An improved intraoral transverse loop coil design for high‐resolution dental MRI

To improve intraoral transverse loop coil design for high‐resolution dental MRI.


INTRODUCTION
The value of using MRI for dental applications has been shown for various pathologies in endodontics, [1][2][3] orthodontics, 4 craniomaxillofacial surgery, 5 and implantology. 6 In dentistry, submillimeter structures need to be imaged such as the canals and cracks 7 in teeth. To achieve this high resolution in clinically acceptable measurement times, it is advantageous to restrict the imaging FOV to the target volume. FOV restriction can be realized by coils that offer local RF excitation (B 1 + ), limited receive (Rx) sensitivity (B 1 − ) tailored to the anatomy, or both. So far, mainly extraoral surface coils have been used, [8][9][10] which are positioned at distances of about 30-50 mm to the target region (e.g., a molar tooth). Hence, Rx sensitivity around teeth is significantly reduced, and a restricted FOV setting is impracticle. For this reason, intraoral coils are preferable because they can be placed closer to the teeth. [11][12][13][14][15] Recently, an intraoral coil (IOC) design has been proposed that uses a transmit (Tx)/Rx loop coil with the coil plane orthogonal to B 0. 12 This transverse loop coil (tLoop) can image the whole dental arch, including the upper and the lower teeth as well as the jaw, while minimizing the unwanted signals from the cheek and the tongue. However, there are several limitations of the tLoop design: use of a Tx/Rx coil leads to a steep sensitivity gradient; the sensitivity is reduced at the feed port; the straight geometry makes it uncomfortable to use; the necessary wire insulation introduces additional signals; the long wires show wavelength effects at higher field strengths; the tongue is visible at the back of the mouth; and the design is susceptible to gradient-induced eddy currents.
In this work, a modified loop design (mtLoop) is proposed to overcome these limitations. The mtLoop design is compared with the tLoop in terms of sensitivity distribution, image SNR, and eddy currents using electromagnetic (EM) simulations and MRI measurements at 3T.

METHODS
In Figure 1, the design of the tLoop from 12 is shown together with the new mtLoop coil. Compared to the tLoop, the mtLoop uses a bent coil conductor that has a higher inductance (L = 110.5 nH and 124.3 nH @ 123.2 MHz; measured using 4396B Impedance Analyzer (Agilent Technologies, Santa Clara, CA) and 16092A Spring Clip Fixture) due to its length. Capacitive tuning and matching were applied, and coil conductors were laser-cut from 100 μm-thick copper plates having a track width of 8 mm. At the mtLoop, a 100 μm-thick dielectric tape (Kapton) was used to form a parallel-plate capacitor of 9.6 pF at the posterior section, which interrupts and overlaps the conducting wire of the tLoop design. Instead of a single long conductor with a single tuning capacitor at the feed port as in tLoop, separating the conductor in two shorter conductors, extending one of the conductors and overlapping them to form a distributed capacitor at the posterior section, that is, opposite to the feed port, has two advantages: first, the voltage across the capacitor at the feed port is reduced and a more homogeneous current distribution is achieved by shortening the conductor segment; second, the potential eddy currents are decreased.
Although tLoop consists of a single continuous conductor of 17 cm, mtLoop is built of two 10.4 cm-long conductors that overlap at the posterior section, as demonstrated in Figure 1A,B. Table 1 summarizes the modifications and the resulting improvements. A shorter conductor length is expected to reduce the eddy currents generated on the coil. To verify this assumption, eddy currents were computed for tLoop and mtLoop using a finite element method solver (The COM-SOL AB, Stockholm, Sweden). The eddy currents generated by a switching a gradient coil were calculated based on the quasi-static Maxwell's equations. 16 The input current waveform for the gradient coil was modeled as a Heaviside function. The gradient coil was simulated on a cylindrical surface with a diameter of 70 cm and a height of 120 cm by the stream function method. 17 In Ref. [12] the tLoop was designed as a Tx/Rx coil, which leads to a steep sensitivity gradient because the inhomogeneity of the Tx and Rx sensitivities (i.e., B 1 + and B 1 − ) are multiplied. To extend the sensitivity toward the apices of the roots, in the mtLoop design the coil is operated in Rx-only mode, which offers a more homogeneous sensitivity profile along the z axis. Moreover, the combination with the Tx body coil with its homogeneous B 1 + facilitates the use of spin echo-based pulse sequences and magnetization preparation pulses. To realize Rx-only mode, an active detuning network was added as shown in Figure 1B. During RF transmission, the detuning diode is switched on via a direct current (DC) bias (10 V @ 100 mA) supplied through the Rx cable. DC and RF paths are separated by RF-choke inductors (22 μH) and DC-blocking capacitors (two 10 nF in parallel). tLoop was driven both in Tx/Rx and Rx-only modes and compared to the mtLoop.
With the tLoop, a signal void was seen along the incisors 12 due to the gap between the differential feed ports. In the mtLoop, the coil conductor was made of a plain copper sheet, which allowed an overlapped feed port construction ( Figure 1A). Therefore, a two-layer interface circuit was constructed to be able to connect the feed ports from both sides. The capacitance formed by the overlapped conductors (20 × 8 mm 2 ) at the feed port was ignored due to 1.6 mm distance in between.
In the tLoop design, the straight posterior section of the coil compresses the tongue and restraints the tongue  movement, which inhibits the swallowing reflex and causes discomfort during MRI. In the mTLoop, use of a thin copper sheet as the coil conductor allowed bending the posterior section and increasing the comfort. Intraoral coils require insulation and coating of the electrically conducting parts. Insulation materials must be biocompatible, so it is preferable to use compounds that are already established in dental practice: impression putty (mostly based on silicon), dental resin including 3D printing resin, X-ray film sleeves, bonding fluids, gutta percha, dental cement, and other filling compounds. The materials need to offer good electrical insulation; be waterproof, mechanically and chemically stable (e.g., against bite), and comfortable to wear; and should not be visible in the MR images (i.e., MR silent). In the tLoop, a modified polytetrafluoroethylene bag (Welch Fluorocarbon Inc., Dover, NH) and a sticky foam unfortunately visible in MRI were used to cover the coil. To find a more suitable material, image artifacts and MR signals of the above-mentioned insulation materials were evaluated using conventional sequences and a modified "zero echo time" sequence 18,19 (TE = 20 μs) (see Supporting Information). Based on the results of this evaluation, a biocompatible insulating composite was made for the mtLoop, which consisted of a dental resin used to 3D-print denture bases (Denture Base OP Resin 1, Formlabs Inc., Somerville, MA) and dental resin powder (Coldpac, Yates Motloid, Elmhurst, IL). This composite material was artifact-free, MR-silent, and provided the required electrical and mechanical insulation. The composite was applied manually and cured for 4 min under UV exposure. A layer thickness of 0.95 ± 0.17 mm at each side was verified. Then, the coil safety was evaluated by hotspot detection and temperature measurements as described in Ref. [20] Finite-difference Time-domain simulations were performed (Sim4Life v7.0, ZMT, Zurich, Switzerland) to assess the sensitivity along the coil conductor, and the results were compared to the measurements. To find the optimal insulation thickness, 4 mtLoop designs were simulated covered by different thicknesses of the insulating layer (d = 0.2, 0.5, 1.0, 2.0 mm). The results were normalized to 1 W input power. The simulation settings are given in Table S1. The 3D CAD models of the coil and the phantom ( r = 78; = 0.47 S/m) used in the simulations are provided in https://github.com/ozenEEE/UKF_ IntraOralLoop.
MRI measurements were performed at a clinical 3T MRI system (Prisma-Fit, Siemens, Erlangen, Germany). To study the different sensitivity profiles, both tLoop and mtLoop were tested in Rx-only mode, and the tLoop was additionally assessed in Tx/Rx mode. 12 A phantom was 3D-printed from an optical intraoral scan using the insulation material of the intraoral coils.
The phantom was developed as hollow containers both for mandible and maxilla, which was filled with 1 g/L Agar-agar (CARL Roth AG, Karlsruhe, Germany, Art.-No. 4508.1, standardized carrageenan mixture, containing sugar), and 1 g/L CuSO4 (E. Merck, Darmstadt, Germany, Art.-No. 2790), and finally sealed with a 3D-printed plate. Source files for the 3D prints can be downloaded from https://github.com/ozenEEE/UKF_IntraOralLoop. All methods were carried out in accordance with relevant guidelines and regulations; healthy volunteer scanning was approved by the institutional review board (Ethikkommission) of the University Medical Center Freiburg (No. 160/2000); and informed written consent was obtained before imaging. In vivo images of the whole dental arch were acquired at 0.25-mm-isotropic resolution using a custom-developed 3D UTE sequence (XTE, https://webclient.eu.api.teamplay.siemens-healthineers. com/c2p) with TR/TE = 3.4/0.03 ms, α = 5T RF = 20 μs, number of samples per spoke = 720, T encoding = 1.28 ms, N spokes = 33 000, G max = 40 mT/m, FOV read = 180 mm. T 2 -SPACE was applied with TR/TE = 1500/247 ms, α = 110 • , acquisition matrix = 320 × 306, slice thickness = 1 mm, T encoding = 3.03 ms, N average = 2, fat saturation: ON, echo train = 85, FOV read = 112 mm. This parameter set is similar to the one used for imaging canals. 15 Finally, panoramic slices were created by using the "Straighten" function 21 of ImageJ, which is a Java-based image processing program. 22 Temperature measurements were performed during an RF-pulse-only pulse sequence with system-reported whole-body specific absorption rate (SAR) value of 4 W/kg. An Hydroxyethylcellulose-gel phantom with r (124 MHz) = 79.2, (124 MHz) = 0.61 S/m (measured with DAKS12; Zurich Med Tech, Zurich, Switzerland) was prepared. 23 The mtLoop was immersed in the phantom with the coil plane orthogonal to B 0 , and temperature rise near the coil was monitored at different locations using four fiber-optic temperature probes (FOTEMP6-19, Optocon AG, Dresden, Germany). At least one fiber-optic temperature probe was positioned at the hotspot determined using E-field mapping as described in Refs. [20,[24][25][26].

RESULTS
Simulation results show the improvements of the mtLoop originating from the overlapped feed port and the parallel-plate capacitor at the posterior section ( Figure 2): a 3.7-fold sensitivity increase was seen along the incisors, and a 76% signal reduction at the back of the tongue was achieved compared to the tLoop design. Simulation also showed that an insulation thickness of 1 mm was optimal for the given coating material properties

F I G U R E 2
Simulations: Comparison of tLoop versus mtLoop. (Upper row) The effect of modifying the feed port from adjacent conductors to overlapping conductors improves the sensitivity along the incisors by 3.7-fold. (Middle row) The effect of modifying the posterior section of the coil by cutting the loop; introducing a parallel-plate capacitance reduces signal from the tongue near the coil conductor by 76%. The sensitivity distribution in the transverse plane selected 10 mm above the coil conductor resembles the dental arch; therefore, it is more suitable for dental applications. (Bottom row) Eddy current simulations show reduced surface current density in the mtLoop design, as expected. The direction of the current indicates that a mechanical force along y direction might be experienced. The simulations were performed for a gradient activity along the y axis at 128 MHz ( Figure S1). The line profiles represent the sensitivity profile along the z direction at the position of the third molar. A thicker insulation layer increased the distance between the tissue and the coil leading to a reduced sensitivity, whereas a thinner insulation caused increased parasitic impedances due to the losses in the tissue. For the EM properties of the MR-silent insulation material developed in this study, 1-mm thickness yielded higher The UV-cured 3D-printing resin mixed with resin powder was MR-silent and did not show image artifacts ( Figure 1C and Figure S2). Dental cement was also artifact-free but had an unfavorably high conductivity of 0.014S/m (measured with DAKS12, ZMT) that would lead to losses.
The phantom provided a loading very similar to the in vivo situation: Q Loaded,phantom = 34.2; Q Loaded,invivo = 35.2, Q Unloaded,bare = 60.8, Q Unloaded,coated = 60.6, measured from S 11 -curves. The frequency, at which S 11 is minimum was observed at 123.194 MHz, and 123.188 MHz for phantom and in vivo, respectively. Active detuning provided 41 dB isolation (obtained using an S21 probe with the detuning diode forward-and reverse-biased), and no resonance was identified in detuned state within 200 MHz band.
Transverse and coronal phantom images showed a higher SNR in at the incisors (SNR = 6721) for the mtLoop compared to the Rx-tLoop (SNR = 602) and Tx/Rx-tLoop (SNR = 72) (Figure 3), and the SNR ratio of the incisors over the third molars were 1.04, 0.48, and 1.04, respectively. However, SNR ratios between the tongue and third molars were 0.56, 0.60, and 0.84 for mtLoop, Rx-only, and Tx/Rx tLoops, respectively.
At the hotspots, a maximum temperature rise of 0.4 • C was measured with the high-SAR protocol. In Figure 4, in vivo panoramic images were reconstructed from T 2 -SPACE (350 μm-in-plane-resolution, 1 mm-slice thickness) and UTE (250 μm-isotropic) sequences. In the transverse slices of mandible, a missing tooth (lower left-/right first molar) and a porcelain-fused-to-metal crown (lower right/left first molar) can be observed. As expected, the artifact around the metallic restoration is larger in the T 2 -SPACE images than the UTE data. In T 2 -SPACE images, the canals in the teeth are better visualized than in the UTE images. No canal is visible along the upper-right first incisor due to a canal treatment.

DISCUSSION
In this study, we introduced a modified loop coil design to improve sensitivity, homogeneity, and comfort compared to a previous loop design 12 for intraoral MRI. Additionally, we used a biocompatible, artifact-free, and MR-silent coating for the new intraoral coils. B 1 − increase of 3.7-fold along the incisors and decrease of 76% at the back of the tongue in simulations are different than measured elevenfold increase and 2.3-fold decrease in SNR, respectively. The major difference between the simulations were the following: the coils were embedded in a homogenous phantom volume, leading to different loading conditions than measurements; the dielectric separating the overlapping conductors at the posterior section of the tongue had 1 mm thickness instead of 0.1 mm to enable meshing; ideal matching conditions were assumed; and the conductors were modeled as perfect electric conductors. Finite-difference Time-domain simulations were performed using a homogeneous head model. Although this oversimplifies the anatomy, the simulation results provide insights into the SNR changes across the feed port and along posterior section of the coil. A human model might provide more realistic simulation results, but available human models have two limitations: the teeth are not modeled properly because most models include only crowns of teeth, and they do not allow to open the mouth.
Another limitation of the simulation is the geometry for eddy current simulations: here, the coil was modeled

F I G U R E 4
In vivo images acquired with a water-excited UTE implemented using binomial pulses (left) and fat-saturated T 2 -SPACE (right). The panoramic reconstructions were generated along a vertical axis and therefore do not cover the whole teeth in the same slice plane. T 2 -space provides better visualization of canals than UTE. However, UTE has reduced metal artifacts as a plane geometry without the bent posterior section. A more realistic geometry and more complicated gradient coil design will increase the accuracy. Yet, the simplified model was sufficient to demonstrate the reduced eddy current sensitivity with the parallel plate capacitor at the posterior section. Eddy currents might also cause mechanical force or vibrations on the patient similar to an actuator. The forces calculated from the eddy current simulations were, however, less than 1.33 × 10 −4 N, and are not expected to cause sensible motion.
Even though the intraoral region presents a high loading factor, the Q-ratios (Q Unloaded /Q Loaded ) of the coils were close to 2 and were not significantly affected by the parallel plate capacitor. Therefore, sample noise was dominant for both tLoop and mtLoop. The distance between coil conductor and tissue determines the amount of parasitic losses. Here, we optimized the insulation thickness based on simulations, and the same approach could also be applied for other coil sizes and applications. The parallel plate capacitor formed by the overlapping conductors in mtLoop design improved the performance of the coil. In this study, we overlapped only the posterior section, and the resulting capacitance of 9.6 pF allowed tuning and capacitive matching. The length, geometry, and dielectric properties of the substrate between the plates can be further optimized to improve the performance by achieving a higher Q-factor or reducing the signal from tongue.
Overlapped feed port design was a significant step to achieve full dental arch coverage. Bending of the posterior section improved patient comfort and decreased the tongue signal. To further reduce the tongue signal, the posterior section of the mtLoop can be isolated with a thicker insulation layer. Another method to modify the field distribution along the coil conductor could be coating with a high-permittivity material similar to Ref. [14] Although reduced in the mtLoop design, there is still a signal contribution from the tongue. However, if MRI of the tongue is also clinically relevant for specific applications, 27 a dedicated tongue coil would be preferred. 28 Although current MR safety standards and guidelines do not explicitly mention IOCs, the guidelines for active implanted devices can be used to assess the safety of IOCs. 29 Here, we used a hotspot detection followed by temperature measurements as in Ref. [20] In general, tiers 3 and 4 of ISO 10974 guidelines could also be applied to assess MR safety of IOCs. However, in tier 4, electromagnetic simulations are performed, which is time-consuming for IOCs because small components like capacitors increase the computational burden. Unless the capacitors are modeled as lumped elements, the tier 4 is less feasible. If capacitors are modeled as lumped elements, their influence on the accuracy of the absolute SAR values is unknown. Alternatively, a transfer function approach could be applied using the experimentally detected hotspot as the focal point of the scattered electric fields. A comparison between the different safety assessments will be studied in the future.
The mtLoop was manually coated with a liquid UV-sensitive resin and resin powder mixture. To form more uniform insulation layers, the viscosity of the insulation material can be lowered such that the coil can be dipped into a resin bath; however, this imposes a further constraint on the choice of the coating material and mixture. Combination of a core insulation layer with partial patient-specific coating could result in more comfortable and stable coil placement. For example, if a registration of the patient's bite as a mold to apply the insulation layer, the teeth could be engaged perfectly to the coil surface, thus eliminating the need for biting on the coil to hold it and ensuring the relationship between the teeth and coil element are appropriately positioned. The current MR properties of the insulation were satisfactory; however, an insulation material with a lower dielectric loss could further improve coil performance.
This study was performed at 3T because it is a moderately accessible clinical field strength. It should also be applicable at lower field strengths, for example, at the clinically most available B 0 = 1.5T. However, optimal field strength for dental applications is yet to be determined because lower fields provide the advantages of longer T 2 *, shorter T 1 , and reduced metal artifacts, as well as potentially higher accessibility. [30][31][32][33] tLoop has been tested successfully at 4T (12), 34 3T (14), and 1.5T (4). Although mTLoop can be adapted in principle to other field strengths, optimization of the insulation thickness and the parallel plate capacitor formed by overlapping the coil conductors at the posterior section is required. Especially at 1.5T and lower field strengths, if the parallel plate capacitor has a low capacitance, tuning will be limited. To overcome this, the capacitance can be increased by placing thinner substrates with higher permittivity between the conductors.

CONCLUSION
Intraoral coils are indispensable components of imaging teeth and supporting structures with MRI. The modified design offers higher SNR by elevenfold and 2.5-fold at the incisors and apices of the molar roots, respectively. The modifications introduced in this study overcome anatomical constraints and improve sensitivity distribution and patient comfort.

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of the article at the publisher's website.  Figure S1: Insulation thickness optimization. Intraoral coils need to be properly isolated to prevent direct tissue contact. A thicker insulation layer cause increased distance between the tissue and the coil, reducing efficiency. Thinner insulation, however, cause increased parasitic impedance due to the losses in the tissue. For the electromagnetic properties of the MR-silent insulation material developed in this study, 1 mm thickness yielded higher sensitivity in the simulations. The line plots represent the sensitivity profile along the black line along z direction. Reduced losses in the insulation improves coil performance. Figure S2: Comparison of impression material, Gutta-percha, 3D-printed dental resin and cement phantom measurements with TSE and ZTE sequences.