Introduction

Magnetic resonance spectroscopic imaging provides metabolic information with high diagnostic value, but scan times are often too long for routine clinical application. One technique to reduce acquisition time is echoplanar spectroscopic imaging (EPSI)^1-3. There is great interest in using this technique at 7T, where there is higher SNR and improved spectral resolution. However, the spectral width (SW) in EPSI is determined by gradient performance and can be insufficient to match the higher value required at 7T. In addition, the EPSI trajectory suffers from inconsistencies between readout lines acquired with alternating positive and negative gradients⁴. In this ¹H study at 7T, we combine two methods: (1) readout segmentation⁵ to increase the spectral width, and (2) COKE (COnsistent K-t space Epsi)^6,7 to produce a consistent temporal phase evolution.

Methods

Fig. 1 shows the Readout-Segmented COKE (RS-COKE) pulse sequence⁸, which was implemented on a 7T MAGNETOM Terra (Siemens Healthcare, Erlangen, Germany). The sequence has a VAPOR water-suppression module⁹ and a pair of adiabatic pulses for spin-echo preparation¹⁰. The COKE gradient-encoding scheme interleaves blipped phase-encoding (PE) gradients between odd and even readout lobes. This displaces phase discontinuities caused by readout-gradient polarity from the temporal to the PE direction, thereby shifting the Nyquist artefacts from the spectral to the image domain, where their effect is more benign. The resulting image-domain ghost is minimized by phase correction using standard techniques¹¹.

Three phantoms were used to examine and optimize the sequence – 1) a spherical oil phantom, 2) a 3D-printed phantom with a “brain” compartment mimicking brain metabolite content and 3) a 3D-shaped phantom with brain-like compounds and a superficial compartment with a lipid-like compound. The 3D head-shaped phantom was designed to resemble the brain with respect to B₀ and B₁ distributions, metabolite content, and the subcutaneous lipid layer (see Ref. 12). The metabolite content included 10mM L-Glutamic acid, 10mM Creatine, 8mM myo-Inositol, 2mM GABA, 2mM Choline chloride, 5mM Choline chloride, 5mM Sodium lactate, and 12.5mM NAA. Sequence parameters for the 3D printed phantom scans were: TR/TE 1200/35 ms, FOV 200x250 mm², slice thickness 15 mm, in-plane resolution 3.7x4 mm² (54x64 acquisition matrix), 3 readout segments, SW 2778 Hz, echo spacing 0.36 ms, 296 echoes, and scan duration 3:50 min.

Data were acquired from the brain of a healthy subject using the following parameters: TR/TE 2000/35 ms, FOV 200x250 mm², slice thickness 15 mm, in-plane resolution 4.3x4.5 mm² (46x56 acquisition matrix), 3 readout segments, SW 2778 Hz, echo spacing 0.36 ms, 296 echoes, and scan duration 5:36 min. Simulations were also performed to investigate image-domain artifacts caused by the effect of residual lipid signals on readout-segmented spatial encoding.

Results and Discussion

Fig. 2 compares single-segment COKE and EPSI, demonstrating that the COKE version can provide a full spectrum without ghosting artifacts following a navigator-based even/odd phase correction¹¹. Fig. 3 shows summed spectra from three human brain regions (5x5 voxels, 0.3mL per voxel). The spectra demonstrate well-resolved metabolite peaks, but are contaminated by signal from subcutaneous lipids. This contamination is partly caused by imperfect k_x trajectories causing a ringing artefact in the readout direction of the image due to mismatch at the interfaces between readout segments. Fig. 4a shows that for a “brain” only compartment the 3-segment RS-EPSI and RS-COKE provide similar results without significant artifacts. Fig. 4b compares single-segment and three-segment acquisitions with a lipid layer. A ringing artefact can be observed in the three-segment water image. One method to reduce the ringing artifact is to apply saturation bands to reduce the lipid signal contribution, as shown in Fig.4b. It shows clear improvement of the metabolite spectra, however, the saturation bands also reduce the water suppression efficiency by a factor of four. Fig. 5 shows simulated results for an annulus-shaped object, considering only the first echo of each echo-train. In the current implementation, each readout segment is acquired at the same time point, relative to the excitation, resulting in phase discontinuities at segment interfaces for off-resonant signals. Such an acquisition can result in a ringing artifact when the annulus signal is off-resonance (Fig.5a). The simulation results in fig. 5b show how time-shifting the outer segments by ±echo-spacing removes the phase discontinuities and the off-resonance artifacts are suppressed. The much smaller residual ringing artifact probably corresponds to Gibbs ringing due to the low spatial resolution. Fig. 5d shows simulated data that demonstrate the effect of k_x trajectory imperfections, which cause similar ringing artifacts, but which are not specific to off-resonant signals. In practice, these artifacts are minimised by empirically scaling the readout gradients to improve the continuity of the k_x trajectory at segment interfaces.

Conclusion

The readout-segmented COKE method allows fast spectroscopic imaging and high spatial resolution for human application at 7T. The method is capable of removing the spectral-width limitation at 7T and provides a coherent phase evolution that avoids Nyquist artefacts in the spectra. However, before the method can be used for routine studies in vivo, further work is required to minimize the effect of lipid contamination on the readout-segmented spatial encoding scheme. The phantom experiments and simulations in this study have led to an increased understanding of these artifacts and provided potential mitigations that can be explored in future work.

Acknowledgements

No acknowledgement found.

References

[1] Mansfield, MRM 1984; [2] Posse et.al, MRM 1997; [3] Lin et.al, MRM 2007; [4] Posse et.al JMRI 2013; [5] Keith et al, ISMRM 2019; [6] Schmidt et al MRM 2019; [7] Webb et.al, MRM 1989; [8] Keith et al, ESMRMB 2020 [9] Tkac, MRM 1999; [10]Connolly MRM 1991; [11] Chen. et. Al, MRM 2004; [12] Jona et. al, NMR BioMed 2020.