1853

The Goldilocks zone for 3T MRS studies using semi-LASER: Determining the optimal balance between repetition time and scan time using FSL-MRS

Alex Ensworth^1,2, Laura Barlow^3,4, Piotr Kozlowski^1,2,3,4, Erin MacMillan^3,4,5, and Cornelia Laule^1,2,3,6
¹Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada, ²International Collaboration on Repair Discoveries (ICORD), University of British Columbia, Vancouver, BC, Canada, ³Radiology, University of British Columbia, Vancouver, BC, Canada, ⁴UBC MRI Research Centre, University of British Columbia, Vancouver, BC, Canada, ⁵Philips Canada, Mississauga, ON, Canada, ⁶Pathology and Laboratory Medicine, University of British Columbia, Vancouver, BC, Canada

Synopsis

Keywords: Spectroscopy, Spectroscopy, MRS, T1 weighting, FSL-MRS, repetition time, TR, semi-LASER, 3T, metabolite concentration, brain, grey matter

Motivation: A short TR is often used in clinical MRS brain studies, leading to T₁-weighting of metabolites and inaccurate T₁ correction.

Goal(s): To determine the optimal balance between scan time and TR that minimizes T₁-weighting effects when using semi-LASER MRS.

Approach: SNR and metabolite concentrations were compared for TRs of 2, 5 and 8s using FSL-MRS.

Results: A TR of 5s provides a good balance of scan time, SNR and signal recovery. For a similar scan time, TR of 2s leads to incomplete signal recovery and thus heavy T₁ weighting, while a TR of 8s results in complete signal recovery but reduced SNR.

Impact: For MRS studies using semi-LASER, our work informs scientists and clinicians on the issues of using a short TR, and recommends the optimal scan parameters to use while implementing the newly available FSL-MRS analysis package.

Introduction

To reduce scan time, ¹H-MRS studies often use a short repetition time (TR)^1-4 to increase the signal to noise ratio (SNR) with more acquisitions. However, this leads to significant T₁-weighting effects on metabolite concentrations at 3T, where T₁ relaxation times are longer than at 1.5T^5,6. Some methods have been proposed on how to correct for these T₁ effects, but they rely on literature-reported metabolite T₁s which typically have large uncertainty^6-8. Additionally, T₁ times are only reported for metabolites with high SNR, are largely unavailable for metabolites with low SNR, and do not span all brain regions⁹, ages, pathologies and MR field strengths. Therefore, accurate T₁-weighting correction of all metabolites is nearly impossible.
Our goal was to determine the optimal balance between scan time and TR that minimizes T₁-weighting effects when using semi-LASER¹⁰ MRS and the newly developed MRS analysis software by FSL (FSL-MRS¹¹).

Methods

Data collection: Data was collected in 5 healthy volunteers (2M/3F, mean age: 25±2 years) at 3T (Philips Ingenia Elition X, 32-channel head coil). The MRI protocol included a 3D-T₁w scan and ¹H-MRS (semi-LASER, 7.8cm³ voxel, posterior cingulate cortex, Fig.1). The TR and number of acquisitions were varied (Fig.2). Individual transients were exported to enable post-processing comparisons of similar scan times (in multiples of 32 phase cycle steps¹²), and same number of acquisitions.
Data analysis: 3DT₁’s were segmented using FSL FAST¹³. The semi-LASER basis set was simulated using FID-A¹⁴. Processing, fitting and quantification of data was completed using FSL-MRS (v2.1.12)¹¹, involving frequency alignment, eddy current correction, phase correction and water removal via HLSVD. Metabolite T₁ and T₂ parameters were changed to 0.001 and 100.0, respectively, to remove any relaxation corrections. Fig.1 shows an example spectrum with fit. SNR and metabolite concentrations from FSL-MRS were compared for the cases of similar scan times and same number of acquisitions.

Results

Fig.3 demonstrates the effect of TR on SNR for the brain metabolite n-acetyl-aspartate (NAA), where all SNR values were normalized to TR=8s (~99% relaxed). Average SNR is reported along with each individuals’ SNR. For similar scan times, SNR at TR=2s was 29% larger than SNR at TR=8s, similar to SNR at TR=5s which was 34% larger than at TR=8s. For constant number of acquisitions, SNR at TR=2s was 29% less than SNR at TR=8s, while SNR at TR=5s was only 2% lower than at TR=8s.
Fig.4 evaluates the SNR per unit scan time (shown in grey). Using TR=8s as a reference, the expected SNR per scan time was calculated for TR=2s and 5s, shown in red. At TR=5s, the measured SNR/minute was larger than expected.
Fig.5 demonstrates the trends of average metabolite concentrations for five metabolites (NAA, total creatine (tCr), total choline (tCho), myoinositol (mI) and glutamate (Glu)) normalized to TR=8s. Metabolite concentrations were 15-30% lower at TR=2s than at TR=8s, while they were only 10-15% lower at TR=5s.

Discussion

¹H-MRS measured with a clinically feasible scan time and a TR=5s demonstrated a similar SNR as spectra measured with TR=2s, and achieved the optimal SNR per unit time as compared to TR=2s or 8s. While there is more SNR/minute for TR=2s, it underperforms compared to TR=5s due to T₁-weighting effects. In addition, spectra measured with TR=2s need to be corrected for T₁ relaxation, however the metabolite T₁ times are not known for every metabolite, brain region, age and condition. Based on our results, TR=5s appears to improve the balance between T₁ relaxation and scan time compared to TR=2s at 3T.
We used TR=8s as a reference for full relaxation. However, TR=8s underperforms in SNR because of too few acquisitions in 5min, and also a reduced gain in SNR/time as magnetization recovery approaches thermal equilibrium. Participant motion during long scan times can also reduce SNR. Comparing the SNR for the same number of acquisitions demonstrates that most of the signal has recovered after TR=5s (95%) compared to TR=8s (theoretically 99% recovery for a T₁=1.5s).
The change in metabolite concentrations across TRs demonstrate the effects of T₁ weighting on metabolites when using short TRs, resulting in an underestimation of the true concentration. The amount of T₁-weighting varies greatly depending on the T₁ time of each metabolite. Using a longer TR avoids these issues and leads to greater reproducibility between ¹H-MRS studies.

Conclusion

Using semi-LASER MRS brain acquisition at 3T and FSL-MRS analysis, a TR=2s does not provide the optimal balance between metabolite relaxation and scan time, despite popular belief. Instead, with a longer TR=5s, one can achieve a more quantitative measurement of metabolite concentration without T₁-weighting or the need for corrections based on generic assumptions.

Acknowledgements

AE extends appreciation to MS Canada for a Doctoral Studentship Award. CL and PK gratefully acknowledge funding from the NSERC Discovery grants program. ELM received salary support from Philips Canada. This work was conducted on the traditional, ancestral, and unceded territories of Coast Salish Peoples, including the territories of the xwməθkwəy̓əm (Musqueam), Skwxwú7mesh (Squamish), Stó:lō and Səl̓ílwətaʔ/Selilwitulh (Tsleil- Waututh) Nations.

References

1. Landheer K, Gajdošík M, Juchem C. A semi-LASER, single-voxel spectroscopic sequence with a minimal echo time of 20.1 ms in the human brain at 3 T. NMR Biomed. 2020;33(9):1-12. doi:10.1002/nbm.4324

2. Camicioli RM, Korzan JR, Foster SL, et al. Posterior cingulate metabolic changes occur in Parkinson’s disease patients without dementia. Neurosci Lett. 2004;354(3):177-180. doi:10.1016/j.neulet.2003.09.076

3. Kirov II, Liu S, Tal A, et al. Proton MR spectroscopy of lesion evolution in multiple sclerosis: Steady-state metabolism and its relationship to conventional imaging. Hum Brain Mapp. 2017;38(8):4047-4063. doi:10.1002/hbm.23647

4. G. Hale Hatay, E. Ozturk-Isik, Optimized Multi-Voxel TE-Averaged PRESS for Glutamate Detection in the Human Brain at 3T, Journal of Magnetic Resonance (2023), doi: https://doi.org/10.1016/j.jmr.2023.107574

5. Li Y. T1 and T2 Metabolite Relaxation Times in Normal Brain at 3T and 7T. J Mol Imaging Dyn. 2013;02(02):1-5. doi:10.4172/2155-9937.s1-002

6. Ethofer T, Mader I, Seeger U, et al. Comparison of Longitudinal Metabolite Relaxation Times in Different Regions of the Human Brain at 1.5 and 3 Tesla. Magn Reson Med. 2003;50(6):1296-1301. doi:10.1002/mrm.10640

7. Träber F, Block W, Lamerichs R, Gieseke J, Schild HH. 1H Metabolite Relaxation Times at 3.0 Tesla: Measurements of T1 and T2 Values in Normal Brain and Determination of Regional Differences in Transverse Relaxation. J Magn Reson Imaging. 2004;19(5):537-545. doi:10.1002/jmri.20053

8. Knight-Scott J, Brennan P, Palasis S, Zhong X. Effect of repetition time on metabolite quantification in the human brain in 1H MR spectroscopy at 3 tesla. J Magn Reson Imaging. 2017;45(3):710-721. doi:10.1002/jmri.25403

9. Sporn L, MacMillan EL, Ge R, Greenway K, Vila-Rodriguez F, Laule C. Longer Repetition Time Proton MR Spectroscopy Shows Increasing Hippocampal and Parahippocampal Metabolite Concentrations with Aging. J Neuroimaging. 2019;29(5):592-597. doi:10.1111/jon.12648

10. Öz G, Deelchand DK, Wijnen JP, et al. Advanced single voxel 1H magnetic resonance spectroscopy techniques in humans: Experts’ consensus recommendations. NMR Biomed. 2021;34(5):1-18. doi:10.1002/nbm.4236

11. Clarke WT, Stagg CJ, Jbabdi S. FSL-MRS: An end-to-end spectroscopy analysis package. Magnetic Resonance in Medicine 2021;85:2950–2964 doi: 10.1002/mrm.28630.

12. Deelchand DK, Berrington A, Noeske R, et al. Across-vendor standardization of semi-LASER for single-voxel MRS at 3T. NMR Biomed. 2021;34(5):1-11. doi:10.1002/nbm.4218

13. Zhang, Y; Brady, M; Smith S. Segmentation of brain MR images through a hidden Markov random field model and the Expectation-Maximization Algorithm. IEEE Trans Med Imaging. 2001;20(1):45-57.

14. Simpson R, Devenyi GA, Jezzard P, Hennessy TJ, Near J. Advanced processing and simulation of MRS data using the FID appliance (FID-A)—An open source, MATLAB-based toolkit. Magn Reson Med. 2017;77(1):23-33. doi:10.1002/mrm.26091

Figures

A sample MRS spectrum at TR=8s. Localization via semi-LASER in the posterior cingulate cortex, done with a 32-step phase cycle¹², water suppression, a TE=32ms and 2048 samples at a frequency of 2000 Hz. Data was processed and analyzed using FSL-MRS¹¹.

A summary of the MRS acquisition parameters used. Each volunteer was scanned according to the first three rows with the order of TRs randomized from volunteer to volunteer. Each transient was saved separately so the data could be separated and organized into categories of similar scan time (given multiples of 32 acquisitions) or the same number of acquisitions.

The SNR for NAA was normalized to a TR=8s for each volunteer and was then averaged together. Data was grouped into cases of similar scan time (top), and the same number of acquisitions (bottom). The magnitude of SNR values corresponding to a TR=5s and 64 acquisitions for NAA was 57.

The SNR/minute for NAA is calculated for each TR investigated and is plotted in grey. For each TR, the largest number of acquisitions and the associated scan time was used in determining the SNR/minute. The expected SNR is shown in red and is calculated for a TR=2s and 5s with reference to TR=8s. Expected SNR for TR=5s is calculated as (SNR at TR5) = (SNR at TR8)$$$\times\sqrt{(8/5)}$$$, and similarly for TR=2s. This stems from SNR being proportional to the square root of the number of acquisitions.

The metabolite concentrations for 5 different metabolites (N-acetylaspartate (NAA), total creatine (tCr), total choline (tCho), myoinositol (mI) and glutamate (Glu)) were normalized to their concentration values at TR=8s. For each metabolite, the 5 volunteers’ normalized metabolite concentrations were averaged together and plotted here. This was done for the data when the scan time was similar (top) and when the number of acquisitions was kept constant at 64 (bottom).

Proc. Intl. Soc. Mag. Reson. Med. 32 (2024)

1853

DOI: https://doi.org/10.58530/2024/1853