Purpose

In an earlier report, the relative metabolite ratios obtained using short TE whole brain MR spectroscopic imaging (wbMRSI) of human brain, with data analysis using the MIDAS program ¹, were compared with those obtained using Single-Voxel Spectroscopy (SVS) with data analysis using LCModel ². In this study, the wbMRSI and SVS acquisitions are compared using LCModel, with evaluation at three brain structures and in a larger volunteer sample.

Methods

60 healthy volunteers aged between 20 and 70 years underwent MR examinations at 3T (Verio, Siemens, Erlangen) with a 12-channel phased-array head coil. The scan protocol included an EPSI (TR/TE = 1550/17.6 ms) acquisition over a field-of-view (FOV) of 280 x 280 x 180 mm with a scan time of 16 minutes ³; an axial T1-weighted 3D MPRAGE with the same angulation as that of EPSI; and three SVS using stimulated echo acquisition (TR/TE = 1550/20 ms, 192 acquisitions, voxel volume 4 ml, scan time 5 min.) at right parietal white matter (pWM), occipital gray matter (oGM), and basal ganglia (BG) (Fig.1). To obtain regional metabolite concentrations from the EPSI data two different methods were used: a) the wbMRSI data were analyzed using MIDAS to obtain brain metabolite maps, from which individual voxel results were averaged over the brain regions corresponding to each SVS measurement (wbMRSI method); b) the spectra from the wbMRSI measurement, after B0 field correction and phase correction, were integrated over each of the volumes corresponding to the SVS measurements and the result fitted using LCModel (hybrid method ⁴). The SVS data were also analyzed using LCModel to obtain relative concentrations of n-acetyl-aspartate (NAA), choline (Cho), total creatine (tCr), glutamine/glutamate (Glx), and myo-inositol (mI). Results were compared as ratios to tCr. The values were compared by using one-way analysis of variance (ANOVA) with Bonferroni corrected post hoc tests. Linear regression analyses were made to estimate age-dependence of the metabolite concentrations obtained with the three methods.

Results/Discussion

No statistically significant difference beween male and female subjects was found, thus the male and female data were combined for further analysis. Metabolite concentrations (in ratios to tCr) and the spectral linewidths with standard deviations (SD) determined with wbMRSI, hybrid, and SVS methods are given in Table 1, in which the results of ANOVA comparison were also included, and in Fig.2. The results of linear regression analysis are given in Table 2. At each of three brain structures, the mean concentrations (in ratios to tCr) derived from three methods were close to each other and showed similar relations between the metabolites (Fig.2), indicating that all the methods could be used reliably to measure local brain metabolites. Spectral linewidths, especially those in basal ganglia with more local field distortions, were narrower for wbMRSI (Table 1) due to the smaller nominal voxel size of the EPSI acquisition and application of B0 correction. Linear regression analysis revealed age-related decreases of NAA in all 3 ROIs measured with wbMRSI, but only in two ROIs measured with SVS (Table 2) – the reason could be the lower variance for NAA determined with wbMRSI (Fig.2, first box vs. third box in each). For the measured volume (4ml) the use of LCModel to fit the integrated spectra from EPSI data did not show benefits concerning the derived results. ANOVA comparison with Bonferroni corrected post hoc test revealed significant differences between individual metabolite concentrations measured with three different methods at each brain structure (Table 1), indicating that it is necessary for each acquisition technique to have its own reference for quantification purpose.

Conclusion

Both short TE wbMRSI and SVS could be used to estimate local brain concentrations of NAA, Cho, tCr, mI, and Glx, with consistent results. By using the same TR and clinically acceptable scan times the wbMRSI method provided better spectral resolutions with narrower linewidth. Data with lower variance seems to benefit further data analysis.

Acknowledgements

This work was supported by Deutsche Forschungsgemeinschaft and by NIH grant R01 EB016064 (A.A.M.).