MR spectroscopy
Magnetic resonance spectroscopy quantifies neurometabolite concentrations — glutamate, GABA, glutamine, NAA and others — within a defined voxel. It answers a different question from fMRI: not where activity changes, but what the local neurochemistry is.
The mrs/ workflows are built on Osprey, an open
MRS analysis toolbox, which must be installed separately and added to the MATLAB path.
The workflow in outline
- Load raw spectroscopy data and the associated structural scan
- Process — coil combination, frequency and phase correction, eddy-current correction, averaging
- Fit the spectrum to a basis set to estimate metabolite concentrations
- Segment and co-register the voxel against the structural image, so tissue composition is known
- Quantify, correcting for the grey matter, white matter and CSF fractions within the voxel
- Export to a table for statistical analysis
Step 5 is not optional. A voxel containing more CSF has less tissue to generate signal, so uncorrected concentrations partly reflect voxel placement rather than neurochemistry — and voxel placement varies systematically with brain morphology, which can differ between the groups being compared.
Quality control
Osprey reports linewidth, signal-to-noise ratio and fit residuals. Set exclusion criteria in advance and apply them uniformly. Spectral quality varies more between subjects than fMRI data quality does, and deciding what to exclude after seeing the group results is not a decision you can defend.
Downstream
MRS output is typically a small table of metabolite concentrations per subject and voxel, which goes into conventional statistical models rather than the imaging pipelines. Where MRS is combined with fMRI or PET in the same study, the metabolite measure usually enters as a predictor or covariate.
Consult Osprey’s own documentation for the processing
and fitting detail; the mrs/ scripts handle our study conventions around it.