Post acquisition frequency correction in GABA editing
Jan Willem van der Veen1, Stefano Marenco2, Karen Berman2, and Jun Shen1

1Magnetic Resonance Spectroacopy Core, NIH, NIMH, Bethesda, MD, United States, 2NIH,NIMH,CTNB, Bethesda, MD, United States

Synopsis

Patient motion and magnetic field drift may shift the frequency of the GABA editing pulse relative to that of metabolites. Using the frequency location of the residual water we corrected for changes caused by frequency variations by using averaged reference signals simulated at specific editing frequency offsets. Our analysis also showed that GABA editing with a top-hat editing pulse is highly robust in the presence of frequency variations.

Introduction

γ-aminobutyric acid (GABA) is the primary inhibitory neurotransmitter in the CNS, believed to be involved in a variety of psychiatric and neurological disorders. GABA can be measured using proton MRS with a PRESS-based two-step editing sequence (1). However due to the low concentration of GABA, 20 minutes are usually needed to achieve a sufficient precision of the GABA value to detect small differences in many psychiatric disorders. During this scan time the frequency offset of the measured voxel can fluctuate because of patient movement and even magnetic field drift. This drift will move the frequency of the editing pulse relative to the metabolite frequencies and editing efficiency can fluctuate accordingly if the performance of the editing pulse is sensitive to frequency fluctuations. Here we describe an editing frequency drift compensation method by post acquisition fitting. The results also demonstrate the remarkable robustness of a GABA editing technique in the presence of frequency fluctuations.

Methods

141 normal volunteers were scanned on a 3 Tesla whole body scanner (GE, Milwaukee, WI, 14M4 platform). The spectroscopy voxel was placed immediately superior to the ventricles in mostly gray matter, NS=768, TR/TE=1500/68 ms, NEX=2. A total of 384 edited and non-edited FID pairs were acquired for a total of 20 minutes. The non-edited, edited and the difference time domain data were fitted simultaneously using a Levenberg-Marquardt non-linear fitting program written in IDL (ITTVIS). The fitting was performed using simulated reference signals for: NAA, NAAG, creatine (CRE), 1/2 creatine and 1/2 phospho-creatine (2), choline, 1/3 PCH and 2/3 GPC (2), myo-inositol (MIO), glutamate (GLU), glutamine (GLN), scyllo-inositol (SCI), glutathione (GSH), and GABA, with GABA2 at 3.0 ppm combined with a macromolecular contribution (MM) modelled by a gauss line (linewidth 14 Hz, GABA/MM = 0.61 (3)). An updated GAMMA library (4) from duke.edu (5) was used to simulate the effects of frequency shifts on the GABA editing sequence. Also effects of RF shapes, crusher gradients and various coherence pathways were fully simulated (6,7). Simulations were made for a range of editing pulse offsets with a resolution of 1Hz around the default position. The editing frequency offset drift during the scan was estimated from the offset of the residual water. An average simulated reference signal was created based on these offsets.

Results and discussion

The amplitude of the NAA peak at 2.01 ppm is in the transition area of the editing pulse (Fig 1), therefore, it is most sensitive to frequency shifts. To demonstrate the effectiveness of our correction method, the effect of frequency offset of the editing pulse on the amplitude of NAA in the edited spectra is shown in the left panel of figure 2. Strong correlation between editing frequency offset and NAA amplitude is visible for theedited spectrum with R2 equal to 0.42. The correlation is absent in the unedited data. Correcting for the editing offset (Fig 2 center panel) eliminates the dependency on frequency offset with R2 equal to 0.002 and results in an equal amplitude and scatter pattern for the edited and unedited data. The GABA amplitude is not affected by the editing pulse frequency offset over the range of frequency fluctuation seen experimentally (right panel of Fig 2). The R2 for GABA is 4e-6. The results of the estimated NAA amplitude show the high accuracy of numerical simulation and the capacity of our method to correct for variations in editing efficiency due to frequency shifts originated from patient motion and scanner frequency drift. For an editing pulse (1) that has a flat top over the edited GABA frequencies (as shown by the dotted line in Fig. 1), the effect of the correction is negligible, demonstrating the robustness of the GABA editing sequence described in ref (1). The correction method described here can also be applied to other editing sequences that are sensitive to frequency fluctuations due to employing an editing pulse with a sharper frequency profile.

Acknowledgements

This work is supported by NIMH, IRP.

References

1. Sailasuta P. et al, Proc ISMRM 9:1011 (2001).

2. R.A. de Graaf, in vivo NMR Spectroscopy,2nd ed.

3 Van der Veen, JW et al, Proc ISMRM 22:2889 (2014)

4. Smith SA et al, J. Magn. Reson. 106a, 75-105, (1994).

5. http://scion.duhs.duke.edu/vespa/gamma

6. de Beer R et al, Meas. Sci. Technol. 22 (2011) 114022 (9pp).

7. Van der Veen, JW et al, Proc ISMRM 21:2032 (2013)

Figures

Figure 1. The average of 141 fits on the edited, unedited, and difference signals with only the unedited signals shown. Black lines are the average, red lines the standard deviation. The dotted line is the editing pulse profile at the default position.

Figure 2. Dependence of NAA and GABA amplitudes on editing pulse frequency offset. Left panel: NAA amplitude from the edited data without correction; middle panel: NAA amplitude after correction. Right panel is the GABA amplitude as a function of editing pulse frequency offset without any correction. Solid lines are the line fit to the data and the P=0.01 confidence limit.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3993