Purpose

Distortion-free multi-shot EPI acquisitions represent a compelling approach to high-fidelity diffusion, relaxometry and/or functional MRI. Point spread function (PSF)-based techniques have been proposed for distortion-free diffusion imaging ^1,2. A similar technique, echo-planar time resolved imaging (EPTI), has also been demonstrated for distortion-free relaxometry imaging with an EPI readout ³. Additionally, a combination of PROPELLER and EPTI (PEPTIDE) has been reported for motion-robust simultaneous diffusion and relaxometry imaging ^4,5. In this study, we aim to develop a self-navigated Cartesian-based EPTI (cEPTI) acquisition for distortion-free diffusion-relaxometry imaging.

Methods

Sequence Design As with other multi-shot EPI acquisitions, inter-shot phase variations represent the main challenge associated with incorporating EPTI into diffusion imaging. In conventional multi-shot diffusion imaging, this can be solved by using an internal or external navigator image ^6-8. With the conventional EPTI implementation ³, not all the shots traverse the center of k-space (Fig. 1a), so external navigators would be needed for inter-shot phase corrections. Here, we first design a self-navigated Cartesian EPTI sequence (Fig. 1b). Meanwhile, in our cEPTI, data is uniformly sampled at each echo time, which maintains the ability to acquire quantitative relaxometry data as in the conventional EPTI ³.
The self-navigated cEPTI is implemented by introducing a rewinding gradient when k_y reaches k_{y, max} during the readout (Fig. 1d). With the rewinding gradient, the sampling trajectory will be reversed back to -k_{y, max} and then moves towards k_{y, max} again until the end of the readout. The area of the rewinding gradient is:
A_rewind=(2π)/(γΔy)
where γ is the gyromagnetic ratio, Δy is the spatial resolution. There are two highlights of the rewinding gradient. First, the reversing from k_{y, max} to -k_{y, max} occurs at different echo times for different shots (blue dashed lines in Fig. 1b). Therefore, the time to add the rewinding gradient also changes over different shots. Second, when adding the rewinding gradient, both k_x and k_y locations will change simultaneously, which indicates that the corresponding sampled data do not lie on a single k_y line. In this study, we simply discard the sampling points during the rewinding gradient and denote them with a hollow circle in Fig. 1b.
Data Acquisition Data were acquired on a GE 3.0T Premier MRI scanner, using a 48-channel head coil. The imaging parameters were: FOV=220×220 mm², in-plane resolution=1.0×1.0 mm², slice thickness=3 mm, 20 slices, echo spacing time=1.0 ms, TR=2.8 s, 30 diffusion encoding directions with b=1000 s/mm². Two under-sampling factors are defined in cEPTI (Fig. 1b): R_pe is the k_y step in each shot to adjust the sampling locations and R_shot is the step of the starting k_y location between two adjacent shots for acceleration. Fig. 1b shows an example when R_pe=1 and R_shot=4, 3 shots are used to sample the k_y-t space (N_shot=3) without applying partial Fourier (PF=1). The actual parameters of the cEPTI acquisition in this study were: R_pe=4 and R_shot=22, N_shot=8 and PF=0.8. CAIPI sampling ^9,10 was used in the k_y-t domain. Forty-eight echoes (N_echo=48) were acquired as a proof-of-concept, with the TE of the spin echo located at the 17th echo (TE_SE=63 ms). The total scan time was 11min40sec.
A fully sampled cEPTI (R_pe=0, R_shot=1) was acquired as the time-resolved 2D calibration scan (k_x-k_y-t) (Fig. 1c). R_pe=0 means that in each shot, all echoes were sampled at the same k_y location. The time-resolved 2D calibration scan contained the coil sensitivity and the field susceptibility induced phase at each echo time.
Reconstruction The preprocessing of the navigator is described in Fig. 2a. Each shot was recovered using GRAPPA ¹¹ and the central k-space of each shot was extracted as the navigator. Then the phase of each shot was calculated by multiplying the diffusion navigators with the conjugate of the b0 navigators from the same shot. In this way, we removed both the coil sensitivity phase and field susceptibility induced phase from the diffusion navigator and only kept the motion-induced phase. The phase was duplicated N_echo times and added to the time-resolved 2D calibration data (k_x-k_y-t) at each echo time, synthesizing a time-resolved navigator with matched inter-shot phase variation, coil sensitivity map and field susceptibility induced phase as the diffusion cEPTI data at each echo time.
GRAPPA with compact kernel (GRAPPA-CK) ¹² was extended for the inter-shot phase correction in diffusion cEPTI, with the weighting matrix estimated from the synthesized time-resolved 2D navigator (k_x-k_y-t) (Fig. 2b). An example of the interpolation process is shown in Fig. 2c.

Results

Fig. 3a shows a comparison of the diffusion images using the proposed cEPTI and the conventional 4-shot EPI acquisition, with a T2 FLAIR image as distortion-free reference. Residual distortion artifacts can still be observed in the 4-shot EPI results but are absent in the cEPTI results.
Fig. 4 displays the MD, FA and color-coded FA maps from the combined images over all the echoes. Fig. 5 shows the b0 image, MD and FA maps at six representative echo times, demonstrating the potential of cEPTI for simultaneous diffusion and relaxometry imaging.

Discussion & Conclusion

The development of a self-navigated Cartesian-based EPTI acquisition method for distortion-free diffusion-relaxometry MRI represents a promising acquisition strategy for a broad range of clinical and research applications including many microstructure imaging techniques.

Acknowledgements

This study is funded by GE Healthcare and the Focused Ultrasound Foundation.