During the last decade, several functional magnetic resonance spectroscopy (fMRS) studies have been performed to characterize the relationship between neuronal activation and energy or neurotransmitter metabolism. Due to higher SNR and chemical shift dispersion offered at 7T, dynamic changes of metabolite concentrations in the human brain were successfully detected in response to visual, motor, or auditory stimulation,^1-4 whereas there are no reports from similar experiments performed at lower, more abundantly available magnetic fields. Goal of the present study was to characterize the detection limits of metabolic changes and their correlation with activation by a visual stimulus at 3T.

7 healthy volunteers (3 female, 22-35 years) from a cohort of initially 15 participants in an fMRS study were selected for further analysis whereas the remaining subjects had to be discarded due to significant head motion or severe lipid contamination of the spectra. Scans were performed on a 3T Verio system (Siemens Healthcare, Erlangen, Germany) using a 32-channel head coil. The stimulus consisted on a moving star-field paradigm.⁵ An additional colored fixation point was changing randomly to track the attention of the volunteers via a response button. A block design (5 repetitions) with alternation of 30sec of rest and 30sec of activation during gradient-echo EPI scanning (TE 30 ms, TR 2 sec) was used as a functional localizer. A standard SPM12 processing scheme was employed to obtain BOLD activation maps, including realignment, slice-time correction, and spatial smoothing using an 8mm FWHM Gaussian kernel. Localization of a single 8ml voxel for fMRS was done by co-registration of the thresholded BOLD activation map to an MP-RAGE anatomical image (figure 1). Subsequently, first- and second-order shims were adjusted using FAST(EST)MAP,^6,7 and PRESS spectra (TE 30 ms, TR 2 sec, 124 repetitions, 4:08 min) were acquired without water suppression while presenting the same paradigm as employed for fMRI to study the BOLD effect on the water resonance as additional verification of a correct voxel positioning. Finally, water-suppressed PRESS spectra (600 repetitions) were acquired during a 20min visual paradigm consisting of alternating 5min intervals with stimulus OFF/ON/OFF/ON.

Spectra quantification was performed with LCModel,^8 (figure 1) with further analysis of both absolute concentration estimates employing the unsuppressed water signal at rest as internal reference and ratios relative to creatine to avoid potential bias from variation of the linewidth due to the BOLD effect. Spectra were frequency- and phase-corrected using FID-A tools^9 and then averaged using the following schemes (a) 150 averages according to the OFF/ON/OFF/ON stimulus intervals to estimate average metabolite concentration changes due to activation and (b) a moving average of 32 repetitions shifted in steps of 6 repetitions to generate smooth time courses consisting of 95 points. In addition, group-averaged time courses were computed from a subset of 5 subjects that showed minimal inter-subject variation of the linewidth within the precision of the LCModel output (figure 2).

A significant intra-individual increase in glutamate+glutamine (Glx) was found in four of the subjects (minimum increase 1.96%, maximum increase 9.86%) as well as in the group analysis (1.83%, p = 0.017, figure 3). The cross-subjects averaged time course also indicated a corresponding 1.98% concentration increase during the ON periods. These changes are similar to activation-induced glutamate increases previously observed at 7T. FSL analyses of the smooth Glx time courses yielded a mean Z-score of 7.18±0.69 for individual subjects as well as a Z-score of 7.72 for the group-averaged data. A similar result was obtained with simple correlation analyses using a boxcar design function (R = 0.5, p = 10−7).

The reliability of a glutamate concentration estimate at 3T might be questioned due to overlap with glutamine resonances. In our data, glutamine could not reliably be quantified (as indicated by large Cramér-Rao lower bounds) in a large portion of the spectra. Therefore, we prefer to report the Glx values, which provide, however, an indirect indication of excitatory neurotransmission.

A direct correlation between subtle changes of Glx concentrations and the application of a visual stimulus could be shown by single-voxel MRS acquired at 3T, which is consistent with previous reports from ultra-high field studies.