Fast Method for 1D Non-Cartesian Parallel Imaging Using
GRAPPA
Robin M. Heidemann,
1,2
*
Mark A. Griswold,
1,3
Nicole Seiberlich,
1
Mathias Nittka,
2
Stephan A.R. Kannengiesser,
2
Berthold Kiefer,
2
and Peter M. Jakob
1
MRI with non-Cartesian sampling schemes can offer inherent
advantages. Radial acquisitions are known to be very robust,
even in the case of vast undersampling. This is also true for 1D
non-Cartesian MRI, in which the center of k-space is over-
sampled or at least sampled at the Nyquist rate. There are two
main reasons for the more relaxed foldover artifact behavior:
First, due to the oversampling of the center, high-energy
foldover artifacts originating from the center of k-space are
avoided. Second, due to the non-equidistant sampling of k-
space, the corresponding field of view (FOV) is no longer well
defined. As a result, foldover artifacts are blurred over a broad
range and appear less severe. The more relaxed foldover arti-
fact behavior and the densely sampled central k-space make
trajectories of this type an ideal complement to autocalibrated
parallel MRI (pMRI) techniques, such as generalized autocali-
brating partially parallel acquisitions (GRAPPA). Although pMRI
can benefit from non-Cartesian trajectories, this combination
has not yet entered routine clinical use. One of the main rea-
sons for this is the need for long reconstruction times due to the
complex calculations necessary for non-Cartesian pMRI. In this
work it is shown that one can significantly reduce the complex-
ity of the calculations by exploiting a few specific properties of
k-space-based pMRI. Magn Reson Med 57:1037–1046, 2007.
© 2007 Wiley-Liss, Inc.
Key words: parallel imaging; GRAPPA; SENSE; variable density
sampling; non-Cartesian trajectories
It has been shown (e.g., in Refs. 1–3) that non-Cartesian
sampling trajectories can offer inherent advantages for
magnetic resonance imaging (MRI) compared to Cartesian
acquisition schemes. Spiral trajectories, for example, make
efficient use of the gradient performance of an MR system,
while radial sampling schemes are known to be very ro-
bust even in the case of vast undersampling. Radial and
spiral trajectories are two-dimensional (2D) non-Cartesian
sampling strategies. However, the use of trajectories that
are non-Cartesian along one dimension in k-space and
Cartesian along the other can also be advantageous. These
so-called one-dimensional (1D) non-Cartesian trajectories
can be designed and optimized for a robust foldover arti-
fact behavior. Even though the advantages of non-Carte-
sian trajectories are well known, these trajectories are not
in common use in clinical MRI. In comparison, it is stan-
dard in clinical routine to use parallel imaging (4,5) to
speed up the acquisition of MR examinations. However,
parallel MRI (pMRI) is not without its limitations. All
pMRI acquisitions are affected by an inherent signal-to-
noise ratio (SNR) loss, which is, as a rule of thumb, at least
proportional to the square root of the acceleration factor
(AF). Additionally, noise introduced by imperfections in
the coil geometry and reconstruction algorithms further
degrades the SNR. In the sensitivity encoding (SENSE)
method (4), reconstruction is performed in the image do-
main on a pixel-by-pixel basis. Therefore, errors in this
procedure are local and appear as a kind of noise enhance-
ment in the image. In contrast, the generalized autocali-
brating partially parallel acquisitions (GRAPPA) recon-
struction (5) is performed in k-space using a fit procedure
between adjacent k-space lines or data points. Since a
single data line or data point in k-space affects the whole
image in the image domain, errors in this procedure are
global and appear as remaining foldover artifacts in the
final image. The more relaxed foldover artifact behavior of
non-Cartesian trajectories is beneficial for reducing poten-
tial remaining foldover artifacts. Non-Cartesian trajectories
have another advantage for pMRI. Whenever they are de-
signed to oversample a certain part or several parts of
k-space, they are inherently autocalibrated. The autocali-
bration approach (6) is a method in which a few k-space
data lines or points are sampled densely enough to fulfill
the Nyquist criterion (and may be sampled so densely that
they surpass the Nyquist criterion). The densely sampled
k-space data (otherwise known as the autocalibration sig-
nal (ACS)) are then used to derive the reconstruction pa-
rameters (i.e., the coil weights). The ACS data can be
acquired before, after, or during the actual accelerated
scan. The acquisition of the ACS data during the acceler-
ated scan as suggested in Ref. 7, the so-called variable
density (VD) approach, has the advantage that the ACS
data can be incorporated into the reconstructed data set.
This can significantly increase the overall image quality of
the reconstruction. From the VD approach proposed in
Ref. 7, which mainly used two different sampling densi-
ties, it is a logical step to move to a continuously VD
sampling scheme using a 1D non-Cartesian trajectory, as
suggested in Ref. 8. The more relaxed foldover artifact
behavior of non-Cartesian imaging makes such a trajectory
an ideal complement to pMRI. Although parallel imaging
can benefit from the advantages of non-Cartesian trajecto-
ries described above, and vice versa, the combination of
these acquisition schemes with parallel imaging has yet
not entered routine use. One of the main reasons for this is
the long reconstruction times required due to the complex
calculations necessary for non-Cartesian parallel imaging.
In this work, which is based on ideas presented at the
11th Annual Meeting of ISMRM in Toronto (9) and the
1
Universita ¨ t Wu ¨ rzburg, Physikalisches Institut, Wu ¨ rzburg, Germany.
2
Siemens Medical Solutions, Erlangen, Germany.
3
Department of Radiology, University Hospitals of Cleveland, Cleveland, USA.
*Correspondence to: Robin M. Heidemann, Siemens AG, Med MREA-Adv-
Neuro, P.O. Box 3260, D-91050 Erlangen, Germany. E-mail:
robin.heidemann@siemens.com
Received 10 October 2006; revised 29 January 2007; accepted 8 February
2007.
DOI 10.1002/mrm.21227
Published online in Wiley InterScience (www.interscience.wiley.com).
Magnetic Resonance in Medicine 57:1037–1046 (2007)
© 2007 Wiley-Liss, Inc. 1037