Assessing the Reliability of MRS Estimates of GABA and Glx
Poster Session D - Monday, March 31, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom
Silvia Abbasi1 (silviaa@umich.edu), Esther Kim1, Bingjie Liu1, Noah Reardon1, Kayla Wyatt1, Thad A. Polk1; 1University of Michigan
Magnetic resonance spectroscopy (MRS) is increasingly used to estimate levels of the inhibitory neurotransmitter gamma-aminobutyric acid (GABA) in the human brain non-invasively. This is challenging because GABA concentrations are relatively low and its signal overlaps with other metabolites’. A common approach uses a MEGA-PRESS sequence, alternating between GABA-selective excitation pulses and non-selective pulses to compute a difference spectrum isolating GABA. GABA estimates derived this way are often corrected for the brain region's tissue composition (tissue correction) and the proportion of GABA expected in gray versus white matter (alpha tissue correction). A similar approach estimates levels of glutamate and glutamine in a combined signal called Glx. This study examines the test-retest reliability of MRS estimates of GABA and Glx in a longitudinal sample of 59 older adults (mean age at initial session = 70.02, second session = 74.20, average interval = 4.21 years) using intraclass correlation coefficients. All participants completed an MRS scan at 3T using MEGA-PRESS, and data were analyzed with Gannet. We placed six 3cm^3 voxels bilaterally in auditory, sensorimotor, and ventrovisual cortices. Uncorrected estimates relative to water and creatine were analyzed, as were tissue-corrected and alpha tissue-corrected estimates. Paired t-tests revealed that average GABA uncorrected, tissue-corrected, and alpha tissue-corrected estimates, as well as estimates relative to creatine, were significantly more reliable than their Glx counterparts. Likewise, average GABA uncorrected estimates were significantly more reliable than GABA tissue-corrected and alpha tissue-corrected estimates, and average GABA estimates relative to water were more reliable than GABA estimates relative to creatine.
Topic Area: METHODS: Neuroimaging