SU-E-I-64: 1H-MRS Quality Assurance in Short and Long Term Using a Phantom Object

To evaluate intrinsic variability of 1H-MRS single voxel technique for short and long periods of time using a metabolic phantom in a clinical protocol (hippocampi).

A solution with several metabolites (NAA, Cre, Cho, Lac, M-Ins, Glu, Gln) was placed in a spherical round glass with 300 mL. MRS data were acquired in a 3 T MRI scanner, with a PRESS sequence of TE/TR=35/1500 ms, 2048 data points, 1725 Hz of bandwidth, 6 cm³ VOI centered in the phantom and NSA of 128. A total of 130 spectra were obtained from 27 acquisition dates (5 spectra/date without removing the phantom) over 15 months. Spectra were processed with LCModel software to reduce the human variability during spectra processing. SNR and FWHM outputs from LCModel are mean calculated values from identified resonances.

Spectra from each acquisition date showed no mean deviations superior to 4% for all metabolites except for Lac (14 %). During 15 months all metabolites showed inferior deviations to 12%, once again except for Lac (20.5%). The worst estimated concentration was from Gln, which represent 60% of the expected value. The mean SNR was 16.3±2.1. There was no correlation between SNR and mean FWHM from resonances, thus wrong water suppression must be the main factor for variable SNR.

NAA was the most stable metabolite in the whole analysis period with a variation of 6.8%, and Gln was the less stable with a 25.4% variation. FWHM was more variable than SNR, possible due to shimming variability, although it showed no influence in SNR. It was observed two types of artifacts in some spectra generated by unbalanced gradients, B0 inhomogeneity and also insufficient amplitude from crusher gradients. Temporal analysis demonstrated the feasibility of compare results obtained on measurements of the same date and for long periods due to low deviations provided by the technique.