The aim of this simulation study was to evaluate the performance of parallel-transmit head transmit RF coils compared with a circularly polarised (quadrature) and dual-drive (pTx) head birdcage coil on a computational human model without and with implanted bilateral DBS, with a focus on excitation fidelity under SAR control for standard brain imaging targets. Our computations compellingly confirm that multi-row pTx coils: (1) can produce more homogeneous transmit fields than the birdcage coil and can operate at lower local SAR when there are no DBS; (2) can maintain significant benefit by both measures in a straight comparison with DBS; and (3) can achieve local SAR levels in the presence of electrodes that are similar to a quadrature birdcage without any implanted electrodes without sacrificing homogeneity.

RF shimming was performed to assess the coils, optimising for B1+ homogeneity over the whole brain (rather than over a single slice as is sometimes done) with a peak local SAR penalty term. L-curves, in which the relative weighting of the SAR term was varied (Fig. 4), demonstrated the trade-off between excitation error for B1+ homogeneity and local SAR for each coil design. In the absence of DBS electrodes, overlapping coil elements, in general, provided a better strategy than using non-overlapping elements, with an exception seen in the single-row pTx coil where the L-curves crossed. In the simulations, while the non-overlapping coil designs utilised planar loops, the overlapping coils were forced to employ a curved cylindrical geometry for all elements. Since the observed performance of the non-overlapping designs was inferior to their corresponding overlapping designs, the non-overlapping designs were not re-modelled using the fully curved geometry and were not modelled with DBS added. Furthermore, it has been verified that the positioning of the ports on the birdcage coil, whether in their current position or in the conventional position offset symmetrically at the back of the head, does not affect the SARmax,1g avg or the COV.

When comparing the L-curves of the coils without DBS, all pTx coils outperformed the pTx and the quadrature birdcage coil. The pTx coils demonstrated improved homogeneity and lower SAR values, enhancing safety and efficiency. Upon introducing the DBS devices, the L-curves for all tested pTx coils shifted upwards to the right, indicating increased inhomogeneity, and heightened local SAR values, although all still achieved more favourable performance than the birdcage. Single-row coil performance was in agreement with the findings by McElcheran et al. [25], who compared an 8-channel single-row coil with the birdcage coil. It was also observed that using the birdcage coil in pTx mode significantly improves the balance between B1+ homogeneity and local SAR compared to the birdcage coil in quadrature mode. In the no-DBS scenario, the pTx birdcage coil performs similarly to the pTx coils in the lower local SAR and higher COV regime, but when the DBS is in situ, the pTx coils significantly outperformed the pTx birdcage coil.

A striking result is that the multi-row pTx coils could produce more homogeneous fields at slightly higher peak SAR with DBS than is achieved with the birdcage without DBS. Although peak local SAR was markedly reduced, the head average SAR was similar for all coils for all conditions tested (~ 0.5 W/kg lower than the birdcage for multi-row coils and ~ 0.1 W/kg higher for the single-row pTx coil). The capability to lower local SAR even when DBS electrodes are present could be transformative for the brain imaging of this patient group. As an aside, we note that after shimming, the local SAR distributions of pTx designs presented here show peak local SAR values elsewhere in the head and not at the tip of the electrodes (Figs. 5,6,8 and 10), which indicates the importance of a whole head volume SAR assessment.

The triple-row array might have been expected to provide superior RF performance due to its larger number of channels but only showed a minor efficiency benefit compared to the double-row array. Its increased complexity may contribute to higher manufacturing costs and implementation challenges.

A previous study of pTx coils with DBS [26] explored whether adding complexity to the modelled electrodes alters the predicted SAR for a double-row coil across various human and DBS models, and did not find this to be an important aspect. To check stability of results for both our deployed DBS and human model, the double-row coil, which provides optimal performance comparable to the triple-row in this study, is simulated in two additional scenarios: (1) on the Duke model, with added extracranial looping, and (2) on the Ella model, using a different trajectory, also with extracranial looping. It is demonstrated that for the double-row coil, the addition of extracranial looping does not affect B1+ homogeneity and SARmax,1g avg values for the Duke model (Fig. 9), while for the Ella model, comparable B1+ homogeneity is achieved, and SARmax,1g avg is reduced (Fig. 10). This helps to reinforce Guerin's findings [26], confirming that the performance of multi-row coils is consistent across different models and unaffected by the specifics of electrode placement or model complexity. In contrast, when we tested this additional trajectory scenario using the birdcage coil, the SARmax,1g avg varied significantly, up to 68%, from one trajectory to another, which aligns with the literature [41]. This unpredictability of the birdcage coil can pose a safety issue, which is mitigated by multi-row pTx coils.

In agreement with the findings presented by McElchearan et al. [23,24,25], Eryaman et al. [22], and Guerin et al. [26], our study demonstrates that single-row pTx coils outperformed the birdcage coil in the tested configurations. Our research is consistent with the study by Guerin et al. [26], which showed the benefit of two-row pTx head coils over single-row pTx and birdcage coils. In their findings, the heating around the electrodes was reduced by a factor of over 100 when using the double-row head coil while keeping the same level of transmit field homogeneity at the selected slice. A complementary aspect is that they emphasised simulating a broader range of body models [26], while our study focussed on simulating a greater variety of head coil designs.

The pTx coil designs considered are complex and it is challenging to achieve decoupling between individual coils with multiple close neighbours. The achieved coupling coefficients, between -6 dB and -8 dB, could raise a concern in this regard. However, the strong localisation and diversity observed in the transmit fields of the individual channels (Online Resource 1) confirms that coil coupling has not undermined performance. Furthermore, the placement of the voltage sources to the top and left-hand side of the coil elements only resulted in changes in Sii, max and Sij,max parameters of less than 1dB. Another potential risk is that errors in the measured in the coil B1+ maps or prescribed shims might lead to uncontrolled deviations in performance. To explore this, we performed a test of the sensitivity of the double-row coil with DBS electrodes in situ to phase and amplitude errors in shim measurement or setting. We performed 100 trials, randomly perturbing the optimised shim values by ± 5% in amplitude and phase independently on all channels. This resulted in a maximal increase in SARmax,1g avg of 0.0625% and a maximal reduction of B1+ homogeneity of 0.3%, supporting a conclusion that performance is robust at least to this level of error.

The simulations performed in this study had some specific limitations. In Figs. 5,6, 8 and 10, with DBS electrodes in place, the optimal shim solutions often have secondary local peaks in SAR at the distal end of the wires in the chest wall, even though the local fields in this location are much lower (~ 70 times) than in the head. We hypothesised that this is an artefact associated with a lack of explicit inclusion of the IPG to provide a termination for the distal end of the DBS lead. To examine this further, one additional simulation for the double-row coil was executed, incorporating an IPG with a conductive case to the inner surface of which the DBS lead's inner conductor is terminated via a 2 kΩ resistor. It was observed that this refinement resulted in only slight changes to field distributions, with a 6% increase in the COV of B1+ and maximum local SAR, combined with complete eradication of a local peak of SAR in the chest. This provides confidence that the use of more refined models that include the IPG would not change the conclusions of this study. While this study shows substantial potential benefits of pTx for controlling the safety of MRI with implanted electrodes, there remain many challenges in realising these benefits in operational practice. Eryaman and others [19,20,21,22,23] have started to address this challenge, but much is yet to be done. Our results add further motivation.

In conclusion, PTx coils with multiple rows can achieve more homogeneous B1+ and lower SAR around DBS electrodes at 3T compared to birdcage coils. Together these can achieve SAR efficiency (mean B1+/\(\surd\) SAR) that could rival performance in the presence of DBS electrodes of existing coils in the absence of such implants (conventional imaging regime). Single-row pTx coils provide much reduced performance gains. These findings indicate that multi-row pTx coils could offer significant benefits for MRI subjects with DBS electrodes. This provides further motivation for exploring the key challenge of achieving robust deployment of this exciting technology so that its demonstrable potential gains can be delivered in practice for patient benefit.