Comparison of high-resolution FID-MRSI in the brain between 3 and 7 Tesla
Eva Heckova1, Stephan Gruber1, Bernhard Strasser1, Michal Povazan1, Gilbert Hangel1, Siegfried Trattnig1,2, and Wolfgang Bogner1

1High Field MR Center, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, 2Christian Doppler Laboratory for Clinical Molecular MR Imaging, Vienna, Austria

Synopsis

Magnetic resonance spectroscopic imaging (MRSI) allows to measure different metabolites in the brain. SNR and spectral resolution increases at higher magnetic fields. We compared FID-MRSI with ultra short acquisition delay (1.5 ms) and a very high spatial resolution in the same group of healthy subjects at 3T and 7T. We found 1.87-fold increased SNR and decreased CRLBs at 7T in comparison with 3T. The higher spectral resolution at 7T allows to distinguish between NAA and NAAG and reliable detect other metabolites like Glx or Tau. Accelerating the acquisition techniques leads to lower SNR, however not to substantially decreased quantification precision.

Purpose

Magnetic resonance spectroscopic imaging (MRSI) of the brain allows to detect several metabolites to achieve complementary information to be added to the conventional MR imaging methods1. High field systems (7T) offer increased SNR and spectral resolution with the prospect to increase the spatial resolution of MRSI and to better quantify adjacent metabolites which are overlapping at lower field strengths (e.g. NAA and NAAG). In addition FID-MRSI will add additional SNR, in particular for J-coupled resonances2. For this reason we compared MRSI at 3 and 7 Tesla in 10 healthy subjects using a sequence with ultra short acquisition delay (TE*) and high spatial resolution.

Methods

Ten healthy subjects (8m/2f; age: 30.6±5.52) were measured at 3T and 7T scanners (3T Trio, 7T Magnetom; Siemens Healthcare, Erlangen, Germany) using a 32-channel head coil. A FID-MRSI sequence2 with 64×64 phase encoding steps, 2048 spectral points, 6000 Hz bandwidth, FOV=220×220×10 mm2, voxel size 3.4×3.4×10 mm3,TR=600 ms, TE*=1.5 ms resulting in an acquisition time (TAQ) of 30 min was used. Moreover we obtained data with the same sequence and the same parameters accelerated with (2+1)D Caipirinha3 by a factor of 5 (TAQ=6 min). Spectra were processed using LCModel based on a simulated basis set. Metabolic maps were created using a fully automated script based on Matlab4 and MINC (Minc tools; v2.0; McConnell Brain Imaging Center, Montreal, Canada). This included also the removal of lipid contamination from the spectra using L2-regularisation5. Furthermore the SNR was computed using the pseudo-replica method in frequency domain.

Results

Good data quality was achieved from all subjects measured at 3T and 7T. Representative spectra of 3T and 7T are displayed in Figure 1. SNR was 1.87/1.74 times higher at 7T compared to that at 3T for non-accelerated/accelerated data. CRLBs (Table 1) were below 11% / 5% at 3T / 7T for the main metabolites (tNAA, tCr, tCho, Ins). The faster acquisition technique led to slightly increased CRLB values at both field strength, however the estimation of metabolites was still reliable. Furthermore NAAG, Glx, Glu and Tau could be quantified at 7T with CRLBs < 20 (except of Tau using R=5), which was not possible at 3T (Table 2, Figure 3). The increased spectral resolution at 7T allows to distinguish between NAA and NAAG (Figure 2).

Conclusion

FID-MRSI allows to acquire whole slices without fat contamination from the scalp, because of the high matrix size which was used together with hamming filtering and lipid decontamination. With the high in-plane resolution of 3.4×3.4 mm2 metabolic maps showing anatomical details could be constructed at both field strengths. The increased SNR at 7T compared to that of 3T results in satisfactory CRLBs of (j-coupled) metabolites such as Glx, Glu and Tau and high-resolution metabolic maps. NAAG could be separated from NAA at 7T but not at 3T. Therefore MRSI improves substantially at 7T allowing to obtain data with high spatial resolution and high quantification accuracy. MRSI at 7T will be further improved using fast acquisition techniques and a larger brain coverage (i.e. 3D-MRSI).

Acknowledgements

This study was supported by the Austrian Science Fund (FWF): KLI-61 and the FFG Bridge Early Stage Grant #846505.

References

1. Öz et al., Radiology 2014; 270(3):658-79.

2. Bogner et al., NMR in Biomed. 2012; 25(6):873-82.

3. Strasser et al., Proc. Intl. Soc. MRM 2014; 22:5048.

4. Považan et al., Proc. Intl. Soc. MRM 2015; 23: 1973.

5. Bilgic et al., JMRI 2014; 40(1):181-91.

Figures

Figure 1: Data processed with LCModel at 3T (left) and 7T (right).

Figure 2: The visible separation of NAA and NAAG at 7T. Figure shows metabolic maps of NAA/tCr, NAAG/tCr and NAA+NAAG/tCr at both field strengths.

Figure 3: Metabolic maps of Glx/tCr (top) and Ins/tCr (bottom) at 3T and 7T. Note that the reliable Glx quantification over the whole slice was possible at 7T but not at 3T.

Table 1: The amplitudes (Amp) and Cramer-Rao lower bounds (CRLB) of 4 metabolites at 3 and 7T without and with acceleration (acc.). In addition SNR of NAA is given.

Table 2: The amplitudes and CRLB of NAAG, Glx, Glu and Tau measured at 7T without and with acceleration (acc.)



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
2358