LC Model

LCModel was utilized to quantify the spectral data (in the range of 1.8–4.0 ppm) (Provencher, 2001), which consist of default macromolecule components and 14 basis metabolite functions (N-acetylaspartate (NAA), N-acetylaspartylglutamate (NAAG), aspartate, lactate, creatine (Cre), γ-aminobutyric acid (GABA), glucose, glutamate (Glu), glutamine, glutathione, glycerophosphorylcholine (GPC), phosphorylcholine (PCh), myoinositol, and taurine). LCModel outputs contain Cramer-Rao Lower Bounds (CRLB), which provide an expression of the uncertainty of the estimate for each metabolite. The CRLB provides an estimate of the lowest achievable variance for an estimator and is used as an indicator of how well the measurement of a given metabolite fits the ideal spectrum for that metabolite, providing an assessment of data quality. Additionally, outputs were used to visually inspect spectra and model fit, which we used as a first-pass exclusion criteria for data, along with the CRLB.

Brain neurotransmitter levels are given as metabolite ratios of the neurotransmitter of interest to creatine (Cr) (Glu/Cr or GABA/Cr). This allows for a shorter acquisition time (as compared to using a water reference) while also providing for a means to control for inter-subject variability due to the factors such as an amount of cerebrospinal fluid in the voxel. Cr is used as the denominator in metabolite ratios because it is a metabolite with a strong signal and reliable chemical shift, in addition to being relatively stable under healthy conditions (Rackayova et al., 2017) and with age (Saunders et al., 1999).

LCModel fits were first visually inspected, and data were excluded for bad model fits due to obvious artifacts, lipid contamination or baseline distortion. Data were then excluded if any of the three major metabolite peaks, GPC/Cho, NAA/NAAG, and Cr had a CRLB greater than 10% or if the macromolecule/Cr content in their spectra was greater than 3, in order to remove distorted data that would lead to improper estimates of harder to detect metabolites.

MRSI modeling. previously expensive software, now free to compile. estimates metabolite concentrations for MRSI in HC @ 7T and PFC ROI Placement. Requires basis functions specific to scanner (from MRRC, also stored on rhea)