Elsevier

NeuroImage

Volume 73, June 2013, Pages 59-70
NeuroImage

Single shot whole brain imaging using spherical stack of spirals trajectories

https://doi.org/10.1016/j.neuroimage.2013.01.065Get rights and content

Abstract

MR-encephalography allows the observation of functional signal in the brain at a frequency of 10 Hz, permitting filtering of physiological “noise” and the detection of single event activations. High temporal resolution is achieved by the use of undersampled non-Cartesian trajectories, parallel imaging and regularized image reconstruction. MR-encephalography is based on 3D-encoding, allowing undersampling in two dimensions and providing advantages in terms of signal to noise ratio. Long readout times, which are necessary for single shot whole brain imaging (up to 75 ms), cause off-resonance artifacts. To meet this issue, a spherical stack of spirals trajectory is proposed in this work. By examining the trajectories in local k-space, it is shown that in areas of strong susceptibility gradients spatial information is fundamentally lost, making a meaningful image reconstruction impossible in the affected areas. It is shown that the loss of spatial information is reduced when using a stack of spirals trajectory compared to concentric shells.

Highlights

► A method for full brain imaging within 100 ms is presented. ► The off-resonance behavior of a stack of spirals trajectory is analyzed. ► The effect of off-resonance gradients is analyzed in local k-space. ► The dependency of artifacts on the acquisition direction is examined. ► Stack of spirals trajectories are compared to concentric shells.

Introduction

Echo planar imaging (EPI) (Mansfield, 1977) is the established technique for functional magnetic resonance imaging (fMRI). For whole brain imaging EPI has a temporal resolution of typically 2–3 s. This is sufficiently fast compared to the blood oxygenation level dependent response (BOLD) (Ogawa et al., 1990), but there are several issues making it desirable to achieve high temporal resolution for fMRI. A single event hemodynamic response function (HRF) has significant signal fluctuations lasting approximately 20 s. Therefore about 10 data points are measured in the meantime when acquiring with a standard EPI protocol. With MR-encephalography (MREG) 200 data points are supplied for the same period of time, allowing a better analysis of the onset and shape of the HRF (Zahneisen et al., 2011). Furthermore the increased amount of data points improves the statistical power from which single event fMRI can benefit (LeVan et al., 2012), as well as the study of functional connectivity of networks (Lee et al., 2013, Lin et al., 2008). Fast acquisition also allows direct filtering of respiration and cardiac artifacts, since they are not aliased in the frequency domain anymore (Hennig et al., 2007, Posse et al., 2012).

While in the original implementation of MREG (Hennig et al., 2007) only coil sensitivities were used for spatial encoding, there have been several approaches for combining a small amount of gradient encoding with the spatial information multi coil arrays provide. An early attempt was to perform a fully sampled sagittal 2D-EPI measurement, while encoding the third dimension purely by coil sensitivities, which is referred to as inverse imaging (Lin et al., 2006). Further approaches were to use a small number of projections for 2D functional imaging (Grotz et al., 2009) or a rosette k-space trajectory for single shot 3D imaging (Zahneisen et al., 2011).

For all single shot acquisition techniques (and this includes standard EPI) the image quality is strongly affected by susceptibility induced local field inhomogeneities. For EPI this is reasonably well understood and leads to the well-known susceptibility artifacts — primarily signal attenuation and geometric distortions. Distortions predominantly occur along the phase encoding direction and can be corrected with suitable methods (Jezzard and Balaban, 1995).

For non-Cartesian trajectories the off-resonance behavior is more complex. As outlined by (Zahneisen et al., 2012), self-intersecting trajectories like rosettes (Zahneisen et al., 2011) and single shot radial trajectories (Hugger et al., 2011) suffer from a high sensitivity to off-resonance, T2-decay and gradient imperfections. Concentric shells (Zahneisen et al., 2012) are designed to have no intersections. They have a more benign off-resonance behavior, allowing longer readout times and therefore higher spatial resolution, while keeping the temporal resolution below 100 ms.

Like rosettes, concentric shells sample k-space symmetrically. As a consequence of the symmetry their point spread functions (PSF) do not show any off-resonance dependent shifts and therefore no image distortions. On the downside off-resonance leads to blurring and signal dropout. In practice geometric distortions are easier to handle compared to blurring since they can be corrected later on with suitable coregistration procedures.

Therefore and in order to maintain the comparatively beneficial off-resonance behavior of EPI, a single shot stack of spirals (SoS) technique is proposed. Sampling the k-space monotonously from one end to the other in variable sized steps, while acquiring variable density spirals at each step, the slow encoding direction follows one of the Cartesian k-space coordinate axes. Each individual spiral is acquired sufficiently fast, avoiding appreciable artifacts in the other directions. Therefore the main effects occur along a single direction, similar to EPI.

In the following methodology, implementation and first results of SoS will be presented together with a more detailed discussion of B0 inhomogeneity effects in non-Cartesian imaging.

Section snippets

Trajectory design

With the polar coordinates of k-space r and ϑ, a variable density spiral (Spielman et al., 1995) can be expressed with the complex functionkxyt=kxt+ikyt=rtet.kx and ky denote the Cartesian coordinates of the trajectory in 2D k-space. Time is indicated by t.

The Nyquist constraint in the radial direction is given bydr=RrrkzFOVxy,where Rr is the radial undersampling factor of the variable density spiral. It is chosen to increase with |kz| as well in order to shorten the trajectory. Nyquist

Methods

All experiments were performed on a 3 T TIM Trio scanner (Siemens Healthcare, Erlangen, Germany) with approval of the local ethics committee. Excitation was done with the whole body transmit coil. For signal reception a 32-channel head coil array (Siemens Healthcare, Erlangen, Germany) was used.

The trajectories were designed offline using MatLab (The Mathworks, Natick, MA, USA). The corresponding gradient shapes were exported to a text file. A gradient-spoiled gradient echo sequence was modified

Results

The spatial distribution of Larmor frequencies (Fig. 2a) can be approximated as piecewise constant. The effect of the off-resonance in this approximation is analyzed using point spread functions. However, some artifacts like signal dropout cannot be explained in this approximation. Therefore, the concept of local k-space in employed in order to analyze the influence of off-resonance gradients. This corresponds to a piecewise linear approximation. The cutoff gradient strength for signal

Discussion

The focus of this work was to analyze the off-resonance behavior of non-Cartesian 3D trajectories and to optimize the data acquisition scheme in this regard. The feasibility of a single shot stack of spirals trajectory for MREG has been demonstrated.

The present method allows single shot whole brain imaging within 100 ms. High temporal resolution is achieved by using highly undersampled non-Cartesian trajectories. The stack of spirals trajectories are undersampled in all three Cartesian

Acknowledgments

This work was supported by the European Research Council Advanced Grant ‘OVOC’ grant agreement 232908.

References (40)

  • D. Gallichan et al.

    Simultaneously driven linear and nonlinear spatial encoding fields in MRI

    Magn. Reson. Med.

    (2011)
  • G.H. Glover

    Simple analytic spiral K-space algorithm

    Magn. Reson. Med.

    (1999)
  • T. Grotz et al.

    Fast functional brain imaging using constrained reconstruction based on regularization using arbitrary projections

    Magn. Reson. Med.

    (2009)
  • B.A. Hargreaves et al.

    Time-optimal multidimensional gradient waveform design for rapid imaging

    Magn. Reson. Med.

    (2004)
  • F.X. Hebrank et al.

    SAFE-model — a new method for predicting peripheral nerve stimulations in MRI

    Proc. Int. Soc. Magn. Reson. Med.

    (2000)
  • J. Hennig et al.

    Parallel imaging in non-bijective, curvilinear magnetic field gradients: a concept study

    MAGMA

    (2008)
  • M.R. Hestenes et al.

    Methods of conjugate gradients for solving linear systems

    J. Res. Natl. Bur. Stand.

    (1952)
  • T. Hugger et al.

    Fast undersampled functional magnetic resonance imaging using nonlinear regularized parallel image reconstruction

    PLoS One

    (2011)
  • M. Jenkinson

    Fast, automated, N-dimensional phase-unwrapping algorithm

    Magn. Reson. Med.

    (2003)
  • P. Jezzard et al.

    Correction for geometric distortion in echo planar images from B0 field variations

    Magn. Reson. Med.

    (1995)
  • Cited by (0)

    View full text