RF-Encoding for Simultaneous Multi Slab Imaging
Benjamin Zahneisen1, Murat Aksoy1, Julian Maclaren1, Christian Wuerslin1, and Roland Bammer1

1Stanford University, Stanford, CA, United States

Synopsis

Simultaneous-Multi-Slab-Imaging is a promising approach for high resolution DW-EPI (<1mm) where the slice profile of a typical multi-band RF-pulse limits the spatial resolution for single-shot DW-EPI. Like every multi-band technique, slab separation during a parallel imaging reconstruction requires a combination of coil sensitivity variation and k-space encoding. The most prominent way is blipped-CAIPIRINHA where k-space encoding along the slice direction is achieved by adding small gradient blips in z-direction to the EPI readout. Here we demonstrate how multi-band RF phase modulation can be used as an alternative way to encode along the slab dimension without affecting intra-slab phase encoding. The use of phase modulated RF-pulses decouples encoding between the logical intra- and inter-slab directions that are otherwise linked because there is only one physical z-gradient axis.

Introduction

Simultaneous-Multi-Slab-Imaging1,2,3 is a promising approach for high resolution DW-EPI (<1mm) where the slice profile of a typical multi-band rf-pulse limits the spatial resolution for single-shot DW-EPI. Like every multi-band technique, slab separation during a parallel imaging reconstruction requires a combination of coil sensitivity variation and k-space encoding. The most prominent way, and only way in case of single-shot acquisition, is blipped-CAIPIRINHA4 where k-space encoding along the slice direction is achieved by adding small gradient blips in z-direction to the EPI readout. Here we demonstrate how multi-band RF phase modulation5,6 can be used as an alternative way to encode along the slab dimension without affecting intra-slab phase encoding. The use of phase modulated RF-pulses decouples encoding between the logical intra- and inter-slab directions that are otherwise linked because there is only one physical z-gradient axis.

Theory

Every pixel within a group of simultaneously excited slabs (Fig.1) can be described by 4 coordinates r=(x,y,z',s), where x and y are the in-plane coordinates, z’ is the distance relative to the center of each slab and s is the position of each slab relative to the gradient isocenter. The resolution and the intra-slab voxel dimensions Δx,Δy,Δz define the intra-slab k-space. Δzslab is the distance between the centers of Nslab simultaneously excited, non-contiguous slabs. It determines the FOV along the slice direction by FOVslab=NslabxΔzslab. Full k-space sampling requires NxxNyxNzxNslabs samples spanning a 4-dimensional k-space (Fig. 1 b). Each k-space location corresponds to a multi-directional (i.e. 4d) and mutually orthogonal phase modulation in image space. In practice, however, there are only 3 independent physical gradient directions to encode 4 logical directions. The slab dimension and the high-resolution intra slab dimension are both represented by the physical z-gradient axis and therefore cannot be varied independently. The effective wave vector along the slice direction is always a sum of both logical k-space contributions. Since encoding is solely based on the wave length (k-space value) both dimensions (i.e. multi-slab and intra-slab) are modulated with an effective k-space wave k'=kz+kslab. For the logical z-axis that means that additional slab encoding (CAIPI-blips) results in small to moderate deviations from the nominal k-space location (see Figure 2b where the red dots are slightly shifted relative to the Cartesian grid depicted as gray circles). Imposing the phase modulation along slabs with RF- instead of gradient pulses overcomes this limitation and all four dimensions can be varied independently.

Methods

Images were acquired at 3T using an 8-channel head coil array (GE Healthcare). Multi-band RF excitation was performed by cosine modulation of a sinc-pulse (3.2ms, bw= 950Hz) and adding a slice dependent phase prior to complex summation across slices. Each phase modulation corresponds to a k-space sample along the logical slab dimension (ksms in Figure 1). For m slabs, m different RF-pulses were generated online, corresponding to full sampling along the slab direction. RF-waveforms were then updated in real-time between sequence TRs. Sequence parameters for the 3D-EPI sequence were FOVxy=224mm, Nxy=64x64, slab thickness=FOVz=4mm, Nz=4, multi-band factor=3, Δzslab=12mm Reconstruction was performed off-line using a 3D-SENSE method or in case of fully sampled scans by 4d-iFFT.

Results

Figures 2a,b display the magnitude and phase of a multi-band RF-pulse. In c,d the phase modulation relative to b along the slab direction corresponds to k-space coordinates -kmax and kmax + Δk, respectively. Figure 4 displays the reconstruction of 3 simultaneously excited slabs with Nz=4 intra-slab resolution acquired with a 3D-EPI readout using (Nz=4 kz phase encoded) x (m=3 different RF pulses) TR's. The blue rectangles highlight the first group of simultaneously excited slabs, each containing Nz=4 slices. The reconstruction was performed using a 4 dimensional inverse FT. No coil sensitivity information was used. Under sampling can be performed by either skipping RF-pulses or lines/planes of the remaining gradient directions.

Discussion & Comparison between RF-encoding and blipped-gradient encoding

Using RF-pulses (POMP5, CAIPIRINHA6) to encode the slab dimension in a simultaneous multi slab experiment is an alternative to blipped gradient encoding in combination with a 3d-EPI trajectory. It allows sampling on a Cartesian grid for all 4 encoding dimension and can be used with an unmodified 3d-EPI readout. It further has the potential to optimize the acquisition by providing more flexibility in terms of multi-band k-space traversal especially with respect to non-Cartesian trajectories. It also might help to reduce slice/slab leakage because all k-space values along the slab dimension have identical T2* weighting. The downside of the method is that peak power reduction using PINS pulses is not feasible for RF-encoding because one needs full control over the phase of each slice.

Acknowledgements

NIH (2R01 EB002711 , 5R01 EB008706, 5R01 EB011654), the Center of Advanced MR Technology at Stanford (P41 RR009784), Lucas Foundation.

References

[1] Frost R et al., MRM, 2013. doi: 10.1002/mrm.25062.

[2] Bruce I. et al., Proc. Intl. Soc. Mag. Reson. Med. 23 (2015)

[3] Van AT et al., MRM, 2014.doi: 10.1002/mrm.25169.

[4] Setsompop K. et al. MRM 2012; 67:1210-1224.

[5] Glover G., Phase-offset multiplanar (POMP) volume imaging: A new technique, JMRI 1991

[6] Breuer F. et al., MRM 2006

Figures

Figure 1: Multi-slab geometry and corresponding 4D k-space (read direction collapsed) for m=4 simultaneously excited slabs and Nz=4 intra-slab resolution.

Figure 2: RF phase modulation for slab encoding. Magnitude (a) and phase (b) of m=4 simultaneously excited slabs. c) slice dependent phase modulation relative to (b) corresponding to k = -kmax. d) phase modulation corresponding to k = -kmax + Δk.

Figure 3: a) Blipped gradient 3D-EPI k-space sampling on a hyperplane in kslab-kz-space. Gray circles represent samples in logical k-space for full sampling. b) After re-arranging the samples by subtracting integer multiples of 2*kmax we see that the actual sample locations are slightly shifted between neighboring kz-planes.

Figure 4: Reconstruction from fully sampled simultaneous multi-slab acquisition (m=3) with RF-encoding of the slab direction.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3257