2.3. GABA MRS acquisition and analysis

MRI and MRS exams were performed using a 3T GE Signa MRI scanner equipped with an 8-channel head coil. A high-resolution 3D T1-weighted FSPGR sequence (TR/TE=6.26/2.67 ms, resolution: 1×1×1 mm³) was followed by a fast T2-weighted sequence, acquiring sagittal and axial images aligned with the anterior and posterior commissure (AC/PC) for accurate placement of the MRS volumes of interest (VOIs). GABA spectra were acquired using the MEGA-PRESS sequence (TR/TE=2000/68 ms) (Mullins et al., 2014) with the editing pulses centered on 1.9 ppm (edit ON) and 7.5 ppm (edit OFF). 256 averages were acquired for edit ON and edit OFF spectra in an interleaved fashion. 8 averages of water reference scans without water suppression were acquired for frequency and phase correction. To determine the thalamic GABA level the MRS VOI was centered on the right thalamic region (25 mm × 30 mm × 25 mm) as shown in Fig. 2. The same GABA-edited MRS measurement was also performed on the right corpus striatal region (25 mm × 30 mm × 25 mm), which is referred to as the striatal region in this paper, including globus pallidus and a part of putamen and caudate nucleus. This VOI intentionally overlapped partly with the thalamus VOI to detect or rule out a possible influence from striatal GABA signal to the large voxel size of the thalamic VOI.

Thalamus a) and striatum b) VOI for GABA measurements shown on sagittal and AC-PC aligned views on T2-weighted images, and on an axial view of a brain tissue map (red: white matter, blue: grey matter, green: CSF). Please note the tissue maps have a slightly different angulation from the AC-PC aligned T2-weighted images. Also shown are typical GABA-edited spectra from thalamus (top) and striatum (bottom) fitted with LCModel (red lines).

Edit ON and OFF spectra were first post-processed with a Matlab 2013a (The MathWorks, Natick, 2013) tool to perform coil combination, zero and first order phase correction, as well as frame-by-frame spectral alignment. Difference spectra obtained from subtracting the averaged OFF spectra from the averaged ON spectra, showing a GABA peak at 3 ppm, were quantified using LCModel V6.3-1B (Provencher, 1993). A basis set generated by density matrix simulation using GABA coupling constants from Kaiser et al. (2008) was used. The examples of GABA-edited spectra from the thalamus and the striatum fitted by LCModel are shown in Fig. 2. 3D T1-weighted images were segmented into grey matter (GM), white matter (WM) and cerebrospinal fluid (CSF) using statistical parametric mapping (SPM8, Wellcome Department of Imaging Neuroscience, London, United Kingdom). The percentages of GM, WM and CSF for each VOI were then calculated from the tissue probability maps with a home-made Matlab (The MathWorks, Natick, MA) code. Since CSF has a near-zero concentration of GABA and other metabolites, GABA levels were corrected for CSF to obtain tissue GABA concentrations. Examples of tissue maps of GM, WM and CSF overlaid with the GABA VOIs are shown in Fig. 2. A widely-used approach in the MRS literature expressed by the following equation (Chowdhury et al., 2015) was applied to obtain the CSF-corrected tissue GABA levels from each VOI: