Bracoud, Luc

AbstractBackgroundMagnetic Resonance Spectroscopy (MRS) is a non‐invasive technique assessing the concentration of various metabolites within the brain, both cross‐sectionally to characterize disease type and severity, and longitudinally to assess disease progression and potential treatment efficacy in the context of disease‐modifying interventions. In such context, the involvement of multiple imaging facilities inevitably introduces variability, whose amount must be known to inform on comparability between sites and sensitivity to change. Our objective is to study variability in the concentration of various brain metabolites using a dedicated MRS phantom and assess impact of field strength.Methods30 sites are being considered to participate in an on‐going clinical trial in metachromatic leukodystrophy. Site qualification mandates the scanning of a dedicated MRS phantom. 30 of these HD‐polyethylene spheres were simultaneously manufactured and filled with the same mixture of chemicals mimicking main metabolite concentrations. To date, 14 sites provided data (5 1.5T scanners and 9 3T). All sites implemented a standardized single‐voxel MRS sequence. MRS voxel was placed within the phantom and acquisition repeated 3 times. Spectra were centrally analyzed using LCModel v6.3 and quality controlled by Bioclinica. Repeat acquisitions were requested for parameter deviations, voxel positioning and/or shimming issues. Results only include passing datasets. Coefficient of Variation (CV) was calculated for NAA/Cr, Cho/Cr, mI/Cr, Glx/Cr and Lac/Cr ratios for each site and overall. LCModel quality metrics (FHWM and S/N) were also extracted to assess spectra quality. Comparison between field strengths was carried out using t‐tests.ResultsAverage CV by site was 2.0% for NAA/Cr (2.5% at 1.5T and 1.8% at 3T), 2.4% for Cho/Cr, 3.0% for mI/Cr, 4.5% for Glx/Cr and 6.1% for Lac/Cr. CVs were generally lower at 3T (except for Lac/Cr) although difference was not significant. As expected, Signal‐to‐Noise was higher at 3T (p=0.05). Overall CV by metabolite across scanners was much higher (ranging from 4.6% for Cho/Cr up to 12.1% for Lac/Cr).ConclusionMRS measures of metabolite concentrations have low variability at scanner‐level, insensitive to field strength. Nevertheless, variability across sites is much higher and would require, if possible, cross‐calibration. These phantom data make such calibration possible and will improve sensitivity of subject data analyses.