Synopsis

We investigated the effect of outer volume saturation (OVS) using very selective saturation pulses (VSS) on metabolite concentrations in PRESS MRS. Our results show that if the PRESS excitation voxel isn’t increased in size, metabolite concentrations are overestimated compared with not using VSS-OVS. These results will be of interest to other researchers planning or conducting multi-centre MRS studies.

Introduction

According to a recent¹ 3T multi-centre magnetic resonance spectroscopy (MRS) study, higher metabolite concentrations were reported when using a General Electric (GE) scanner compared with Siemens and Philips scanners. While the repetition times (TR), echo times (TE), receive bandwidth (BW), number of data points and centre frequency (2.7 ppm) were the same across vendors, some acquisition parameters could not be matched. These include the RF pulse shapes and timings for water suppression and PRESS localisation. Additionally, the GE site used outer volume saturation (OVS) using very selective saturation (VSS)² pulses (unavailable on Siemens and Philips platforms) with the PRESS localised voxel equal in size to the nominal voxel.

We investigated the effect of VSS-OVS in combination with chemical shift displacement (CSD) of the unsuppressed water. Our hypothesis was that these would lead to underestimation of the area of the unsuppressed water peak, hence overestimation of the metabolite concentrations. We also investigated the effects of overprescribing the PRESS localisation voxel in conjunction with OVS.

Methods

A phantom and healthy volunteer were scanned on a 3T GE MR750 scanner (General Electric, Chicago, IL), using a body coil for RF transmission and a 32-channel head receive coil. Volunteer scans had ethical approval and informed consent was obtained. For phantom scans, the GE “braino” phantom was used, containing a solution including 12.5 mM NAA.

For volunteer and phantom scans, axial voxels were prescribed in the anterior cingulate cortex and phantom centre respectively. PRESS³ spectra were acquired with TR/TE = 3000 ms/30 ms; CHESS⁴ water suppression; receiver bandwidth = 5 kHz; # points = 4096; voxel size = 2×2×2 cm³; centre frequency = −2.0 ppm offset from water; 96 water suppressed acquisitions for volunteer scans and 32 for phantom scans; 16 unsuppressed water acquisitions. RF pulse bandwidths were 2.36667 kHz, 1.38462 kHz and 11 kHz for the PRESS excitation pulse, PRESS refocusing pulses and VSS pulses respectively. Axial images of the voxels were also acquired using the PRESS sequence (GE’s “scan mode” = 0) with FOV = 240 mm; matrix size = 128×128, TR = 225 ms. The centre frequency was adjusted to −255 Hz for the voxel image acquisitions to match the MRS acquisitions. The size of the PRESS localisation box (overpress multiplier) was adjusted from 1.0× to 1.5× in steps of 0.1, for both phantom and volunteer scans. Assuming perfect RF pulse profiles, the water fraction in the nominal voxel excited by the RF pulses with OVS on and PRESS localised region equal to the nominal voxel is:

$$ \large(\small1-(\frac{\Delta\delta}{\Delta\nu_{exc}}-\frac{\Delta\delta}{\Delta\nu_{vss}})\large)\cdot \large(\small1-(\frac{\Delta\delta}{\Delta\nu_{ref}}-\frac{\Delta\delta}{\Delta\nu_{vss}})\small^2\large) $$

where $$$\Delta\delta$$$ is the offset between water and the centre frequency (255 Hz) and $$$\Delta\nu_{exc}$$$, $$$\Delta\nu_{ref}$$$ & $$$\Delta\nu_{vss}$$$ are the bandwidths of the excitation, refocusing and VSS pulses respectively. Figure 1 shows the relative positions of the nominal voxel and the shifted voxels for the PRESS localisation and VSS-OVS pulses.

Spectra were analysed with LC model (6.3-1L)⁵, and the NAA concentration for each overpress value compared with the concentration without OVS and overpress = 1.0, the closest match to the Siemens and Philips protocols.

Results

Figure 2 shows voxel images for the volunteer with different overpress values and combinations of OVS on/off.

From the equation above, the fraction of the voxel occupied with water but not affected by the OVS pulses was 64.4%. Table 1 shows NAA concentrations for OVS off and on with different overpress settings. LCModel fits and concentrations are illustrated in figure 3.

For human and phantom data, the NAA concentrations were 40% & 42% respectively higher with OVS on (overpress = 1).

Discussion

Since water-scaled metabolite concentrations are inversely proportional to the area of the unsuppressed water peak, the calculated fraction of the voxel occupied by water when using OVS pulses would lead to overestimation of metabolite concentrations by 55%. This is higher than for the experimental data but doesn’t account for real slice selection profiles which may include more unsuppressed water signal. The overpress value where NAA concentrations were closest to those without OVS was 1.4 for both volunteer and phantom data.

One consequence of the shift for the unsuppressed water voxel is that the shifted voxel should be used for partial volume correction of tissue water content when measuring water-scaled metabolite concentrations.

Conclusions

Acknowledgements

References

1) Naaijen J et al, Neuropsychopharmacology, 2017, 42, 2456–2465.

2) Tran T-HC et al, Magn Reson Med, 2000, 43, 23–33.

3) Bottomley PA, Ann N Y Acad Sci, 1987, 508, 333–348.

4) Haase A et al, Phys Med Biol,1985, 30, 341–344.

5) Provencher S, Magn Reson Med,30, 672–679.