Robust quantification of short echo time ¹H magnetic resonance spectra using the Padé approximant

Abstract

Accurate quantification of in vivo short echo time spectra is hampered by the presence of overlapping peaks and a significant baseline. In this work the Padé approximant in conjunction with Monte Carlo simulation is used to extract peak areas from short echo time 1H spectra. We exploit the fact that the Padé approximant is known to model broad non-Lorentzian signals as arbitrary sums of Lorentzian components to separate baseline components from sharper metabolite signals by combining the Padé approximant with Monte Carlo simulation. The simulation results demonstrate that the Padé approximant-Monte Carlo hybrid analysis is able to separate the metabolite signals from the baseline, while a least squares fitting of a time domain model may result in significant bias of the peak area estimations. For the in vivo data the estimates of the peak areas using the Padé approximant and AMARES compare well, with the exception of the NAA peak at 2.02 ppm. We suggest that the discrepancy may be due to the baseline contamination as supported by the simulation results; however, without an in vivo gold standard this remains difficult to demonstrate.