the following people for
their valuable help: Gabor Mizsei, Jian Xu, Abram
Voorhees, Jean Reid, and Joseph Reaume.
H. E. Moller, X. J. Chen, B. Saam, K. D. Hagspiel,
G. A. Johnson, T. A. Altes, E. E. de Lange, and H.-
U. Kauczor, "MRI of the Lungs Using
Hyperpolarized Noble Gases," Magnetic
Resonance in Medicine, vol. 47, pp. 1029-1051,
E. J. R. van Beek, J. M. Vild, H.-U. Kauczor, W.
Schreiber, J. P. Mugler III, and E. E. de Lange,
"Functinal MRI of the Lung Using Hyperpolarized
3-Helium Gas," Journal of Magnetic Resonance
Imaging, vol. 20, pp. 540-554, 2004.
D. K. Sodickson and W. J. Manning,
"Simultaneous Acquisition of Spatial Harmonics
(SMASH): Fast Imaging with Radiofrequency Coil
Array," Magnetic Resonance in Medicine, vol. 38,
pp. 591-603, 1997.
K. P. Pruessmann, M. Weiger, M. B. Scheidegger,
and P. Boesiger, "SENSE: Sensitivity Encoding for
Fast MRI," Magnetic Resonance in Medicine, vol.
42, pp. 952-962, 1999.
P. B. Roemer, W. A. Edelstein, C. E. Hayes, S. P.
Souza, and O. M. Mueller, "The NMR Phased
Array," Magnetic Resonance in Medicine, pp. 192-
M. Weiger, K. P. Pruessmann, C. Leussler, P.
Roschmann, and P. Boesiger, "Specific Coil
Design for SENSE: A Six-Element Cardiac Array,"
Magnetic Resonance in Medicine, vol. 45, pp. 495-
R. F. Lee, R. Giaquinto, and C. J. Hardy,
"Coupling and Decoupling Theory and Its
Applications to the MRI Phased-Array," Magnetic
Resonance in Medicine, vol. 48, pp. 203-213, 2002.
The gains and noise figures of these preamplifiers are
measured by noise figure analyzer (Agilent N8973A). A
typical measurement is shown in Fig. 12. The noise figures
of all the preamplifiers are about 0.4dB. The gains of all
preamplifiers are about 31dB. Notice that the gain and S21
are slightly different in this case due to the mismatch at
input of preamplifier.
The in vivo 3He parallel imaging was conducted with
the new phased array coil. Both 2D and 3D data acquisitions
were tested. Figure 13 (A-C) are three slices in a 3D parallel
imaging MRI, where TR=6ms, TE=2.4ms, the data
acquisition matrix is 256x256, and the reduction factor
(iPAT) is 8 (4x2). No obvious SNR reduction was observed