Simultaneous editing of GABA and GSH with Hadamard-encoded MR spectroscopic imaging

Abstract

Purpose: To evaluate the feasibility of simultaneous MR spectroscopic imaging (MRSI) of gamma-aminobutyric acid (GABA) and glutathione (GSH) in the human brain using Hadamard Encoding and Reconstruction of MEGA-Edited Spectroscopy (HERMES).

Methods: Point RESolved Spectroscopy (PRESS)-localized MRSI was performed in GABA and GSH phantoms and in the human brain (n = 3) using HERMES editing and compared to conventional MEGA editing of each metabolite. Multiplet patterns, signal intensities, and metabolite crosstalk were compared between methods. GABA+ and GSH levels were compared between methods for bias and variability. Linear regression of HERMES-MRSI GABA+/H₂ O and GSH/H₂ O versus gray matter (GM) fraction were performed to assess differences between GM and white matter (WM).

Results: Phantom HERMES-MRSI scans gave comparable GABA+ and GSH signals to MEGA-MRSI across the PRESS-localized volume. In vivo, HERMES-reconstructed GABA+ and GSH values had minimal measurement bias and variability relative to MEGA-MRSI. Intersubject coefficients of variation (CV) from two regions within the PRESS-localized volume for HERMES and MEGA were 6-12% for GABA+ and 6-19% for GSH. Interregion CVs were 5-15% for GABA+ and 3-17% for GSH. The GABA+/H₂ O and GSH/H₂ O ratios were ~1.8 times higher and ~1.9 times higher, respectively, in GM than in WM.

Conclusion: HERMES-MRSI of GABA+ and GSH was found to be practical in the human brain with minimal measurement bias and comparable variability to separate MEGA-edited acquisitions of each metabolite performed in double the scan time. The HERMES-MRSI is a promising method for simultaneously mapping the distribution of multiple low-concentration metabolites.

Figure1.

Frequencies and inversion envelope of the editing pulses for HERMES editing of GABA+ and GSH in vivo with symmetrical lipid suppression lobes applied in GABA-OFF scans. These lobes (indicated in red) are placed at 0.7 ppm in sub-acquisitions B and D so as to be symmetrical about the 1.3 ppm lipid resonance relative to the 1.9 ppm GABA ‘ON’ editing pulse.

Figure2.

(a) PRESS excitation volume placements of in vivo HERMES and MEGA experiments overlaid on triplanar T₁-weighted images. (b) Two 3 × 3 voxel areas overlaid on an axial T₁-weighted image from which inter-region and inter-subject CVs were calculated.

Figure3.

GABA and GSH multiplets from HERMES phantom experiments. (a) The first row shows the results from the combined GABA-GSH phantom. HERMES and MEGA-PRESS have equivalent GABA and GSH signal intensities and multiplets across the entire volume of interest. (b) In the GABA-only and GSH-only phantoms (second row) GSH and GABA are segregated into their respective reconstructions with minimal metabolite crossover across the volume of interest.

Figure4.

PRESS excitation volume overlaid on an axial T₁-weighted image (top). Representative MEGA- and HERMES-encoded GABA+ spectra plotted from 2.2 to 4 ppm across the entire volume of interest shown top from subject 2. The MEGA- and HERMES-encoded GABA+ spectra are in good agreement between methods.

Figure5.

PRESS excitation volume overlaid on an axial T₁-weighted image (top). Representative MEGA- and HERMES-encoded GSH spectra plotted from 2.2 to 4 ppm across the entire volume of interest shown top from subject 2. The MEGA- and HERMES-encoded GSH spectra are in good agreement between methods.

Figure6.

Scattergrams plotted on top of box plots of the voxel-by-voxel GSH and GABA+ peak area ratios (HERMES/MEGA) for each subject. The number of voxels included for each subject ranged from 85 to 88 voxels and were located at the center of the VOI (excluding voxels at the edges). Median ratio values for both GABA+ and GSH were close to one, indicating that HERMES and MEGA give comparable results.

Figure7.

(a) Linear regression of HERMES MRSI GABA+/H₂O ratios versus gray matter fraction for all three subjects. GABA+/H₂O was estimated to be 1.84 times higher in gray matter than in white matter (p = 0.0002). (b) Linear regression of HERMES MRSI GSH/H₂O ratios versus gray matter fraction showing a correlation with 1.91 times greater GSH/ H₂O in gray matter than in white matter (p = 0.005).