This work describes a methodology to perform functional MR spectroscopy (
Data acquisition: The fMRS auditory experiment (Fig-1) was performed in five healthy volunteers at MR 7T scanner (Terra; Siemens, Erlangen, Germany). A 20x40x20mm3 volume located on the left superior temporal gyrus (LSTG) was acquired using STEAM localization and the following parameters: TE/TR=20/3000ms, TM=10ms, 768 signal averages during the full experiment, and a total acquisition time of 38min. Likewise, a visual fMRS experiment was also performed on a healthy volunteer for validation of the methodology using the setup described in Fig-4 and reported in literature3.
Auditory stimuli and fMRS paradigm: The fMRS auditory paradigm (Fig-1) consisted in 8 blocks of STIM frequent (STIM freq), eight blocks of STIM count and 8 blocks with no stimulus (REST), each of 48 MRS averages, total acquisition duration was 38 min. In the STIM freq blocks, 1 kHz pure tones were presented to the subject, in STIM count blocks, 2 kHz pure tones added (30%) among the 1kHz frequent tones (70%). Subjects were asked to keep eyes open during the entire experiment and to silently count the number of 2 kHz tones per STIM count block.
Data Processing and Quantification: The pipeline used for reconstruction of the spectra (Fig-2) was implemented in python using OpenMRSLab4 and included the following steps: (1) coil combination, (2) frequency alignment to water, (3) zero-order phase removal, (4) combination of 4 signal averages (sub-blocks) to increase SNR for processing, (5) water removal at each sub-block using HSVD, (6) frequency alignment to NAA, (7) zero-order phase removal, (8) filtering of the dynamic temporal signal (e.g. sliding window average, Fourier thresholding), and (9) combination of sub-blocks to obtain the desired temporal resolution to match the paradigm. Frequency/phase corrections eliminated distortions caused by external effects that affect the measurement, such as B0-drifts, eddy currents, temperature, or the BOLD effect (Fig-2a and Fig-5a). The reconstructed spectra, corresponding to consecutive time points, were fitted using LCModel with a simulated basis set.
The analysis of the auditory fMRS data showed the presence of BOLD signal when subtracting each STIM block with its nearest REST block (Fig-3a). Variations in metabolite concentrations were observed and enhanced using dynamic filtering. The quantified metabolite concentrations processed with the different methods superimposed with the auditory paradigm are shown (Fig-3b). Similarly, improvements in the signal obtained for the fMRS experiment in the visual cortex were obtained (Fig‑5b). The visual experiment served as a validation for the methodology as it was able to reproduce previously published experiments3.
[1] Taylor S F, et al. GABA abnormalities in schizophrenia: A methodological review of in vivo studies. Schizophr Res. (2015).
[2] Javitt D C, et al. Auditory dysfunction in schizophrenia: integrating clinical and basic features. Nat Rev Neurosci. (2015).
[3] Bednařík P, et al. Neurochemical and BOLD responses during neuronal activation measured in the human visual cortex at 7 Tesla. Journal of Cerebral Blood Flow & Metabolism (2015)
[4] Rowland B, et al. An open-source software repository for magnetic resonance spectroscopy data analysis tools. ISMRM MR Spectroscopy Workshop (2016).
[5] Provencher S. W. Automatic quantitation of localized in vivo 1H spectra with LCModel. NMR in Biomedicine 14, 260–264 (2001).