In MRS and MR spectroscopic imaging (MRSI) the struggle between signal-to-noise ratio (SNR) and voxel size seems never ending. With abundant water and lipid signals present, but with metabolite concentrations in the millimolar range, it remains a challenge to localize and quantify metabolite signals in small voxels. Compared with traditional MRI, the sizes of single voxels or voxels in an MRSI matrix appear disappointingly large for the non-involved scientist or clinician.

Although people with experience in spectroscopy acknowledge these shortcomings, we often do have a final gap to bridge: we cannot expect the spectra of all voxels in an MRSI matrix to be of the same quality as a spectrum from a single-voxel experiment. This is a somewhat self-inflicted shortcoming, since there is a tendency in the scientific literature for authors to show the best spectra from a matrix as representative of the whole dataset, and characteristic spectra for different brain tissues are still often obtained from single voxel acquisitions.

In this issue of NMR in Biomedicine, Strasser et al. (2021) bring the spectral quality of multiple voxels closer to the quality of a single voxel spectroscopy acquisition. They combine two independent ways to improve the SNR and accuracy of the detection of a specific metabolite in brain tumours into a single setup: usage of multiple local coils for both MR signal reception (alternating current, AC) and B₀ shimming (direct current, DC). Receiving AC MR signals with multiple closely fitting loops provides a high SNR, and using the same (and additional) loops with DC can locally homogenize the main magnetic field.

The improvements in spectral quality that dedicated B₀ shimming can provide have been shown over the years with automation and higher order shims (eg in References 1, 2) but also with the use of multiple additional shim coils around the area of interest.³ Strasser et al elegantly show, with simulations representing ideal cases, that for D-2-hydroxyglutarate (2HG), an oncometabolite that is present in gliomas with the isocitrate dehydrogenase 1 (IDH1) mutation,⁴ these simulations predict a strong decrease in ability to properly fit and accurately estimate 2HG concentrations at linewidths above about 0.12 ppm (~15 Hz). It is thus of crucial importance to obtain spectra with narrow linewidths across large parts of the brain.

In their paper, the authors illustrate the unique capabilities of the instrumental setup they achieved over the years^5-7 by combining it with an elegant method of MRSI data acquisition with navigators, adiabatic excitation and refocusing, spiral readouts and switching B₀-shim settings for improved lipid suppression. The application of their protocol for direct detection of 2HG has created a method that can truly make a difference in the non-invasive characterization of glioma, a dreadful disease that still has a very poor prognosis.

Taking the quality of the vast majority of MRSI voxels closer to the level of multiple single voxels is a major achievement. We look forward to seeing this coil and its accompanying acquisition and postprocessing protocol progressing towards an integrated product for clinical use.