Full-wave Simulation of Complete MRI RF Coils
Many simulations of MRI RF coils have been published that provide useful insights into some aspects of the MRI RF coil optimization problem, but all involve assumptions or approximations that also limit their utility. Most published full-wave simulations have utilized codes based on the Finite Difference Time Domain (FDTD) method that generally do a poor job of accounting for the effects of real impedance matching networks, losses in slotted shields, capacitor and coil-element losses, interactions with local receive coils and (usually) radiation. Moreover, the computer resource requirements are such that exploring the effects of tuning errors or coil asymmetries is prohibitively expensive. Some have utilized codes based on Method of Moments (MoM), Finite Elements (FE), or Finite Integration Technique (FIT) that generally do a better job with coil-element losses and radiation but may have more difficulty with realistic 3D inhomogeneous tissue models or (at least with the FE approach) unrealistic computer resource requirements unless perhaps if there are extremely sophisticated mesh-refinement capabilities. The more highly developed commercial codes generally involve some sort of hybridization of several methods.
We, very recently, performed rather extensive evaluations of three of the most advanced commercial codes. (We didn’t evaluate FEKO, Vector Fields, IES, AWR, Fidelity, and many other possibilities). We have come to the conclusion that of those we evaluated, CST MWS 4.2 really stands out in front, especially for cases where a very high aspect ratio (ratio of largest to smallest cell-size) will be required and in situations where it is important to include the effects of tuning and matching circuits, coil losses, slotted HF shields, radiation, real samples, and coil asymmetries. All of these effects are often quite important, especially in high-field coils for small-animal research.
We are still in the very early stages of applying the powerful simulation and optimization capabilities of CST MWS 4.2 to our CP litzcages and linear litz coils. One early conclusion is that the (proprietary) rf-augmented Biot-Savart software COILS, which we developed in-house (mostly, six years ago, that allowed us to develop our novel linear litz coils and crescent gradient coils) is remarkably accurate and powerful for complex coils with frequency-diameter products ( fd ) up to ~15 MHz-m (e.g., a 50 mm coil at 300 MHz) with saline phantoms.
Figures below compare some field calculations from CST MWS 4.2 and COILS 6.1 for a litzcage where fd =24 MHz-m. The arrows in the CST output (top) and the directed line segments in the COILS output (lower) show the expected H field for the m=1 mode for one quadrant of the transverse (x-y) plane about 2 cm above center. (The phases here differed by 90° ). The small circles in the COILS output are approximate locations of the current elements, on the surfaces of the litzcage rungs and external shield. (In the COILS case, the external cylindrical shield was solid.) Although not evident here, the CST simulations include all details in the matching network, slotted shield, insulated crossovers, tuning capacitors, etc. The primary modes (m0, m1, m2, and any parasitics between m0 and m2) are normally predicted within ~1%, QL is generally correct within ~20% (which is comparable to experimental accuracy), and B1 is usually accurate within 10%.