Purporse

MEGA-PRESS ¹H-MRS measurements of inhibitory neurotransmitter GABA allow investigations of neurochemical processes modulating neuronal activity in the brain¹. Applying a standard MEGA-PRESS sequence with alternating editing pulses at 1.9 ppm (edited spectra, ed) and 7.5 ppm (non-edited spectra, ned), the GABA intensities at 3 ppm in the difference spectrum (ed–ned) are, however, overlapped by co-edited macromolecular resonances (MM), which vary depending on physiologic factors or disease state. The MM resonances can be suppressed by alternating narrow-band editing pulses at 1.9 ppm and 1.5 ppm, providing similar MM modulation in both ed and ned spectra, whereas GABA will be edited only in ed scans (‘Henry method’²). The editing efficiency for MM (and also for GABA) can, however, be compromised by gradient heating or movement related rf frequency drifts. In this study, we present a novel MM suppression approach relying on an adiabatic inversion of the longitudinal magnetization of both metabolites and MMs prior to playing-out the standard MEGA-PRESS editing scheme, which is applied after an inversion time delay (TI) corresponding to the zero-crossing of the MM magnetization.

Material and Methods

A conventional MEGA-PRESS sequence (Siemens IDEA, VE11B) was modified by inserting a spatially non-selective, adiabatic inversion pulse (BIR4). The TI time for MM nulling (TI_0,MM(T_1,MM) = T_1,MM × ln(2)) was set to 0.173 s, according to the previously reported T₁ of 0.25 s for MM in the human brain³. Six healthy males (32.5±5.8 years) were measured on a whole-body 3T MR scanner (Magnetom PRISMA, 64-channel head-matrix coil). The protocol comprised a 3D T₁-weighted MP-RAGE scan, followed by three ¹H-MEGA-PRESS water-suppressed scans (TR: 4.5 s, 64 ned/ed averages) in the posterior cortex (PC, 18 ml, see Fig. 1). The first scan was performed with a conventional editing scheme to determine the MM contaminated GABA+ intensities (‘ConvMEGA’, TE: 68 ms, Gaussian editing pulse duration: 17 ms), whereas the second scan was conducted by using the Henry Method to quantitate ‘pure’ GABA intensities (‘HenryMEGA’, TE: 80ms, alternating editing pulses at 1.5 and 1.9 ppm, editing pulse duration: 23ms). Finally, the novel MEGA-PRESS sequence with adiabatic MM nulling was applied during the third scan (‘MMnulMEGA’, TE/TI: 68/173 ms, editing pulse duration: 17ms). GABA+ (ConvMEGA) and GABA (HenryMEGA and MMnullMEGA) intensities were quantified from the particular difference spectra (jMRUI⁴) and normalised with the creatine intensities obtained from corresponding ned spectra (GABA+/tCr, GABA/tCr and Glx/tCr ratios).

Results

Figs.1 and 2 show representative ned and difference spectra acquired with HenryMEGA (blue graphs), MMnulMEGA (red graphs) and ConvMEGA (black graphs) scans. Due to progressing signal recovery during the TI delay, the tNAA, tCr and tCho resonances are distinctly lower in the MMnullMEGA compared to the ConvMEGA spectrum. On the other hand, the ned and difference MMnullMEGA spectra yield the most flat baseline between 0.5 and 1.9 ppm, which evidences appropriate MM suppression. Similar to the MMnullMEGA ned spectrum, the corresponding difference spectrum reveals an apparent, T₁-weighting related decrease of the GABA and Glx peaks. The GABA peaks in the MMnulMEGA and HenryMEGA difference spectra are lower than in the ConvMEGA spectrum due to suppressed MM multiplet fraction. Interestingly, the MMnullMEGA GABA resonance is higher than in the HenryMEGA spectrum, where the editing at 1.5 ppm particularly co-edits GABA at 1.9 ppm, leading to a GABA signal reduction in the difference spectrum. These method-specific GABA signal modifications are also replicated in the GABA/tCr subject group distributions (Fig. 3). Fig. 4 shows associations between GABA and GABA+ intensities obtained in scans with and without MM suppression. Considering the Henry method to be the gold standard approach for pure GABA measurements, the significant correlation between the MMnulMEGA and HenryMEGA GABA levels (Fig. 4a) indicates appropriate GABA quantitation with the novel MMnulMEGA method. On the other hand, the weaker associations between the GABA+ (ConvGABA) and pure GABA levels (MMnulMEGA or HenryMEGA, Fig. 4b) illustrate the GABA+ uncertainty arising due to inter-individually varying MM levels.

Discussion and Conclusion

We propose a novel method, which combines MEGA-PRESS ¹H-MRS with adiabatic MM nulling and provides appropriate GABA quantitation without macromolecular contaminations. Besides being less dependent on rf-frequency drifts, our method avoids GABA co-editing and provides a GABA signal gain compared to the Henry method. In addition, our method allows more flexible editing pulse frequency selection in ned scans, which can be exploited for advanced editing strategies as they are applied, for example, in the HERMES approach⁵.

Acknowledgements

Alexander Gussew and Martin Krämer acknowledge funding from the German Research Foundation (DFG, GU 1108/3-1; RE 1123/22-1). Andreas Masek is supported by a graduate scholarship from the Friedrich‐Schiller‐University Jena (Landesgraduiertenstipendium).