Introduction

Among several concepts to utilize nonlinear gradients to accelerate MRI acquisitions^1–5, we push a recent idea of spread spectrum MRI forward, which is based on rapid modulation of localized magnetic fields superimposed with the linear gradients, to further reduce the scan time. The modulation is produced by a set of current loops, and thus imposes unique and localized spin phase evolution during signal readout. Consequently, the k-space data is sampled by a “stamp” instead of a point at a time⁶ and therefore, can be fully collected within a shorter time duration.

With a recently built 8-channel local B₀ coil array compatible with our parallel imaging setup at 9.4T, highly accelerated imaging can be achieved jointly with SENSE⁷. Therefore, the goal of this study is to experimentally test the feasibility of imaging acceleration with our newly built system, and further investigate the advantages of the local modulation by simulation, as significant steps towards highly accelerated in-vivo scans.

Methods

A new local B₀ coil array (8 square loops, each 10cm x 10cm/14 windings) was built with geometry compatible with a shielded 18/32 transmit/receive RF coil for in-vivo head imaging at 9.4T⁸. With the temperature constantly monitored, each coil has its current independently driven and monitored by the power amplifiers (IECO, Finland), which were controlled by a high-speed signal generator/monitor device (ADwin, Germany) synchronized with the scanner. The coil setup was filled with epoxy to increase the weight for reducing the mechanical vibration. The experiments (Figure 1) were carried out on a 9.4T whole-body MR scanner (Siemens Healthineers, Erlangen, Germany).

An algorithm was proposed to estimate the phase modulations by local B₀ coils with monitored currents in real-time (Figure 2). First, eight high-resolution maps of accumulated phase offsets, contributed from each coil per unit current and time, were estimated with the ESPIRiT algorithm⁹, by comparing each of the eight FLASH scans with an additional current blip in a local coil and a reference scan, all in low-resolution. Second, the phase evolution of spins was calculated from these unit phase maps and the monitored currents in the coils during later modulation experiments.

Experimentally, 2D FLASH scans of a head-shoulder phantom with 10x readout oversampling were performed with a basic sinusoidal modulation scheme of the local B₀ coils (10kHz/40A_pk in all coils, 45° phase shift between neighboring coils)⁵. The receiver sensitivity maps were estimated by the ESPIRiT⁹. The final images were reconstructed similar to the Wave-CAIPI¹⁰, with and without SENSE.

In one simulation, a 2D FLASH scan with similar modulation but slower frequency (5kHz/40A_pk) was investigated, which should introduce less current distortions and slower magnetic field switching with our amplifiers. In addition to the sinusoidal modulation used above, a phase shift of π was applied for currents in all channels during even phase encoding steps, given 2-fold undersampling.

In the other simulation, the local coils modulation was combined with a single shot spiral-out trajectory acquisition(~50 ms)¹¹. The basic sinusoidal modulation (7.5kHz/45A_pk, 45° phase shift between coils)⁵ was tested, and compared with another one, in which the sinusoidal amplitude was reduced (7.5kHz/30A_pk) but DC currents randomly switched between ±15A every 3ms were added in each coil. The latter aims to oscillate the Larmor frequency of local regions of the object in a randomly shifted bandwidth to increase the orthogonality of spin signals from voxels.

Both simulations share the same FOV/matrix size (i.e., 220mm/128) and the unit phase maps with the experiments, and were mainly examined by calculating the G-map⁷ and the noise amplification map in k-space¹². However, in simulation, the reconstruction matrix and the noise maps were directly calculated from the encoding matrix on a high memory compute node (1T-RAM) for simplicity.

Results

In experiments with the basic sinusoidal modulation, a rotating magnetic field was superimposed on the linear gradient during readout, similar to the Wave-CAIPI/FRONSAC but with a local-coils setup (equiv. 6.6 and 7.7 mT/m approx. linear gradient). The images were successfully reconstructed from undersampled (i.e., reduced phase encoding steps) k-space data (Figure 3). With local coil modulation alone, 3-fold acceleration can be reached with negligible artifacts. Furthermore, 6-fold acceleration can be jointly achieved with SENSE, which provides higher SNR (i.e., lower G-factor) and less residual artifacts than the SENSE reconstruction without modulation.

In the simulated FLASH scan (Figure 4), the additional π shift of modulation currents during even phase encoding steps can achieve substantially noise reduction (i.e., lower noise maps values) by sampling the k-space more uniformly (i.e., based on k-space noise amplification patterns).

In the simulated spiral acquisition (Figure 5), the reconstructed image in all sampling schemes can approximate the object well, but only with the sinusoidal modulation with random DC offsets, the object can be sampled uniformly in k-space and thus has much lower noise amplification.

Conclusion/Discussion

The experimental results have validated the feasibility of our new system to perform local phase modulation for image acceleration jointly with SENSE at 9.4T. In the simulation, the two newly designed schemes provide much more uniform and efficient k-space sampling, showing the advantages of having B₀ modulation in a local manner. These results make our system promising for the highly accelerated in-vivo experiments in the near future.

References

1. Patz S, Hrovat MI, Pulyer YM, Rybicki FJ. Novel encoding technology for ultrafast MRI in a limited spatial region. Int J Imaging Syst Technol. 1999;10(3):216–24.
2. Hennig J, Welz AM, Schultz G, Korvink J, Liu Z, Speck O, et al. Parallel imaging in non-bijective, curvilinear magnetic field gradients: a concept study. Magn Reson Mater Phys Biol Med. 2008 Mar;21(1–2):5.
3. Stockmann JP, Ciris PA, Galiana G, Tam L, Constable RT. O ‐space imaging: Highly efficient parallel imaging using second‐order nonlinear fields as encoding gradients with no phase encoding. Magn Reson Med. 2010 Aug;64(2):447–56.
4. Wang H, Tam LK, Constable RT, Galiana G. Fast rotary nonlinear spatial acquisition (FRONSAC) imaging. Magn Reson Med. 2016 Mar;75(3):1154–65.
5. Scheffler K, Loktyushin A, Bause J, Aghaeifar A, Steffen T, Schölkopf B. Spread‐spectrum magnetic resonance imaging. Magn Reson Med. 2019 Apr 26;mrm.27766.
6. Galiana G, Stockmann JP, Tam L, Peters D, Tagare H, Constable RT. The role of nonlinear gradients in parallel imaging: A k-space based analysis. Concepts Magn Reson Part A. 2012 Sep;40A(5):253–67.
7. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med. 1999;42(5):952–62.
8. Avdievich NI, Giapitzakis I-A, Bause J, Shajan G, Scheffler K, Henning A. Double-row 18-loop transceive-32-loop receive tight-fit array provides for whole-brain coverage, high transmit performance, and SNR improvement near the brain center at 9.4T. Magn Reson Med. 2019 May;81(5):3392–405.
9. Uecker M, Lai P, Murphy MJ, Virtue P, Elad M, Pauly JM, Vasanawala SS, Lustig M. ESPIRiT-an eigenvalue approach to autocalibrating parallel MRI: Where SENSE meets GRAPPA. Magn Reson Med. 2014 Mar;71(3):990–1001.
10. Bilgic B, Gagoski BA, Cauley SF, Fan AP, Polimeni JR, Grant PE, Wald LL, Setsompop K. Wave-CAIPI for highly accelerated 3D imaging: Wave-CAIPI for Highly Accelerated 3D Imaging. Magn Reson Med. 2015 Jun;73(6):2152–62.
11. Lustig M, Kim S-J, Pauly JM. A fast method for designing time-optimal gradient waveforms for arbitrary k-space trajectories. IEEE Trans Med Imaging. 2008 Jun;27(6):866–73.
12. Athalye V, Lustig M, Martin Uecker. Parallel magnetic resonance imaging as approximation in a reproducing kernel Hilbert space. Inverse Probl. 2015 Apr 1;31(4):045008.