Marcus J. Couch1, Lumeng Cui1, and Sinyeob Ahn2
1Siemens Healthcare Limited, Oakville, ON, Canada, 2Siemens Medical Solutions, Malvern, PA, United States
Synopsis
Keywords: System Imperfections, Spectroscopy
Motivation: Spectroscopy is sensitive to frequency drift. After running sequences that have a high gradient duty-cycle, frequency continues to drift due to gradient cooling.
Goal(s): To propose a generalized prospective real-time frequency correction that can be applied to imaging and spectroscopic sequences, where TR-variant and -invariant events are separable in time.
Approach: Real-time frequency adjustment was implemented for CSI, using a navigator placed where the spin phase is consistent across TR cycles prior to the CSI readout.
Results: Phantom measurements with the generalized navigator provide accurate estimates of the frequency drift, thereby minimizing spectral distortion and providing an improved baseline in the final spectra.
Impact: Prospective real-time frequency adjustment using a generalized
navigator approach, where the spin phase is consistent across TR cycles, can be
applied to spectroscopic imaging to correct for frequency drift caused by
gradient heating and cooling.
Introduction
A stable B0
field is important for maintaining spectral quality; however, frequency drifts
can be induced by imaging protocols that make use of a high gradient duty
cycle, such as fMRI and diffusion imaging. When the RF pulse frequency drifts
away from the resonance frequency, water suppression and metabolite signal
quality is affected. A recent multi center study has assessed the frequency
drift on 3T platforms from multiple scanner vendors, where the median drift
after a 5-minute fMRI scan was 1 Hz, and as high as 10 Hz in some cases (1).
Various
prospective frequency correction techniques have been developed, such as an
interleaved approach that can switch between acquiring a localized PRESS water
reference signal and water suppressed MEGA-PRESS (2). Other prospective
approaches use an image-based EPI navigator before the single voxel
spectroscopy (SVS) readout, which can additionally be used for motion
correction (3). A recent prospective frequency correction technique, involving
no additional pulse events, inserted a 1 ms FID acquisition window after the
first CHESS water suppression RF pulse in a MEGA-PRESS SVS acquisition (4).
We propose a more generalized approach, termed a
generalized real-time frequency adjustment (RFA) navigator, that can be applied
to imaging and spectroscopic imaging sequences using an asymmetric non-echo
forming FID acquisition window.Methods
A CSI sequence was modified to insert a 1 ms FID acquisition
window for RFA after spatial location RF pulses and subsequent gradient crushers
(TR-invariant events) and before the phase-encode gradients (TR-variant events).
The FID signal must be acquired when the 0th order gradient moments
are refocused. This condition is consistent and TR-invariant. This assumes any
incidental phase change is attributed to non-sequence invoked phases such as from
temperature changes. In this study, we assume there is no motion contributing
to phase changes. The frequency is calculated in real-time over an adjustable
sliding window (typically 7 TRs), and the system frequency is updated every TR.
For comparison, the method used in (4) was also implemented with a CSI sequence
where the navigator is non-selective.
All phantom measurements were performed using a 3T system (Siemens
MAGNETOM Prisma, Erlangen, Germany) with a 2-L Agar metabolite phantom. Three
separate CSI scans were acquired, with each following a diffusion acquisition
(1.7x1.7x4 mm3 with 64 directions, b=3000 sec/mm2, 5 averages,
16:43 minutes) to induce gradient heating and cooling. The CSI acquisitions
used TE/TR=40/1500 msec, 13x13x15 mm3
voxel, 12x12 grid, 6 averages, 12:15 minutes, FWHM=13.7 Hz, with i) no RFA,
ii) non-selective RFA (prototype sequence), and iii) generalized RFA (prototype
sequence). The measured frequency was saved to a text file while the scan was
running and plotted in Matlab R2019a.Results
Figure 1 shows the frequency drift measured in real-time for
two CSI acquisitions. The non-selective RFA navigator corrected for an 11 Hz
shift in the system frequency, while the generalized RFA navigator corrected
for a 13 Hz shift in the system frequency. The non-selective RFA navigator appeared
to include more noise, with slightly higher mean absolute frequency jumps
between acquisitions (0.128±0.096 Hz, compared to 0.094±0.069
Hz for the generalized RFA).
Figure 2 shows the comparison of spectral maps acquired with
and without RFA, where the prospective frequency correction shows better
signals (see lateral and lower side in the spectral map) than the uncorrected. The
upper part of the map suffers from field inhomogeneity due to phantom-air
interface and the central part experiences a poor B1 efficiency.
Figure 3 shows the comparison of spectral signals from two
representative voxels, one from a reasonable location (V1) and one near the
field-inhomogeneous location (V2). At V1, spectral signal zoomed in around NAA shows
a better baseline and a slightly narrower spectrum in the frequency corrected
spectra, with the generalized navigator (h) being slightly better than the
non-selective RFA navigator (e). At V2, the uncorrected NAA signal shows a split
peak, whereas the corrected ones maintain the NAA peak shape. Discussion
This work has shown that a generalized RFA can be applied to
CSI to correct for frequency drift caused by gradient heating and cooling,
where the frequency estimation is done only on the imaging ROI. The non-echo
forming navigator with a short readout (high readout bandwidth) of water
suppressed signal has sufficient SNR with the sliding window approach. Minimum
TE allowable may increase slightly but the proposed technique may be considered
for cases where the benefit outweighs the min TE restriction. The generalized RFA
navigator had, on average, smaller frequency jumps between acquisitions
compared to the non-selective RFA navigator which includes signal from outside
of the shim-optimized volume of interest.Acknowledgements
No acknowledgement found.References
- Steve C N
Hui et al. Frequency drift in
MR spectroscopy at 3T.
NeuroImage (2021) 241: 118430.
- Richard A E Edden, Georg
Oeltzschner, Ashley D Harris, Nicolaas A J Puts, Kimberly L Chan, Vincent O
Boer, Michael Schär, and Peter B Barker. Prospective frequency correction
for macromolecule-suppressed GABA editing at 3T. Journal of Magnetic
Resonance Imaging (2016) 44:1474–1482.
- Aaron T. Hess, M. Dylan
Tisdall, Ovidiu C. Andronesi, Ernesta M. Meintjes, and Andre J. W. van der
Kouwe. Real-Time Motion and B0 Corrected Single Voxel Spectroscopy Using
Volumetric Navigators. Magnetic Resonance in Medicine (2011) 66:314–323.
- Sinyeob Ahn, Tongbai Meng,
Dieter J. Meyerhoff, and Gerhard Laub. MEGA-PRESS Single-voxel Spectroscopy
for GABA J-editing with Real-time Frequency Adjustment. Proceedings of
ISMRM (2016) 4008.