Effects of noise and linewidth on in vivo analysis of glutamate at 3 T

Abstract

Magnetic resonance spectroscopy (MRS) can noninvasively detect metabolites in vivo, including glutamate (Glu). However, quantification is known to be affected by the overlaps among metabolite resonance lines and background macromolecule signals. We found that adding a moderate amount of noise or line broadening (2 Hz) caused large variations in concentration of Glu and other metabolites, when determined by LCModel analysis of in vivo short-echo time (TE) spectra. Theses variations were largely attributed to strong spectral baselines in short TE spectra, especially near 2.35 ppm, as well as overlapping metabolite resonance lines. To address this issue, we acquired in vivo data at 3 T using both short-TE and the multiple echo time J-resolved point-resolved spectroscopy (JPRESS) MRS techniques. We found that one-dimensional (1D) JPRESS, by simultaneously fitting the two cross-sections of JPRESS at J = 0 and J = 7.5 Hz, was highly resistant to variations in noise levels and spectral linewidths. Our results demonstrate that LCModel analysis of short-TE data is highly sensitive to variations in noise levels and spectral linewidths and this sensitivity is greatly reduced by 1D JPRESS given its substantially reduced baselines and enhanced spectral resolution.

Keywords: Baseline of macromolecule background; Glutamate; JPRESS; Magnetic resonance spectroscopy; Short-echo time.

Fig. 1.

Typical placement of MRS voxel (2×2×2 cm³) and image segmentation (blue: gray matter, red: white matter, yellow: cerebrospinal fluid). The 2×2×2 cm³ voxel was prescribed for short-TE MRS.

Fig. 2.

Fitting of 1D JPRESS of a female participant (age = 21 years old; voxel size = 8 mL; scan time = 5.3 mins). (a): J = 0. (b): J = 7.5 Hz. The fitted spectra (red) were vertically shifted from the in vivo spectra (black) for better visualization. The fit residuals (grey) and the baseline (blue) are placed at the top and bottom, respectively. Individual components of fitted metabolite spectra of N-acetylaspartate (NAA), N-acetylaspartylglutamate (NAAG), creatine (Cr), choline (tCho), glutamate (Glu), myo-inositol (mI), glutamine (Gln), and glutathione (GSH) are displayed in (c) and (d) for both cross-sections at J = 0 and J = 7.5 Hz. The cross-sections at J = 7.5 Hz, (b) and (d), were scaled up by 2.5 folds.

Fig. 3.

LCModel fitting of a short TE spectrum, with TE = 35 ms, voxel size = 8 mL, average number = 128. (a): original data; (b): in vivo data with the noise level increased by 55%. An increase in estimated glutamate (Glu) concentration from the noise-increased data is accompanied by a decrease in the estimated baseline intensity near 2.35 ppm. The arrows show the change in baseline intensity near 2.35 ppm.

Fig. 4.

Fit comparison between the original (Fig. 3a and 3b) and the noise-increased data (3c and 3d) for 1D JPRESS. Cross-sections at J =0 and J =7.5 Hz were placed on the left and right, respectively. The JPRESS data were collected from the same voxel and participant as for short TE data (Fig. 3). The noise level in 3c) and 3d) was 55% higher than that of the original data (3a and 3b). The estimated metabolite concentrations for the noised-increased data agreed well with the results of the original data, and the estimated baselines (blue lines) showed no significant differences between the original data and the noise-increased data.

Fig. 5.

Estimated concentrations of N-acetylaspartate (NAA), glutamine (Glu), and creatine (Cr) as a function of SNR. (a): a short-TE dataset (35 ms) with LCModel quantification. (b): 1D JPRESS data acquired from the same voxel, measuring 2.5×2.5×2.5 cm³. The SNR of the original data was decreased progressively by noise injection in the time domain. With the short TE data, Glu concentrations were negatively correlated with SNR. This trend was not observed for 1D JPRESS.

Fig. 6.

Fit of a short echo time (TE) (35 ms) spectrum with LCModel before (a) and after (b) 2 Hz linewidth broadening. The estimated concentrations of glutamate (Glu) and N-acetylaspartate (NAA) were significantly reduced for the line-broadened data, and the reduction was strongly correlated with the elevation of baseline around 2.35 ppm, as shown by the arrow in (b).

Fig. 7.

Comparison of 1D JPRESS fit of the original data (a, b) and the data created by 2 Hz line-broadening (c, d). The cross-sections at J = 0 and J = 7.5 Hz are displayed on the left (a, c) and right (b, d), respectively. No significant differences in quantification results (Table 1) were observed between the original and line-broadened spectra.