–Design of an Intraoral Dipole Antenna for Dental Applications

Abstract

Objective:

In dental MRI, intraoral coils provide higher signal-to-noise ratio (SNR) than coils placed outside the mouth. This study aims to design an intraoral dipole antenna and demonstrates the feasibility of combining it with an extraoral coil.

Methods:

Dipole antenna design was chosen over loop design, as it is open toward the distal; therefore, it does not restrain tongue movement. The dipole design offers also an increased depth-of-sensitivity that allows for MRI of dental roots. Different dipole antenna designs were simulated using a finite-difference-time-domain approach. Ribbon, wire, and multi-wire arms were compared. The best design was improved further by covering the ends of the dipole arms with a high-permittivity material. Phantom and in vivo measurements were conducted on a 3T clinical MRI system.

Results:

The best transmit efficiency and homogeneity was achieved with a multi-wire curved dipole antenna with 7 wires for each arm. With an additional high-permittivity cap the transmit field inhomogeneity was further reduced from 20% to 5% along the dipole arm. When combined with extraoral flexible surface-coil, the coupling between the coils was less than −32dB and SNR was increased.

Conclusion:

Using intraoral dipole design instead of loop improves patient comfort. We demonstrated feasibility of the intraoral dipole combined with an extraoral flexible coil-array for dental MRI. Dipole antenna enabled decreasing imaging field-of-view, and reduced the prevalent signal from tongue.

Significance:

This study highlights the advantages and the main challenges of the intraoral RF coils and describes a novel RF coil that addresses those challenges.

I. Introduction

IN CLINICAL practice, imaging of the teeth and the surrounding tissue is usually performed with digital two-dimensional radiography, multi-detector computed tomography (MDCT), or cone-beam CT (CBCT) [1]–[3]. The potential of dental MRI was already discussed in 1981 [4], as MRI does not use ionizing radiation and offers excellent and adjustable contrast, but it has been clinically implemented only in the recent years due to the increasing number of applications where high resolution soft tissue contrast is needed, and because it is becoming technically and economically more feasible. The feasibility and advantages of dental MRI have been demonstrated in various applications including: pre-surgical planning, implant planning, assessment of dental and periapical anatomy and pathology, localization of impacted teeth, evaluation of dental vitality, tumor detection, and imaging of the morphology and function of the temporomandibular joint [5], [6], [15]–[20], [7]–[14].

Dental MRI is also potentially useful for investigating hard tissues, such as cortical bone and dentin [21], [22]. Unfortunately, in bone the rapid decay of the MR signal requires an increase in the encoding bandwidth. For bone MRI FID-based imaging methods have been proposed, which are limited by finite acquisition delays, i.e., dead time, and bandwidth due to limited radio frequency (RF) power amplifier performance, and the gradient slew rates (for reviews and comparison [23]–[28]). Imaging techniques, such as SWIFT and its variations, use a frequency-modulated excitation and a simultaneous acquisition to acquire signal from tissues with very short T2 or T2* [29]–[34]. All these pulse sequences are based on radial encoding, and cannot use sliceor slab-selective pulses to preserve signal from tissues with ultra-short T2. Thus, the acquired field-of-view (FOV) must include the entire sensitive volume of the RF coil to avoid aliasing. The resolution of an image voxel depends on the FOV and the reconstructed matrix size. Increased FOV results in highly increased scan times to achieve the desired image resolution with similar SNR [35]. Therefore, to achieve higher resolution in reasonable scan times, smaller and more localized RF coils must be used.

The SNR provided by a coil is essentially proportional to the geometric filling factor, η, which is a measure of the fraction of the RF coil volume occupied by the sample; and the quality factor, Q, of the coil [36].Although there is a clinical need for dental MRI, and advanced short-T2*-sensitive pulse sequences for bone imaging can be implemented in most of the clinical MRI systems, commercially available RF coils are often not compatible with the dental anatomy. A coil configured to provide a limited and sculpted FOV is highly desirable to solve this problem.

In dental MRI, submillimeter structures (e.g., small fractures within dental roots) must be resolved. To achieve this high resolution within clinically acceptable measurement times, the imaging FOV must be restricted to the target region which can be realized by a small RF excitation field, $B_1^{+}$, a limited receive sensitivity B₁, or a combination of both. So far mainly extraoral surface coils and coil arrays have been used [37]–[39]. For an average-sized patient, the distance between an extraoral coil element and a molar tooth is about 30–50 mm, which reduces the sensitivity significantly. In addition, with an extraoral configuration, the cheek and buccal fat produce intense signal. As a result, the signal from these unwanted tissues dominates the images. To overcome this limitation an intraoral coil was proposed that is positioned between the teeth and cheek, i.e., in the buccal vestibule which increases both resolution and SNR [40]. By sacrificing patient comfort, as well as SNR, the intense signal from the cheek could be eliminated in a similar intraoral configuration with shielding [41]. In [42], an inductively coupled intraoral coil was developed. The inductively coupled coils were designed to image a single tooth and to be used with external coil elements. However, the tuning of the coil for each patient is challenging, as it cannot be modified once the coil is sealed. Furthermore, these coils can only acquire MR signal from the apices of posterior teeth due to anatomical constraints.

For better coverage and patient comfort, an intraoral loop coil with the coil plane orthogonal to B₀ was developed, where the transverse B₁ fields are sensitive to MR signal [43]. In most standard uses of surface loop coils, the coil plane is oriented parallel to B₀, and B₁ field components orthogonal to the coil plane are used for RF excitation and detection, whereas in [43], the unconventional placement of the loop coil intraorally and orthogonal to B₀, resulted in transverse components of B₁ which were effectively used for imaging of teeth and associated structures. The sensitive volume of the loop coil included the most important dental structures, such as the teeth and jaws, and mostly excluded cheeks and lips. However, the sensitive depth did not reach the root apices of molar teeth, and the loop obstructed the tongue movement resulting in patient discomfort. Although there are standard dental procedures where the tongue is restricted, the patient is always supervised regarding gagging and management of saliva. Nevertheless, during MRI measurements, patients are not supervised and the dipole antenna design, proposed below, is advantageous because it does not have the posterior component restricting the tongue.

In this study, we developed a novel intraoral dipole antenna and improved its performance based on the B₁ field simulations. The coil was designed to achieve a high transmit field homogeneity and penetration depth to be able to image the dental roots. For the best-performing dipole, a prototype was constructed and compared to the reference loop coil in phantom and in vivo measurements. The dipole was also combined with a flexible surface coil receive (Rx) array to increase sensitivity.

II. Methods

A new intraoral coil design was realized with the following target features: the new coil should allow for tongue movements which are impossible with a loop design, it should provide a uniform sensitivity over the teeth and surrounding tissues, it should have a penetration depth which allows sensitive detection of the signal from dental roots, and it should minimize coil coupling with exterior Rx coil elements. A dipole design was chosen, and different dipole designs were simulated using an electromagnetic (EM) simulation tool based on finite-difference-time-domain (FDTD) approach (Sim4Life v5.0, ZMT, Zurich, Switzerland). The prototype coils were tested on a clinical 3T MRI system (Prisma, Siemens, Erlangen, Germany) in a homogeneous phantom and a human volunteer.

A. FDTD Simulations

5 dipoles with planar structure (Fig. 1(b)–(f)) and 8 with multi-planar structure (5 of which are shown in Fig. 2(a)–(e)) were designed based on average human adult mandibular dimensions [44]. A list of the dipoles is given in Table I. For comparison, the loop coil in [43] was modeled as the reference coil (Fig. 1(a)). The size of the dipole is not a free design parameter, because of the fixed size of the dental arch. In our design, we chose a size that corresponds to size “M” impression tray, which is a commonly used size for adults in the USA. This size also fits the Duke anatomical model. The loop and the dipoles have the same footprint, i.e., width of 60 mm. The dipole arms are formed either by a 10 mm-wide ribbon, a single wire (Ø1.2 mm), or multiple wires separated by 5 mm. Multi-planar dipoles have additional wires (Ø1.2 mm) at a distance of 8.6 mm to the central plane. Total length of the wire and ribbon dipoles are 123 mm, except the meandered wire, which is 188 mm-long. The resonant length of a half-wave dipole immersed inside a medium of ε_r = 78 would be 138 mm. To increase the effective wavelength of the dipole, tuning inductors are used as shown in Fig. 1.

Fig. 1.

Comparison of the intraoral loop and planar dipole coils. (a) Reference loop coil model with the line profile tracks and levels marked. Planar dipole models with (b) ribbon arms, (c) wire arms, (d) meandered wire arms, (e) 3-wire arms, and (f) 3-wire arms with shorted ends. Corresponding field maps are shown in the same column as the coil models. (g)–(h) Sensitivity within the tongue region is lower in dipoles compared to the loop coil. (m)–(r) Ribbon and multi-wire conductors result in a wider sensitivity across the dipole arms.

Fig. 2.

Multi-planar dipole models and corresponding field maps. (a) 7-wire, (b) 7-wire with shorted ends, (c) ribbon-based, (d) ribbon-based with shorted ends, and (e) 7-wire dipole with high-ε_r cap. Corresponding field maps are shown in the same column as the coil models. (f)–(j) High-permittivity cap at the end of the dipole arms results in a more homogeneous sensitivity distribution along the arms. (k)–(o) Fields on the coronal slice at the end of the dipole arm also shows the superior performance of the multi-planar dipole with high-permittivity cap in transmit efficiency and homogeneity.

Table I.

Results of the Line Profile Comparison for Simulations of 13 Dipole Antenna Designs and the Reference Loop Coil Design. Values That Represent the Intensity Are Color-Coded With Green, Whereas the Inhomogeneity, i.e., Deviation of the Intensity, Is Rated With Red Color Scale. The Columns Are Grouped Into Two for Color Coding

Coil Form		Maximum $\Vert B_1{}^{+}\Vert /P_{\text{in}}$ along axial profile (μT/W)		$\Vert B_1{}^{+}\Vert /P_{\text{in}}$ mean absolute deviation along axial profile (%)		Peak $\Vert B_1{}^{+}\Vert /P_{\text{in}}$ along longitudinal profile (μT/W), $\Vert B_1{}^{+}\left(z=11\text{mm}\right)\Vert$ for multi-planar		$\Vert B_1{}^{+}\Vert /P_{\text{in}}$ variation along longitudinal profile, i.e., $100*\left(1-\Vert B_1{}^{+}\left(z_2\right)\Vert /\Vert B_1{}^{+}\left(z_1\right)\Vert \right)$
		z1 = 10 mm	z2 = 25 mm	z1 = 10 mm	z2 = 25 mm	y1	y2	y1	y2
Planar coils	Loop	1.453	0.512	9.7	14.6	0.735	0.928	76.9	70.5
	Ribbon	1.680	0.531	17.8	14.4	1.482	0.774	66.1	62.6
	Wire	1.703	0.497	18.8	15.6	1.510	0.724	69.4	64.4
	Meander	1.620	0.477	17.4	13.7	1.423	0.755	68.2	64.7
	3 wire	1.692	0.531	17.4	14.6	1.455	0.762	65.8	62.2
Multi-planar dipoles	3 wire-short	1.687	0.539	17.3	14.6	1.477	0.787	65.7	62.1
	7 wire	1.718	0.736	21.3	19.7	1.377	0.608	53.8	49.2
	7 wire short	1.678	0.734	16.8	17.2	1.373	0.762	52.3	52.8
	Ribbon	1.667	0.725	21.2	19.7	1.322	0.600	52.7	49.4
	Ribbon-short	1.661	0.731	16.5	17.2	1.361	0.772	52.1	53.6
Multi-planar dipoles with high εr	7 wire	2.589	0.778	4.9	8.1	1.605	1.507	52.9	60.9
	7 wire short	2.565	0.787	4.8	7.8	1.638	1.527	53.5	60.2
	Ribbon	2.487	0.773	5.0	8.1	1.602	1.484	53.2	61.0
	Ribbon-short	2.526	0.786	4.9	7.8	1.636	1.507	53.6	60.2

The transmit field $B_1^{+}\left(x,y,z\right)$ efficiency, i.e., transmit field amplitude over conducted input power, $B_1^{+}/P_{in}$, were compared between the reference loop coil and dipole antennae with arms formed by ribbon, wire, meandered wire, and multiple wires in parallel with open and shorted ends (Fig. 1(a)–(f)). In all simulations, the direction of the static field B₀ is assumed to be in z direction. Tuning was achieved by a parallel capacitance, and series inductance for the loop and the dipoles, respectively (Fig. 1). Tuning inductance required for ribbon and single-wire dipoles are 116 nH and 226 nH, respectively. Meandered wire required 168 nH inductance due to the longer conductor compared to single wire, as expected. 3-wire dipole required 93 nH, whereas multi-planar dipoles required only around 40 nH. Shorted-ends did not change the resonant length significantly, whereas multi-planar dipoles with high-permittivity cap at the ends required only 15 nH for tuning. Impedance matching was done by adjusting the input impedance of the edge source defined between the poles of the input port.

To assess the field homogeneity of the coils, the transmit efficiency was plotted along a curve along the center of the coil’s conductor. This curve covers a full circular path that starts at the feed port of the coil and tracks the coil conductor in clockwise direction. Two curve profiles were extracted at transverse planes that are 10 mm and 25 mm distant from the coil (Fig. 1(a)). Note that the average human adult tooth length is 22 mm thus, the 25 mm-plane covers the apical section of the roots. The relative change of $B_1^{+}\left(z\right)/P_{in}$ at z₁ = 10 mm and z₂ = 25 mm was calculated as a figure of merit of the coil homogeneity. Similarly, longitudinal profiles were determined at two different levels along a line orthogonal to the coil plane (Fig. 1(a), marked as y₁ and y₂).

To increase the longitudinal depth-of-sensitivity, additional wires were added to the multiple-wire design on superior and inferior planes so that the dipole has a multi-planar structure similar to dental impression trays (Fig. 2(a)–(d)). As both the upper and the lower dental arches are imaged in dental MRI, an “H” shaped cross-section was used compared to “U” shaped cross-sections of impression trays, which cover only one arch. Again, 4 different versions of the dipoles were simulated: ribbon-based central arm, multiple-wire-based central arm, and shorted-end versions of those (Fig. 2(a)–(d)).

In general, dipole antennae have a very high impedance at the end of the arms resulting in a steep decrease in RF current. The rapid current change causes $B_1^{+}$ inhomogeneities along the dipole arms that can be reduced if a smooth transition of the impedance is set up along the arms. In [45] high dielectric constant (i.e., permittivity, ε_r) materials were attached to the end of the arms to reduce electric fields, thus to prevent tissue heating. In this study, we applied dielectric coating to improve $B_1^{+}$ homogeneity – therefore, dipole antennae were simulated with open and short-end multi-wire designs (Fig. 2(e)) where lengths of 10, 15, 20, 30 mm of the tip of the arms were coated with barium titanate (BaTiO₃, ε_r = 300) to study the potential increase in field homogeneity.

From all simulated designs, the most favorable design was chosen according to the highest penetration depth and field homogeneity criteria. Full simulation protocols and CAD models of the coils can be found in: https://github.com/alibaz/DentalMRI.

To ensure safe coil operation, 10g-averaged specific absorption rate (SAR_10g) values were simulated for the loop and the multi-planar dipole coils in homogenous phantom and the Duke anatomical model (Duke cV3.1, IT’IS Foundation, Zurich, Switzerland). The intraoral coil models were placed in the mouth of the anatomical model after slight modifications. The teeth in the upper and lower jaws of the anatomical model were shifted by 2 mm and −2 mm, respectively, leaving a gap of 4 mm to accommodate the intraoral coil with an insulation layer of 2 mm and a total thickness of 3.2 mm. In case of overlap of the intraoral coil and the tissue (e.g., gums), a prioritization was implemented during the voxelization, where, the coil structure was set to supersede the tissue voxels. For high-permittivity cap at the end of dipole arms, ε_r = 270 was set.

B. Coil Construction

For the imaging experiments the best-performing design with the high-ε_rcap and the reference loop coil were built. The reference coil was laser-cut from a 1 mm-thick copper plate and was then soldered on a single-sided PCB for capacitive tuning and matching (Fig. 3(a), Supporting Information Fig. S1). For the dipole, a form with wire tracks was constructed from UV-cured resin using a 3D printer (Form 3, Formlabs Inc., Somerville, MA). Enameled copper wires (Ø1.2 mm) were inserted and glued within the tracks. Wires were connected at the input side of each arm, and then soldered to a single-sided PCB for inductive tuning and capacitive matching. For the implementation with a high-ε_r cap, the base form was shortened, an H-shaped hollow container was filled with BaTiO₃ and D₂O slurry (4:1 mixture - ε_r = 274 was measured with DAKS 12 – SPEAG, Zurich, Switzerland) and attached to the base form. Note that, the slurry was prepared at 40 °C, and the measured permittivity change was less than 3.6% at room temperature. The dipole arm ends were shorted by soldering short wire segments between the neighboring wires. The conducting parts were then isolated with an orthodontic resin and powder (Caulk, Dentsply Sirona Global, Charlotte, NC). Finally, the whole dipole was sealed with a silicon putty (Rapid Soft, Coltène/Whaledent AG, Altstaetten, Switzerland) (Fig. 3(b)). Photos from the construction steps are shown in Supporting Information Fig. S2. Schematics of the loop and dipole are shown in Supporting Information Fig. S1a, b.

Fig. 3.

A photo of (a) the reference loop coil, (b) 7-wire dipole with shorted ends and high ε_r cap (not visible), and (c) the S-parameter measurements. (d) An illustration of the coil positioning for dipole combined with SLR array, (e) S-parameter and (f) Noise correlation matrices of the SLR array.

In ultra-high field MRI the performance of dipole antennae is improved by combining them with loop receive coils [46], [47]. To apply this concept to dental MRI at high fields, a 4 channel flexible transmission line resonator (TLR, i.e., shielded loop resonators-SLR) was designed. Similar to loops and dipoles, SLRs and dipoles are geometrically decoupled when coil planes are parallel to each other. We propose to use the dipole with high-ε_r cap as the Tx element, and the SLR array as the Rx element for optimal performance. Details of the SLR design and construction can be found in Supplementary Material, Supporting Information Fig. S3.

S-parameter measurements were performed in loaded and unloaded conditions of the coils using a network analyzer (ZVB4; Rhode & Schwarz GmbH & Co. KG, Munich, Germany). For loading, the intraoral coils were immersed in a phantom made of distilled water with 2g/l NaCl, and 5g/l NiSO₄.

C. MRI Measurements

Loop coil and dipole antenna were interfaced to a clinical 3T MRI system (Prisma Fit, Siemens, Erlangen, Germany) via a custom made Tx/Rx switch [27], [48], and the 4 channel SLR array was connected to the system via commercially available preamplifiers (Stark Contrast, Erlangen, Germany).

To compare the imaging results with the simulations, at first the phantom experiments were performed. A PVC bridge was fabricated and attached to the MRI patient table to fix the intraoral coils in the phantom (see Methods, B. Coil Construction) – the same bridge was later used in the volunteer experiments to fıx the intraoral coils and minimize head motion. The dipole was placed through an opening on the fabric of the flex SLR array to avoid displacement of the flex coil and the phantom.

To compare the transmit efficiency of the intraoral loop and dipole to the system’s body resonator, a flip angle mapping measurement using double angle method [49] was performed, where the flexible SLR array was used as the Rx element for all three Tx coils. The flip angle was then converted to $B_1^{+}$ value, which is then normalized to the supplied input power as reported by the MRI system.

Then, SNR maps were acquired as described in [50] for loop and dipole in Tx/Rx mode, SLR only, and dipole as Tx and SLR as the extraoral Rx element. A 3D GRE sequence was used (Please see Supplementary Material for full sequence protocol). In vivo measurements were performed with the reference loop coil, the dipole antenna with the best-performing design, SLR array alone, and the SLR array combined with the loop coil and the dipole antenna. For imaging, a 3D spoiled gradient echo (GRE) sequence was used with the following parameters: 0.6 mm-isotropic resolution, α = 20°, TR = 25 ms and TE = 3.14 ms. The FOV could be set as 160 × 160 86.4 mm³ with 144 slices and the total acquisition time amounted to 4:19 min:s. The panoramic slices were created by using the “Straighten” function [51] of ImageJ, which is a Java based image processing program [52]. To demonstrate the compatibility with short T2_* imaging sequences, an ultra-short TE (UTE) sequence protocol was also implemented with TR = 7.96 ms and TE = 60 μs. The nominal resolution was 0.9 mm-isotropic, FOV was 110 × 110 × 110 mm³, α = 6°, and 128 points per radial spoke were acquired resulting in a total scan time of 8:41 min:s. Full protocols are available in the Supplementary Material. All measurements were performed in accordance with the institutional guidelines and the local ethical review board (University of Minnesota, Human Research Protection Program, STUDY00005298).

Based on the SAR_10g simulation results, the transmit power limits for the RF coil were set such that the corresponding maximum SAR_10g value was at least 20-fold less than the input power that generates maximum SAR_10g of 10 W/kg (averaged over 10 g tissue and 6 min [53]). Note also that the minimum distance between the volunteers’ lips and the closest capacitor, which is the highest local E field, was 14 mm and 37 mm for the loop and the dipole, respectively. Peak 10g-averaged SAR calculator of Sim4Life 5.2 software was used to calculate the spatial-average of SAR with moving constant-mass cubes as recommended in [54]. SAR_10g maps and SAR efficiency maps (i.e., $B_1^{+}/SAR$ in μT kg^0.5/W^0.5) of the loop coil and the dipole are shown in Supporting Information Figs. S11 and S12, respectively for a homogeneous phantom.

III. Results

A. FDTD Simulations

In Fig. 1, the coil designs (a-f) are shown together with transverse (g-l) and coronal (M-R) transmit efficiency $B_1^{+}/P_{in}$ maps. The loop coil shows a homogeneous field distribution along the coil conductor with an increased $B_1^{+}/P_{in}$ in the conductor section opposite to the feed port. For dental imaging, axial line profile between [0°–130°] and [230°–360°] are relevant (Fig. 1(b)). In Table I, the maximum $B_1^{+}/P_{in}$ along the axial profiles at z₁ = 10 mm and z₂ = 25 mm planes, mean absolute deviation along the axial profile, peak $B_1^{+}$ along the longitudinal profile at levels y₁ and y₂, and longitudinal variation calculated as $100\left(1B_1^{+}\left(z_2\right)/B_1^{+}\left(z_1\right)\right)$ are given. Corresponding line profile plots for planar dipoles, multi-planar dipoles, and multi-planar dipoles with high-ε_r cap are shown in Supporting Information Figs. S4, S5, and S6, respectively.

In Table I, for the planar dipoles, 3-wire designs have the highest $B_1^{+}$, yet mean absolute deviation is lowest for the meander dipole. However, due to the near field (observed at transverse planes at z < 5 mm) inhomogeneities of the meander dipole (Supporting Information Video S1), this design was left out. 3-wire and ribbon dipoles have comparable deviation and maximum $B_1^{+}$ values.

In Fig. 2(f)–(j), transverse $B_1^{+}/P_{in}$ maps of the multi-plane dipoles are shown, which is superior to the planar dipoles in terms of maximum $B_1^{+}/P_{in}$. Unlike the planar 3-wire dipole with shorted ends, 7-wire dipole with shorted ends has a more homogeneous axial profile than other multi-planar designs (Table I). 7-wire dipoles have higher maximum $B_1^{+}/P_{in}$ values, whereas the ribbon based multi-planar design performs slightly better in terms of longitudinal homogeneity.

The problem of the inhomogeneity along the axial profile is resolved with the addition of a high-ε_r cap, as shown in Table I. According to the simulations, high-ε_r cap of 15 mm length results in a more flat axial profile along the dipole arms (Fig. 4). Fig. 2(k)–(o) shows longitudinal profiles along a plane cutting the 15 mm-long high-permittivity cap at the center. Here 7-wire dipole with high-ε_r cap has higher intensity compared to other 7-wire and ribbon-based multi-planar designs. From Table I, a 7-wire dipole with shorted ends and 15 mm-long high-ε_r cap was selected as the best-performing design as it has slightly higher homogeneity and comparable maximum $B_1^{+}/P_{in}$ as the open-end design, and was constructed for MRI measurements. Axial and longitudinal line profiles of the selected dipole designs and the loop design are shown in Supporting Information Fig. S7.

Fig. 4.

Line profile comparison along the dipole arm for different lengths of the high-permittivity cap at the end of the multi-planar 7-wire dipole arm shows that the field homogeneity along the dipole arm can be improved by adjusting the length of the high-permittivity cap.

SAR_10g maps for loop and dipole coil in the Duke anatomical model are shown in Fig. 5. The values are normalized to 0.02W input power, which is calculated from the peak voltage level of the RF pulse used in the measurements. Slices with the highest SAR_10g are shown in transverse and coronal planes. The color maps are scaled to the maximum SAR_10g. In the loop coil, power deposition around the feed port is higher than the other regions (Fig. 5(c)–(d)). In the dipole, power deposition increases around the high-ε_r cap of the dipole arms (Fig. 5(a)–(b)). To investigate the effect of the high-ε_r cap further, another set of simulations, where the high-ε_r cap was excluded was performed (Supporting Information Fig. S13). The power deposition was comparable for dipoles with high and low ε_r, i.e., 0.17 W/kg and 0.20 W/kg, respectively. However, the power deposition in low ε_r dipole was higher around the feed port and the sections of the dipole arms close the feed port.

Fig. 5.

SAR_10g maps in Duke anatomical model for the intraoral dipole in (a) transverse and (b) coronal planes. SAR_10g maps for the intraoral loop in (c) transverse and (d) coronal planes – highest SAR is observed around the lips. The planes are selected with the maximum values; and the color bars represent also the dynamic range of the SAR_10g results. Local SAR tends to increase around the feed ports and the high-permittivity cap.

B. Coil Construction

S-parameter measurements are shown in Fig. 3(c). Unloaded and loaded Q factors for loop, dipole, and SLR elements are Q_unloaded/Q_loaded = 70.8/6.2, 8.2/3.9, 64.5/20.3, respectively. SLR elements and dipole are decoupled by 33±4 dB when positioned as illustrated in Fig. 3(d), and the reflection coefficient of the dipole is not affected from the nearby SLR elements. S-parameter and noise covariance matrices for the SLR array are shown in Figs. 3(e)–(f). The range of coupling between overlapping/non neighboring elements was −21.3–13.5 dB/−18.3–13.5 dB, to which preamplifier decoupling added another 23±5 dB isolation. The active detuning efficiency was measured as 48±4 dB for all elements. Noise correlation values ranged from 3% to 9% (Fig. 3(f)).

C. MRI Measurements

$B_1^{+}$ efficiency maps in μT/W are shown in Fig. 6 for the transverse slices 10 mm away from the coil planes. The intraoral dipole and loop have comparable maximum $B_1^{+}$ efficiency of 388.2, 384.6, 6.9 μT/W, respectively. The dipole has a smaller excited volume than the loop; however, both intraoral coils have higher $B_1^{+}$ efficiency and smaller excitation volume compared to the body resonator of the system.

Fig. 6.

Measured $B_1^{+}$ efficiency of the intraoral dipole (left), loop (center) and system’s body resonator (right) normalized to the output power of the RFPA. As expected, local coils are more efficient than the body coil. Dipole sensitivity drops significantly towards the back and at the medial sulcus of the tongue.

Transverse slices at z = 10 mm of the phantom images are shown in Supporting Information Fig. S8b–f. Mean SNR in the marked region of interest shown on the images were 271.9/32.3/440.1/861.7/645.7 for the loop, the dipole, SLR only, and the SLR combined with the loop and the dipole, respectively. At z = 20 mm, the SNR values for the SLR combined with loop and dipole are 287.8 and 924.2, respectively, due to the inhomogeneous flip angle distribution along the longitudinal direction.

In Fig. 7(b)–(f), in vivo images of the human mandible are shown in a transverse and a coronal plane. The transverse plane was positioned close to the intraoral coil plane (z = 4 mm), and the SNR values from dentine at the right first molar were measured and found to be 14.8/4.6/40.5/141.7/189.6 for the loop, the dipole, SLR only, and the SLR combined with the loop and the dipole, respectively. The dipole has an inferior SNR when used as a Tx/Rx element, which is expected from the low Q factor. Coronal profiles, on the other hand, show that the dipole combined with SLR has a higher depth–of-sensitivity than the loop, in agreement with the phantom SNR comparison of the transverse planes at z = 10 mm and z = 20 mm. The combination of the dipole with extraoral SLR Rx array improved the SNR without the need for increasing the FOV. Signal from tongue, which can cause unwanted artifacts, was also reduced in the dipole compared to the loop coil. When the intraoral coils are combined with the extraoral flexible SLR array, there is a signal enhancement around the feed port, close to the phantom surface, which is more noticeable in the dipole and SLR combination (Supporting Information Fig. S8f). A signal enhancement around the feed port is visible also in Fig. 7(f), where image intensity of lips and frontal gums is higher than the other tissue. Note that signal from tongue could not be suppressed completely in the dipole. Line profiles of the contrast-to-noise ratio (CNR) of the tongue to the first molar teeth reveal that the edges of the tongue are still prominent compared to signal at the medial sulcus of the tongue, since the sensitivity of the dipole decreases gradually towards the center of the tongue (Supporting Information Fig. S14).

Fig. 7.

A photo of the in vivo measurement setup (a). SNR maps at z = 10 mm away from the coil planes for intraoral (b) loop, (c) dipole, (d) extraoral SLR only, and SLR combined with (e) loop and (f) dipole in Tx mode. The image intensities are scaled to the same range to allow a visual SNR comparison. SLR combined with the intraoral dipole provides higher SNR and less signal from the central region of the tongue.

In the GRE images, the dental restoration in the left first molar is seen as a susceptibility artefact (Fig. 7(f)), which is reduced in the corresponding UTE images (Supporting Information Fig. S9). A panoramic view of the in vivo images acquired using the SLR and dipole combination are shown in Fig. 8. An animated overview covering all the slices of the GRE and UTE images is presented as Supporting Video S2.

Fig. 8.

A panoramic view of the in vivo GRE images acquired using the SLR and dipole combination. Yellow line shows the cross-section of the folding plane (a). Dipole’s sensitivity reaches the roots as shown in the coronal view (c).

IV. Discussion

In this study, a dipole antenna was designed to overcome the limitations of current intraoral coil designs based on loop coils. The results show that a dipole antenna with 7 wires and a high-ε_r cap can be potentially useful in dental MRI at 3 T. Although dipoles have lower SNR compared to loop coils, it was demonstrated that when they are used as a Tx-only element in combination with an extraoral SLR array, SNR can be improved, patient comfort can be improved, and small FOVs can be used with conventional pulse sequence protocols.

From the outset of the coil designs, planar designs with various arm shapes were compared in EM simulations. The multi-wire approach was inspired by UHF antenna designs; however, a unique design was developed to generate a region with homogeneous sensitivity that covers molar teeth. Other unique design features are the use of additional wires around the teeth that extend the depth-of-sensitivity toward the roots, and the use of high-ε_r cap at the tip of the dipole arms. In the comparison, the wire type, the conductivity, the distance between the upper wires and the main arm at the central plane, and the dielectric constant of the high-ε_r cap were not varied. A higher conductivity and a lower wire resistance would lead to a higher Q for the coil but would not substantially change the $B_1^{+}$ profiles. The distance between the wires is limited by the oral anatomy. The permittivity is already very high in barium titanate and is difficult to increase with biocompatible materials. Thus, the current comparison should already provide a good approximation for the desired sensitivity profile of the coil. Note that, in the simulations, for an 8.7% change in ε_r, absolute changes are within 2% and 3% $B_1^{+}$ and SAR_10g, respectively. Therefore, for a substantial modification of the field distribution, materials with substantially higher permittivity should be investigated [55]–[58].

The 3D UTE imaging with (110 m)³ FOV proves that the coil system with the intraoral dipole as Tx element and the extraoral SLR array as the Rx element can be used with high-resolution radial imaging sequences in vivo. The low Q of the dipole is even advantageous for the ultra-short TE measurements as the Q is proportional to the ring down time. Ring-down time might be a limiting factor in minimizing the acquisition delay in ultra-short TE sequences such as UTE and SWIFT.

Direct imaging of calcified tissues by using echo-based conventional MRI, such as GRE, is not possible due to the ultrashort relaxation time of transverse magnetization (for enamel, T2_* ~0.07 ms), which is usually shorter than the echo time (TE = 3.1 ms, in this study). However, studies have shown that conventional MRI can be used clinically to visualize, for example, pulp structure and vitality [8], to reveal the anatomy and pathology of the dento-alveolar [59], to detect osteomyelitis in the mandible [60], and to indirectly image highly mineralized tissue through contrast produced by tissue with long relaxation times [10], [40], [59], [61]–[63]. On the other hand, the performances of MRI methods for obtaining images of densely calcified dental tissues with positive contrast such as UTE and SWIFT are still limited by the electronics and digitization chain of the clinical MRI systems. For the UTE images acquired with 0.06 ms echo time, a certain signal is expected from the dentin with T2_* of about 70 μs [27], [64], yet, due to the blurring from rapid decay of magnetization and the high signal intensity of the surrounding tissue, the available contrast is hiding these parts of the teeth. Alternatively, two UTE images acquired with short and long TEs could be subtracted to highlight the densely calcified tissues [65]. Furthermore, patient motion must be addressed in clinical dental imaging by adaptive retrospective correction of motion artefacts when high-resolution imaging of small structures with ultra-short T2_* is desired. MR signal from the intraoral coil housing can also obscure dental tissues. Signal from coil former and insulation materials should either be minimized by using 1H-free compounds or be adjusted using contrast agent mixtures.

In general, the Tx homogeneity of the dipole at 3 T is inferior to a volume coil such as a birdcage or solenoid. However, larger volume coils would also lead to large FOVs and unwanted signal contributions from the tongue and other surrounding structures, so that the combination of the inhomogeneous dipole with an extraoral receive array is still advantageous. The Tx inhomogeneity of the dipole can be partially overcome by using $B_1^{+}$ insensitive excitation pulses and encoding methods that improve the homogeneity [29], [66]–[68].

At the feed port, high signal from lips and surrounding tissues can be observed, which might be difficult to remove in clinical imaging. High signal from these spots could be reduced by RF shielding of the field port, or the feed port could be modified to form a strip line, where the electric and magnetic fields are constrained between two grounds, so that the sensitivity to external signals is minimized.

There are several challenges to produce a re-usable intraoral dipole antenna. First, the coil must allow for easy disinfection without damaging the isolation layers. Therefore, the isolating cover needs to be chemically and mechanically more stable than the one used in this prototype implementation. Alternatively, it can be designed as two parts, a dipole part and the circuitry part, that are linked via a connector. The circuitry could be re-usable, whereas the dipole part is disposed after in vivo use, similar to rectal coils for prostate MRI. Another challenge is to reproduce the high-permittivity cap. The current method of filling a hollow H-shaped container with a BaTiO₃ slurry is mechanically fragile, and the wires inside it can be unintentionally bent, which can affect the performance of the coil. In the future, the high-permittivity cap can be made of a non-toxic ceramic or another high-permittivity biocompatible compound.

Human mandibular size and shape can vary depending on several factors including, but not limited to patient age, ethnicity, and gender. Thus, the dipole size is not a free parameter, but must be chosen to fit comfortably inside the mouth and cover the dental arches for proper diagnosis. Dental impression trays, for example, are commonly available in five sizes denoted as XS, S, M, L, and XL, which covers most of the population. A similar range of sizes could also be offered for intraoral coils. Dipole sensitivity is not significantly affected by the size of the arms (see Supporting Information Fig. S10, depth of sensitive region is plot for S, M, and L size coils), so the results of this comparison study can be extended to dipoles with arms that are longer or shorter by about 20%. It is a limitation of the multi-planar dipole design that not only the size, but also the form and the shape of the antenna structure will be determining if the antenna fits the patient’s mouth without stressing the tissue, which could be a major source of discomfort during clinically relevant measurement times. To eliminate this effect, development of multi-planar dipole antennae with using flexible conductors and formers are in the scope of future research. To compensate for the tuning and matching offsets in such a flexible arrangement, automated tuning and matching technologies are favorable [69]–[71].

Most of the electric fields generated by the intraoral dipoles are dissipated within the isolating cover and the surrounding tissue. Even if the total power delivered to the coil is low due to the high filling ratio and the choice of the pulse sequence, local E field hot spots might be formed within the mouth, which could lead to unwanted local heating when RF pulses with high peak power are used. High local SAR was observed at the input port of the loop and at the end of the dipole arms as shown in the preliminary SAR simulation results for the loop and dipole models used in the volunteer measurements (Supporting Information Fig. S11). SAR efficiency of the dipole is therefore decreasing towards the end of the dipole arms (Supporting Information Fig. S12). The simulations performed using the Duke anatomical model also revealed similar distribution of power deposition in the tissue.

High-intensity E fields are observed close to the conducting surfaces and capacitors. Although no temperature increase was observed during the MRI measurements in this study, temperature measurements must be performed at these potential hot spots, and, additionally, the design must be locally changed to minimize direct tissue contact before translating this technology to a clinical study. The thickness and material properties of the isolation layer can be optimized to reduce local SAR. E fields around the capacitors and the feed ports can be reduced either by keeping the tissue away from hotspots, or shielding the feed port similar to an embedded coplanar strip line design, which would keep the E fields within the parallel ground planes. Especially for high power demanding sequences, the SAR distribution needs to be optimized. A limitation of this study is that we evaluated only the sensitivity profile of the dipole by simulating dipoles with different lengths of high-ε_r caps. In future studies, the power deposition should be optimized in parallel, to avoid hotspots. A more comprehensive RF safety analysis including thermal simulations with different computational human models and temperature measurements is beyond scope of this manuscript but needs to be performed before intraoral loop coil and dipole antennae can be used in a larger patient population.

The results of this study might also be interesting for the design of ultra-high field RF coils, as dipoles are frequently used at 7 T and higher field strengths [72], [73]. Although an intraoral coil has a much higher filling factor than dipoles used for body or head MRI, techniques for increasing penetration depth and for improving homogeneity can be adapted to other coils.

Figs. 7 and 8 present the first in vivo MRI images using an intraoral dipole antenna. Combination of a dipole antenna and SLRs were also implemented for the first time, to the best of our knowledge. The spatial resolution can be increased in the expense of acquisition time. Although there is no hardware limit to achieve a higher resolution, the effects of the residual motion and off-resonance blurring needs to be addressed; these can be compensated at the post processing stage for radial acquisition and are in the scope of future research.

We also performed the first SAR simulations for an intraoral coil in an anatomical human model. Intraoral coils needs to be optimized to minimize local power deposition in tissue to extend the range of the applications to power demanding sequences such as RARE [74]. Since the intraoral dipole antenna is set very close to the sensitive organs and tissue such as tongue, RF energy deposition and local SAR values need to be monitored carefully. Limiting the transmit power to ensure safe operation may also limit the maximal flip angle in magnetization prepared and spin-echo-based sequences, which is a critical aspect of the intraoral coils and needs to be addressed in future studies.

V. Conclusion

We demonstrated feasibility of using a dipole antenna as an intraoral coil and improved the transmit efficiency and homogeneity of this coil. The intraoral dipole allows free tongue movement. Transmission efficiency of the intraoral loop and dipole are similar within 15 mm distance to the coil planes, however, dipole outperforms the loop when the distance is larger than 20 mm. The combination of the intraoral dipole and the extraoral SLR Rx array is required to achieve SNR comparable to or higher than the loop coil.