A strategy for using tissue water as a concentration standard in (1)H magnetic resonance spectroscopic imaging studies on the brain is presented, and the potential errors that may arise when the method is used are examined. The sensitivity of the method to errors in estimates of the different water compartment relaxation times is shown to be small at short echo times (TEs). Using data from healthy human subjects, it is shown that different image segmentation approaches that are commonly used to account for partial volume effects (SPM2, FSL's FAST, and K-means) lead to different estimates of metabolite levels, particularly in gray matter (GM), owing primarily to variability in the estimates of the cerebrospinal fluid (CSF) fraction. While consistency does not necessarily validate a method, a multispectral segmentation approach using FAST yielded the lowest intersubject variability in the estimates of GM metabolites. The mean GM and white matter (WM) levels of N-acetyl groups (NAc, primarily N-acetylaspartate), choline (Ch), and creatine (Cr) obtained in these subjects using the described method with FAST multispectral segmentation are reported: GM [NAc] = 17.16 +/- 1.19 mM; WM [NAc] = 14.26 +/- 1.38 mM; GM [Ch] = 3.27 +/- 0.47 mM; WM [Ch] = 2.65 +/- 0.25 mM; GM [Cr] = 13.98 +/- 1.20 mM; and WM [Cr] = 7.10 +/- 0.67 mM.
Copyright 2006 Wiley-Liss, Inc.