31P Single Voxel MRS in the Human Brain at 9.4T: ISIS versus SemiLASER
Johanna Dorst1, Loreen Ruhm1, Nikolai Avdievich1, Wolfgang Bogner2, and Anke Henning1

1High-Field MR Center, Max Planck Institute for Biological Cybernetics, Tübingen, Germany, 2High-Field MR Center, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria


In vivo phosphorus MR spectroscopy has been established to be a powerful tool for studies of energy metabolism. To provide clinically relevant information about physiologic function, accurate spatial localization as well as sufficient spectral quality are fundamental requirements. Therefore, in this study, first steps to optimize a multi-shot ISIS sequence and a single-shot semiLASER sequence for phosphorus MR spectroscopy in the human brain at 9.4T have been taken. The sequences were compared in terms of localization accuracy and SNR.


Phosphorus MR Spectroscopy enables observation of cellular energy metabolism noninvasively. However, 31P MRS is hampered by low 31P metabolite concentrations, low intrinsic sensitivity, and short T2 and long T1 relaxation times. As a result, large voxel sizes and long scan durations are required, which makes functional 31P MRS studies difficult. To mitigate this shortcoming, 31P MRS can benefit from the introduction of ultra-high field strength, which causes increased SNR and enhanced spectral dispersion as well as shorter T11. In this study, initial steps to optimize single voxel localization sequences for the use in 31P spectroscopy in the human brain at 9.4T were taken for Image Selected In vivo Spectroscopy (ISIS) and for a conventional slice selective excitation combined with Localization by Adiabatic Selective Refocusing (semiLASER). Both sequences were compared in terms of localization accuracy and SNR and applied in the human brain at 9.4T.


For ISIS2 (Figure 1a), Gradient Offset Independent Adiabatic (GOIA) pulses (TP:5ms) based on WURST amplitude (16th order) and gradient (4th order) modulations were used for inversion3 and a rectangular pulse (TP:350µs) for excitation. The used GOIA pulses have an inversion bandwidth of 21.6kHz and 30kHz for inversion thicknesses of 5cm and 7cm, respectively4. SemiLASER (Figure 1b) was optimized in terms of adiabaticity and localization accuracy. For slice-selective excitation, a 90° hamming windowed sinc pulse (3.35ms, TBWP:8.7) was used. For inversion, trapezoid-shaped adiabatic full passage pulses5 were optimized for phosphorus using Bloch simulations (Figure 2). To suppress unwanted coherence pathways, the crusher scheme was optimized6.
All measurements were performed on a 9.4T whole-body MRI scanner (Siemens) using a home-built double-tuned 31P/1H human head array with 8 transceiver and 2 receive only 31P channels. Localization performance was tested on a two-compartment phantom filled with equally concentrated phosphate buffer solution with different pH; afterwards, it was evaluated according to the equations for selectivity and outer volume suppression7. For in vivo spectra, 4 healthy subjects were measured with VOIs of 7cm/5cm isotropic, 4096 samples with an acquisition bandwidth of 10kHz, 64/128 averages, TRISIS:5s, TEISIS:0.3ms, TRsemiLASER:7.5s, TEsemiLASER:33ms. Voxels were placed in the center of the brain to exploit highest possible B1. For analysis, data were averaged, truncated after 75ms, filtered to calculate weights for coil elements combination employing SVD8 based on PCr, coil elements were combined using the calculated weights, then data were zero-order phase corrected and missing points at the beginning of the FIDs were predicted9. Finally, an exponential filter of 5Hz was applied. SNR was defined in the frequency domain as the ratio between the metabolite peak amplitudes and the standard deviation of the spectral noise.

Results and Discussion

Bloch simulations of a trapezoid-shaped AFP pulse (TP:6ms) with a bandwidth of 3.2kHz are shown in Figure 2. To obtain inversion of at least 90%, a minimum B1 of 35µT is needed for 31P. The transition bandwidth of this pulse is considerably broad.
Figures 3b-d show spectra acquired from the localization phantom (Figure 3a) for ISIS, semiLASER without the use of crusher gradients and semiLASER with crusher gradients, respectively. When measuring with ISIS, there is little contamination (signal at 2.4ppm) from the outer chamber. The measured selectivity was 96% and the outer volume suppression 99%. For semiLASER without crusher gradients (Figure 3c), considerably more signal is detected from the outer compartment than from the inner compartment as intended. When using the optimized crusher scheme, signal from outside the VOI is substantially spoiled, but semiLASER still provides higher contamination than ISIS. While the outer volume suppression is also measured to be 99%, the selectivity is only 53%.
Representative 31P-MRS spectra of the human brain at 9.4T are depicted in Figure 4. The corresponding SNR are listed in Table 1. Besides phosphocreatine (PCr), the metabolites phosphomonoesters (PE, PC), inorganic phosphate (Pi), phosphodiesters (GPE, GPC), γ-ATP and α-ATP are clearly visible. In addition, NAD is identifiable in ISIS spectra. Spectral quality of ISIS spectra is distinctly higher even though it is prone to motion artifacts. Due to higher TE, semiLASER localized spectra experience tremendous T2 loss, especially for ATP. Furthermore, ATP, phosphomonoesters and phosphodiesters undergo homonuclear and heteronuclear scalar coupling evolution, respectively10. These effects are also reflected in terms of SNR.


At 9.4T, single voxel localized 31P spectra with excellent spectral quality can be acquired in the human brain. Single-shot spectra measured with semiLASER experience J-coupling evolution and SNR loss due to short T2 relaxation times. FID based multi-shot ISIS spectra provide higher SNR, but are prone to motion artifacts. In this study, voxel size and positioning is limited by relatively low B1 of the multichannel coil array, especially in peripheral regions.


Funding by the European Union (ERC Starting Grant, SYNAPLAST MR, Grant Number: 679927) is gratefully acknowledged.


1Pohmann R, Raju S, Scheffler K. T1 values of phosphorus metabolites in the human visual cortex at 9.4T. Proc. Intl. Soc. Mag. Reson Med. 2018;26:3994.
2Bogner W, Chmelik M, Andronesi OC, et al. In Vivo 31P Spectroscopy by Fully Adiabatic Extended Image Selected In Vivo Spectroscopy: A Comparison Between 3T and 7T. Magn Reson Med. 2011;66(4):923-930.
3Andronesi OC, Ramadan S, Ratai E, et al. Spectroscopic imaging with improved gradient modulated constant adiabaticity pulses on high-field clinical scanners. J Magn Reson. 2010;203(2):283-293.
4Chmelík M, Just Kurková I, Gruber S, et al. Fully adiabatic 31P 2D-CSI with reduced chemical shift displacement error at 7T – GOIA-1D-ISIS/2D-CSI. Magn Reson Med. 2013;69(5):1233-1244.
5Boer VO, van Lier A, Hoogduin JM, et al. 7-T 1H MRS with adiabatic refocusing at short TE using radiofrequency focusing with a dual-channel volume transmit coil. NMR Biomed. 2011;24(9):1038-1046.
6Landheer K, Juchem C. Optimized Crusher Design for Magnetic Resonance Spectroscopy. Proc. Intl. Soc. Mag. Reson. Med. 2018;26:1288.
7Keevil SF, Porter DA, Smith MA. A method for characterizing localization techniques in volume selected nuclear magnetic resonance spectroscopy. Phys Med Biol. 1990;35(7):821-834.
8Bydder M, Hamilton G, Yokoo T, et al. Optimal phased-array combination for spectroscopy. Magn Reson Imaging. 2008;26(6):847-850.
9Nassirpour S, Chang P, Henning A. High and ultra-high resolution metabolite mapping of the human brain using 1H FID MRSI at 9.4T. 2018;168:211-212.
10De Graaf RA. In Vivo NMR Spectroscopy: Principles and Techniques. 2nd ed. Chichester: Wiley 2007.


Figure 1: a) ISIS sequence with three GOIA inversion pulses and a rectangular excitation pulse. GOIA gradient strength was fixed to 25mT/m and the gradient modulation factor to 0.9. Ramp up and down times were 1ms each. b) SemiLASER sequence with slice selective sinc excitation and implemented optimized trapezoid-shaped adiabatic full passage pulses. RF pulses are interleaved with crusher gradients with a duration of 1.2ms (400µs ramp time) and varying amplitudes of 8.2mT/m, 21.8mT/m and 30mT/m in Gx, Gy and Gz.

Figure 2: a) Amplitude, Frequency and Phase modulation of trapezoid-shaped adiabatic full passage pulses with 3.2kHz bandwidth and 6ms pulse duration, b) spectral inversion bandwidth as a function of B1 (color bar indicates Mz/M0) and c) inversion profile at B1 = 35µT.

Figure 3: a) The localizer images show the two-compartment phantom overlaid with the localization volume (48x48x19mm3) of the sequence. Spectra measured with b) ISIS, c) semiLASER without crushers and d) semiLASER with the optimized crusher scheme. Signal from the inner chamber is visible at 0ppm.

Figure 4: Representative 31P-MRS spectra of the human brain acquired at 9.4T with ISIS (a,b) and semiLASER (c,d) for voxel sizes of 7cm (top row, 64 averages) and 5cm isotropic (bottom row, 128 averages). SemiLASER localized spectra show J-coupling evolution. To guide the eye, spectra were phase first order corrected, so that most of the peaks point upward except of the phosphodiesters (GPE, GPC).

Table 1: SNR values of the shown (Figure 4) in vivo spectra for inorganic phosphate (Pi), phosphocreatine (PCr) and γ- and α-ATP. SNR was defined in the frequency domain as the ratio between the metabolite peak amplitudes and the standard deviation of the spectral noise in the range of +15ppm to +30ppm.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)