High temporal resolution retrospective motion and $B_0$ correction using FIDNavs and segmented FatNavs at 7T.
Frédéric Gretsch1, Tobias Kober2,3,4, Maryna Waszak2,3,4, José P. Marques5, and Daniel Gallichan1

1CIBM, EPFL, Lausanne, Switzerland, 2Advanced Clinical Imaging Technology (HC CMEA SUI DI BM PI), Siemens Healthcare AG, Lausanne, Switzerland, 3Department of Radiology, University Hospital (CHUV), Lausanne, Switzerland, 4LTS5, EPFL, Lausanne, Switzerland, 5Donders Centre for Cognitive Neuroimaging, Radboud University, Netherlands

### Synopsis

FID navigators (FIDNavs) and a few lines of highly accelerated double-echo 3D fat navigators (FatNavs) were measured each TR of a high-resolution 3D-GRE acquisition. High temporal resolution $B_0$ variations and motion parameters could be estimated by the FIDNavs using the lower temporal resolution FatNavs to derive calibration data. The cardiac cycle pattern emerged clearly in these estimates and retrospectively corrected images were of clearly improved quality, thereby demonstrating the potential for this hybrid approach.

### Introduction

MRI at very high spatial resolution is challenging due to several limitations, including lower SNR, increased sensitivity to motion and longer scan-times. For T2*-weighted imaging of the brain there is the further complication of artifacts arising due to changes in the $B_0$-field with the respiratory cycle. Fat navigators (FatNavs) have already demonstrated excellent motion tracking and retrospective correction at 7T [1] and we have also shown their sensitivity to $B_0$-field tracking [2]. Extremely rapid FID navigators (FIDNavs) have previously demonstrated to be sensitive to $B_0$ changes [3] and to motion [4,5], but the mapping between the FID signals and the $B_0$ and motion parameters requires per-session calibration.

Here we propose a GRE sequence with hybrid integration of segmented dual-echo FatNavs and FIDNavs which allows the low temporal resolution dual-echo FatNavs to calibrate the FIDNav mapping for each session. This allows the estimation of high temporal-resolution motion (6 parameters) and zeroth and first-order $B_0$ variations (4 parameters) which can be used for retrospective correction of the T2*-weighted GRE k-space.

### Method

Using a 7T head-only MR scanner (Siemens Healthcare, Germany) and a 32-channel RF coil (Nova Medical Inc.), we scanned a healthy volunteer who was asked to breathe deeply to amplify the $B_0$ variations. His cardiac and respiratory traces were recorded by the external monitoring unit during the high-resolution slab-selective 3D-GRE. An FID readout (FIDNav) was inserted between the excitation pulse and the phase encoding gradients. After each host sequence TR, three lines of a highly accelerated double-echo FatNav were acquired, resulting in an entire FatNav volume every 36 TRs (~2s). Figure 1 summarizes the sequence design and protocol parameters used. The effective TR (time between two host excitation pulses) was 56.2 ms. GRAPPA reference lines for FatNavs were acquired in a 2.3s pre-scan. Total acquisition time: 32min25s, whereas navigator-free would have been 23min2s (although SNR efficiency is similar due to increased effective TR for water).

FatNavs were retrospectively co-registered using SPM to obtain motion-estimates, and a $B_0 = \beta_0 + \vec \beta \cdot \vec x$ fit (constrained to the host sequence FOV) was computed. Each FIDNav was assigned to the nearest FatNav, a linear mapping $$M = \Lambda F$$ was determined, where $M$ are the motion and field parameters and $F$ the mean FIDNav signals for the given FatNav volume for each receive channel. This linear mapping is expected to hold for sufficiently small variations [4]. Once $\Lambda$ was found it was reapplied to each individual FIDNav to obtain higher temporal resolution motion estimates. To observe cardiac related variations in these estimates, each FIDNav and its time-to-peak was mapped to the closest cardiac peak (from external monitoring), and a binning (150 bins) of the global time-to-peak range allowed a bin-wise mean of the estimated motion and $B_0$ field estimates.

For retrospective correction, the motion parameters and field coefficients were temporally smoothed via a sliding-window of 7 FIDNavs. Hence the effective temporal resolution of these estimates was approximately 400 ms. The correction itself was performed coil-wise using the NUFFT adjoint operator [6].

### Results and discussion

Temporally smoothed FIDNav-estimated motion parameters are shown in Figure 2.A. One can readily observe the clear correlation between the cardiac peaks (dashed vertical line) and the motion estimates. Figure 2.B shows the binning result for translations and linear $B_0$ coefficients. As expected, the maximal cardiac effect appears slightly before the peak in the physiological data, as the pulse was measured on the right index finger where its arrival time is later than for the brain. Comparison to a similar measurement of the ballistocardiogram obtained from a camera/marker system [7] is encouraging, although we observe the strongest effect on x and y axes in this example – rather than the z direction. It is likely that the direction and amplitude of the effect is strongly subject-dependent, and also affected by the position of the subject in the scanner. Further studies will allow quantification of the associated variability.

The quality of the corrected image is greatly improved, as exemplified in Figures 3 and 4 (zoomed section), where high-resolution improvements in features are clearly identifiable, implying that the motion estimates are of reasonable accuracy. Global improvement is especially visible in a minimal intensity projection contrast over a 9mm slab (Figure 5) where the diffuse blotches associated with respiratory artifacts are greatly reduced.

### Conclusion

FIDNavs calibrated by segmented FatNavs were able to track both head motion and field variation at high temporal resolution. Retrospective corrections based on to these estimates showed considerable improvement in image quality thereby demonstrating the potential of this new hybrid approach.

### Acknowledgements

This work was supported by CIBM of the UNIL, UNIGE, HUG, CHUV, EPFL and the Leenaards and Jeantet Foundations, as well as SNSF project number 205321_153564.

### References

[1] Gallichan D. et al., “Retrospective correction of involuntary microscopic head movement using highly accelerated fat image navigators (3D FatNavs) at 7T: 3D FatNavs for High-Resolution Retrospective Motion Correction”, Magnetic Resonance in Medicine, Apr. 2015.

[2] Gretsch F. et al, "Investigating the potential of highly accelerated FatNavs for dynamic shimming", ISMRM 23rd Annual Meeting, 2015.

[3] Splitthoff D. N. and Zaitsev M., “SENSE shimming (SSH): A fast approach for determining B 0 field inhomogeneities using sensitivity coding,” Magnetic Resonance in Medicine, vol. 62, no. 5, pp. 1319–1325, Nov. 2009.

[4] Babayeva M. et al., “Accuracy and Precision of Head Motion Information in Multi-Channel Free Induction Decay Navigators for Magnetic Resonance Imaging”, IEEE Transactions on Medical Imaging, vol. 34, no. 9, pp. 1879–1889, Sep. 2015.

[5] Kober T. et al, “Head motion detection using FID navigators”, Magnetic Resonance in Medicine, vol. 66, no. 1, pp. 135–143, Jul. 2011.

[6] Fessler J. A. and Sutton B. P., “Nonuniform fast Fourier transforms using min-max interpolation,” Signal Processing, IEEE Transactions on, vol. 51, no. 2, pp. 560–574, 2003.

[7] Maclaren J. et al., “Measurement and Correction of Microscopic Head Motion during Magnetic Resonance Imaging of the Brain.”, PLoS ONE 7 no. 11, 2012.

### Figures

Figure 1: A. Sequence kernel : an FID readout (FIDNav) was acquired after the host sequence (gradient spoiled GRE) excitation pulse. A segmented FatNav module is inserted before the next host sequence pulse and consisted of three readout lines of a 4x4 GRAPPA accelerated FatNav protocol.

B. Protocol parameters.

Figure 2: A. Temporally smoothed motion parameters (blue, red, yellow lines) estimated from the FIDNavs. Top : translations, bottom: rotations. Physiological data acquired by the monitoring unit is indicated in black : dashed vertical lines: cardiac cycle peaks, black lines in the subplots: respiration trace (in arbitrary units).

B. High-pass filtered translations and B0 variation linear coefficients estimated from the FIDNavs, as a function of their time to the closest cardiac cycle peak.

Figure 3: Two different slices (top row, bottom row) of the uncorrected (left) and the B0 and motion corrected reconstructions (right).

Figure 4: Zoomed in view of both uncorrected (left) and motion and B0 corrected reconstructions. Light-blue arrows indicate clear artifact reduction in the corrected image.

Figure 5: Minimal intensity projection of a 9mm slab: uncorrected reconstruction on the left, motion and B0 corrected reconstruction on the right.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
4251