Validation of in vivo MRS measures of metabolite concentrations in the human brain

Purpose In vivo magnetic resonance spectroscopy (MRS) is the only technique capable of non‐invasively assessing metabolite concentrations in the brain. The lack of alternative methods makes validation of MRS measures challenging. The aim of this study is to assess the validity of MRS measures of human brain metabolite concentrations by comparing multiple MRS measures acquired using different MRS acquisition sequences. Methods Single‐voxel SPECIAL and MEGA‐PRESS MR spectra were acquired from both the dorsolateral prefrontal cortex and primary motor cortices in 15 healthy subjects. The SPECIAL spectrum, as well as both the edit‐off and difference spectra of MEGA‐PRESS were each analyzed in LCModel to obtain estimates of the absolute concentrations of total choline (TCh; glycerophosphocholine + phosphocholine), total creatine (TCr; creatine + phosphocreatine), N‐acetylaspartate (NAA), N‐acetylaspartylglutamate (NAAG), NAA + NAAG, glutamate (Glu), glutamine (Gln), Glu + Gln, scyllo‐inositol (Scyllo), myo‐inositol (Ins), glutathione (GSH), γ‐aminobutyric acid (GABA), lactate (Lac) and aspartate (Asp). Then, having obtained up to three independent measures of each metabolite per brain region per subject, correlations between the different measures were assessed. Results The degree of correlation between measures varied greatly across both the metabolites and sequences tested. As expected, metabolites with the most prominent spectral peaks (TCh, TCr, NAA + NAAG, Ins and Glu) had the most well‐correlated measures between methods, while metabolites with less prominent spectral peaks (Lac, Gln, GABA, Asp, and NAAG) tended to have poorly‐correlated measures between methods. Some metabolites with relatively less prominent spectral peaks (GSH, Scyllo) had fairly well‐correlated measures between some methods. Combining metabolites improved the agreement between methods for measures of NAA + NAAG, but not for Glu + Gln. Conclusions Given that the ground truth for in vivo MRS measures is never known, the method proposed here provides a promising means to assess the validity of in vivo MRS measures, which has not yet been explored widely. In this study, we acquired brain MRS data from healthy participants using three different single‐voxel MRS acquisition techniques. We then correlated these measures against each other as a means of validating the metabolite concentration estimates. We found good agreement between the different acquisition methods for metabolites with the