Modular transmit/receive arrays using very-high permittivity dielectric resonator antennas.

PURPOSE: Dielectric resonator antenna (DRAs) are compact structures that exhibit low coupling between adjacent elements and therefore can be used as MRI transmit arrays. In this study, we use very high permittivity materials to construct modular flexible transceive arrays of a variable numbers of elements for operation at 7T.
METHODS: DRAs were constructed using rectangular blocks of ceramic (lead zirconate titanate, εr  = 1070) with the transverse electric (TE)01 mode tuned to 298 MHz. Finite-difference time-domain simulations were used to determine the B1 and specific absorption rate distributions. B1+ maps were acquired in a phantom to validate the simulations. Performance was compared to an equally sized surface coil. In vivo images were acquired of the wrist (four elements), ankle (seven elements), and calf muscle (16 elements).
RESULTS: Coupling between DRAs spaced 5 mm apart on a phantom was -18.2 dB compared to -9.1 dB for equivalently spaced surface coils. DRAs showed a higher B1+ intensity close to the antenna but a lower penetration depth compared to the surface coil.
CONCLUSION: DRAs show very low coupling compared to equally sized surface coils and can be used in transceive arrays without requiring decoupling networks. The penetration depth of the current DRA geometry means they are ideally suited to imaging of extremities. Magn Reson Med 79:1781-1788, 2018.

INTRODUCTION

Ultrahigh field (UHF) MRI suffers from B1 inhomogeneities due to radiofrequency (RF) interferences that arise when the RF wavelength is of the same order as the imaging region of interest 1, 2. It has been shown that inhomogeneities can be reduced through use of multi‐element transmit array systems 3. Although receive‐only arrays universally are used in UHF, MRI transmit arrays predominantly are used for body imaging using decoupled surface coils 4; microstrip or dipole antenna 5, 6; and to a lesser extent, neuroimaging using decoupled surface coils7 or an array of decoupled transmission line antenna 8, 9. Musculoskeletal (MSK) imaging is less sensitive to inhomogeneities due to the relatively small dimensions of the typical region of interest; as a result, the birdcage remains the dominant design for RF transmission 10, 11, 12. Nevertheless, recent studies have begun to show the utility of transmit arrays for MSK imaging at 7T 13, 14, 15, 16. Array designs utilizing overlapping surface coils—with the surface coils either fixed on a cylindrical housing into which the region of interest (ROI) is inserted 16, or with two separate surface coil arrays that are placed around the ROI 13, 14—have shown promise. An innovative U‐shaped eight‐channel microstrip array using capacitive decoupling has been used for imaging the ankle joint at 7T 15.

One of the main issues in designing large multi‐element arrays is RF coupling between proximal array elements 17. Aside from causing changes in the impedance of individual array elements, RF coupling also reduces the signal‐to‐noise ratio (SNR) in parallel imaging techniques 18, 19. Many system designs have been proposed to reduce inter‐element coupling, including the overlapping of surface coils, preamplifier decoupling 18, resonant inductive decoupling 20, capacitive decoupling 21, inductive decoupling 22, decoupling annexes 23, and induced current elimination 24. The implementation of these systems typically increases the complexity of the antenna arrays. Decoupling methods that introduce additional decoupling structures to the array, such as the aforementioned resonant inductive decoupling and induced current elimination methods, are highly sensitive to geometric changes to the decoupling structures and subsequently to array deformation. Changes in coil loading also can impact the effectivity of decoupling systems 20 and change the inductance/capacitance value needed for optimal decoupling of array elements 25. Furthermore, the introduction of additional decoupling elements can result in significant alterations to the distribution compared to independent antenna 26.

High permittivity materials (also referred to as dielectric materials in other literature) have seen increased usage as the trend toward higher magnetic fields continues. High‐permittivity pads placed between the patient and the transmit coil have been used to tailor fields, with the aim to improving homogeneity in body‐ and neuroimaging at high field 27, 28, 29, 30 and neuroimaging in ultrahigh field MRI 31, 32, or for strong local focusing of to improve signal intensity in the cervical spine at 3T 30 and in the inner ear at 7T 33. The interesting electromagnetic (EM) properties of high‐permittivity materials (HPMs) have seen them integrated into several (UHF) antenna designs. The shortened RF wavelength in HPMs has been exploited to reduce the dimensions of bow‐tie antenna by submerging them in water 34 and to construct dielectric waveguide antenna 35, again using water as the HPM. HPMs also have been used as substrates for dipole antenna to improve penetration and reduce specific absorption rate (SAR) 36, 37.

Recent work has shown that dielectric resonator antenna (DRAs) operating in the transverse electric (TE)01δ mode 38, 39 and hybrid electromagnetic (HEM)11δ 40 can be used as transceive antenna in ultrahigh field MRI. The frequency of the TE01δ mode in the dielectric resonator used in DRAs is determined by the shape, dimensions, and relative permittivity of the material. No expression for the mode frequencies of rectangular dielectric resonators exist, and electromagnetic simulations typically are used to determine the exact design parameters. Both Lu et al. 38 and Aussenhofer and Webb 39 use cylindrical dielectric resonators constructed from water (ɛr = 80) and barium titanate (ɛr = 170), respectively. Although having several advantages compared to equivalently sized surface coils, including much lower interelement coupling (the electric field in the TE01δ mostly is contained within the DRA; and the magnetic field mainly is in the z‐direction, resulting in little EM interaction with adjacent elements), the individual elements reported in both papers were relatively large and correspondingly heavy, which impacts patient comfort in the scanner when used as surface elements.

In this paper, we present a new design of DRAs using rectangular elements with very high relative permittivities (εr ∼1070), which result in much smaller and lighter antennas, thereby improving patient comfort compared to previous designs. The new lightweight DRAs, combined with the lack of need for decoupling networks, enable the construction of a flexible array with an arbitrary number of antennas that can be placed directly on the patient and conform to the particular region of interest. In vivo results are shown from scans using between four and 16 separate elements.

METHODS

Electromagnetic Simulations

EM simulations of the B1 and SAR distributions of the DRAs were performed using the time‐domain solver in CST Microwave Studio 2016 (CST AG, Darmstadt, Germany). A mesh size of 50 cells per wavelength was used for all simulations with open boundaries spaced λ/10 away from the model. The dielectric resonators were simulated with a relative permittivity of 1,070 and a conductivity of 1.5 S/m. A cuboid 120 × 120 × 210 mm3 phantom (εr = 80, σ = 0.40 S/m) was used in all simulations. The SAR distribution was computed in accordance to the Institute of Electrical and Electronics Engineers Standards Association/International Electrotechnical Commission 62704‐1 standard 41. All simulations were normalized to 1W accepted power. The intrinsic signal‐to‐noise ratio of the coils was evaluated using the method specified by Schnell et al. 42. The power loss in the phantom and antenna was determined using the same simulation method as for the B1 and SAR.

Dielectric Resonator Construction

Rectangular ceramic blocks (TRS Technologies, State College, Pennsylvania, USA) with dimensions of 68 × 90 × 5 mm3 were formed from sintered lead zirconate titanate (PZT). EM simulations using an eigenmode solver (CST Microwave Studio 2014, CST AG) were used to confirm the relative permittivity of the material by simulating the frequency of the TE01δ mode for different permittivity values, and then comparing this with an S11 measurement using an unmatched 1‐cm diameter pickup loop placed above the center of the ceramic block and a vector network analyzer (Planar TR1300/1, Copper Mountain Technologies, Indianapolis, Indiana, USA). The conductivity of the dielectric resonators was determined using a quality (Q)‐factor measurement of the resonance peak. The relative permittivity value of 1,070 subsequently was used to determine the dimensions (44 × 90 × 5 mm3) of the ceramic block such that the TE01δ mode was at 298 MHz. The ceramic block was trimmed to these dimensions, and the resonance frequency was experimentally measured to be 298 MHz using the unmatched loop.

Dielectric Resonator Antenna Design

An inductively coupled circular loop, with an inner diameter of 11 mm, an outer diameter of 15 mm, and a balanced matching network was constructed (see Fig. 1). The loop was placed concentrically above the DR to most effectively couple to the magnetic component of the TE01δ mode of the DR. The distance between the loop and the DR to achieve critical coupling was determined. The distance between the loop and the DR was kept constant with a hard plastic separator. Impedance matching of the critically coupled system was performed with the DR placed on a 120 × 120 × 210 mm3 saline phantom (εr = 80, σ = 0.4 S/m). The coupling between two DRAs placed next to one another was measured using the S12 parameter, with the long sides of the elements parallel to one another and placed on one face of a cuboid 120 × 120 × 210 mm3 phantom (εr = 80, σ = 0.4 S/m).

Figure 1

(a) Dielectric resonator made from PZT with a relative permittivity of ∼1,070. The dimensions of the block are 90 × 44 × 5 mm3, such that the frequency of the TE01δ mode is at 298 MHz. (b) A single DRA, the resonant loop is spaced 13 mm from the resonator. This distance is kept constant with a hard plastic separator. (c) Surface loop coil used for comparison to the DRAs. The outer dimensions are 90 × 44 mm, with a track width of 5 mm. (d and e) Circuit diagrams for the two loop coils.

Reference Surface Coil

Two rectangular surface coils with a balanced matching network, four tuning capacitors, outer dimensions of 90 × 44 mm2, and a copper trace width of 5 mm were constructed (see Fig. 1) as reference coils to compare performance with the DRs. The coils were tuned to 298 MHz and impedance matched to 50 Ω on a 120 × 120 × 210 mm3 saline phantom (εr = 80, σ = 0.4 S/m). The coupling between the adjacent loops was measured using the same setup as for the DRAs.

MRI Data Acquisition

All experiments were performed on a 7T whole‐body human MRI scanner (Phillips Achieva, Best, the Netherlands). All in vivo scans were performed on healthy volunteers, and written consent was obtained from all volunteers prior to scanning. For experiments with the four‐element DRA array, two independent transceive channels were split into a total of four channels using two 1‐to‐2 Wilkinson transmission line power dividers. The four element DRA array was driven with a relative phase difference of 0°, 0°, 90°, and 90° between the antenna measured in a clockwise direction. For experiments performed with seven‐ and 16‐element DRA arrays, a custom‐built 16‐channel transmit/receive interface box was used. The interface box consists of two 1‐to‐8 Wilkinson transmission line power dividers, fed with two independent transmit channels, and 16 transmit/receive (TR) switches that provide 16 independent receive channels. The seven element DRA array was driven with a relative phase difference of 0°, 0°, 0°, 0°, 90°, 90°, and 90° between the antenna measured in a clockwise direction. The 16‐element DRA array was driven with no phase difference between the antenna.

Single‐slice maps were obtained using the dual refocusing echo acquisition mode sequence 43 with the following parameters: field of view = 16.3 × 10 cm, slice thickness = 5 mm, spatial resolution = 1.56 × 1.56 mm, stimulated echo acquisition mode (STEAM) flip angle = 60°, imaging tip angle = 10°, TR/TE = 5/1.13 ms, number of signal averages = 256, and acquisition time = 136 s. T1‐weighted 3D gradient recalled echo (GRE) images of the wrist were acquired with four DRA elements with the following parameters: field of view = 10 × 10 × 4 cm, spatial resolution = 0.3 × 0.3 × 2 mm, TR/TE = 20/3.2 ms, flip angle = 10°, echo train length = 30, and acquisition time = 3m 18s. T1‐weighted 3D GRE images of the ankle were acquired with seven DRA elements, with the following parameters: field of view = 12 × 12 × 6 cm, spatial resolution = 0.28 × 0.28 × 2 mm, TR/TE = 20/3.2 ms, echo train length = 30, and acquisition time = 5m 37s. T1‐weighted 3D GRE images of the lower leg were acquired with 16 DRA elements, with the following parameters: field of view = 15 × 15 × 10 cm, spatial resolution = 0.47 × 0.47 × 2mm, TE/TR = 4.9/2.2 ms, flip angle = 20°, echo train length = 352, and acquisition time = 3m 28 s.

RESULTS

Coil Characterization

The reflection coefficient (S11 parameter) of all DRAs was measured to be lower than −30 dB, and that of the two reference surface coils less than −25 dB when loaded with a phantom. Figure 2 shows a plot of the S11 parameter of an unloaded dielectric resonator and surface coil, as well as the case when loaded with a human leg, using the same critically‐coupled 15 mm diameter secondary loop. The separation distance at which critical coupling was achieved, as well as the Q‐factor at the point of critical coupling, are reported in Table 1.

Figure 2

Plot of the measured S11 parameters of (a) a dielectric resonator and (b) an equally sized surface coil, both unloaded and loaded with a human leg when the dielectric resonator and surface coil are critically coupled to a tuned 15‐mm diameter resonant loop.

Table 1

The Required Distance Between a 15‐mm Diameter Tuned Resonant Loop and a Dielectric Resonator or Equally Sized Surface Coil and the Measured Q‐Factor of the System at the Point of Critical Coupling.

Antenna		Critical Coupling Distance (mm)	Q‐Factor
Dielectric resonator	Unloaded	16	34.0
	Loaded	13	37.2
Surface coil	Unloaded	40	99.4
	Loaded	4	34.7

Q‐factor, quality factor.

The interelement coupling (indicated by the S12 parameter) between two DRAs placed 5 mm apart on a phantom was −18.2 dB (see Fig. 3), with minimal change in S11 compared to the individual elements. Placing the DRAs directly against each other increases the coupling to −15.1 dB. The interelement coupling between the two surface coils separated by 5 mm is −9.1 dB (simulated) and −10.6 dB (measured), resulting in a shifted resonance frequency and reduced coil sensitivity. The simulated S‐parameters of the DRA and surface coil show good agreement with measurements.

Figure 3

Simulated (segmented line) and measured (solid line) S11 (blue) and S12 (orange) parameters of (a) two dielectric resonator antennae and (b) two surface coils spaced 5 mm apart on a phantom.

Maximum intensity plots of the 10‐gram average SAR (SAR10g, avg) of a DRA and surface coil are shown in Figure 4. The distribution of the SAR10g, avg of both setups is similar, with the maximum SAR located proximally along the long side of the antenna/coil. The maximum SAR10g, avg of the DRA was 1.62 W/kg compared to a maximum SAR10g, avg of 2.20 W/kg for the surface coil. The simulated and measured distributions across the central slice of the antenna and the coil also are shown in Figure 4. There is good agreement between the simulated and measured distribution, although the very high close to the surface of both the DRA and surface coil is not replicated in the maps. This most likely is due to the limited dynamic range of the mapping method. The overall distribution of the is broadly similar between the DRA and surface coil, although the surface coil shows a slightly higher at greater depth.

Figure 4

(a–b) Maximum intensity plot of the simulated SAR10g, avg of the DRA and surface coil normalized to 1W input power placed on a 120 × 120 × 210 mm3 phantom (εr = 80, σ = 0.40 S/m). (c–d) Simulated distribution normalized to 1W input power in the same phantom. (e–f) distribution measured using the dual refocusing echo acquisition mode sequence normalized to 1W input power. (g–h) Simulated distribution normalized to the maximum SAR10g, avg, 1.62 W/kg, and 2.20 W/kg for the DRA and surface coil, respectively.

DRA, dielectric resonator antenna; SAR, specific absorption rate.

Figure 5 shows the simulated along a line through the maximum of both antennas, marked by a segmented white line in Figures 4c and 4d for the DRA and surface coil, respectively. The DRA produces a stronger at shallow depths but has a stronger dropoff compared to the surface coil, with the latter displaying a higher at depths greater than 1.5 cm. Figure 6 shows the normalized to the maximum SAR10g, avg, of the DRA and surface coil through the same lines as used in Figure 5. In this case, the crossover point is approximately 2 cm. Figure 7a shows a plot of the intrinsic SNR through the point of maximum intrinsic SNR for both the DRA and the surface coil. Figure 7b shows the ratio between the intrinsic SNR of the DRA and surface coil.

Figure 5

Simulated per Watt accepted power for a dielectric resonator and a “loop” coil in a phantom. Both profiles were taking through each antenna's respective maximum , indicated by a white line in Figures 4c and 4d.

DRA, dielectric resonator antenna.

Figure 6

Simulated profile normalized to maximum SAR10g, avg for a dielectric resonator and loop coil placed on a phantom (120 × 120 × 210 mm3, εr = 80, σ = 0.40 S/m).

DRA, dielectric resonator antenna; SAR, specific absorption rate.

Figure 7

(a) A profile of the intrinsic SNR for the DRA and surface coil placed on a phantom (120 × 120 × 210 mm3, εr = 80, σ = 0.40 S/m). (b) The ratio of the intrinsic SNR of the DRA and surface coil on the same phantom.

DRA, dielectric resonator antenna; iSNR, intrinsic signal‐to‐noise ratio; SNR, signal‐to‐noise ratio.

In Vivo Results

Figure 8 shows in vivo T1‐weighted 3D gradient echo images of a wrist, ankle, and lower leg with four, seven, and 16 DRA elements, respectively, as well as the S‐parameter matrix measured for the various setups. Interelement coupling did not exceed −14 dB in any of the configurations of the array. No retuning of the DRA array elements was required for the different imaging configurations. The S11 parameter was below −21 dB for all elements in the wrist array, −18 dB for all elements in the ankle array, and −15 dB for all elements in the leg array. Note that the configurations have not been extensively optimized, but the choice of the respective matrix (4 × 4, 7 × 1, 4 × 1) was made simply to show the versatility of placement of the resonators.

Figure 8

(a–c) DRA array configuration for imaging the wrist (a,d,g), ankle (b,e,h) and calf muscle (c,f,i) using four, seven, and 16 elements. (d–f) S‐parameter matrix of the respective configurations. (g‐i) T1‐weighted 3D gradient recalled echo obtained using the DRA array as a transceive system.

DRA, dielectric resonator antenna.

DISCUSSION

In this study, we have shown that lightweight DRAs consisting of extremely high permittivity materials can be used to construct transceive surface arrays with arbitrary dimensions, without the need for additional decoupling systems due to the inherently high isolation of dielectric resonator antenna. The shape of the dielectric resonators was practical for the conformation of the DRA arrays to highly irregular body areas, but the small effective area of the TE01δ mode contributes to the steep dropoff displayed by the DRAs compared to the equally sized surface coil. As such, this geometry is most suited for studying regions with relatively small dimensions or regions close to the surface of the body. Although this study has not optimized the shape of the dielectric resonators, we anticipate that significant optimizations in the distribution can be achieved by using square or circular resonators (field distribution of the TE01 mode would result in a larger fraction of the surface contributing to the ) as well as by using a larger dielectric resonator with lower relative permittivity. Furthermore, simulations indicate that a higher material conductivity is associated with a more “leaky” resonator, and both B1 (magnetic fields) and SAR (electric fields) increase in magnitude with resonator conductivity, suggesting that transmit efficiency may be optimized at a particular (non‐zero) resonator conductivity.

The PZT blocks are delivered in slabs with a predefined thickness and permittivity due to small interbatch variations in the permittivity (±5%); precise tuning of the resonators should only be done once their exact permittivity has been determined. PZT is a hard and brittle ceramic material with high lead content; therefore, cutting of the resonators should be done using specialized equipment, and waste products must be handled with care. It is interesting to note that PZT most commonly is used for wide‐band ultrasound transducers; therefore, the conductivity of these types of materials tends to be high. However, it is certainly possible to produce materials with high relative permittivity and low conductivity, for example, materials used in dielectric resonators for MR microscopy 44, and this may lead to improved performance.

For the ceramic blocks used in our study, a change of ±1 mm in the width, length, and thickness of the dielectric resonator results in a ∓ 3.3 MHz, ∓0.3 mm, and ∓23 MHz TE 01 mode frequency change, respectively. Because the resonators can be cut with millimeter accuracy, it is highly recommended to first cut the resonators to thickness and then cut the remaining dimensions.

Measurements of the temperature dependence of the relative permittivity of the dielectric resonators showed little variation, corresponding to a −0.3 MHz per degree temperature increase between 8°C and 65°C. No warming of the dielectric resonator was measured during the imaging in vivo imaging sequences; therefore, very minor changes in resonance frequency during imaging can be neglected.

CONCLUSION

In vivo imaging of the lower leg showed some image‐intensity inhomogeneity. The inhomogeneities arise due to several factors: dielectric focusing due to the short RF wavelength in tissue; the fact that the coils are nonoverlapping, meaning they behave as spatially separated surface coils; and constructive and destructive interferences in the field of the individual array elements. This current study was performed using a system equipped with only two independent (in terms of transmit phase and amplitude) RF transmit channels, which severely limits the possibilities of shimming. Other groups have shown that significant improvements in homogeneity can be obtained through use of a higher number of independent transmit channels, and one can anticipate that the same will apply to DRA arrays.