2491
B0 Drift Correction in Proton Chemical Shift Imaging1Psychiatry, University of Utah, Salt Lake City, UT, United States
Synopsis
This study aims to improve 1H CSI data quality, by presenting a novel method for correction of B0 instability (0.9 Hz/min drift) due to gradient system heating produced by application of DTI and fMRI sequences. The method tracks magnetic field drift using three reference lines in the 1H CSI data, which allow misaligned spectral data to be corrected in post-processing. This novel method may be combined with any spectroscopic technique that employs water suppression. Both phantom and in vivo data are presented, to demonstrate improved SNR and spectrum quality, with minimal influence on metabolite data acquisition or added cost.
INTRODUCTION
Proton (1H) Magnetic Resonance Spectroscopy (MRS) is an informative non-invasive method for in-vivo tissue characterization. Over the last three decades, MRS has been increasingly applied to the study of central nervous system (CNS) disorders. In practice, MRS sequences such as Point Resolved Spectroscopy (PRESS), Free Induction Decay (FID) and spin-echo (SE) chemical shift imaging (CSI) are often applied sequentially, along with magnetic resonance imaging (MRI) sequences. Application of high speed MRI sequences such as DTI and multi-echo BOLD imaging may cause mechanical vibration of the gradient system, increasing the gradient system temperature. In turn, this may induce magnetic field drift, which decreases the quality of MRS data.
During MRS acquisitions from the CNS, due to relatively low in-vivo metabolite concentrations, multiple spectroscopic signals are acquired and averaged, to improve signal-to-noise ratios (SNR). However, temporally varying B0 magnetic fields will degrade the SNR of averaged signals due to misalignment of metabolite peaks. The quality of single voxel PRESS data suffering from B0 magnetic field drift can be improved using post-processing procedures [1]. However, B0 field drift correction of CSI data is challenging, due to application of phase-encoding gradients. Researchers have addressed this issue by acquiring additional reference water spectra, for use in B0 drift correction of CSI data [2]. However, this approach extends acquisition time, thereby limiting its use in applications requiring shorter repetition times. Therefore, we propose a novel B0 field drift tracking method that does not require acquisition of additional reference spectra, with minimal impact on the measurement of MRS metabolites.
METHOD
Fig. 1 illustrates a B0 drift tracked sequence diagram with three reference lines for each acquisition. This design utilizes the intact non-selective, unsuppressed water signal to maximize the signal intensity of reference echoes from the whole brain. Frequency drift is computed according to the phase difference of reference line 1 and line 3: $$$ f=\sum_iw_{i}\frac{\phi_{3i}-\phi_{1i}}{2\pi*2*echospace}$$$, where the phase difference of reference line 3 ($$$\phi_{3i}$$$) and line 1 ($$$\phi_{1i}$$$) is divided by two times echospace time, $$$\omega_{i}$$$ is the i-th voxel intensity weight factor and $$$f$$$ is the frequency drift.
All studies were performed on a 3 T Siemens Prisma Scanner. 1H spectra were acquired using modified CSI SE and CSI FID pulse sequences. To illustrate how the proposed method works, phantom data was acquired using a 3D CSI SE sequence with TR/TE 2000/30 ms, 36 average, voxel size 12.5×12.5×12.5 mm, and echospace 1.58 ms.
In addition, to demonstrate how long the B0 drift lasts following a multi-echo BOLD sequence, we collected serial data sets using 3D CSI FID sequence with acquisition parameters: TR/TE 3000/2.3 ms, 23 average, voxel size 2.5×2.5×2.5 cm3, and echospace 0.88 ms. Three phantom data sets were collected at 0, 30, and 60 min after multi-echo BOLD sequence, respectively. B0 drift frequency curves are demonstrated in Fig 3. Drift frequency is filtered by low pass filter to remove variation from noise. To examine the performance of the proposed method for in-vivo brain study, a volunteer was scanned to acquire 2D CSI SE data with TR/TE 2000/30 ms, 14 average, voxel size 12.5x12.5x25 mm3, and echospace 1.58 ms.
RESULTS & DISCUSSION
Fig. 2a illustrates significant B0 frequency drift of 15 Hz after implementation of a BOLD fMRI sequence. The resulting uncorrected raw spectrum (blue curve in Fig. 2b) is dramatically out of alignment. After correction of the data for B0 drift, the quality of the spectrum is much improved (red curve in Fig. 2b). To determine the duration of magnetic field drift once the gradient system is disturbed, phantom data were collected at 30-minute intervals. Frequency drift curves are plotted in Fig. 3. These curves suggest that magnetic field drift resolves after 1 hour. To test the feasibility of the proposed technique in human subjects, a healthy volunteer was scanned. Fig. 4a demonstrates magnetic field drift of approximately 27 Hz during in-vivo scanning. However, the spectral peak’s full-width-at-half-maximum (FWHM) is determined by the field inhomogeneity over a larger volume of interest. The misaligned metabolite peaks may not be visible, because poor field homogeneity broadens the spectral peak width (blue in Fig.4b). However after correction, we observed significantly improved spectral FWHM. By taking advantage of unwanted whole brain water signal to track magnetic field drift, this technique maximizes reference echo intensity and reduces the variation of phase induced by noise. Moreover, metabolite data acquisition is undisturbed. The proposed method can be combined with any spectroscopic technique that employs water suppression, while enabling the spectroscopist to improve the spectral quality at minimal cost.Acknowledgements
No acknowledgement found.References
[1] Rowland BC et al. J Neuroimaging 2017; 27(1):23-28.
[2] Tal and Gonen. Magn Reson Med 2013; 70(4):895-904.
Figures
