Analysis of in vivo 1H MR spectra of normal brain tissue by means of second derivative method

Abstract

Background: As follows from comparative studies conducted in a number of centres, meeting the repeatability and reliability conditions in MRS in vivo measurements depends mainly on system stability and operator-related external factors. The aim of the study was to check the usefulness of the automated fitting method offered by PeakFit 4.0 (by SPSS Inc., Chicago, USA).

Material and methods: Short echo time in vivo PRESS 1H MR spectra (2T, TE = 35 ms) of normal brain were fitted in the frequency domain using the second derivative method. In this approach local maxima and hidden peaks are found as local minima of spectrum second derivative. The Lorentzian robust minimization procedure using Levenburg-Marquardt non-linear fitting engine was applied. Spectral lines were approximated under the assumption of the mixed Lorentzian/Gaussian lineshapes. The same procedure was applied to 18 proton spectra recorded from frontal lobe white matter.

Results: The number of peaks found within the range of 0.74 + 4.2 ppm was 52 +/- 3 and their positions were almost the same. The fitted lines were assigned on the basis of the J-pattern recalculated for the field strength of 2T and by comparing the chemical shifts with the shifts in single compound spectra. The ratios of main metabolites, such as NAA/Cr, Cho/Cr, Cho/NAA and mI/Cr, are in accordance with those obtained previously using the software supplied with the MR imager and the absolute concentrations of NAA, Cho, mI, Glc and Glu obtained from the fit agree with those reported in literature which confirms the usefulness of the second derivative method in routine analyses of 1H MR brain spectra.

Conclusions: The automated fitting enables reasonable metabolite ratios and absolute concentrations to be obtained, however it should be kept in mind that the second derivative follows only the shape of the curve that represents the sum of all metabolites.