1. Correcting for Center Frequency Variations in MRSI Data Using the Partially Suppressed Water Signal Lawrence P Panych, Ph.D., Joseph R Roebuck, Ph.D., M.D., Nan-Kuei Chen, Ph.D. Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, Boston MA This work was supported by NIH U41-RR019703, R21-CA11092, 5R25-CA089017. ISMRM Workshop on Data Processing for MR Spectroscopy and Imaging Warrenton, Virginia 11-13 November, 2006

2. Several authors have reported methods using interleaved navigators to dynamically estimate the center frequency during a spectroscopic acquisition using a short acquisition without water suppression to correct spatial and spectral artifacts that would otherwise occur. Thiel et. al. Magn Reson Med. (2002)47:1077 Interleaved navigator during single voxel 1 H-MRS acquisition Elbel and Maudsley. Magn Reson Med. (2005)53:465. Interleaved navigator during 1 H 3D-EPSI acquisition In preliminary work reported here, we explored using the raw phase-encoded data from a MRSI acquisition with partial water suppression to estimate the center frequency variations. INTRODUCTION

3. INTRODUCTION: Spatial Misregistration of MRSI Signal 1 24 1 x y Reconstructed MRSI of a pre-selected rectangular voxel placed at the center of the field of view (shown in red) within a homogeneous phantom. The signal is largely localized to the pre-selected voxel but spatial misregistration is seen in both directions, x > y.

4. INTRODUCTION: The Magnetic Field Drift Center frequency variation with time estimated by interleaving MRSIs with and without PE. Top: Plots from same magnet, separate acquisitions, without and with water suppression. Bottom: Plots from two 1.5T magnets acquired synchronously with water suppression. Frequency (Hz) Without suppressionWith suppression Stable ω o Fluctuating ω o

5. METHODS Magnet: General Electric Signa 1.5T 12.0 hardware/software. Localization:PRESS TR 1000/TE30 Rectangular voxel pre-selected in the axial plane of magnet isocenter MRSI: 24 x 24 or 16 x 16 sequential phase encoding in x, y 1 x 1 x 1 cm in-plane resolution Navigators:Interleaved with MRSI with G x = G y = 0 Phantom: General Electric MRS brain phantom NameSymbolConc. Myo-inositolmI 7.5 mM 7.5 mM CholineCho 3.0 mM 3.0 mM CreatineCr 10.0 mM GlutamateGlu 12.5 mM N-acetyl-L-aspartateNAA LactateLac 5.0 mM 5.0 mM Post-processing: MatLab, Power Mac

6. METHODS: Interleaved MRSI and Navigators (G x =G y =0) 1 M 1 N kxkx kyky 1 M 1 N n=1 k x =N  t = TR MRSI NAVIGATORS INTERLEAVED m 2 3 2 n=N 3 t = 0 t = 2N×M × TR n k x =1 t = 2N × TR n, k x

7. METHODS: Uncorrected Data kyky kxkx INTERLEAVED MRSI NAVIGATORS y x 2D-FT y x (G x =G y =0)

8. METHODS: Uncorrected Data kyky kxkx INTERLEAVED MRSI NAVIGATORS 2D-FT y x y x ( G x =G y =0 )

9. METHODS:Center Frequency Drift Estimation and Correction Method Using Solvent Suppressed Navigators t = 32 s n, k x INTERLEAVED samples used to estimate  n (t) t = 0

10. METHODS: Center Frequency Drift Estimation and Correction Using MRSI Data t = 0 kyky kxkx MRSI INTERLEAVED k x =2, k y =2 k x =9, k y =7 samples used to estimate  n (t) t = 32 s 1 16 1

11. METHODS: Self Correction Using Unsuppressed Water Shift parameter estimated from MRSI when values close to ‘true’ value (Hz) Maximum signal level in order of phase encoding steps Shift parameter estimated from navigators (Hz) Uncorrected MRSISelf-corrected MRSI k x =8, k y =1, 2, …, 16

12. METHODS: Correction Strategies Using Interleaved Data samples used to estimate  n (t) INTERLEAVED NAVIGATORS MRSI Self-Correction NAVIGATORSMRSI Navigated Correction

13. METHODS: Correction Strategies Using MRSI Data Reordered CorrectionHybrid Correction k x = 8, 9, and 10 reordered in data acquisition order to provide more equally distributed high SNR k-space data for estimating center frequency drift. 5 9 13 kyky kxkx MRSI 5 9 13 kxkx kyky MRSI 1 In addition to reordering k x = 8, 9, and 10, G x and G y were set to 0 for k x = 1 to provide high SNR navigators every 16 TRs for estimating center frequency drift.

14. Water RESULTS: Comparison of Correction Strategies No correction Reordered Without Correction Reordered With Correction Hybrid Correction Navigated Correction

15. RESULTS: Comparison of Correction Strategies No correction Reordered Without Correction Reordered With Correction Hybrid Correction Navigated Correction NAA

16. RESULTS: Comparison of Correction Strategies No correction Reordered Without Correction Reordered With Correction Hybrid Correction Navigated Correction NAA Lac NAA Cr Cho mI Glu Cr mI Lac NAACr Cho mI Glu Cr mI Uncorrected Reordered correction

17. FUTURE WORK: Additional Correction Strategies Using MRSI Data Reordered Correction Using Additional Lines of k x 5 9 13 kyky kxkx MRSI 5 9 13 kyky kxkx MRSI 11 Reordered Correction Using Central Portion of k-space Error in the frequency shift parameter defined as the difference between the parameter estimated from the MRSI and the interleaved navigators. Low error is represented as white and high error as black.

18. We have demonstrated that it is possible to significantly reduce spatial artifact encountered in MRSI data resulting from fluctuations in the center frequency during MRSI data acquisition. These variations may be due to instrumentation, environmental or physiological factors In this work, the MRSI data itself was used to determine parameters used in the correction. This self-navigated correction approach is feasible when there is strong enough signal from water present in the raw data, either because residual water is intentionally left to provide a reference or because the data is acquired without solvent suppression. Future work will focus on improving the temporal resolution of the center frequency estimates by additional reordering of the high SNR points in k-space, validating the concept with multi-compartment phantoms, and evaluating the method using clinically relevant models. SUMMARY

19. Spectrum from uncorrected MRSI data Spectrum from corrected MRSI data Lac NAA Cr Cho mI Glu Cr mI Lac NAACr Cho mI Glu Cr mI