MEGA-PRESS Single-voxel Spectroscopy for GABA J-editing with Real-time Frequency Adjustment
Sinyeob Ahn1, Tongbai Meng2, Dieter J Meyerhoff3,4, and Gerhard Laub1

1Siemens Healthcare, San Francisco, CA, United States, 2Siemens Healthcare, Baltimore, MD, United States, 3Radiology and Biomedical Imaging, UCSF, San Francisco, CA, United States, 4CIND, Veterans Affairs Medical Center, San Francisco, CA, United States


Spectroscopy scan is highly sensitive to frequency drift during the acquisition. Although most spectroscopy sequences do not create large frequency drift themselves, it is problematic when they are run immediately after sequences with high gradient duty-cycle, as the frequency still drifts due to gradient cooling. In this paper, realtime frequency adjustment was implemented for MEGA-PRESS sequence for GABA J-editing. FID signal was read during water suppression and used to calculate and update the system frequency every TR cycle. It was tested on a phantom and in vivo and showed effectiveness by providing insensitivity to the frequency drifts during the acquisition.


Spectral quality and quantitation are highly sensitive to spectrometer frequency drift, which affects water signal suppression and metabolite signal quality. Due to their relatively low gradient duty cycle, most single-voxel or spectroscopic imaging methods do not cause large frequency drifts themselves during data acquisition. However, frequency drifts can be substantial during gradient cool-down after sequences with high gradient duty-cycle, typically used for DTI or fMRI acquisition. Such frequency drifts are problematic especially for J-editing technique (1). It is because J-editing involves spectral subtraction and requires stable editing efficiency during data acquisition only provided by a stable spectrometer frequency (i.e., the latter effect cannot be corrected for by post-acquisition alignment of spectra to be subtracted for yielding the edited signal). Therefore, it is advised that spectroscopy scans not be performed after sequences with high gradient duty-cycle. However, this “solution” is inconvenient, often impractical, inefficient and uneconomical. In this paper, we propose real-time frequency adjustment (RFA) during MRS data acquisition. We demonstrated it for GABA J-editing (2) with single voxel MEGA-PRESS (3) on both a GABA phantom and in-vivo. RFA significantly improved water suppression and J-editing quality and efficiency, thereby increasing GABA quantitation accuracy and reliability.

Methods and Materials

FID signal is read out for about 1 msec immediately after the first CHESS water suppression (4) RF pulse. This provides sufficient FID signal from transverse magnetization as the first CHESS pulse usually has a flip angle close to 90 degree. The required sequence modification does not alter any of the J-editing sequence timing or performance. Water frequency is calculated in real-time over the period of a 30-second sliding window and the system frequency is updated every TR cycle in steps of 1 Hz. The technique was implemented and tested on a 3T system (Skyra, Siemens Erlangen). Diffusion tensor imaging (2 mm isotropic with 64 directions, b=2000 sec/mm2, 9:40 min scan) was run immediately before the J-editing sequence to induce gradient heating and subsequent cooling during MRS data acquisition. GABA was measured on a phantom (GABA 200mM, 50 mM NaAc in saline solution) from 20 mm isotropic voxel with 64 averages (TE/TR=68/2000 msec, 4:41 minutes, fwhm=4.4 Hz) with and without the RFA. Similarly, following a 10:07 min DTI scan (1.6 mm isotropic with the maximum readout gradient switching), MEGA-PRESS was performed to measure GABA signal at parieto-occipital cortex (POC) in human brain (30x25x25 mm voxel, 128 averages, 8:40 minutes, fwhm=12.8 Hz).


Figure 1 shows the detection and correction of frequency drift in realtime (phantom study) that corrected 15 Hz. Long train of zero correction in the beginning is attributed to the initial frequency estimation over the first sliding window period. Frequency drifts were 18 Hz and 14 Hz (frequency increased because MRS was performed immediately after the DTI) in the phantom study for acquisition with and without RFA, respectively. Top row in Figure 2 shows poor editing data quality without RFA: poor inversion (2b) and refocusing (2a) of the 3-ppm GABA resonance. On the contrary, RFA (bottom row) produced good inversion (2e) and refocusing (2d) with consequently good-quality edited GABA signal (2f). There were -48 and -42 Hz frequency drifts (frequency decreased due to gradient cool-down) in-vivo study for acquisition with (RFA corrected -40 Hz) and without RFA, respectively. Top row in Figure 3 shows results of measurements without RFA: We observe poor water suppression (strong residual water signal at 4.5 ppm) and unrecognizable metabolite signal, presumably related to significant frequency drift. Not unexpectedly, the edited spectrum, both the GABA signal at 3 ppm and the co-edited Glx signal at 3.75 ppm, is ill-defined, making spectral fitting/quantitation unreliable. On the contrary, the spectral data in Figure 3c obtained with RFA demonstrates good water suppression and high spectral quality of the off-resonance spectrum. The resulting edited GABA and Glx signals in Figure 3d are of expectedly excellent quality and imminently quantifiable.

Discussion and Conclusion

The novel RFA application in single-volume MRS demonstrated here, improves water suppression and J-editing quality in the circumstances of frequency drifts. Editing efficiency, reliability, and robustness of in-vivo MEGA-PRESS J-editing were largely improved due to RFA. It appears that RFA slightly under-estimates the actual frequency drift by a few Hz. This is believed to be from the fact that RFA did not include any calibration/preparation scan periods while the actual frequency was manually recorded before the scan began. Nevertheless, the observed improvements from relative immunity to common frequency drift are remarkable, providing quality MRS data and increased patient throughput.


No acknowledgement found.


  1. Harris et al, “Impact of Frequency Drift on Gamma-Aminobutyric Acid-Edited MR Spectroscopy”, MRM 2014;72(4):941-948.
  2. Mullins et al, “Current practice in the use of MEGA-PRESS spectroscopy for the detection of GABA”, NeuroImage 2014;86:43-52.
  3. Mescher et al, “Simultaneous in vivo spectral editing and water suppression”, NMR in Biomedicine 1998;11:266-272.
  4. Ogg et al, “WET, a T1- and B1-insensitive water-suppression method for in vivo localized 1H NMR spectroscopy”, JMR(B) 1994;104(1):1-10.


Figure 1. Realtime frequency detection and correction in steps of 1 Hz (correction of 15 Hz is shown).

Figure 2. GABA measurement on the phantom. Left: on resonance, center: off resonance, right: edited spectrum. Top row: without RFA and bottom row: with RFA. RFA provided quality on and off spectra and consequently, edited GABA signal (d~f).

Figure 3. Spectra acquired at parieto-occipital cortex. Left: Off-resonance and right: edited spectrum. Top row: without RFA and bottom row: with RFA. Inset in (a) indicates frequency drift (114th acquisition out of 128) and consequent failure of water suppression. RFA provided quality GABA signal at 3 ppm (circle in d).

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