Gilbert Hangel1, Bernhard Strasser2, Michal Považan2, Lukas Hingerl1, Marek Chmelík2, Stephan Gruber2, Siegfried Trattnig2,3, and Wolfgang Bogner2
1MR Centre of Excellence, Medical University of Vienna, Vienna, Austria, 2MRCE, Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, 3Christian Doppler Laboratory for Clinical Molecular MR Imaging, Vienna, Austria
Synopsis
Increasing the resolution of MRSI is desirable to delineate
small structures and pathologic deviations such as Multiple Sclerosis lesions and
increase local B0-homogeneity per voxel. We show that using an
FID-MRSI sequence with short TR and L2-regularisation for lipid contamination
removal, the major brain metabolites can be mapped with a 128x128 matrix over a
whole brain slice with unprecedented detail, with a nominal voxel volume of
1.7×1.7×8 mm³. The additional application of parallel imaging allows reducing
measurement times enough for potential clinical applications.Purpose
Over the last years, FID-MRSI
1,2 in the
brain was successfully demonstrated for matrix sizes up to 64×64. We propose
that by optimising the sequence design in order to reduce the TR to 200 ms, ultra-high-resolution
(UHR)-MRSI at 7 T with a nominal voxel size of 23 µl and a 128×128 matrix becomes
possible at reasonable measurement times. The benefits of this approach include
a better point-spread function, better local B0-homogeneity and
reduced partial volume effects. Reaching resolutions similar to conventional
MRI, metabolite maps already feature details in brain anatomy such as
visualization of separate gyri and allow the analysis of smaller structures and
lesions, which has a high potential for research as well as for diagnostic
applications.
Methods
Three volunteers (all male, 27, 34 and 34 y) were
measured with a Siemens 7 T
Magnetom scanner and a 32-channel head coil. Written informed consent
was obtained as well as approval of the institutional review board.
In order to reduce the already low minimum TR of the
FID-MRSI sequence with a short acquisition delay of 1.3 ms, the duration of the
WET-water suppression
3 (originally designed for 1.5 T) was shortened
from 180 ms to 64 ms by optimising the spoiler gradient duration making use of
the available gradient performance (gradient strength of 40 mT/m per direction, maximum slew rate of 200 T/m/s).
By reducing the ADC sampling points to 512 and increasing the receiver
bandwidth to 4500 Hz, the ADC duration was further shortened. As displayed in
Figure 1, this combination allowed reducing the TR down to 200 ms.
Based on this, our measurement protocol was centred on
an UHR-MRSI sequence with a 128×128 matrix and an FOV of 220×220×8 mm³,
resulting in a nominal voxel size of 1.7×1.7×8
mm³. The excitation flip angle was set to 27° and the delta frequency to -2.1
ppm. With elliptical encoding, this resulted in a measurement time of ~41 min. For
a comparison to our previous standard, we conducted two further MRSI
measurements with a matrix size of 64×64 accelerated with 2×2 GRAPPA
4,
one with the same TR of 200 ms (2:35 min) and one with a longer TR of 600 ms
(7:38 min). Anatomical reference imaging was provided by an MP2RAGE scan (4:39
min).
Data processing utilised a MATLAB-based routine
developed in-house
5 that included lipid signal removal by
L2-regularisation
6. The resulting metabolite maps were interpolated
to double resolution and then compared together with SNR, FWHM and CRLB values
for NAA. Additionally, to test the limits of measurement acceleration, the UHR
datasets were processed again with artificial GRAPPA-undersampling with
acceleration factors (R) of 2, 4 and 6.
Results
Individual spectra (as
in Figure 2) show a good comparability between the resolutions. As presented in
Figure 3, while the comparison scans hint at anatomical details, the UHR
measurements allow resolving finer structures such as gyri and an improved
GM/WM separation. Examining the metabolite and metabolic ratio maps of all
volunteers (Figure 4), this is true for all measurements.
The SNR loss due to
reduced voxel size is less than linear, as previously reported for MRSI
7,
with an average SNR for UHR of 41±15 over all volunteers compared to an SNR of
42±13 for the comparison scan with the same TR. Reducing the spectral sampling
points affected CRLBs and FWHMs more than the smaller voxel volume, with
average CRLB[%]/FHWMs[Hz] of 10±5/16±7 for UHR-MRSI, 9±4/17±6 for the TR200-comparison
and 4±3/12±6 for the TR600-comparison.
Figure 5 shows that
an R of 2 is achievable without a perceptible quality loss while even an R of 4
should be robust enough. At an R of 6, metabolites different from NAA suffer
from unreliable quantification. This relates to ~20 and ~10 min scan time, respectively.
Discussion/Conclusions
We
successfully showed that UHR-MRSI in the brain at 7 T is feasible in a
reasonable time for healthy subjects. The resulting maps show an unmatched
correspondence to anatomical detail that will be useful in the study of small
lesions or focal regions. Downsides are the reduced spectral vector size and
less SNR, limiting the application of parallel imaging. Still, with an
acceleration down to 10-20 min, clinical applications are possible, with MS as
a prime candidate. Reducing the slice thickness would also be desirable to
reduce the superimposition of different small structures like gyri onto each
other. For the evaluation metabolic ratio maps, the different metabolite weighting
due to the shorter TR must be kept in mind.
Acknowledgements
This study was supported by the Austrian Science
Fund (FWF): KLI-61 and the FFG Bridge Early Stage Grant #846505.References
[1] Bogner et al., NMR Biomed 2012; 25(6):873-82
[2] Henning et al., NMR Biomed 2009; 22(7):683-96
[3] Ogg et al., J. Magn. Reson. B 1994; 104:1–10
[4] Strasser et al., Proc.
Intl. Soc. MRM 21 (2013):2018
[5] Považan et al., Proc. Intl. Soc. MRM 23 (2015): 1973
[6] Bilgic et al., JMRI 2014; 40(1):181-191
[7] Gruber et al.,
Magn. Reson. In Med. 2003 49(2):299-306