The aim of this work was to study and correct the influence of varying coil load and local B(1) field in single volume MR spectroscopy. A simple, precise, and user-independent way to adjust the transmitter gain has been developed and validated. It is based on a fit of the localized signal to flip angle variation around 90 degrees. This method proved to be robust against B(1) gradients and suitable for in vivo applications. Local B(1) correction was combined with an external reference and decomposition of the volume into CSF and tissue to obtain a comprehensive absolute quantification of tissue water content and metabolite concentrations in human brain. STEAM localized spectra of parietal and insular gray matter and subparietal white matter (n = 11, TE = 30 ms) were analyzed using a linear combination of model spectra (LCModel). Coefficients of variation (CV) between 1.5% and 4% were obtained for the tissue water content (1-2% in a single subject). The CVs of major metabolite concentrations (4-21%) were dominated by the errors of the spectral analysis. The largest B(1) variation in the in vivo experiments (range 30%) was due to changes in coil load. Differences in regional sensitivity due to B(1) inhomogeneity (parietal: 8% and 9%; insular: 16%) were found to be the second largest source of variation. Correction for local B(1) improved standard deviations and intra-subject reproducibility. On average, sensitivity was 9% less in insular than in parietal gray matter. If ignored, significant differences were introduced for water and N-acetyl-aspartate or were obscured for creatine and cholines. Hence, local sensitivity correction proved to be necessary for regional comparison of absolute metabolite concentrations.
Copyright 2000 John Wiley & Sons, Ltd.