Reproducibility of in vivo metabolite quantification with proton magnetic resonance spectroscopic imaging

Purpose: To investigate intra- and interscanner in vivo reproducibility of brain metabolite quantification using 1H magnetic resonance spectroscopic imaging (1H-MRSI) (PRESS localization, TE = 30 msec, voxel volume = 2.3 mL) and the linear combination model (LCModel). Materials and Methods: One subject had a total of nine scans on three occasions at a single site, and three subjects had single scans at two sites. Coefficients of variation (CVs) were estimated using different statistical models applied to intra- and interscanner data; therefore, only qualitative comparisons may be made between results. Results:: CV (intra-/interscanner) for metabolite quantifications were choline, 12.3%/10.1%; creatine, 9.9%/10.6%; glutamate + glutamine, 15.8%/13.6%; myo-inositol, 18.5%/14.7%; and N-acetyl-aspartate + N-acetyl-aspartyl-glutamate, 6.1%/7.0%. Overall, total intra- and intersubject variability was greater than intra- and interscanner variability. Conclusion: When quantifying metabolic concentrations using the methods employed in this study, biological factors contribute a greater proportion to measurement variability than measurement errors. Using this technique, intra- and intersite measurement errors are of the same order. © 2002 Wiley-Liss, Inc.