Quantification of brain tumor metabolites in magnetic resonance spectroscopy using singular value decomposition

Abstract

This study aims to get the quantification results from Magnetic Resonance Spectroscopy (MRS) spectra of metabolites from subjects with brain tumors. MRS data were acquired using Magnetic Resonance Imaging 1.5 Tesla with spin-echo sequence from single-voxel and multiple voxels of 9 regions-of-interest (ROIs). The spectra were processed using the Hankel Lanczos Singular Value Decomposition (HLSVD) quantification method. The pre-processing steps were conducted on metabolite spectra in the range of 0-5 ppm using lipid signal positioned at 1.4 ppm as a reference. The HLSVD quantification of 50 components was conducted to obtain the processed spectra and the concentration of metabolites, such as myo-inositol (mI), choline (Cho), creatine (Cr), N-Acetylaspartic Acid (NAA), and lipid. The metabolite concentrations from the voxels of interest obtained using this method were compared to concentrations obtained in other studies. The metabolite concentrations obtained in this study showed changes in the concentration of metabolites in ROIs with brain tumors compared to healthy tissues. The corresponding ROIs with brain tumors showed the dominant Cho concentrations. The metabolite ratios were also calculated, namely NAA/Cr, NAA/Cho, and Cho/Cr. The MRS quantification using HLSVD method in this study was comparable to other studies and could be used to characterize metabolites in brain pathologies.