Reproducibility of H MRS measurements as a function of SNR

I. Hancu, S. Adak, E. Hutchison, R. Hurd GE GRC, Niskayuna, NY, United States, GE GRC, Bangalore, India, Union College, Schenectady, NY, United States, GE ASL, Menlo Park, CA, United States Introduction Longitudinal MRS studies to assess the progress of a disease [1] or the efficacy of a drug [2] are increasingly common. For these studies it is of utmost importance that the MRS measurements are extremely reproducible, in order for any small deviation from the baseline to be associated with the effect of the disease or drug. Literature documents factors contributing to the measurement variability, such as biological variation or voxel repositioning [3]. Since any quantitative MRS measurement involves curve fitting for the determination of metabolite concentration, it has usually been assumed that the (limited) SNR of a spectrum is also contributing to the overall variability [4]. We present here the results of a study designed to help understand the impact of the SNR on MRS data reproducibility.