Introduction

Development of low-field MRIs (with B₀<0.1T) for point-of-care (POC) applications has become increasingly widespread¹. Magnetic resonance spectroscopy (MRS) at low field poses significant challenges. Besides the low signal-to-noise ratio (SNR) and the low concentration of biologically significant molecules compared to water, the entire spectral region containing the proton resonances spans only few Hz. When B₀ inhomogeneity exceeds this spectral range, spectral separability is essentially impossible. One way to address this relies on the evolution of the spin-spin coupling among coupled protons in a Carr-Purcell-Meiboom-Gill (CPMG) echo train². The possibility of obtaining J-spectra of lactate, whose concentration is significantly elevated in several pathological conditions, including stroke, was recently demonstrated³. Here we show our initial results of 2D J-spectroscopic imaging (J-MRSI) sequence in a phantom. We also demonstrate the benefit of using composite radiofrequency (RF) refocusing pulses to reduce potential B₁ imperfections.

Materials and Methods

Hardware: We used a 46mT Halbach array-based magnet (outer/inner diameter=60/30.1cm, length=49.2cm, weight=111kg), with custom-built RF and Bruker gradient amplifiers and a Magritek Kea2 spectrometer⁴. Imaging was performed using a custom-build solenoid RF coil^3,4.
Phantom: Solutions of lithium lactate (Sigma Aldrich, St. Louis MO, USA) with concentration ranging from 250 mM to 1 M, water mixed with D₂O and water mix with CuSO₄ were prepared.
Sequence: The J-MRSI sequence is shown in Fig.1. The original CPMG sequence provided by the spectrometer manufacturer was modified to allow setting of the first echo time (2τ1) to be independent of subsequent echo times (2τ). This was done to minimize the phase accrual between the excitation pulse and the first echo. The non-localized CPMG sequence was then further modified to include 2 phase encoding (PE) gradients after each 180º pulses (2D CSI-CPMG). The PE amplitude remained constant across the CPMG train and incremented for the following encoding. Acquisition parameters were: bandwidth: 5kHz, number of complex data points: 64, inter-pulse delays (2τ): 62.5ms and 83.3ms, 128 or 256 echoes, field-of-view=160x160mm, and matrix size=12x12 or 16x16.
RF pulses: We evaluated the use of composite refocusing pulses to boost signal in conditions of inhomogeneous B₁ and B₀⁵. We eventually chose the TPG composite pulse suggested by Tycko and Pines⁶ (Fig.1B), which provided the highest SNR.
Analysis: Data were processed with in-house Matlab and Python routines. The integral over each echo yielded a single complex value and the series subsequently underwent Fourier transformation to yield the J-spectrum. Peaks were fitted using linear prediction and singular value decomposition (LPSVD). Images were created by integrating signal from side peaks.

Results and Discussion

Spatially-localized signal pattern: Fig.2 shows data acquired with the J-MRSI sequence with two phantoms, one with 1M lactate and one with 8% H2O in D2O. Fig.2A shows the spatially-localized signal pattern from both phantoms. Fig.2B shows spectral features of lactate across two inter-pulse delays (62.5ms and 83.3ms), similar to those previously published for the non-localized version of J-MRS³.
Improvements with composite refocusing pulses: The non-localized CPMG sequence using water mixed with CuSO₄ phantom was used to evaluate the sensitivity of hard and composite pulses to changes in B₁. Fig.3A illustrates that both the amplitude and T2 values obtained using TPG composite pulses are relatively insensitive to B₁.
J-Spectroscopic imaging of lactate at 4 different concentrations: Panels B to D in Fig.3 illustrate the lactate results obtained using both hard or composite pulses. Differences in signal intensity across four different lactate concentrations are clearly visible. The improvement in signal thanks to the TPG composite pulses is also shown. Averaging signal in voxels (n=9) within each phantom resulted in the following values: 53±24/132±68 (1M phantom), 28±10/77±30 (0.75M phantom), 14±9/42±27 (0.5M phantom) and 9±5/24±15 (0.25M phantom) with hard/composite pulses, respectively. As shown in panel B, we obtained a non-linear relationship between lactate concentration and MR signal (independently of the choice of RF pulses). This is different to what was previously reported using a non-localized CPMG sequence⁷. We hypothesize that this could be due to (1) the coarse quantification method and (2) the need for B₀-insensitive RF pulses.

Conclusion

Here, we showed the first steps towards lactate imaging at low-field (B₀<0.1T). We have demonstrated the feasibility of obtaining unique spectral information of lactate with spatial encoding at low-field. The ultimate goal is to apply this methodology first in phantom with lactate concentrations close to physiological levels, and then apply the sequence in vivo. Quantifying metabolites in vivo should be properly assessed in realistic conditions that reflect the complexity of acquiring data in the human brain and other anatomies and challenges such as suppression of the non-J coupled signal from water and other tissue constituents should be implemented.

Acknowledgements

This project has received funding from Horizon 2020 ERC Advanced PASMAR 101021218 and the Dutch Science Foundation Open Technology 18981.

References

[1] Saracanie et al., Front Phys 2020; [2] Meiboom et al., Rev Sci Instrum 1958; [3] Ronen et al., J Magn Reson 2020; [4] O’Reilly et al., MRM 2020 ; [5] Odedra, Wimperis, J Magn Reson 2012; [6] Tycko et al., J Chem Phys 1985; [7] Webb et al., Applied Magn Res 2023