Ultra-short echo/acquisition delay MRS spectra have a strong characteristic background consisting of macromolecule (MM) resonances superimposed on the signal of metabolites. Omission of MM contribution in the quantification of such spectra may yield substantial errors in the quantified metabolite levels¹. As shown previously, the simulation of MM background is insufficient at ultra-high fields and a separate acquisition of MM was found to be more accurate². Since regional differences of MM exist even in the healthy brain and a further change of distinct MM resonances^3,4 was observed in various diseases, this single MM background should be replaced by individual MM peaks to account for more prominent regional and pathologic changes.An average MM spectrum of the full spectral range was acquired using double inversion MRSI as described previously⁵ from 6 healthy volunteers (28±2 years). Metabolite residuals were removed from the averaged spectrum and parameterized by fitting using HLSVD implemented in jMRUI 5.2. Individual components (altogether nine separate peaks - 0.91 ppm (MM1), 1.21 ppm (MM2), 1.43 ppm (MM3), 1.67 ppm (MM4), 2.04 ppm (MM5), 2.26 ppm (MM6), 2.99 ppm (MM7), 3.21 ppm (MM8), and 3.77 ppm(MM9)) were saved separately and included into the basis set. Additionally, to assess the risk of over-parameterization several peaks were combined into altogether 5 groups (MM1 (0.9ppm); MM2-MM4 (1.2-1.6ppm); MM5-MM6 (2.04-2.26ppm); MM7-MM8 (2.99-3.21ppm) and MM9 (3.77ppm)) and included in the basis set. Data from 3 volunteers (32±3 years) were measured with a Magnetom 7T (Siemens Healthcare, Erlangen, Germany) scanner and 32 channel head coil (Nova Medical, Wilmington, USA) with the following parameters: TR, 600 ms; TE*, 1.3 ms; flip angle, 45°; FoV, 220 × 220 mm²; matrix size, 64 × 64; nominal voxel size, 3.4 × 3.4 × 12mm³; scan time ~30 min; one average with an elliptically sampled k-space acquired in a pseudo-spiral pattern; Data from one Multiple Sclerosis patient were acquired using the same protocol and acceleration using 2D Caipirinha⁶ with acceleration factor R=6 and scan time of ~5 min. These data were quantified in LCModel 6.3 using three different basis sets: a) a full measured MM spectrum, b) parameterized MM peaks and c) MM groups (Fig.1) and compared between white matter (WM) and gray matter (GM). A small region of GM and WM (min. 80% of the tissue) was defined in every volunteer.The data from three volunteers are displayed in form of boxplots (Fig.4). The paired t-test showed a significant difference between GM and WM levels for MM1, MM2, MM3, MM4, MM6, MM7 and MM8. The MM1 through MM6 peaks ranging from approximately 2.3 ppm to 0.5ppm were found to be higher in GM compared to WM. On the contrary, the region from 3.2 ppm to 2.8 ppm showed to be significantly higher in WM than in GM.We have quantified MM levels of individual MM peaks, MM groups and the full MM spectrum as well as created maps of their spatial distribution (Fig.2). Although the overall MM background in healthy volunteers was found to be higher in GM than in WM^5,7, according to our findings this applies only to MM resonances from 2.26 ppm to 0.9 ppm (MM1-MM6), whereas the peaks at 2.9 ppm and 3.2 ppm (MM7-MM8) were higher in WM (Fig.2 and Fig.4). These MM resonances play an important role in GABA quantification (the so called GABA+ signal). A possible over-parameterization was excluded based on the comparison of metabolite levels for all quantification methods (Fig.3). Results obtained from a patient (Fig. 5) show that the basis set with parameterized MM can be used also for MM mapping in diseases which was not possible with unparameterized MM spectra due to a wild baseline. These results are however preliminary and need further investigation.Parameterization of an averaged MM spectrum allowed us to detect subtle changes of individual MM peaks and provided additional information on MM spatial distribution in brain without biasing the metabolite quantification by over-parameterization.