Improvement of Aspartate-Signal Fitting Accuracy in Asp-Edited MEGA-PRESS Spectra

Abstract

Changes in the aspartate (Asp) concentration are potential biomarkers of malate–aspartate shuttle dysfunction as well as the disturbances in the synthesis of NAA. J-difference editing can be utilized to detect a resolved Asp signal at δ_Asp ≈ 2.72 ppm. However, as this resonance has a complex shape, fitting of the Asp signal is challenging. In our study, we compare the performance of four Asp fitting models applied to data from three brain regions: the anterior cingulate gyrus (ACC), frontal white matter (DLPFA), and early visual cortex (VC). As a result, fitting with the phantom and simulated spectra can be recommended for robust Asp quantification spectra. This conclusion based on flexibility and suitability of these approaches for Asp fitting for all range of SNR. One gauss model shows the worst performance and could not be recommended for total Asp fitting. The absolute concentrations of Asp are quantified using a modified phantom calibration method. As a result, the estimated Asp concentrations are as close as possible to the biochemical estimations. The quantified Asp and tCr concentrations are, respectively, 1.69 ± 0.18 and 7.43 ± 0.71 for the ACC, 0.93 ± 0.14 and 5.86 ± 0.41 for the DLPFA, and 1.33 ± 0.35 and 8.4 ± 0.64 for the VC. Regional Asp variations in the brain are discovered; the Asp gray matter (GM) concentrations are higher than that in the white matter, and the Asp concentrations in the VC are significantly lower than that in the ACC, although the GM ratio is equal in all these regions.

Appendices

Appendix 1

Quantification of Asp Concentration

The cerebral Asp concentrations were calculated for the best fitting approach (simulated model spectral line) using a modified phantom calibration method [24]:

$$\left[ {{\text{Asp}}} \right] = \frac{{S_{{{\text{Asp}}}} }}{{S_{{{\text{Cr}}}} }} \times \frac{1}{a} \times \left[ {{\text{Cr}}} \right],$$

(3)

where $S_{{{\text{Asp}}}}$ and $S_{{{\text{Cr}}}}$ are the integral intensities of cerebral Asp and Cr, fitted by simulated model and Lorenz line, respectively, $\left[ {{\text{Cr}}} \right]$ is the absolute Cr concentration calculated using formula (3), ${\text{and}} a$ is the linear regression coefficient of the calibration curve that forces the line to go through the (0,0) point (formula 2).

$$S_{{{\text{Asp }}\left( {{\text{phantom}}} \right)}} /S_{{{\text{Cr }}\left( {{\text{phantom}}} \right)}} \times \left[ {{\text{Cr}}} \right]_{{{\text{phantom}}}} = a \times \left[ {{\text{Asp}}} \right]_{{{\text{phantom}}}} .$$

(4)

Linear regression was performed based on Phantom study data. The y-axis represents the product of the known Cr concentration $\left[ {{\text{Cr}}} \right]_{{\text{phantom }}}$ and the Asp-to-Cr integral intensity ratio $S_{{{\text{Asp }}\left( {{\text{phantom}}} \right)}} /S_{{{\text{Cr }}\left( {{\text{phantom}}} \right)}}$ signal intensity, and the x-axis represents the known concentration of Asp in the Phantom $\left[ {{\text{Asp}}} \right]_{{{\text{phantom}}}}$, as shown in (Fig. 4). The Cr concentration is constant at 18 mmol, whereas the Asp concentrations are 2, 3, 4, 5, 6, 8, 10, 12, 16, and 20 mM, forming a set of 11 experimental points. The spectra were collected using a 32-channel SENSE head quadrature coil. The edited Asp signal with δ ≈ 2.55–2.85 ppm in the phantom spectra was fitted with the simulated model. Substitution of the value of a value in Eq. (1) yields the final expression for Asp quantification:

$$\left[ {{\text{Asp}}} \right]_{{\text{in vivo}}} = \frac{{S_{{{\text{Asp}}\_{\text{in vivo}}}} }}{{S_{{{\text{Cr}}\_{\text{in vivo}}}} }} \times 3.82 \times \left[ {{\text{Cr}}} \right]_{{\text{in vivo}}} .$$

(5)

Fig. 4

Ratio of aspartate (Asp) over the total creatine (tCr) multiplied by the known tCr concentration versus the known Asp phantom concentration. The coefficient of determination R² = 0.987, p < 0.05