Introduction

Geometric distortions of echo-planar imaging (EPI)¹ can be reduced with inner-field-of-view techniques^2-3 based on 2D-selective RF (2DRF) excitations^4-5, which has been shown to improve diffusion-weighted imaging of the human spinal cord^2-3. A comparison of two inner-field-of-view approaches^2-3 revealed a lower signal-to-noise ratio (SNR) for 2DRF excitations being collinear to the image plane if a large volume is covered⁶. Here, it is shown that this SNR reduction is related to the larger sensitivity of the corresponding 2DRF excitation to frequency offsets caused by field inhomogeneities that are particularly pronounced for the large fields-of-excitation needed for tall slice stacks.

Methods

Figure 1 and 2 present the EPI pulse sequences and geometric setups used in the present study. Both sequences involve a 2DRF excitation based on a fly-back blipped-planar trajectory as the initial excitation but differ regarding its orientation relative to the image plane and the position of the side excitations occurring in the blip direction. In the “collinear” setup (Fig. 1a and 2a), the line and blip direction of the 2DRF trajectory coincide with the imaging phase and slice directions², respectively. Thus, to avoid interferences, the side excitations must be positioned above and below the slice stack which requires a large field-of-excitation (FOE) and 2DRF pulse duration for tall slice stacks (cf. Fig. 2a). For the “tilted” setup (Fig. 1b and 2b), the 2DRF trajectory is rotated compared to slice and phase-encoding direction such that the side excitations neither are refocussed nor interfere with the slice stack to measure³. Thus, a small FOE is sufficient, independent of the number of slices, yielding short 2DRF pulses. Experiments were performed on a 3T whole-body MR system (PrismaFit, Siemens Healthineers, Erlangen, Germany) using a standard 64-channel head-neck coil in combination with a 32-channel spine array coil. Only coil elements with significant signal contributions were selected for the acquisitions. A water phantom and healthy volunteers that gave their informed consent prior to the examination, were investigated. 2DRF excitations were based on a trajectory resolution of 4.0×10.0mm² in the line×blip (tilted) and blip×line directions (collinear) and were designed to excite rectangular profile(s) with a size of 4×56mm² (slice×phase-encoding direction). 15 to 62 slices were targeted yielding FOEs of 60mm to 248mm for the collinear setup (2DRF pulse duration between (33.8 ms and 7.9 ms) and a fixed FOEs of 70mm for the tilted setup (6.4 ms). An inner FOV of 32×128 mm² with 16 mm oversampling in the phase encoding direction to account for profile imperfection was acquired with a resolution of 1.0×1.0×4.0mm³. Diffusion-tensor imaging was performed with six non-collinear directions of the diffusion weighting (b value 625 s mm^-2) yielding echo times between 64 and 70ms and repetition times between 5600ms. Additionally simultaneous multi-slice (SMS) acquisitions⁷ with an acceleration factor of 2 (TR = 3400ms) and a correspondingly reduced FOE were performed for comparison.

Results and Discussion

Figure 3 demonstrates the influence of frequency and gradient offsets on the SNR in a water phantom. For the collinear setup with a small FOE (15 slices) and the tilted setup (15-62 slices), no significant impact on the SNR is observed. However, for the collinear setup with larger FOEs (31 and 62 slices), the frequency (100Hz) and z gradient offset (0.05mT m^-1) clearly show a reduced SNR due to the low bandwidth of the 2DRF excitation in the blip, i.e. slice, direction: the offsets shift the excited profile in the slice direction such that it is not fully covered by the refocussing RF pulse. In contrast, the tilted setup exhibits a shift mainly in the phase-direction making it much more robust to frequency offsets and field inhomogeneities. Example images of the human spinal cord are shown in Fig. 4. For larger slice stacks and FOEs, the signal intensity is reduced in some sections of the spinal cord for the collinear setup due to field inhomogeneities. In contrast, the tilted setup shows a no SNR reduction in the cord even for a tall slice stack. In Fig. 5, conventional and SMS DTI acquisitions obtained with the collinear setup are presented. In the SMS acquisitions, the FOE is halved because one of the side excitations is positioned to cover the other slice to be acquired simultaneously which reduces the SNR loss caused by field inhomogeneities.

Conclusion

Compared to the tilted setup, the 2DRF excitation of the collinear setup suffer from an increased sensitivity to field inhomogeneities for tall slice stacks that can reduce the SNR in the spinal cord; however, simultaneous multi-slice acquisitions can help to ameliorate the problem.

Acknowledgements

This work was supported by Wings for Life.