0207

Experimental Assessment of the Effects of Subject Motion on Local SAR and pTx Pulse Performance at 7T

Sydney Nicole Williams¹, Paul McElhinney¹, Belinda Ding^1,2, Sarah Allwood-Spiers³, David A. Porter¹, Shajan Gunamony^1,4, and Wyger Brink⁵
¹Imaging Centre of Excellence, University of Glasgow, Glasgow, Scotland, ²Siemens Healthcare Ltd., Frimley, United Kingdom, ³NHS GGC MRI Physics, NHS Greater Glasgow & Clyde, Glasgow, Scotland, ⁴MR CoilTech Ltd., Glasgow, Scotland, ⁵Magnetic Detection & Imaging Group, University of Twente, TechMed Centre, Netherlands

Synopsis

Keywords: Parallel Transmit & Multiband, Safety, RF Pulse Design, RF Coils

Previous simulation studies of motion in parallel transmission (pTx) have identified large increases in specific absorption rate (SAR) and pulse performance error. This abstract extends this by evaluating the effects of motion on RF shimming and dynamic pTx in vivo at 7T. The T1w image of a healthy volunteer was segmented to create a body model for electromagnetic simulation in a custom pTx coil. The model was simulated at six head positions matched to experimental measurements. Motion had a detrimental effect on pTx pulse performance and also caused local SAR changes, though not as severe as seen previously in simulation.

Introduction

With increasing prevalence of ultra-high field (UHF) MRI, parallel transmission (pTx) is gaining popularity as an important tool for mitigating RF field (B₁⁺) inhomogeneity. Recently, fully-integrated or calibration-free methods such as Universal Pulses (UPs)¹, Fast Online CUStomized (FOCUS) pulses², and Direct Signal Control (DSC)³ have improved the cumbersome pTx workflow and broadened usage.

The effects of subject motion on pTx performance have recently been investigated in comprehensive simulation studies.^4,5,6,7,8 These works reported substantial degradation of pTx pulse fidelity with increases in normalized root mean squared error (NRMSE) of up to ~35%⁷ and increases in local specific absorption rate (SAR) of up to ~300%.⁸ Currently, such SAR variations need to be accounted for in the SAR model, for example by assuming the worst-case eigenvalue or as virtual observation points (VOPs)⁹ in conjunction with further safety factors. Despite improvements in B₁⁺ estimation via deep learning⁸, updating the pTx pulse design to address motion-induced degradation remains an open challenge.

So far, the UHF community has relied heavily on generic body models and relatively wide motion ranges to assess local SAR variability due to head position variations. Alternatively, introducing improved segmentation schemes tolerant to B₁⁺ image shading can provide accurate subject-specific SAR estimates in the head¹⁰. Using a subject-specific electromagnetic (EM) body model, this abstract presents preliminary work that improves upon previous simulation-only studies by assessing pTx performance with actual subject motion for the first time in vivo at 7T.

Methods

Following informed consent, a healthy volunteer was scanned in a 7T MRI system (MAGNETOM Terra, Siemens Healthineers, Germany) using a custom 8Tx/32Rx head coil¹¹ and with the subject positioned in the center of the coil. First, a T1w 3D MPRAGE scan (TR/TE/TI=2500/2.2/1050ms, FA/GRAPPA/FOV/Res=5॰/R=2/250x250x192mm³/1mm³, TA=5:45min) was used to create an individualized EM body model of the head following the method in Ref. 7. Additionally, single-channel B₁⁺ maps were collected with a presaturated TurboFLASH method¹² (TR/TE =7340/1.6ms, FA/FOV/Res/Slice Thickness=5॰/250x250mm²/5mm²/4mm, TA=1:31min).

The subject was then asked to move along six prescribed motion directions, i.e. along the roll, yaw, and pitch rotation axes (shown in Figure 1b), to the maximum capacity allowed by the tight-fitting receive helmet. In each position, a shorter T1w-MPRAGE scan was acquired (TA=2:00min, Res=2mm³, all other parameters constant) along with single-channel B₁⁺ maps. All anatomical scans were performed in CP mode.

Rigid registration of T1w images was performed in FSL¹³ using six degrees-of-freedom to identify rotations and translations with respect to the reference data acquired in center position. These transformations were then imposed on the subject-specific EM body model in CST Microwave Studio Suite (Dassault Systems, France) to simulate B₁⁺ and Q-matrices¹⁴ for all head positions.

Experimental B₁⁺ maps were used to assess motion effects on pTx pulse performance in terms of NRMSE. First, slice-specific RF shims for all transverse slices were optimized for the center position and then evaluated on all head positions, along with CP excitation and a position-specific RF shims set. Next, non-selective excitation and inversion pulses were designed for the center position with the FOCUS method² and again evaluated for all head positions.

Finally, local SAR effects were evaluated in all head positions for both the RF shims and non-selective pTx pulses described above.

Results

Figure 2 compares the simulated and measured B₁⁺ maps for the subject positioned in the center of the coil.

Figure 3 shows NRMSE performance of slice-specific shimming in the center and 6 offset positions, while Figure 4 shows the corresponding results for the 3D excitation and inversion pulses.

Figure 5 shows the simulated subject-specific local SAR for the center position and offset head positions A, C, and E, for the same slice-specific shims and the 3D dynamic FOCUS pulses evaluated in Figures 3 and 4.

Discussion & Conclusion

This experimental analysis of motion effects on pTx with a subject-specific EM body model showed similar results to previous simulations with substantial degradation of pTx pulse performance⁷. For slice-by-slice shimming, each offset head position produced NRMSE values higher than CP excitation at center position. Meanwhile, the NRMSE in corresponding dynamic pTx pulses increased by 2-5x (Fig. 4). With the continued effort in pTx pulse design, this detriment is an important consideration for all pTx users. With the increased usage of pTx and likelihood of less compliant subjects and patient exams, these results warrant further simulation and in vivo studies.

Changes in SAR were seen for varying head positions (Fig. 5). Importantly, the local SAR distribution decreased when slice-specific RF shims were re-optimized. This is likely due to the reduced activation of elements closer to the head, compared to using center-optimized settings. For dynamic pTx, different head positions exhibited different local SAR across time with the worst position and time point creating 16% higher SAR for excitation and 17% for inversion. Encouragingly, this is well-below the SAR increases reported previously⁸, which might reflect the more realistic motion ranges considered here.

One limitation is the match between the simulated and measured B₁⁺ maps, yet future work will refine this agreement with additional segmentation model training. Furthermore, the computation time required for EM simulations, currently amounting to several 10s of hours, is a substantial bottleneck in evaluating additional body models and head positions.

Acknowledgements

*SNW and PM contributed equally to this work.We would also like to thank the radiographers from the Imaging Centre of Excellence, University of Glasgow.

References

1. Gras V, Vignaud A, Amadon A, et al. Universal pulses: a new concept for calibration-free parallel transmission. Mag. Reson. Med. Feb. 2017. Vol 77:2, p. 635-643.

2. Herrler J, Liebig P, Gumbrecht R, et al. Fast online-customized (FOCUS) parallel transmission pulses: a combination of universal pulses and individual optimization. Mag. Reson. Med. Jun. 2021. Vol 85:6, p. 3140-3153.

3. Tomi-Tricot R, Sedlacik J, Endres J, et al. Fully integrated Scanner implementation of direct signal Control for 2D T2-weighted TSE at ultra-high Field. Proc Int Soc Mag Reson. May 2021. p. 0621.

4. de Greef M, Ipek O, Raaijmakers A, et al. Specific absorption rate intersubject variability in 7T parallel transmit MRI of the head. Mag. Reson. Med. May. 2020. Vol 69:5, p.1476-1485.

5. Kopanoglu E, Deniz C, Erturk MA, and Wise RG.Specific absorption rate implications of within-scan patient head motion for ultra-high field MRI. Mag. Reson. Med. Nov. 2020. Med. Vol 84:5, p. 2724-2738.

6. Ajanovic A, Brackenier Y, Tomi-Tricot R et al. Motion and pose variability of SAR estimation with parallel transmission at 7T. Proc Int Soc Mag Reson. May 2021. p. 2487.

7. Plumley A, Watkins L, Treder M, et al. Rigid motion-resolved B₁⁺ prediction using deep learning for real-time parallel-transmission pulse design. Mag. Reson. Med. Dec. 2021. Vol 87, p. 2254–2270.

8. Plumley A, Goodwin N, and Kopanoglu E. Inter-subject differences in SAR sensitivity to motion with parallel-transmit at 7T. Proc Int Soc Mag Reson. May 2022. p. 0457.

9. Eichfelder G and Gerbhardt G. Local specific absorption rate control for parallel transmission by virtual observation points. Mag. Reson. Med. Nov. 2011. Vol 66:5, p. 1468-1476.

10. Brink WM, Yousefi S, Bhatnagar P, et al. Personalized local SAR prediction for parallel transmit neuroimaging at 7T from a single T1-weighted dataset. Mag. Reson. Med. Mar. 2022. Vol 8, p. 464-475.

11. Williams SN, Allwood-Spiers S, McElhinney P, et al. A nested eight-channel transmit array with open-face concept for human brain imaging at 7 tesla. Front Phys. Jul 2021. Vol 9.

12. Fautz HP, Vogel M, Gross P, et al. B₁ mapping of coil arrays for parallel transmission. Proc Int Soc Mag Reson. May 2008. p. 1247.

13. Jenkinson M, Bannister P, Brady, JM and Smith SM. Improved optimisation for the robust and accurate linear registration and motion correction of brain images. NeuroImage. Oct. 2002. Vol 17:2, p. 825-841.

14. Graesslin I, Homann H, Biederer S, et al. A specific absorption rate prediction concept for parallel transmission MR. Mag. Reson. Med. Jan. 2012. Vol 68, p. 1664-1674.

Figures

Figure 1. A) Visualization of the measured subject data and the registration to their model in the EM solver CST Microwave Studio Suite for placement in the pTx coil. B) In vivo measurements of head positions (A-F) used for pTx performance and SAR analysis compared to the center position within the coil.

Figure 2. Comparison of simulated and measured B₁⁺ maps from original head position. After correction using previous Actual Flip angle Imaging³ calibrations, the measured peak B₁⁺ values are comparable, yet spatial B₁⁺ pattern faces discrepancies. Further refinement of the segmented subject model can improve the match.

Figure 3. Effect of motion on slice-specific RF shim fidelity. Each boxplot represents the spread of NRMSE values across all slices with either slice-specific RF shims or CP excitation. Different color groups represent different head positions, with center position on the far left. For all positions, both the center-optimized shims and re-optimized shims were simulated.

Figure 4. Excitation fidelity for dynamic pTx pulses designed on the center position and then simulated on all positions. Left): Excitation pulse RF magnitude and SPINS k-space trajectory. Right): Inversion pulse RF magnitude and kT-points k-space trajectory. Bottom row): Simulated NRMSE for all positions for both excitation and inversion pulses, normalized to the NRMSE obtained in the centered position (0.11 and 0.07 for excitation and inversion, respectively). Position E was worst for excitation (5.25x higher NRMSE), and Position F was worst for inversion (2.08x higher NRMSE).

Figure 5. Local SAR evaluations for slice-by-slice shimming and non-selective pTx excitation and inversion pulses for center position and positions A, C, and E. A) Local SAR for 1 W input power across all EM voxels for all slice-specific shim optimizations. X-labels are of the form [SAR MODEL]/[SHIM SET]. B) Local SAR across time for the evaluated positions for the 3D excitation (left) and inversion (right) pulses. Here, local SAR is normalized by the max value of the center position for a given pulse.

Proc. Intl. Soc. Mag. Reson. Med. 31 (2023)

0207

DOI: https://doi.org/10.58530/2023/0207