J. Nuno Teixeira^{1}, Felipe Godinez^{1}, Shaihan Malik^{1}, and Jo V. Hajnal^{1}

MRI with non-MR compatible objects, namely wires, poses several challenges as risks of high Local SAR and reduced image quality due to B1 enhancement. In this work we looked at previously proved current nulling techniques and at their relationship with the reduction of worst case SAR scenarios. Results showed that current nulling leads to large reductions in worst case SAR, but that performance varies with coil design and sensor location. The analysis could be used to aid design of experimental setup for maximum safety.

Interventional MR and the use of non-MR compatible invasive objects (hereafter referred to as ‘wires’) are usually associated with risks of high local SAR and poor image quality due to B1 enhancement.

Both risk of local hotspots and imaging disruption originate from uncontrolled currents on the wire which are minimised if the wire is decoupled from the transmit system. This is achievable with a parallel transmit architecture combined with sensing wire currents so that transmit weights to apply to each channel can be found which decouple the wire from the RF fields, thus nulling the current^{1}.

Previous demonstrations of the technique have found that full nulling often requires sensing of currents at multiple locations along the wire. Complex current behaviour has previously been investigated experimentally. In this work we used EM simulations to investigate how the nulling approach correlates to the worst case local SAR at the tip of the inserted wire.

Current sensor ‘nulling’ approaches have used singular value decomposition (SVD) on coupling matrices constructed from one or more sensors; the singular vectors identify modes of the transmit coil that couple to the wire, and the singular values reflect their strength. Meanwhile SAR simulations are usually analyzed using the ‘Q-matrix’^{2,3} approach. In this case the maximum eigenvalue of a given Q-matrix reflects the worst case SAR at the corresponding spatial location. We used EM simulations (Sim4Life; Zurich MedTech, Zurich, Switzerland) to explore commonalities between these two approaches.

Three scenarios were investigated using two RF coils: 1) short 8-channel Tx/Rx coil on a rectangular phantom (32x10x80cm; $$$\sigma=0.4 S/m$$$; $$$\epsilon=30$$$); 2) long 8-channel Tx/Rx coil - same phantom; 3) long coil on a human model (Duke - ITIS ViP3.1), all at 64MHz (Fig.1). Previous work at 3T suggested that a long coil would be better for nulling currents. In each simulation a wire (core: $$$\varnothing=1mm$$$; insulating layer: 0.1mm thickness) with the distal 5mm stripped of insulation was inserted into the phantom/body. To reduce computation time, the Huygens Box method, implemented in Sim4Life, was applied to all simulations in a region surrounding the wire. Analysis was performed in MATLAB.

Pointwise Q-matrices and surface current on the wire (J) were extracted from the models. Q-matrices were normalised to the average B1+ under each coil when driven in a nominal quadrature configuration (to account for differences in efficiency). Eigenvalue decomposition of a Q-matrix from a point near the tip of the wire was used to determine worst case SAR configuration.J was sampled at points along the wire in order to synthesize current sensor readings. A number of different sensing configurations were tested, including different combinations of one or two sensing locations inside and outside the phantom/body. For each configuration, the coupling modes were identified using SVD, and then the Q-matrix was projected onto the sub-space defined by excluding one or more of the strongest coupling modes. Eigenvalue decomposition of this projected Q-matrix helps to quantify the remaining worst case SAR at the wire tip (rwcSAR_{TIP}).

1. Etezadi-Amoli M, Stang P, Kerr A, Pauly J, Scott G. Controlling radiofrequency-induced currents in guidewires using parallel transmit. Magn Reson Med. 2015;74(6):1790-1802. doi:10.1002/mrm.25543.

2. Graesslin, I., Homann, H., Biederer, S., Börnert, P., Nehrke, K., Vernickel, P., Mens, G., Harvey, P. and Katscher, U. A specific absorption rate prediction concept for parallel transmission MR. Magn Reson in Med. 2012; 68(5):1664-1674. doi: 10.1002/mrm.24138

3. Eichfelder, G. and Gebhardt, M. Local specific absorption rate control for parallel transmission by virtual observation points. Magn Reson in Med. 2011; 66(5): 1468-1476. doi: 10.1002/mrm.22927

Figure 1 - Geometric disposition of the three scenarios simulated with Sim4Life. (top) Short Coil on a phantom; (middle) Long Coil on a phantom; (bottom) Long Coil on Duke.

Figure 2 - Short Coil on the Dielectric Phantom: (top) Relative position of the sensing points (used in the SVD) and the Wire and Coil locations; (bottom) Remaining rwcSAR_{TIP} after projecting the normalised Q-Matrix on the several decoupling mode sub-spaces for each sensing point.

Figure 3 - Long Coil on the Dielectric Phantom: (top) Relative position of the sensing points (used in the SVD) and the Wire and Coil locations; (bottom) Remaining rwcSAR_{TIP} after projecting the normalised Q-Matrix on the several decoupling mode sub-spaces for each sensing point.

Figure 4 - Long Coil on Duke: (top) Relative position of the sensing points (used in the SVD) and the Wire and Coil locations; (bottom) Remaining rwcSAR_{TIP} after projecting the normalised Q-Matrix on the several decoupling mode sub-spaces for each sensing point.

Figure 5 - Comparison for the three scenarios simulated of the efficiency of the two relevant sensing method for rwcSAR_{TIP} reduction. (top) using one sensing point and considering the first mode as a coupling mode; (bottom) using two sensing points and considering the first two modes as coupling modes.