Influence of basis set composition on metabolite quantification using 1H-MRS at 3T: combining in-silico, in-vivo, and in-vitro evidence.

Motivation: MR spectra are quantified by model-fitting in time or frequency domains using a metabolite basis set. Research on basis set selection is limited making it unclear which components should be included, impacting quantification and reproducibility. Goal(s): To determine an optimal basis-set for quantification of metabolites of interest depending on SNR. Approach: We evaluate the impact of basis set choice on accuracy and precision drawing evidence from synthetic, human brain and phantom data. Results: A reduced number of spectral components can improve precision with limited impact on accuracy. It is important to consider the noise level when selecting basis sets. Impact: We demonstrate reliable metabolite quantification with reduced-basis sets (<16 metabolites) for data with typical SNR at 3T, with an 11-metabolite basis set appear to offer optimal accuracy and low SD. These findings have a potential to enhance reproducibility across studies.