Introduction

An interleaved reverse gradient fMRI (RG-fMRI) with a point-spread function (PSF) mapping-based approach (1) was recently proposed to minimize geometric distortion and signal dropout in echo-planar-imaging (EPI) by measuring a pair of EPI acquisitions with the opposite (forward and reverse) phase-encoding gradient polarity, and then combining them after distortion correction. The effectiveness of RG-fMRI was demonstrated both in animal¹ and human studies². Here we show that the effective echo-shift map (ESM) can be reliably obtained from PSF mapping to predict signal dropouts for a given protocol. The predicted echo-shifting can provide guidance for protocol optimization for RG-fMRI using partial Fourier (PF) acquisition.

Methods

PSF mapping-based ESM (PSF-ESM): Local susceptibility change leads to variations in the effective echo time (or echo-shift). Instead of an EPI-based ESM (EPI-ESM) directly calculated from the phase information of the EPI data³, this work proposes to calculate the corresponding phase map from the PSF mapping-based shift (or distortion) map in the phase-encoding dimension, which is commonly used for distortion correction in EPI⁴. The phase map can be converted from the shift map as follows:
ϕ(r)=(2π∙∆y(r)∙TE_app)/(N_y∙ESP) [1]
where r indexes a particular voxel, ϕ(r) is the unwrapped phase for each voxel at the applied TE, TE_app, ∆y(r) is the voxel shift in the phase-encoding direction, ESP is the echo spacing, and N_y is the number of voxels in the phase-encoding direction. Then, ∆ϕ(r) is obtained by calculating the phase deviation between two adjacent lines along the phase-encoding direction³, which can be translated into a local TE map, TE_local, and ESM in k-space, ∆k_y, respectively with
TE_local=TE_app+(TE_app-t₀)∙∆ϕ/π [2]
and
∆k_y=Ny/2∙∆ϕ/π [3]
where t₀ is a time delay from the isocenter of RF excitation pulse to the beginning of EPI readout. While EPI-ESM can measure the echo shift within the EPI readout acquisition window, the measuring range of PSF-ESM is twice as wide due to the additional spin-warp phase-encoding gradient for PSF mapping, and can even measure the TE_local outside of the readout acquisition window.
Simulation of standard and reversed PF acquisition scheme for RG-fMRI: Regardless of the phase-encoding polarity in gradient-echo EPI, the TE_local is always shorter than TE_app in the stretched areas, and vice versa longer in the compressed area (Fig. 1). For the RG distortion correction approach, when images with opposite geometric distortions (both compressed and stretched) are combined, spatial information in stretched region is more reliable than the information in the compressed regions, because spatial information is lost in the compressed regions¹. Compared to standard PF omitting the early echoes, reversed PF omitting the late echoes can preserve the signal from stretched region better than the compressed region as shorter TE_local components are preserved. In this regard, reversed PF yields less signal loss and suits the RG method better than standard PF. To test this, simulations of both the standard and reversed PF acquisition scheme were performed using breath-holding fMRI data with full k-space acquisition, which was obtained in a previous study² on a compact 3T scanner with high-performance gradients^5-7. Three different PF factors (5/8, 6/8, 7/8) were applied and omitted k-space areas were zero-filled before reconstruction. In the high-susceptibility areas, image intensity and functional contrast in distortion-corrected forward (DF), reverse (DR), and combined (DW) data with different PF schemes and factors were compared with the full acquisition data.

Results and Discussion

Strong echo-shifts surpassing half of the k-space acquisition window (i.e., >±48) were measured in highly compressed regions of the PSF-ESM, which explained the regional signal dropouts in forward EPI (Fig. 1B). In distinction, the EPI-ESM provided the echo-shift estimates only within half of the k-space acquisition window even in the regions with complete signal loss (Fig. 1B). The overall patterns of the intensity difference between the distortion-corrected EPI pair (DF and DR) were matched well with PSF-ESM in the non-distorted (i.e., spin-warp phase-encoding) dimension (see contours in Fig. 2), which demonstrated that the signal dropouts in DF and DR are dominantly caused by echo shifts in the phase-encoding direction. Compared to standard PF, the reversed PF acquisition displayed reduced signal loss, and effectively resolved with the RG scheme (Fig. 3). While TR was not changed for the standard PF method as TE is fixed at 30 ms, the TR for the reversed PF method could be reduced from 934 ms to 861 ms (7.9%), 860 ms (15.7%), and 788 ms (23.6%), with PF factors of 7/8, 6/8, and 5/8, respectively. Consequently, while the functional activation in the combined EPI was either partially or entirely lost with standard PF, there was no noticeable difference in the functional activation map even with a 5/8 PF factor compared to the map without PF acquisition (Fig. 4).

Conclusion

The effective echo-shift (or TE) map obtained from PSF mapping can be useful to predict signal dropouts for a given protocol and thus applied to minimize the signal loss in a target region of interest for fMRI. Upon this investigation, we demonstrate that the use of the reversed PF acquisition rather than the standard PF would be an alternative viable option to further improve the temporal resolution for RG-fMRI.

Acknowledgements

This work was supported by NIH U01 EB024450 and NHI U01 EB026979.