Bias-Reduced LCModel Quantification of 1H Brain MRS with Baseline-and-MM Removal: Application in a Circulatory Arrest Study using Neonatal Pigs

Quantification of in-vivo short-TE MRS using LCModel suffers from biases due to varying baseline and macro-molecule components, especially for metabolites with coupled resonances. This problem exacerbates at low SNR as the baseline estimated becomes less accurate. To improve quantification, baseline and macro-molecule components were estimated from a high SNR line-broadened spectrum and then subtracted from the original spectrum. Using the baseline and macro-molecule removed spectrum, significantly reduced biases for Gln/tCr and GABA/tCr were achieved at different SNRs. Using this method, LCModel quantification revealed a more prominent trend of increasing Gln/tCr and increased GABA/tCr during circulatory arrest in neonatal pig brain.