Introduction

Sodium (²³Na) MRI can reveal valuable metabolic information¹. However, due to their low resolution, ²³Na images are usually paired with high resolution proton (¹H) MRI to reveal the underlying morphology. Recently, a technique to simultaneously acquire sodium images during a proton magnetic resonance fingerprinting (MRF) scan was demonstrated², which can ease the burden of sequential ²³Na and ¹H scans. This method simultaneously acquires quantitative ¹H density (PD), T₁, T₂, B₁⁺ maps and a ²³Na density image, all in 3D with a 20 min total scan time³. In this work, we focus on the optimization of the 3D acquisition technique to obtain parametric maps with reduced artifacts.

Methods

Our sequence uses a stack-of-stars sampling scheme⁴. Each TR (7.5ms), a none-selective ¹H and a non-selective ²³Na pulses are applied, followed by one simultaneous readout for both nuclei (Fig.1). We strategically divide the partition gradient moment between the ¹H pulse and ²³Na pulse, such that images from both nuclei will have the same slice thickness³. Similarly, the frequency encoding gradient moments are distributed such that we can acquire a full radial trajectory for ¹H while also obtain a center-out readout for ²³Na, which allows for a factor ~2 between the ²³Na and ¹H resolutions, instead of the factor ~4 (γ_Na/ γ_H) as for the 2D version.

The previous 3D implementation inherited the flip angle train from the 2D implementation, which is sub-optimal for fast 3D imaging without full relaxation between repetitions. Therefore, we empirically redesigned the flip-angle train with a short delay in mind. The new train starts with an adiabatic inversion, followed by 275 excitations of variable flip angles (Fig1), instead of 1000 excitations from the original sequence³. The first segment (TR 1-100) includes flips angles up to 20° providing signal to readout the T₁ recovery while contributing minimal saturation. The second segment (TR 101-200) contains higher flip-angles (up to 90°) to help encode T₂. A 50-TR delay is inserted to recover some magnetization before the 3^rd segment (TR 251 ~ 325), which is RF spoiled to help encode B₁⁺, with flip-angle up to 60°. A short delay (TR 326 ~ 500) is also added before the next inversion.

On the sodium side, a constant flip angle (30°) is applied at every other TR to reduce SAR and obtain a larger effective gradient spoiling moment for sodium (TR for sodium is thus 15 ms). The sodium acquisition continues throughout the whole scan without interruption (see Fig. 1).For the proton MRF reconstruction, CG-SENSE is applied to eliminate the radial artifacts⁵. After removing all the delays, radial samples from 275 TRs are reconstructed with a sliding window covering 11 TRs (Fig. 2a). The MRF dictionary is also compressed/averaged with the same sliding window along the time domain.

To enable the simultaneous acquisition, the received chain was modified as described in Meyerspeer et al⁶. Phase correction for the sodium signal is necessary due to the asynchronization between the MR system and the external frequency generator. As shown in Figure 2b, there is a linear phase drift between the samples along each shot. In addition, for all shots in the same partition, another layer of phase correction is in need to align all shots. A dictionary matching scheme is applied to correct these phase drifts³.

All experiments were performed at 7T (MAGNETOM, Siemens, Erlangen, Germany) using a 16-channel-Tx/Rx dual-tuned coil developed in-house⁷. All images were reconstructed in MATLAB (Mathworks, Natick, MA, USA) using the CG-SENSE⁵ and Fessler NUFFT toolbox⁸.

A seven-tubes cylindrical phantom (¹H T₁: 600~1600ms, T₂: 40~180ms) was used to verify the accuracy of the multiparametric ¹H maps, using the previous 3D sequence as a reference. In addition, one healthy volunteer (39yo male) was scanned in accordance with our IRB protocol. The 3D sequence parameters for phantom and invivo experiments were: ¹H 160×160 / ²³Na 84×84 matrix, ¹H 1.5×1.5 mm² / ²³Na 3.0×3.0 mm² in-plane resolution, 3.0 mm slice thickness, 56 slabs, 6 shots per slab, total scan time 21 min. In this case, the sodium 3D data has isotropic resolution.

Results & Discussion

The T₁ and T₂ values in phantoms obtained with the new flip-angle train show an excellent correlation with the reference sequence (Fig. 3). The 3D invivo results are shown in Fig. 4, where sodium and proton data were acquired simultaneously in one perfectly co-registered scan. After the phase drift correction, clean sodium images were obtained. For proton MRF, the shorter train length and delay enabled us to acquire more radial samples per time-frame without extending the scan time, and thus reduce radial aliasing artifacts (under-sampling factor per time frame is reduced). Combined with the sliding window CG-SENSE reconstruction, residual streak-like artifacts in the T₁ and T₂ maps were significantly reduced.

Conclusion

The optimized 3D implementation provides the opportunity to simultaneously obtain high-quality volumetric ¹H MRF and ²³Na MRI of the brain. We hope that this technique will make sodium MRI more applicable for future clinical research, as it can now be acquired during the ¹H scan without extending the MRI protocol.

Acknowledgements

The research reported in this publication was supported by the NIH/NIBIB grant R01 EB026456, and performed under the rubric of the Center for Advanced Imaging Innovation and Research, a NIBIB Biomedical Technology Resource Center (P41 EB017183).