OA018] Improved absolute metabolite quantification by localized magnetic resonance spectroscopy simulations

Abstract

Purpose Magnetic resonance spectroscopy (MRS) provides a non-invasive method for assessment of the in vivo abundance of metabolites and serves as a useful diagnostic tool for several pathologies. The most common method for quantification of metabolite abundance includes fitting a linear combination of individual a priori acquired metabolite spectra. The set of individual metabolite spectra (basis set) can either be acquired through phantom measurements or simulations. While simulation has recently become more popular, it is not clear at what complexity level the simulations have to be run. The aim of this work was to study the impact of simulation complexity on absolute concentration estimates. Specifically, results based on non-localized and localized simulations were compared. Methods The Braino phantom (General Electric) containing the most common brain metabolites was used in this study. It included choline (Cho, 3.0 mM), creatine (Cr, 10.0 mM), lactate (Lac, 5.0 mM), myoinositole (mI, 7.5 mM), glutamic acid (Glu, 12.5 mM) and N-acetylactic acid (NAA, 12.5 mM). MR spectra were acquired on a Philips Achieva 3 T scanner with the PRESS pulse sequence. Parameters: echo time 35 ms, repetition time 5000 ms, number of acquisitions 128, volume of interest size 20 × 20 × 20 mm3. Basis sets were generated in MATLAB with the same sequence timing as on the scanner both with a non-localized simulation (using ideal RF pulses) and with a localized simulation (using the actual RF pulses, gradient waveforms and phase cycling scheme as on the scanner) using full density matrix calculations. The acquired spectra were analyzed with LCModel. Metabolite concentrations were separately estimated based on the two basis sets. Results The estimated concentrations based on localized simulations were closer to true ones for all metabolites: Cho 3.6/3.5 mM, Cr 10.8/10.5 mM, Lac 3.8/5.3 mM, mI 6.5/6.9 mM, Glu 12.7/12.6 mM, NAA 13.3/13.1 mM (ideal/localized). A substantial gain in accuracy was especially seen for lactate which was underestimated by 24% with the non-localized simulation, but overestimated by only 6% by the localized simulation. Conclusions Metabolite quantification was improved by inclusion of actual RF pulses, gradient waveforms and phase cycling scheme. This was especially seen for lactate.