Reliable and precise measurement of the primary inhibitory neurotransmitter, γ-aminobutyric acid (GABA), in the human brain is important for the research in a wide variety of neuro-psychiatric disorders¹. The resonances of this low-concentration metabolite are extensively overlapped with other large signals, thus J-difference editing approaches^2,3 are commonly used, in which the GABA 3.0-ppm resonance is obtained via subtraction between two spectra. Given that difference editing may suffer from potential subtraction errors, the capability of detecting GABA in a single-shot manner may be highly beneficial. Here we report a novel triple-refocusing approach that can completely resolve the GABA 2.29-ppm resonance from neighboring resonances at 3T.

The ¹H MRS sequence used had three 180° RF pulses following a 90° excitation pulse, as shown in Fig.1a. The first and third 180° pulses were slice selective (13.2ms; bandwidth 1.3kHz) and the second 180° pulse was non-slice selective. Volume-localized density-matrix simulations were conducted for optimizing the second 180° pulse duration and the triple-refocusing subecho times TE­₁, TE₂ and TE₃, incorporating experimental RF and gradient pulses. Spectra of GABA, Glu and Gln were numerically calculated for various non-slice-selective 180° pulse durations (14 - 32ms) and various subecho times TE­₁, TE₂ and TE₃ (16 - 60ms). Among ~150,000 simulated spectra, an optimal pulse duration and sub-echo time set was obtained with criteria; 1) high amplitude and narrow GABA signal at 2.29-ppm, and 2) good separation of GABA from the Glu and Gln C4-proton resonances.

In-vitro test of the GABA-optimized triple-refocusing sequence was conducted on a phantom solution with GABA (1mM), Glu (12mM) and creatine (10mM), at pH=7.0 and temperature 37°C. Seven normal adult subjects (4 male and 3 female) were recruited. In-vivo ¹H MRS data were obtained, with triple-refocusing MRS, from occipital gray-matter (OG) and occipital white-matter (OW) dominant regions. The voxel size was 23×25×25mm³ for OG and 28×20×20mm³ for OW. Data acquisition parameters included NEX=192 and TR=2s (scan time 6.4 min). Data were acquired with a 32-channel head coil in a 3T whole-body scanner (Philips Medical Systems). Spectral fitting was performed, with LCModel software⁴, using in-house calculated basis spectra of 20 metabolites. Metabolites were quantified with reference to water at 45 M. T2 relaxation effects were corrected using published T2 values⁵.

Numerical simulations indicated that the GABA 2.29-ppm resonance was temporally maximum and can be resolved from adjacent signals of Glu and Gln, with a triple-refocusing subecho time set of (TE₁,TE₂,TE₃) = (34,14,34) ms and a 14-ms non-slice selective 180° pulse (1.2kHz bandwidth, Fig.1b), tuned to 2.5-ppm. In phantom validation (Fig.2), the GABA signal at 2.29-ppm was clearly separated from the Glu 2.35-ppm signal, which was not the case in short-TE MRS. Figure 3 presents representative in-vivo spectra from OG and OW regions, together with LCModel fit and individual metabolite spectra. A signal was clearly discernible at 2.29-ppm, well separated from the large Glu signal at 2.35-ppm. The 2.29-ppm signal was large in the OG spectrum and relatively low in the OW data. GABA was estimated to be 0.95 and 0.49mM with CRLBs of 8% and 11% for OG and OW, respectively. Similar results were observed in all 7 subjects (Fig.4). The mean concentration of GABA was significantly higher in OG than in OW (0.86 vs. 0.44mM; p < 0.001) (Fig.5). The mean GABA CRLB was 7% and 13% in OG and OW, respectively.We report a novel triple-refocusing MRS method that provides complete separation of the GABA 2.29-ppm signal from the Glu C4-proton resonance in a single-shot manner. The major mechanism for the signal separation is narrowing of GABA and Glu multiplets at the optimized subecho times and with non-slice selective 180° pulse. Specifically, the Glu C4-proton resonances are strongly coupled and the signal intensity and patterns are sensitive to individual subecho times while the signal from the weakly-coupled GABA spins depends on the total TE. The signal yield of the new method is ~2 fold higher than that of J-difference editing, reducing the scan time substantially (e.g., 6.4min vs. 14min⁶ for 14mL). Our data show that the GABA level is significantly higher in OG than in OW, in agreement with prior studies⁷. Small CRLBs of GABA and small residuals in spectral fitting support precise measurements of GABA. In conclusion, GABA in the human brain can be measured reliably and precisely using an optimized triple-refocusing sequence at 3T. The method may provide an effective tool for studying potential alterations in GABA levels in neuro-psychiatric disorders.