INTRODUCTION

Recently, B0-inhomogeneity-insensitive GABA editing with macromolecule (MM) suppression has been developed using an improved MEGA-SPECIAL sequence.¹ Because of the effective MM suppression, a pseudo-doublet GABA-only peak at 3 ppm can be obtained in the edited spectrum. Since the edited GABA peak at 3 ppm is obtained by eliminating the overwhelmingly larger Cre peak using subtraction, any discrepancies in Cre peak amplitude due to physiological noise, i.e. slight voxel movement caused by breathing or cardiac motion, will result in variations in edited GABA. As the physiological noise is proportional to the signal strength, variations in Cre peak amplitude constitutes the major physiological noise source in the edited GABA peak. Using principle component analysis (PCA), the dominant Cre signal can be eliminated from the outer peaks of GABA triplet resonance at 3 ppm.

METHODS

Cre was simulated as a singlet at 3.0 ppm with a peak amplitude of 40. GABA for editing-off was simulated as a triplet at 2.95 ppm, 3.0 ppm and 3.05 ppm with amplitude of 1:2:1 while GABA for editing-on was simulated by flipping the outer peaks of the triplet as shown in Fig 1. Principle component analysis was performed for spectra ranging from 2.9 ppm to 3.1 ppm for both editing-on and -off. For the PCA editing, Cre-only peaks for both editing-on and -off were reconstructed using the largest principle component and subtracted from the original spectra before subtracting the editing-off spectra from the editing-on spectra. To simulate physiological noise elimination using the PCA editing, a 20% standard deviation in Cre amplitude was introduced to each acquisition frame for a total of 32 editing-on and 32 editing-off frames, and the entire acquisition was simulated 20 times. GABA levels were estimated by integration of the edited spectra from 2.9 to 3.1 ppm for both regular and PCA editing. For phantom studies, MRS data were acquired from a 20x20x20mm³ voxel in a GE braino phantom with 3 mM GABA using a 48-channel coil on a GE Premiere 3T scanner using an improved MEGA-SPECIAL sequence with TE/TR=80ms/2s and 128 transients. For in vivo studies, MRS data were acquired from a 30x30x30mm3 voxel in occipital lobe from a healthy subject for eight times with the same setting as the phantom study. GABA levels were estimated by integrating the edited peak from 2.9 to 3.1 ppm and referenced to Cre for quantification for both regular editing and the PCA editing.

RESULTS

Cre-only peaks obtained from the largest principle component for both editing-on and -off are shown in Fig 2(a), and the edited GABA peaks using both regular and PCA editing are shown in Fig 2(b). As shown in Fig 3, the addition of simulated physiological noise resulted in the standard deviation of the GABA signal estimated using PCA editing dropping significantly to 173 from 20769 using regular editing. At the same time, there was no loss in edited GABA signal. Given no presence of the physiological noise, near identical edited GABA peaks were obtained from a GE braino phantom with 3mM GABA using both regular editing and the PCA editing as demonstrated in Fig 4. Representative in vivo edited spectra using both PCA and regular editing were shown in Fig 5. From the eight in vivo acquisitions, GABA/Cre measurement using the PCA editing was unbiased and showed a standard deviation reduction of 26% as compared to regular editing.

CONCLUSION

Using the PCA-based GABA editing, Cre and its associated physiological noise can be effectively eliminated without GABA signal loss. Compared to regular GABA editing, significantly better signal-to-noise ratio can be achieved using this post-processing technique with no increase in acquisition time. Because residual macromolecule signal may share signal characteristics with the GABA signal, this PCA editing may not be directly applicable to the regular MEGA-PRESS due to coediting of the macromolecules.

Acknowledgements

Lucas foundation, GE Health Care, NIH R01 EB015891, NIH R01 MH110683, Stanford Center for Cognitive and Neurobiological Imaging

References

1. Gu M, Hurd R, Noeske R, Baltusis L, Hancock R, Sacchet MD, Gotlib IH, Chin FT, Spielman DM. GABA editing with macromolecule suppression using an improved MEGA-SPECIAL sequence. Magn Reson Med. 2017 Mar 31. doi: 10.1002/mrm.26691.