Multi-voxel MR spectroscopic imaging of the brain: utility in clinical setting-initial results

Compared to single voxel methods, MR spectroscopic imaging (MRSI) of the brain provides metabolic information with improved anatomical coverage and spectral resolution, but may be difficult to perform in the clinical setting. We evaluate the factors influencing spectral quality in MRSI using a semi-automated method, focussing on lipid contamination, and phase correction errors related to magnetic field inhomogeneity.

We retrospectively analysed MRSI studies planned by radiologists and radiographers. Two-dimensional MRSI studies using point-resolved spectroscopy (PRESS) localisation, at long echo time (135 or 144 ms) were acquired on a 1.5 T scanner. Studies that contained lipid contamination and abnormally inverted spectra were reviewed and the latter correlated with anatomic location at the base of skull, and with the area of the region of interest (ROI) studied.

Of 128 consecutive MRSI studies, six showed abnormal inverted spectra, of which four were acquired at the base of skull. Multivariate logistic regression analysis showed that study location at the base of skull, but not larger ROI, was a significant predictor for the risk of being affected by inverted spectra (RR for base of skull: 11.76, 95% CI: 1.86-74.18, P=0.009. RR for area of ROI: 3.68, 95% CI: 0.57-23.67, P=0.170). Seven studies showed lipid contamination; all were in close proximity to the overlying scalp.

Using a semi-automated acquisition and post-processing method, MRSI can be successfully applied in the clinical setting. However, care should be taken to avoid regions of high magnetic field inhomogeneity at the base of skull, and lipid contamination in voxels prescribed near the scalp.