Simultaneous Determination of Metabolite Concentrations, Longitudinal and Transverse Relaxation Times
Li An1, Shizhe Li1, and Jun Shen1

1National Institute of Mental Health, National Institutes of Health, Bethesda, MD, United States


A data acquisition and quantification approach is proposed for simultaneous measurement of concentration, T1, and T2 of metabolites. The concentration, T1, and T2 of NAA, creatine, choline, glutamate and myo-inositol were determined reliably in less than 9 min using single voxel MRS at 7 Tesla.


Knowledge of metabolite T1 and T2 relaxation times can be very valuable in the investigation of intracellular micro-environment changes caused by diseases (1). In this work, we propose to simultaneously measure metabolite concentrations and relaxation times using a point resolved spectroscopy (PRESS) sequence with multiple echo times (TEs) and inversion-recovery times (TIs). The whole set of data acquired with different TIs and TEs are modeled by linear combinations of a single set of basis functions to compute metabolite concentrations and relaxation times (T1 and T2).


Two broadband hyperbolic secant pulses were added to a single-voxel PRESS sequence with a variable TE. The first hyperbolic secant pulse had a fixed TI of 2500 ms and a flip angle of 90°, while the second one has a variable flip angle and TI. The following nine TE values were used: 58, 68, 78, 88, 98, 108, 178, 198, and 218 ms. At TE of 68 ms, the second hyperbolic secant pulse was varied using six different settings: (1) no inversion; (2) TI = 1500 ms, flip angle = 160.2°; (3) TI = 900 ms, flip angle = 160.2°; (4) TI = 471 ms, flip angle = 160.2°; (5) TI = 244 ms, flip angle = 142.6°; and (6) TI = 81 ms, flip angle = 160.2°. The position and flip angle of the second hyperbolic secant pulse were chosen to provide sufficiently different T1 weighting and to optimize water suppression at the same time. For TEs other than 68 ms, the second inversion pulse was turned off. The first 90° hyperbolic secant pulse was used to ensure that the longitudinal magnetization was zero at 2500 ms prior to the excitation pulse of the PRESS sequence for all repetitions.

Two healthy volunteers were recruited for the study. MRS data were collected using the proposed pulse sequence from the medial prefrontal cortex of the two volunteers. A repetition time of 3 s and 12 signal averages were used for each (TE, TI) setting. The total scan time was 8.9 minutes. Basis functions of ten brain metabolites (see Table 1) at the nine selected TE values were numerically computed using an in-house developed program based on the GAMMA C++ library. The acquired time domain data were Fourier transformed into the frequency domain to obtain spectra for all 14 different (TE, TI) settings (see Fig. 1). These 14 spectra were put into one large vector and were fitted together by linear combinations of the basis functions using an in-house developed fitting program which models the baseline using a spline function. In the fitting process, mono-exponential functions were used to model T1 and T2 relaxations.


The reconstructed spectra and fitted spectra for all 14 (TE, TI) settings are displayed in Fig. 1. Metabolite concentrations and the corresponding Cramer-Rao lower bounds (CRLB) are given in Table 1. Metabolite T1 and T2 values are given in Table 2. Unreliable metabolite T1 and T2 values are not listed. The spectrum corresponding to TE of 68 ms without the second hyperbolic secant pulse was also quantified for comparison. The corresponding metabolite concentrations and normalized CRLB values are given in Table 3. As shown by the CRLB values in Table 1 and Table 3, the proposed method resulted in much lower or comparable CRLBs for most major metabolites (NAA, tCr, tCho, Glu, Gln, GSH, GABA, NAAG, and mI). Most importantly, the proposed method allows simultaneous determination of T1 and T2 of the major metabolites which is not possible for a single TE data acquisition.

Discussion and Conclusions

A new approach for acquiring and quantifying single voxel MRS data was proposed to allow simultaneous measurement of metabolite concentrations and relaxation times. By changing the sequence parameters, different spectral information is generated with different metabolite-metabolite and metabolite-baseline interactions. Therefore, the overall information content is increased over a single sequence with fixed TE and TI. As a result, it is possible to improve the reliability of measurement for metabolites of interest without increasing the total scan time. In the meantime, metabolite T1 and T2 are simultaneously determined. To our best knowledge, this is the first time that the concentration, T1 and T2 of metabolites are simultaneously measured in vivo. Demonstration of this unique feasibility opens the possibilities of further improving this technique and measuring all three parameters in patient studies.


This research was supported by the Intramural Research Program of the National Institute of Mental Health.


1. Bracken et al., PsychiatryResearch:Neuroimaging 213 (2013):142–153.


Fig 1. Spectra collected from the medial prefrontal cortex of a healthy volunteer (voxel size: 8 cm3). TE and TI values are listed on the left, where the TI values are the TIs for the second hyperbolic secant pulse.

Table 1. Metabolite ratios in the medial prefrontal cortex of two healthy volunteers obtained by modeling the multi-TE and multi-TI data. Number of signal averages = 12 × 14 = 168.

Table 2. Metabolite T1 and T2 relaxation times in the medial prefrontal cortex of two healthy volunteers.

Table 3. Metabolite ratios computed using only the data acquired with TE of 68 ms and without inversion (number of signal averages = 12). The listed CRLB values have been normalized by a scaling factor of 1/sqrt(14) to make them comparable to the CRLB values listed in Table 1.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)