They confirm the post-mortem results in the SN region as well as with findings from PD rat model based on proteasome inhibition. It should be noted that the FWHM values were similar in both groups Salvianolic-acid-B allowing comparison of PD patients and controls. Emir et al. also found no differences in FWHM values determined by LCModel between PD patients and controls in the SN region at 7 T. Differences in SNR are due to reduced main metabolite concentrations in PD patients compared to controls but are not a marker for lower spectral quality. We found significant differences between the rostral part and the caudal part in both groups, as expected from anatomical features. Additionally, these differences are 
also in good agreement with post-mortem results by Gerlach et al.. Moreover, the gradients of metabolite concentrations between the rostral and caudal regions have the same direction in PD patients and controls in both individual data and averaged spectra. This applies to all metabolites, regardless of whether they are increased or decreased for pathological reasons. This study gives the necessary biochemical explanation for our phenomenological findings published recently,. However, the advantage of the phenomenological investigation using the Syngo software in previous studies, is that the spectra evaluation can performed directly on the MR scanner, which is important for clinical application because it is the only approved method. It is reproducible, user-independent, very fast, and robust. Moreover, it could be shown that significant differentiation between patients with idiopathic PD and atypical Parkinsonism was possible in individual subjects. However, problems in quantification are caused by complex multiplet resonance patterns, overlapping metabolites, variations in line shape and underlying baseline variations, which cannot be fitted accurately using only a single peak, as was done with the Syngo software package. Therefore, the estimated rostral-to-caudal NAA/Cr ratios in previous studies, could only be evaluated as phenomenological values and did not allow pathophysiological interpretation. Furthermore, in previous studies, the standardized automatic baseline correction was performed using a 6th-order polynomial in the calculation range between 0 to 4.3 ppm. However, metabolites such as NAA, GSH, Glu, Gln, and GABA show complex multiplet resonance patterns, especially in the range from 2.1 to 2.9 ppm with low signal Praeruptorin-B amplitudes caused by J couplings and low metabolite concentrations. The automatic baseline correction draws these multiplets partially to the zero line. On the other hand, this 6th order polynomial raises the Cr signal and possibly also the Cho signal, resulting in overestimation due to integration of the signal at 3.0 ppm. With the new findings presented in this work, we can provide a pathophysiological explanation for this. However, the presented method is not applicable for routine clinical application due to the unacceptable time and effort required. Using the optimized basis data set for the LCModel analysis yielded robust and reliable values for the metabolite concentrations in both groups. The improved fitting quality was reflected by the closely reproduced the in-vivo spectra, flatness of the resulting baseline and negligible residuals at the noise level without a considerable chemical-shift dependence. The main limitation of our study is that the clinical diagnosis could not be confirmed by post-mortem histological examinations. Additionally, the very small voxel size necessitated by the anatomy of the SN resulted in low SNR of individual spectra. The relatively high iron content of SN caused broader intrinsic line widths.