Toward quantitative CEST imaging of glutamate in the mouse brain using a multi-pool exchange model calibrated by 1H-MRS

Abstract

Purpose: To develop a CEST quantification model to map glutamate concentration in the mouse brain at 11.7 T, overcoming the limitations of conventional glutamate-weighted CEST (gluCEST) contrast (magnetization transfer ratio with asymmetric analysis).

Methods: ¹H-MRS was used as a gold standard for glutamate quantification to calibrate a CEST-based quantitative pipeline. Joint localized measurements of Z-spectra at B₁ = 5 μT and quantitative ¹H-MRS were carried out in two voxels of interest in the mouse brain. A six-pool Bloch-McConnell model was found appropriate to fit experimental data. Glutamate exchange rate was estimated in both regions with this dedicated multi-pool fitting model and using glutamate concentration determined by ¹H-MRS.

Results: Glutamate exchange rate was estimated to be ˜1300 Hz in the mouse brain. Using this calibrated value, maps of glutamate concentration in the mouse brain were obtained by pixel-by-pixel fitting of Z-spectra at B₁ = 5 μT. A complementary study of simulations, however, showed that the quantitative model has high sensitivity to noise, and therefore, requires high-SNR acquisitions. Interestingly, fitted [Glu] seemed to be overestimated compared to ¹H-MRS measurements, although it was estimated with simulations that the model has no intrinsic fitting bias with our experimental level of noise. The hypothesis of an unknown proton-exchanging pool contributing to gluCEST signal is discussed.

Conclusion: High-resolution mapping of glutamate in the brain was made possible using the proposed calibrated quantification model of gluCEST data. Further studying of the in vivo molecular contributions to gluCEST signal could improve modeling.

Keywords: 1H‐MRS; Bloch‐McConnell fitting; CEST; glutamate; quantification; quantitative CEST.

FIGURE 1

Quantitative ¹H‐MRS acquisitions in the mouse brain. (A) Positioning of voxels of interest in striatum and corpus callosum. Typical fractions of gray matter, white matter, and CSF in the voxel are indicated. (B) Example of fit using LCModel toolbox of ¹H‐MRS spectra acquired on two different animals at TE = 20 ms. Macromolecules (MM) spectrum was experimentally acquired, while metabolite spectra were simulated. (C) Typical plots of normalized metabolite concentrations fitted by LCModel and area under water peak as a function of TE. Solid lines indicate mono‐exponential fits for T₂ estimation.

FIGURE 2

Mean Z‐spectra_localized measurements at B₁ = 5 μT in the striatum and in the corpus callosum. (A) Average Z‐spectrum_localized measured at B₁ = 5 μT in the striatum of five mice and the measurement of the magnetization transfer ratio with asymmetric analysis (MTR_asym) (dashed line). (B) Average raw Z‐spectrum_localized (dashed fine line) and CSF‐corrected Z‐spectrum_localized (solid line) at B₁ = 5 μT in the corpus callosum of seven mice. The gray line indicates the typical CSF Z‐spectrum used as reference for correction, and MTR_asym calculated on the corrected Z‐spectrum in bold dashed line. (C) Typical fit of Z‐spectra_localized acquired in striatum and corpus callosum of the same mouse using the optimized multi‐pool model.

FIGURE 3

Selection of a suitable multi‐pool model for glutamate‐weighted CEST (gluCEST) modeling. Modeling of Z‐spectra_localized data with several models including different proton‐exchanging pools, with [Glu] fixed to the ¹H‐MRS m easured value. Typical fits (solid line) of experimental data points (red crosses) acquired in the striatum of a mouse at B₁ = 5 μT using several models. Residuals (absolute values) are plotted under each fit.

FIGURE 4

Quantitative glutamate mapping in the mouse brain in several mice. (A) M₀ images acquired at −100 ppm and (B) measurement of the magnetization transfer ratio with asymmetric analysis (MTR_asym) (3 ppm) maps corrected with the water saturation shift referencing and normalized by M_Z(−3 ppm). (C) [Glu] maps obtained by pixel‐by‐pixel fit of Z‐spectra_imaging data acquired at B₁ = 5 μT.

FIGURE 5

Accuracy, precision and robustness of the fitting model. (A) Histogram of estimation errors of the fitting model on a set of 2000 simulations (no noise). (B–F) Robustness of the fitting model to deviations in exchange rate values varying “true values” of exchange rate k_ex ^Glu, k_ex ^APT, k_ex ^Guan, k_ex ^OH, k_ex ^NOE respectively, with additionally the mean absolute error made on estimations of [APT], [Guan], [OH] and [NOE], respectively.

FIGURE 6

Prediction of [Glu] by quantitative CEST model versus ¹H‐MRS measurements on localized LASER data. (A) Plot of the measurement of the magnetization transfer ratio with asymmetric analysis (MTR_asym) (3 ppm) versus the measured [Glu] with ¹H‐MRS for n = 5 mice in the striatum voxel of interest (VOI). Error bars indicated in the x‐axis direction correspond to the standard error of the mean estimated for individual ¹H‐MRS measurements. (B) [Glu] predicted by the quantitative glutamate‐weighted CEST (gluCEST) fitting model of Table S1 versus the measured [Glu] with ¹H‐MRS for n = 5 mice in the striatum VOI. On average, the predicted [Glu] was overestimated by 4.8 ± 1.6 mM. Error bars indicated in the y‐axis direction correspond to the 95% CI estimated for [Glu] predictions. (C) Plot of the measurement of the magnetization transfer ratio with asymmetric analysis (MTR_asym) (3 ppm) versus the measured [Glu] with ¹H‐MRS for n = 8 mice in the corpus callosum VOI. (D) [Glu] predicted by the quantitative gluCEST fitting model of Table S1 versus the measured [Glu] with ¹H‐MRS for n = 8 mice in the corpus callosum VOI. On average, the predicted [Glu] was overestimated by 8.8 ± 2.3 mM.