Characterization of the full ¹H spectrum may allow for better monitoring of pathologies and metabolism in humans than is currently possible using upfield spectra only. The focus is usually on characterization of the upfield part of the spectrum, even at very high fields^[1]. The downfield part at 5-10ppm is less well characterized, and a confirmation of peak or metabolite identification as well as determination of relaxation and exchange parameters may substantially aid in clinical and basic research. Some data is available on downfield peaks for animal brain at high fields^[2], as well as human brain at 3T^[3]. The current work aims to elucidate the downfield spectrum in human brain and to quantify T₁ and T₂ in grey matter at 7T using series of spectra with variable TE and inversion recovery (IR) delays.

Acquisition methods were similar to those used in Ref.[4]; i.e. a Philips 7T scanner; quadrature transmit/receive surface coil (Rapid Biomedical); STEAM localization with VAPOR water presaturation; ROI size of 16 cm³; second-order B₀ shimming; TM=26.0ms, TR=4000/7000ms for the TE/IR series, respectively. A series of TEs at 13, 23, 35, 47, and 60ms was acquired in 12 healthy volunteers (two data sets discarded due to excessive line broadening). For the IR series, an adiabatic inversion pulse was interleaved with the VAPOR sequence, and a series of IR times was acquired for 10 healthy volunteers at: 45, 150, 290, 580, 800, 1200, 2500, 4000, and 6000ms IR delay time.

The overall averaged data sets from both series were combined to develop a spectral model of partially overlapping signals in FiTAID^[5]; prior knowledge was defined with twelve peaks in the 5 to 9ppm region. The N-acetylaspartate (NAA) and α-glucose (Glc) peaks were defined as binary patterns, as modeled in VESPA^[6]. Homo­carnosine (hCs) was modeled as 2 peaks at well-defined chemical shifts^[7]. The TE and IR data sets were then fitted simultaneously for each case with T₂ and T₁ as fitting variables, and water T₁ and T₂ values were also calculated. Errors were estimated as Cramer-Rao lower bounds (CRB). Results from very poor fits were removed such that the data for the 5.8 and 6.1 ppm peaks are not well defined.

Individual and average TE and IR spectra are shown in Fig. 1. The average short-TE spectrum overlaid with the derived heuristic model and the individual peaks is presented in Fig. 2. The relaxation parameter estimates are listed in Table 1, including the average CRB.

The heuristic model describes the experimental data well and the results for many of the peaks are very consistent across subjects (e.g. NAA and peaks around 6.8-7.3ppm, as seen from the SD and mean CRB in the table). Other peaks were more difficult to fit; although clearly visible in the averaged spectrum used to create the model, SNR is limited in the individual spectra, or they are partially affected by water suppression (Glc).

T₁ values found at 7T are mostly substantially higher than those found at 3T^[3], in particular for the NAA peak. Several peaks show a particularly short T₁ in comparison to the others, indicating that they predominantly originate from macromolecules, which is confirmed visually in Fig. 1d, where traces with IR delays of 0.29-0.80s correspond to metabolite nulling (in particular the 6.8 ppm peak (blue arrow), which features short T₁ and T₂). The T₂ values are in general much shorter than those found for upfield peaks^[2]. For NAA, it is interesting to note that both the T₁ and T₂ experiments show that it is composed of more than one peak with different resonance frequency and linewidth (Fig. 1b,d red arrows, comparing the spectra at 1.20s recovery time, the one at TE=60ms, and those at shorter TE and other inversion times); strong inhomogeneous broadening likely contributed to the broad peak. Exchange might account for some of the shorter T₂’s, but it is unlikely to affect the NAA value, as the exchange rate of NAA is quite low^[3].

For the peaks with confirmed assignment, i.e. NAA, homocarnosine, and α-glucose, the concentrations were found to match those of the literature very well (data not shown)^[7,8].

Assignments for the other peaks are still unconfirmed; however, some suggestions have been made based on literature in previous work^[8,9]. Peaks with moderate to fast exchange are not expected to be visible in these spectra due to the use of VAPOR water suppression. In particular, this includes those from amides at 8.2-8.5ppm and even more so for amine protons which are a basis for CEST contrast^[10].