4157

Repeatability of FDTD simulations for determining local SAR and B1+ distributions for UHF abdominal imaging.
Emma Doran1, Stephen Bawden1, Richard Bowtell1, Penny A. Gowland2, and Paul Glover1

1SPMIC, Physics, University of Nottingham, Nottingham, United Kingdom, 2University of Nottingham, Nottingham, United Kingdom

Synopsis

This study assessed the repeatability of FDTD simulations used to calculate local specific absorption rate (SAR) and B1+ distributions for abdominal imaging on a 7T (UHF) parallel transmit (pTx) system when subjected to typical small displacements of the an eight channel transmit dipole body array experienced when repositioning subjects. Simulations were repeated on two subject-based models and the maximum local 10g averaged SAR (SAR10g,av), B1+ distributions and scattering matrices (S-matrices) were compared to assess variation. The consequences that this has for MR safety, RF shimming and coil loading is discussed.

Introduction

The electromagnetic (EM) fields produced by individual transmit elements vary spatially and temporally on a pTx system [1] and can interact with each other in a way that leads to unexpected SAR10g,av hotspots and inhomogeneous B1+ distributions. It’s not possible to predict these distributions analytically, therefore numerical methods are employed to simulate them when assessing the safety of abdominal imaging at UHF.

Typically, simulations are run with a body array in a fixed position [2] using a pre-defined body model [3]. However, in real life the body array (and subject) do not always remain in the same position throughout the scan session due to respiration [4] and movement. In addition, the coil position will vary between separate scan sessions (see Figure 1). As such, investigating how sensitive the SAR10g,av results are to realistic displacements will give insight into the repeatability of maximum SAR10g,av and increase confidence in the values that ordinarily form the final limit for safety protocols. This study looked at the repeatability of SAR results when the body array is displaced in five separate simulations for two different subject-based models.

Methods

Simulations were performed using REMCOM XFdtd software (v7.7.1) with a NVIDIA GPU (Titan Xp). Optimised gridding was used with a minimum [maximum] cell size of 3.4x1.8x0.4 [7.8x7.9x22.8] mm3. Simulations were run to a convergence of -30 dB and took approximately 20 minutes per transmit element. A broadband excitation waveform was used. Each simulation was scaled to 1 W input power per transmit channel for SAR calculation and the B1+ fields were scaled to 3W input power per channel to match the maximum available power allowed experimentally.

One male and one female on a 3T Philips Ingenia system using mDIXON acquisition and images were manually segmented into different tissue types using Analyze software. These were labelled and constructed into 3D body models (resolution 1.5x1.5x5 mm3) encompassing the torso and arms of each subject in Matlab, then imported into the EM software.

A wire-based model [5] was used to simulate an 8TX/32RX fractionated dipole array (MRCoils). Simulations were performed with each model to generate EM fields from individual transmit elements, and were repeated four times with the body array displaced in ±x directions and conformed around the body at the new position. The EM fields and additional information were exported to Matlab to calculate the SAR10g,av and B1+ profiles for the worst-case scenario where all fields are in-phase and there is maximum constructive interference. S-matrices were also exported for each simulation. An example of a simulation is shown in Figure 2(a) and array displacements are summarised in Figure 2(b).

Results

Figure 3(a) shows the final SAR results from each simulation. Figure 3(b) shows the location of the SAR10g,av hotspots from each simulation as they varied. The mean maximum SAR10g,av value was 2.0±0.3Wkg-1 (16% variation) and 1.2±0.1Wkg-1 (13% variation) for male/female simulations respectively. For the location of the hotspot in each model the other models were all within 15% of their maximum SAR10g,av value. Figure 4 displays the B1+ in the slice containing the maximum SAR10g,av value. Figure 5 shows the S-matrices from each simulation and the absolute difference matrices showing the change between consecutive simulations. The maximum coupling differences for each simulation were 8.2 and 7.3dB.

Discussion and Conclusion

The SAR results showed good repeatability except for the changes in location of the SAR10g,av hotspots. This is to be expected if the array was reoriented around the geometry of each body model. The variation in maximum SAR10g,av values due to array displacement suggests that repeatability should be taken into account as an additional safety factor [6] when performing SAR assessments based on FDTD simulations. The B1+ profiles didn’t vary significantly between each simulation, although intensity did increase on the model surface in some cases where the array was repositioned. The similarity of B1+ interference patterns provides confidence that RF shimming calculations are robust against small movements. The maximum differences in S-matrices implies there is limited variation in coupling between array movements and the loading of the body array is not significant. The variation in the S-parameters due to array movement are nonetheless much larger than changes due to physiological motion measured by Hess et al, and Jaeschke et al [7, 8].

Acknowledgements

We gratefully acknowledge the support of NVIDIA Corporation with the donation of the Titan Xp GPU used for this research.

References

  1. Padormo et al. Parallel transmission for ultrahigh-field imaging, NMR Biomed. 2015
  2. Raaijmakers et al. The Fractionated Dipole Antenna: A New Antenna for Body Imaging at 7 Tesla, MRM. 2016;75:1366-1374
  3. Homann et al. Toward Individualized SAR Models and In Vivo Validation, MRM. 2011;66:1767-1776
  4. Doran et al., Effects of Respiration on B1+ and SAR in Whole-Body Imaging at 7 Tesla, In Proc. ISMRM (Paris). 2018
  5. Bawden et al. Electrical lengthening to improve electromagnetic simulations and SAR calculations of meandered body dipole elements at 7 T, In Proc. ISMRM (Paris). 2018
  6. Le Garrec et al. Probabilistic Analysis of the Specific Absorption Rate Intersubject Variability Safety Factor in Parallel Transmission MRI, MRM. 2017;78:1217-1223
  7. Hess et al. Diaphragm Position Can Be Accurately Estimated From the Scattering of a Parallel Transmit RF Coil at 7 T, MRM. 2018;79:2164-2169
  8. Jaeschke et al. Cardiac Gating Using Scattering of an 8-Channel Parallel Transmit Coil at 7T, MRM. 2018;80:633-640

Figures

In vivo images acquired at 7 Tesla (Philips Acheiva System) with an 8 TX 32 RX transmit body array. Image on left acquired during renal scanning session, image on right acquired during liver magnetisation transfer session. Fiducial markers placed in the body array casing indicate the placement of the transmit elements around the subject. Note the subtle difference in location of each element between the two images.

(a) Female model in XFdtd with body array. Red crosses mark the dipole locations identified in the table below. (b) Displacements in x (∆x) vary between 26.0 and 30.8 mm. Displacements in y (∆y) vary between 2.8 and 5.1 mm. There was no displacement of the body array in the z-direction (∆z = 0).

(a) Male model results are in rows 1 and 2. These display the raw SAR and local 10g averaged SAR respectively. Female model results are in rows 3 and 4. These display the raw SAR and local 10g averaged SAR respectively. (b) Maximum intensity projections of the local 10g averaged SAR for each model (male on left, female on right) shown in transverse, coronal and sagittal orientations respectively. The location of the SAR hotspot from each of the five simulation repeats is shown on each model.

B1+ profiles shown for slice containing the maximum local SAR value in each simulation. Results from male model are on the top row, and the bottom row shows results from the female model.

S-matrices from male and female simulations are shown in rows 1 and 3 respectively. There is limited coupling between posterior (1-4) and anterior (5-8) elements. However, coupling exists between elements within the same array. This is consistent with S-matrices acquired in vivo. The differences between consecutive simulations are shown in rows 2 and 4. The greatest change consistently comes from anterior elements. These tended to experience the greatest reorientation when re-adjusting elements around the body model after displacing the array by small increments.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)
4157