Introduction

Patient-induced B₀ inhomogeneities is a significant source of image and spectroscopic data degradation at 7T. To counter this, scanners are typically equipped with additional shim coils, designed to generate spherical harmonic-varying fields across the entire imaging volume. However, for whole-brain MR, adequate shimming of the fronto-temporal region for low-bandwidth sequences (e.g EPI readout used for fMRI and DWI), T₂^* weighted sequences and spectroscopy remain a challenge.
Shim coils designed to locally improve B₀ homogeneity are attractive as a supplement, as they facilitate specifically addressing areas such as described above. Dynamic updating of the generated shim field is in addition feasible without inducing noteworthy eddy currents. Several local-shim approaches have been shown for improved brain shimming, including a six channel, variable positioning setup [1], a 48-channel combined RF and shim coil [2], and a 32-channel combined RF and shim coil [3] with additional face loops [4]. These setups showed great shimming capabilities but relied on dedicated and complex RF coil setups.
Here we present a shim coil array specifically designed to improve shimming in the fronto-temporal brain region. The array is designed to be mounted on top of a typical 7T head coil (Nova Medical, Wilmington, MA, U.S). The design is freely accessible and can be manufacturing with limited electronics experience and equipment.

Methods

A shim coil array was simulated by varying position, number of coils, coil radius, inter-coil distance, rotation in a coronal plane, offset in the feet-head direction, and limited to a single curved surface to fit the outer diameter of the RF coil’s transmit component (Tx-coil). Based on the simulations a coil layout as shown in figure 1 was chosen (8 coils, inner coil diameters: 70mm, 50 windings, 90mm from the center of the central coil to the center of 7 peripheral coils). Winded coils were attached to a flexible acrylic sheet using zip ties. The sheet was bend into the shape of the outer surface of the Tx-coil, and the coil was embedded in epoxy to minimize vibration when operated. The shim array was positioned relative to the head coil by plastic screws going through the acrylic plate and into 3D printed brackets screwed onto the Tx-coil. The shim array before being embedded in epoxy (top) and mounted on the head coil (bottom) is depicted in figure 2. A printable coil-placement guide, 3D-printable brackets and a 3D-printable coil-winder is accessible at https://resources.drcmr.dk/BrainShimArray/.
To drive the shim-array, we used an open-source, 8-channel, class-A amplifier [5]. Control of the amplifier was done by fiber optic cabling, facilitated by a Teensy 3.5 (PJRC, Sherwood, OR, U.S.). Current-carrying cables from the amplifier to the shim coils were pair-wise twisted and passed through the cable-management system of the scanner. The amplifier was supplied 12V, facilitating up to ±2.5A on each shim array element. B₀ shim fields were acquired using a large balloon containing salt water as phantom.
A B₀ map of a healthy volunteer was acquired using the scanner’s (Philips 7T Achieva, Best, Netherlands) 1^st and 2^nd order shimming. A current-constrained shimming algorithm generated settings for 2 scenarios: (1) Static shim by the scanner, and static shim by the shim array (“static shimming”). (2) Static shim by the scanner, but dynamic slice-wise shimming by the shim array and updating of the scanner's reference frequency, f₀ (“dynamic shimming”). Simulated settings were transferred to the Teensy, which updated shims when triggered by scanner-provided TTL-pulses. Performance of the shimming was evaluated by the standard deviation (std) across a brain-masked B₀ map.

Results

Figure 3 depicts measured B₀ maps when using only the scanner’s 1^st and 2^nd order shims (top), static shimming (mid) and dynamic shimming (bottom). A high degree of similarity between simulated and measured B₀ fields was generally observed (<1Hz difference in std), verifying the applicability of the shimming algorithm. Static shimming improved the std from 39Hz to 31Hz, which is similar performance to [3] (verified by simulations). Dynamic shimming further improved the B₀ homogeneity, resulting in a std of 25Hz.

Discussion

We optimized a multi-coil shim array specically for improving whole-brain shimming at 7T, making a compromise between construction complexity and shimming performance. Placing 8 multi-turn shim coils outside the RF shield of the Tx-coil allows for low interaction with the RF coil, but still provides significant improvements over static 1^st and 2^nd order shimming. Simulations suggest slightly improved performance if the shim array is placed between the RF transmit and receive coil, however this would require a more complex design to address coupling to the transmit and receive coil. While other hardware designs have shown better performance than here presented, these feature a vastly more complex design. Simulations predict further improvement is obtainable from using the presented shim array in conjunction both f₀ updating and 1^st order shimming, but requires extensive pulse programming to verify in measurements.

Conclusion

We have presented an easily reproduced, effective local shim array design, specifically designed to address fronto-temporal B₀ inhomogeneites at 7T. In a healthy volunteer the shim array improved whole-brain standard deviation from 35Hz to 31Hz using static shim settings, and to 25Hz using dynamic slice-wise shimming.

Acknowledgements

The 7T scanner was donated by the John and Birthe Meyer Foundation and The Danish Agency for Science, Technology and Innovation (grant no. 0601-01370B).