An endovaginal MRI array with a forward-looking coil for advanced gynecological cancer brachytherapy procedures: Design and initial results

Abstract

Purpose:

To develop an endovaginal MRI array that provides signal enhancement forward into the posterior parametrium and sideways into the vaginal wall, accelerating multiple-contrast detection of residual tumors that survive external beam radiation. The array’s enclosure should form an obturator for cervical cancer brachytherapy, allowing integration with MRI-guided catheter placement, CT, and interstitial radiation dose delivery.

Methods:

The endovaginal array consisted of forward-looking and sideways-looking components. The forward-looking element imaged the cervix and posterior endometrium, and the sideways-looking elements imaged the vaginal wall. Electromagnetic simulation was performed to optimize the geometry of a forward-looking coil placed on a conductive-metallic substrate, extending the forward penetration above the coil’s tip. Thereafter, an endovaginal array with one forward-looking coil and four sideways-looking elements was constructed and tested at 1.5 Tesla in saline and gel phantoms, and three sexually mature swine. Each coil’s tuning, matching, and decoupling were optimized theoretically, implemented with electronic circuits, and validated with network-analyzer measurements. The array enclosure emulates a conventional brachytherapy obturator, allowing use of the internal imaging array together with tandem coils and interstitial catheters, as well as use of the enclosure alone during CT and radiation delivery. To evaluate the receive magnetic field ($B_1^{-}$) spatial profile, the endovaginal array’s specific absorption-rate (SAR) distribution was simulated inside a gel ASTM phantom to determine extreme heating locations in advance of a heating test. Heating tests were then performed during high SAR imaging in a gel phantom at the predetermined locations, testing compliance with MRI safety standards. To assess array imaging performance, signal-to-noise-ratios (SNR) were calculated in a saline phantom and in vivo. Swine images were acquired with the endovaginal array combined with the scanner’s body and spine arrays.

Results:

Simulated $B_1^{-}$ profiles for the forward-looking lobe pattern, obtained while varying several geometric parameters, disclosed that a forward-looking coil placed on a metal-backed substrate could double the effective forward penetration from approximately 25 to ~40 mm. An endovaginal array, enclosed in an obturator enclosure was then constructed, with all coils tuned, matched, and decoupled. The ASTM gel-phantom SAR test showed that peak local SAR was 1.2 W/kg in the forward-looking coil and 0.3 W/kg in the sideways-looking elements, well within ASTM/FDA/IEC guidelines. A 15-min 4 W/kg average SAR imaging experiment resulted in less than 2°C temperature increase, also within ASTM/FDA/IEC heating limits. In a saline phantom, the forward-looking coil and sideways-looking array’s SNR was four to eight times, over a 20–30 mm field-of-view (FOV), and five to eight times, over a 15–25 mm FOV, relative to the spine array’s SNR, respectively. In three sexually mature swine, the forward-looking coil provided a 5 + 0.2 SNR enhancement factor within the cervix and posterior endometrium, and the sideways-looking array provided a 4 + 0.2 SNR gain factor in the vaginal wall, relative to the Siemens spine array, demonstrating that the array could significantly reduce imaging time.

Conclusions:

Higher SNR gynecological imaging is supported by forward-looking and sideways-looking coils. A forward-looking endovaginal coil for cervix and parametrium imaging was built with optimized metal backing. Array placement within an obturator enhanced integration with the brachytherapy procedure and accelerated imaging for detecting postexternal-beam residual tumors.

1 |. INTRODUCTION

A total of 110 000 US women per year are diagnosed with a gynecologic malignancy, of which 40% with pelvic cancers who are unresectable and require radiation therapy.¹ Worldwide, over 600 000 women per year are diagnosed with cervical cancer.² In later stage cervical cancer, the cancer has spread from its primary cervix location, and is found in the vaginal wall and the parametrium, so surgical resection becomes impractical. Treatment for patients with unresectable cervical cancer includes external beam radiation therapy (EBRT), which provides 50–80% of the required dose to eradicate the tumor. Thereafter, brachytherapy (BT) is performed, in which 50% higher radiation dose is provided to focal regions where residual tumor persists.^3–7 For interstitial BT, a Syed-Neblett template is anchored to the perineum, and afterwards an obturator is placed into the vaginal canal.⁸ Multiple BT catheters are added for dose escalation to precise tumor regions. They are pushed through designated holes in the template and through grooves in the obturator’s exterior wall, and advanced into the tumor.

MRI is the preferred modality for evaluating the local extent of cervical cancer during radiation therapy.^9,10 The post-EBRT soft-tissue environment encountered prior to the BT procedure is complex. The large tumors present before EBRT have shrunk,^11,12 leaving interspersed residual tumor cells, diffuse fibrotic regions, and denser scar tissue. As a result, conventional methods for separating between tumor and surrounding soft tissue may be insufficient.¹³

Multiparametric MR imaging, such as diffusion-weighted imaging (DWI),¹⁴ dynamic contract enhanced (DCE) perfusion imaging,¹⁵ and T₂*-weighted gradient-echo¹⁶ hypoxia scans are increasingly being pursued to detect residual tumors. Multiple high spatial-resolution sequences are acquired, with sequences that suffer from lower intrinsic signal-to-noise ratio (SNR) requiring longer scan times. Therefore, it is desirable to increase the local SNR to improve the detection of residual tumor and decrease the imaging time.

Our intent was to construct an imaging array for insertion into the vaginal canal, with its superior tip resting just below the cervix (Figure 1A). The array should improve the imaging of regions into which advanced cervical cancer has spread and are not well served with surface arrays.

FIGURE 1.

(A) A sagittal T₂-weighted MRI image of the female reproductive organs, highlighting organs of interest during the treatment of advanced cervical cancer. Dotted regions 1 and 2 highlight the regions intended to be imaged with the sideways-looking and forward-looking elements, respectively, of the endovaginal array, which is placed entirely within the vaginal canal. The endovaginal array’s enclosure was designed to act as an active obturator, fulfilling both imaging and catheter guiding tasks. (B) Schematic of the endovaginal array. The array consists of two parts, a forward-looking and a sideways-looking element. (C) Schematic layout of the forward-looking coil: The coil was built of two concentric cones. The inner plastic cone had a thin metallic surface on its exterior, which focused the magnetic field in the forward direction (towards the left). (C) Schematic diagram of the four-element sideways-looking array which was wrapped around a cylindrical polymer former. The elements were critically overlapped to minimize mutual inductive coupling

Array design required knowledge of the mean organ dimensions in adult women in the reproductive or postreproductive stages of life.^17–19 The cervix is 40 mm long and 25 mm wide. The “pear” shaped uterus, which extends primarily in the anterior-posterior direction, is 80–90 mm long and 40–50 mm wide, with the endometrium comprising its central 20 mm. The vaginal canal is 100 mm in length, and 26 mm in diameter at its narrowest region.

The MRI surface arrays used for pelvic imaging are the spine array, positioned below the back, and the abdominal array, placed on the abdominal surface. These arrays are most sensitive at distances of 80–100 mm from the posterior and anterior subcutaneous fat layers, respectively. If the anterior-posterior pelvis extends for 300–350 mm in adult women,¹⁷ a region of 100–150 mm in diameter remains, roughly in the center of the pelvic cavity, that is imaged at lower SNR. The uterus’s anterior 40–60 mm is frequently found close to the abdominal wall, and is well served by the abdominal array, so it does not need to be covered by the endovaginal array.

A solenoidal endovaginal receive-only coil, placed tightly around the cervix, improved the detection of early-stage cervical cancer by enhancing the SNR in the cervix and the upper vaginal canal.^20,21 A family of radio-frequency (RF) coils, when inserted into narrow body orifices, have RF lobe patterns that are effective at visualizing tissues around them, primarily situated parallel to the coil shaft, which we term “sideways-looking coils.” Examples are the curved-loop endorectal prostate coil^22,23 and endourethral coil.²⁴ Such coils are less appropriate for imaging regions in front of the coil.

In a study by Anderson,²⁵ a “forward-looking”coil (also referred to as a “flashlight” coil) was designed to visualize regions above the coil plane. The coil was constructed from two loop coils placed orthogonal to each other at the tip of a cardiovascular catheter and used for visualizing blocked stenotic regions of blood vessels in front of the catheter’s tip. Other forward-looking coils were developed for applications where the imaged region is located outside of the imaging array’s volume. Examples are coils for MR imaging around oil exploration boreholes²⁶ and coils for scanners with split-magnet configurations such as the General Electric Signa-SP interventional scanner.²⁷

In this study, we introduced an endovaginal receive-only coil array, intended for imaging the vaginal wall, cervix, and posterior parametrium. Due to the difficulty in inserting a coil into the cervix and parametrium, a special forward-looking RF coil was designed for imaging the cervix and posterior endometrium from within the vaginal canal. Electromagnetic (EM) simulations were used to optimize the metallic-backed forward-looking coil design for 1.5 Tesla (T) MRI scanners, utilizing specific geometric arrangements of the metal backing that can increase the forward RF lobe penetration.

The array’s plastic enclosure was configured to replace a conventional BT Syed-Neblett obturator, providing facilities for tandem coil and BT catheter placement. Note that with minor changes, we believe this enclosure can conform to other obturators. The combined array and its enclosure, therefore, form an “active” BT vaginal obturator which doubles as a gynecological imaging array. The coils and associated electronics were configurated for easy removal from within the obturator, while leaving the external obturator in place, so that the obturator can be employed, leaving the anatomy, catheters, and tandem coil unperturbed during subsequent CT/X-ray imaging as well as during radiation dose delivery. The article concludes with array validation testing in phantoms and swine models.

We see the main clinical benefits for the endovaginal active obturator for MRI-guided interventional BT procedures, if successful, as; (1) imaging with the vagina and cervix in the same shape and location as during the subsequent insertion of catheters into the tumors, which is important for accurately targeting residual gynecological tumors, while sparing adjacent viable tissues from radiation damage. (2) Several times the image SNR obtained when using commercial surface array coils, which can be used to obtain (a) shorter scan times with equivalent sequence parameters or (b) improved sequence parameters (higher spatial resolution, higher DWI B values, etc.) at equivalent imaging times, which can be very important for both detecting and characterizing residual tumors that survived external beam radiation or previous BT fractions. (3) Local arrays that have higher SNR but have lobe patterns that see limited field-of-view (FOV) have distinct advantages in MRI imaging; imaging can be performed in small regions at very high spatial resolution without folding, and shimming of the magnetic field homogeneity can be better performed over this restricted FOV, namely (a1) for the same spatial resolution, far fewer phase encodes must be collected during image acquisition, so scan times can be accelerated dramatically and (a2) there is less image geometric distortion in sequences that are sensitive to magnetic field inhomogeneity, such as all fat suppressed sequences, and all DWI, Susceptibility Weighted Imaging sequences, and Gradient Recalled Echo (GRE) sequences. This advantage increases at 3 Tesla, since establishing excellent magnetic field homogeneity is even more important there. (4) In the modern methods that exploit parallel imaging using minimally spatially overlapping coils²⁸ and their extension into compressed sensing²⁹ (e.g., methods that perform severe subsampling of k-space and produce images without artifacts), the availability of high SNR in all regions of the image is key to the large reductions in scan time that these methods afford. In regions of the pelvis that are far from the abdominal surface or spine, this endovaginal array can provide this increased SNR.

2 |. MATERIALS AND METHODS

2.1 |. Endovaginal array design and modeling

The proposed endovaginal MRI receiver array, positioned entirely inside the vaginal canal (Figure 1A), included two components (Figure 1B); (1) a cone-shaped forward-looking coil (Figure 1C), and (2) the sideways-looking array (Figure 1D). The forward-looking coil was intended for imaging the cervix and posterior endometrium, with the coil’s tip positioned at the top of the vaginal canal, just below the cervix. The sideways-looking array was designed for imaging the vaginal wall.

The cone-shaped forward-looking coil consisted of two metallic layers with an insulator in-between. The outer metallic layer consisted of windings of a spiral-shaped copper coil, and the inner metallic layer consisted of a thin cone-shaped copper coating. The nine-windings spiral coil had a winding density which was higher toward the forward direction, to concentrate the magnetic field toward the forward direction.

The forward-looking coil presented the largest challenge, due to the restricted diameter of the vaginal canal, which limited the maximal diameter of the coil windings. It was constructed utilizing the image magnetic-field concept,^30,31 wherein metallic surfaces can be used to force magnetic fields to project along desired directions. Image magnetic fields are created by RF magnetic field (the primary field), by current elements of an RF coil, impinges on a metal surface within the coil’s physical volume. Since the net magnetic field on the metal surface must be zero, the metallic surface generates a surface electrical current that, in turn, generates an opposing-polarity magnet field, the image magnetic field. Within the coil’s physical volume, the magnetic fields created by the primary and image fields lower the net magnetic field, while outside the coil volume, the two are oriented in the same direction, creating a larger net field. The size of the image field, and the region in which it significantly enhances the primary field, depend on specific design parameters. For the endovaginal array’s forward-looking coil, we aimed to maximize the RF lobe pattern and direct it forward.

The sideways-looking array consisted of four loop elements wrapped at 90° increments around a cylindrically shaped shaft. Each coil was segmented with ceramic capacitors to maintain RF-phase coherency and connected to a separate receiver.

2.2 |. Numerical EM simulations

All simulations were performed using Microwave Studio (Computer Simulation Technology, Darmstadt, Germany), a full-wave EM simulator based on the finite integration method.

The forward-looking coil parameters were evaluated using EM simulations. The parameters that were varied were; (a) the thickness of the insulator layer (T_i) at the forward-looking coil’s distal end, while keeping the metallic-cone diameter, geometry, and the solenoid windings distribution fixed; and (b) the diameter of the metallic cone (D) at the top, while keeping the insulator thickness and the solenoidal coil winding-distribution fixed.

To assess the receive RF magnetic field ($B_1^{-}$) distribution, the endovaginal array was simulated inside a phantom model (relative permittivity ε = 80, conductivity б = 0.5 S/m) with the boundary conditions being a perfect-electrical-conductor coil winding, a perfect-magnetic-conductor for the metallic cone and a perfectly matched layer in the x, y, and z directions, respectively. The coil was surrounded by the phantom, with all the endovaginal-array’s electrical conductors electrically insulated.

Maximum 10-g average specific-absorption-rate (SAR) was calculated in a phantom based on SAR maps (in W/kg). The phantom was centered inside an MRI’s bird-cage body-transmit coil, with the endovaginal array located at the corner of the phantom. This location was predicted to be the region with the highest RF electric field and, therefore, a region which would incur maximal heating during imaging. The resulting simulated SAR distribution and a subsequent experimental temperature test using the actual endovaginal array were used to determine the safety factor for in-vivo experiments.

2.3 |. Endovaginal array fabrications

In the forward-looking coil, 16-AWG tin-plated copper wire was used to form the spiral-shaped coil, which had one 5 mm diameter winding at the distal tip (e.g., inside the cone’s gap), five loosely wound windings centered at 1.5, 3.0, 4.5, 6.0, and 7.5 mm from the tip, and three windings centered at 10.5, 13.5, and 18.0 mm from the tip. The coil was tuned to the 1.5 T Larmor frequency (64 MHz) using two-series capacitors. Specifics of the insulator thickness along the coil, and the geometry of the cone-shaped copper former were determined by EM modeling.

Each element (length = 60 mm, width = 27 mm) of the sideways-looking array was wrapped around the cylindrical plastic former (length = 140 mm, outer diameter = 20 mm). The overlapping distance between neighboring coils served to reduce mutual inductive coupling. The overlap width was based on the critical loop center-to-center distance of 0.75 times the loop diameter.³² Each element had one 20 pF series capacitor at its center to improve RF-phase coherence.

The internal assembly (Figure 2A) was placed within a waterproof external enclosure (Figure 2B) that contained external groves for placement of six BT catheters and an internal 14-mm diameter lumen for placement of a tandem coil, and was keyed to fit within the standard Syed-Neblett template. As such, the “active” obturator was configured to replace the conventional cervical-cancer obturator during BT procedures. In addition, the enclosure supported easy removal of the internal assembly from within the enclosure, while leaving the enclosure unperturbed within the vaginal canal. This feature allows for the BT procedure’s phases that usually follow MR imaging and MR-guided catheter placement; X-Ray or CT imaging, and then radiation dose delivery, to be performed without the imaging coils in place.

FIGURE 2.

Prototype of the endovaginal array which functioned as an active obturator and its corresponding imaging optimization circuit. (A) Photograph of the inner section of the obturator which contained the five-channel array, showing the forward-looking coil, sideways-looking array, and the detachable connector. (B) The external enclosure contained a waterproof enclosure for the internal section. This enclosure also included external groves for placement of six high-dose-rate interstitial catheters (a single catheter is shown for illustration purposes). (C) Photograph of the five-channel tuning, matching, decoupling, and gain-equalization circuitry in the circuit box, along with that half-wavelength length cabling that connected the assembly to a receiver box (not shown). The cables were shielded and mounted with three floating Baluns. (D) Expanded view photograph of the circuitry shown in the orange box in (C). A test point allowed acquiring the signals before the gain-equalization circuit, as required for tuning and matching the coil signals

2.4 |. Imaging optimization circuit box

A 10 cm length, 0.5 mm diameter microcoaxial cable connected each element to a five-channel box (one channel for the forward-looking coil and four channels for the four-element sideway-looking array) which combined a tuning, matching, and passive-decoupling circuit, followed by a gain-equalization circuit (Figure 2C). To perform tuning and matching, SMA connectors were placed immediately after this circuit, followed by a jumper that allowed disconnecting the gain-equalization circuit. This jumper also allowed connecting the five channels directly to the MRI scanner, while bypassing the equalization circuit.

The signal from each endovaginal array element (numbered 1 to 5) entered from the right side, and then passed through the tuning and matching and decoupling circuit (red square), which consisted of aπ circuit; two parallel adjustable capacitors (Knowles JZ400HV, Cazenovia, IL) surrounding a series variable inductor (Coilcraft 164–06A06L, Cary, IL), followed by an antiparallel diode (Macom MADP 011048, Lowell, MA). It then went through the gain-equalization circuit (blue square), which consisted of a low noise preamplifier (LNA) (Wantcom WMA1R1, Chanhassen, MN) followed by an attenuator (Minicircuits Gat-10+ to Gat-15+, Brooklyn, NY).

When the endovaginal array was used together with commercial surface arrays to produce combined images, the stronger endovaginal array SNR created regions of saturated image intensity that complicated image reading. To resolve this issue, while retaining the improved SNR of the endovaginal array, it was necessary to reduce this larger intensity without significantly reducing the SNR. This was accomplished by a gain-equalization circuit, consisting of a dedicated LNA at the output of each endovaginal element, followed by a low-oise attenuator. The attenuator value was chosen so as to equalize the amplitude from each endovaginal element (i = 1, 2.5) at a location where it was strongest (e.g., close to its center) relative to the surface-array amplitudes at their strongest locations (close to the surface-array elements).

After the gain-equalization circuit, a half-wavelength coaxial cable connects each coil to a custom preamplifier box (Stark Contrast, Erlangen, Germany). The assembly was enclosed in a conductive shield, and mounted with floating resonant RF traps³³ at 30 cm increments to reduce body-transmit-coil-induced common-mode currents in the cables.

The most proximal elements in the Circuit Box were (5 KV) 0.1 μF blocking capacitors, placed on both the conductor and ground circuits. These ensured that the system satisfied the IEC/ANSI 60601–1-11:2015 electrical patient-isolation requirements of (i) <10 μA DC leakage current and (ii) >2 KV 60-Hz blocking limits.³²

2.5 |. Bench tests

A cylindrically shaped (diameter = 20 cm, length = 30 cm) saline phantom (0.9% NaCl, 0.09% gadolinium-DTPA) was used for bench testing of the assembled coil.

The body-unloaded to body-loaded quality-factor (Q) ratio was assessed with a single receive-only coil.

Each element’s tuning, matching, and decoupling were tested by reflection (S₁₁) and transmission (S₂₁) measurements using a vector network analyzer (Agilent HP 8751A). Since the S₂₁ transmission between two elements depended on element tuning and matching, S₂₁ measurements were performed after tuning and matching individual coils in the saline phantom. Coupling measurements were conducted on nearest-neighbor overlapping elements by using an S₂₁ measurement with the cables directly connected to the preamplifier sockets of the two elements being tested, with all other coils detuned.

2.6 |. Phantom MRI experiments

To perform SNR comparisons between the endovaginal array and commercial surface coils, Fast Spin Echo (FSE) images (TR/TE/flip angle (θ) = 1500 ms/80 ms/180°, slice thickness = 3 mm, matrix = 256 × 256, FOV = 220 mm, 1 avg) were obtained in the sagittal, axial, and coronal planes in a 1.5 T Siemens Avanto, with the endovaginal array placed at the center of a CuSO₄ doped-solution phantom. For comparison purposes, the SNR was calculated in the vicinity of the endovaginal array, relative to the Avanto’s 16-channel spine array.

To assess the coil performance, SNR maps were acquired. The SNR map was generated by comparing images obtained from two FSE acquisitions; one with RF power, and the second without RF power.

2.7 |. MRI safety test

The internal section of the endovaginal array (Figure 2A) was coated with paraffin film and immersed in an ASTM gel phantom made with 1.32 g/L NaCl and 10 g/L polyacrylic acid in water.³⁴ With this formulation, room temperature conductivity was 0.47 S/m and the viscosity was sufficient to prevent convective heat transport. The specific heat of the gel phantom was 4150 J/(kg.°C), at 21°C (ASTM-F2182). EM simulations were used to identify the phantom regions with the highest SAR distribution (e.g.,“Hot spots”) in the presence of the endovaginal array. Four fiber-optic temperature sensors were placed in the phantom at locations near the array that were predicted to exhibit maximal heating. Note that these measurements were performed without the scanner’s surface coils in place, and neither the surface coils nor the interstitial catheters were simulated.

The assembly was placed in the isocenter of the scanner’s body coil. The test procedure was divided into two steps. In Step 1, the temperature rise at several locations around the array, and at a reference location far from the array, were measured using fiber-optic thermometry probes (Neoptix RF-04–1, Quebec, Canada) during imaging. In Step 2, the array was removed from the phantom and imaging was repeated while temperature measurements were obtained at the same probe locations. Temperature measurements were performed during 15 min of continuous scanning with a high SAR sequence at a 11 μT excitation magnetic field strength ($B_1^{+}$) (True Fisp: TR/TE/θ = 30 ms/1.3 ms/66°, slice thickness = 20 mm, 128 × 128, FOV = 40 × 40 cm², slices = 25, avg = 32, repetitions = 32). The scanner’s reported whole-body average SAR was 4W/kg. In accordance with the ASTM standard test method F2182–11a, the actual SAR value around the endovaginal array was calculated based on the heating measurements. SAR was calculated according to the equation:

$$SAR=C\frac{\text{Δ}T}{\text{Δ}t},$$ (1)

where C is the heat capacity of the gel phantom (4150 J/(kg.°C)), ΔT is the change in temperature, and Δt is the time increment over which the temperature changes occurred. We calculated the dT/dt using a linear fit over an initial temperature rise. The SAR enhancement factor was calculated by dividing each calculated SAR by the background SAR.

2.8 |. In-vivo experiments

In-vivo experiments were performed in the 1.5 T scanner using three sexually mature (>6-month old) 70–75 kg potbellied minipigs, which have well-developed vaginal and endometrial anatomy.³⁵ Experiments were conducted in accordance with a Johns Hopkins University Institutional Animal Care and Use Committee (IACUC) approved protocol. The swine were intubated and ventilated during the entire procedure.

The endovaginal array was inserted into the swine’s vaginal canal, up to the cervix. 2D and 3D T2-W, and T1-W FSE images (2D T2-W: TR/TE/θ = 2000/99 ms/120°, FOV = 25 × 25cm², 512 × 512, slice thickness = 2 mm, resolution = 0.47 × 0.47 × 2.0 mm³, GRAPPA = 2, 1 avg, 2:20 min/scan; 2D T1-W: TR/TE/θ = 600/12 ms/180°, ETL = 6, FOV = 26 × 26 cm², 512 × 512, slice thickness = 3 mm, resolution = 0.6 × 0.6 × 3.0 mm³,1 avg, 1:50 min/scan ; 3D T2-W: TR/TE/θ = 2500/102 ms/90°, FOV = 26 × 26 cm², 512 × 512, slice thickness = 1.2 mm, resolution = 0.56 × 0.56 × 1.2 mm³, GRAPPA = 3, 1 avg, 2:50 min/scan) were acquired in the coronal, sagittal, and axial planes. Note that each of these acquisitions is approximately half the duration of equivalent scans performed at 1.5 Tesla.

Images were acquired with the endovaginal array together with the Siemens body and spine arrays. In-vivo SNR was calculated by dividing a selected region-of-interest’s (ROI’s) mean pixel intensity by the standard deviation of air regions (e.g., noise). Individual-coil reconstructed images were used, averaging 10 ROI’s within a prescribed FOV. The endovaginal array SNR was divided by the spine-array SNR in the same ROIs, and an SNR gain factor was computed.

We performed an N4 ITK image intensity bias correction³⁶ to account for the MRI signal intensity gradient which occurs in anatomical regions which are very close to the endovaginal array, and may complicate reading the images.

The intensity bias corrections were performed separately on each of the five coils that form the endovaginal array as well as the seven active spine- and body-array elements, and then the images were combined in the conventional sum of squares manner used for MRI array coils on a 3D Slicer workstation. N4 ITK correction parameters: B-Spline grid size: 10 × 10 × 10 pixels³, B-Spline order: 3, Number iterations; (FSE T1-W: 150, 120, 90), (FSE T2-W: 100, 80, 60), convergence threshold: 0.0001, shrink factor: 4.

The complete N4 ITK correction requires 10 s/110 images on a laptop Personal Computer with a standard intel i7 9th generation processor. This filter is substantially equivalent to other homomorphic filters,³⁷ such as the contextvision filter (https://www.contextvision.com), that is FDA-cleared.

Note that the N4 ITK intensity correction does not change the image SNR, but only normalizes the image intensity to the largest intensities present, which makes reading the images easier, although the lower SNR regions do appear grainier.

3 |. RESULT

3.1 |. Electromagnetic modeling and simulation

The insulator thickness (Ti) of the forward-looking coil varied from 0.5 to 5.0 mm. Ti = 5.0 mm was the largest possible thickness wherein the coil outer diameter remained at the predetermined 20 mm limit. The $B_1^{-}$ profile along the forward-looking lobe pattern showed that $B_1^{-}$ could be doubled by changing the insulator thick ness from 0.5 to 5 mm (Figure 3A).

FIGURE 3.

(A) Magnetic field ($B_1^{-}$) simulation along a direction straight in front of the forward-looking coil while varying the coil’s insulator thickness (Ti) from 0.5 to 5.0 mm, while keeping the conducting surface’s shape, as well as the coil-winding distribution fixed. Comparing the $B_1^{-}$ field profile at the 2.0 and 5.0 mm insulation thicknesses demonstrated that the forward lobe pattern was almost doubled along the forward direction (from 10 to 20 mm) by this change in parameters. (B) $B_1^{-}$-simulated profile in the forward direction, while varying the metallic cone (inner cone) diameters (D) from 6 to 18 mm, while keeping the insulator thickness, as well as the coil-winding distribution fixed. The black dotted line shows that the forward penetration, at the same $B_1^{-}$ intensity, could be increased dramatically by increasing the metallic-cone diameter

Results for four different diameter values, varied from 6 to 18 mm, showed (Figure 3B) that increasing the diameter of the inner metallic cone at the coil’s forward end increased the $B_1^{-}$ penetration in the forward direction dramatically (from 5 to 20 mm). A 14-mm metallic-cone diameter was chosen for the final array construction to allow for a 20 mm outer diameter.

We compared the $B_1^{-}$-field distribution produced by a metal-backed forward-looking coil to a nonmetal-backed coil with the same geometric shape and the same coil-windings pattern. It can be seen that the metal-backed coil had its lobe pattern greatly extended along both the forward-looking and sideways-looking directions (Figure 4A and B). The directionality of the field was controlled by the angulation of the metallic layer in the cone, and the angulation and thickness of the insulator. Another notable difference was the total absence of the RF magnetic field inside the cone region (Figure 4A). The total $B_1^{-}$ field produced by the endovaginal array, including the forward and sideways fields, is shown in Figure 4C.

FIGURE 4.

EM-simulation based comparison of the forward-looking coil’s receive-magnetic field ($B_1^{-}$) with (A) versus without (B) a metal-backing, which was placed along the inner wall of the cone, onto which the spiral-coil windings were then mounted. The dotted black cross is overlaid on the field maps to aid in a visual comparison of the field penetration depths. (C) Total $B_1^{-}$ produced by the endovaginal array, including contributions from the forward and four sideways coils. Since only the forward-looking coil had an internal conducting layer, the RF magnetic field penetrated into the interior in regions covered by the sideways-looking array. (D) Ten-gram SAR distribution in the sagittal plane for the five-channel endovaginal array. The numbers 1–4, with adjacent orange arrows, indicate the locations of the fiber-optical temperature sensors. A maximum peak SAR of approximately 1.2 W/kg was calculated for the forward-looking coil, in regions where the spiral windings were denser, which can be explained by the higher electric field confinement at those locations

The resulting maximum 10-g SAR value (Figure 4D) of the entire array showed that the peak local SAR was higher in the forward-looking coil (1.2 W/kg) than in the four sideways-looking coils (0.3 W/kg). This SAR distribution was used to estimate the possible hot spots for the heating test. A negligible SAR enhancement was observed around the sideways-looking array.

3.2 |. Array characteristics

All coil elements were tuned and matched to 64 MHz with a measured reflection coefficient of −21 dB in the forward-looking coil (Figure 5A) and ≥−16 dB in the sideways-looking elements (Figure 5B). The ratio of unloaded-to-loaded Q in a phantom was 10.4, indicating near-complete body-load dominance. The coupling between neighboring sideway-looking elements (S21) ranged from −18 to 21 dB and −14 to −17 dB in simulation and experimental evaluations, respectively (Figure 5C). The forward-looking coil and sideways-looking array were geometrically decoupled.

FIGURE 5.

(A) Simulated and experimentally measured reflection coefficient (S₁₁) of the loaded forward-looking coil. The minor differences in S₁₁ values between simulation and experiment may be due to cable losses. (B) Experimentally measured reflection coefficients of the four elements (1, 2, 3, 4) in the sideways-looking array, measured while the array was loaded. All coils demonstrated a reflection coefficient of less than −16 dB. (C) Simulated and experimentally measured S-parameter matrix of the loaded four-element sideways-looking array, reflecting the coupling between the various (i = 1, 2, 3, 4) coils in the array. The simulated and measured coupling between coils was less than −18 dB and −14 dB, respectively. All elements of the forward-looking and sideways-looking sides were tuned and matched to operate at 64 MHz

3.3 |. Phantom MR imaging

Experimental SNR maps in the saline phantom revealed an overall SNR distribution change due to the presence of the endovaginal array. The sideways-looking array demonstrated an SNR enhancement of two to four times over a 15–25 mm region, relative to the spine surface coil (Figure 6A and B). The forward-looking coil improved the SNR by three to five times over a 20–30 mm region along the forward direction (Figure 6C).

FIGURE 6.

Experimental SNR mapping results in the gel phantom. Coronal plane normalized SNR maps obtained using (A) only the spine array (endovaginal array off) and (B) the spine array and the endovaginal array. Placing the endovaginal array in the plane improved SNR along the sideways-looking directions. An SNR improvement factor of two to four times was obtained with the endovaginal array in the phantom, as outlined in the region surrounded by the dashed black line. (C) Transversal SNR map of the entire endovaginal array. SNR was enhanced by three to five times over a 20–30 mm region in the forward direction. (D) Experimental SNR profile of the forward-looking coil at different orientations relative to the B₀ (Z) direction, with (E) demonstrating the orientations of the endovaginal array with respect to the B₀ (Z) direction.

As expected for many RF coils, the SNR of the endovaginal array varied relative to its orientation with respect to the static magnetic field (B₀). Therefore, SNR profiles of the forward-looking coil at various orientations relative to the B₀ direction were evaluated (Figure 6D and E). The forward-looking coil showed optimal performance when the shaft was aligned perpendicular to the B₀ direction. This suggests that improved patient imaging will result from tilting the subject’s vaginal cavity upwards from the MRI’s superior-inferior (B₀) direction, which entails placing an object below the lower pelvis. The sideway-looking array showed optimal performance when the shaft was aligned parallel to the B₀ direction, where the normal vector of the elements lies perpendicular to the B₀ direction.

We also studied in gel phantoms the use of the endovaginal array during MR-guided needle insertion procedures. We demonstrated a lack of interaction between the endovaginal imaging array and the insertion of both passively and actively tracked³⁸ BT needles through holes in the array’s enclosure.

3.4 |. MRI safety analysis

Measured temperatures at different locations around the endovaginal array (Figure 7A) showed the highest temperature rise of <2°C observed beside the forward-looking coil at a location close to the highest solenoidal winding density (Figure 7B). A corresponding SAR gain of approximately 2.22 was calculated at this location. Table 1 provides calculated SAR values at different locations along with the corresponding SAR enhancement factors. All calculated SAR values were within the corresponding limits (10 W/kg for maximum local SAR) recommended by the FDA and IEC.^39,40

FIGURE 7.

MRI endovaginal array thermal safety test in the ASTM gel phantom: (A) The location of the fiber-optic sensors (blue dots) along the sides of the endovaginal array. Note that the array was coated in a thin paraffin film for this test (i.e., the external enclosure was not used). (B) Temperature was recorded at five different points in the phantom: by the densest spiral windings (tip winding), by the tip, by the array’s edge, along the body, and at a reference (a background reference far from the array) point. The largest temperature increase of 2°C, with a corresponding SAR enhancement factor of 2.22, was observed at the tip spiral winding of the forward-looking coil during a 15 min acquisition of a 4 W/kg SAR sequence

TABLE 1.

Calculated local SAR values from experimental heating test at five different locations

In the presence of the coil			Without the coil
	SAR (W/kg)	dT/dt (°C/s) × 10−4		SAR (W/kg)	dT/dt (°C/s) × 10−4
Tip winding	0.80	57	Tip winding	0.36	26
Tip	0.66	47	Tip	0.38	27
Edge	0.48	35	Edge	0.35	25
Body	0.38	27	Body	0.38	27
Background	0.35	25	Background	0.35	25
SAR enhancement factor
Location	Tip winding	Tip	Edge	Body	Background
SAR gain	2.22	1.73	1.37	0.1	1.00

3.5 |. In-vivo experiments

Figure 8 shows 2D FSE T2-W and T1-W images obtained in two sexually mature swine in the coronal, sagittal, and axial directions, with the active obturator in the vaginal canal. The equalization circuit was not used to demonstrate the endovaginal-array’s SNR gain. The sexually mature swine had endometria that extended upwards and behind the bladder. The forward-looking coil provided a 5 + 0.2 SNR enhancement factor in cervix and posterior endometrium FOVs 30–40 mm above the coil’s superior tip, relative to the Siemens spine array. The sideways-looking array provided a 4 + 0.2 SNR gain factor in vaginal-wall FOVs distanced 30–40 mm from the array enclosure, relative to the spine array.

FIGURE 8.

Images obtained in two sexually mature minipigs (swine #1, #2) using the endovaginal array together with the scanner’s spine and body-surface arrays. In these images, the gain-equalization circuit in the variable gain receiver box (Figure 3C) was bypassed, to demonstrate the SNR advantages of the forward-looking coil and sideways-looking coils of the array, which appear as strongly hyperintense regions. Swine #1 Oblique 3D T2-w sagittal (A) and (B) coronal images demonstrate the image quality available above the strongly hyperintense region of the forward-looking coil. (C) A 0.7-mm slice-width reformatted axial image taken from the yellow rectangular region of (B). (D) Swine #2 oblique axial T1-W slices i-iv obtained at various locations. The slice locations are shown as yellow-line overlays on (E), a T2-W sagittal image. Blue dashed overlays show location of endovaginal array, and Green, Red, and Yellow overlays indicate anatomic organs

Oblique 3D T2-W sagittal and coronal images demonstrated the strongly hyperintense signal in the forward-looking and sideways-looking regions (Figure 8A and B). A 0.7 mm slice-width reformatted axial image from the yellow rectangular region of Figure 8B. Also, Figure 8C demonstrated the achievable endometrial detail. Oblique axial T1-W slices i-iv (Figure 8D), obtained at slice locations, shown as yellow-line overlays on a T2-W sagittal image (Figure 8E), demonstrated the hyperenhanced region size.

Figure 9 shows the results of applying the N4 ITK filtering³⁶ on 2D axial T1-W images (A) and sagittal T2-W images (B) acquired in swine #1. With the filtering, the strong intensity gradients close to the endovaginal array (Figure 8) are removed and it is, therefore, possible to see details such as of the cervix in (A). Use of the filtering does increase the noise in regions of lower SNR, such as toward the abdominal surface in (A or B), or where only the surface coils are active.

FIGURE 9.

Images obtained in sexually mature mini-pig #1 using the endovaginal array together with the scanner’s spine and body surface arrays. In images (A, B), the effect of the N4 ITK image bias correction filter is demonstrated by comparing the raw acquired images on the left column and N4 ITK image-processed images on the right column. Contours of the cervix are shown as dotted red lines. (A) T1-W axial FSE slices taken at 3 mm increments around the cervix, spanning from anterior to superior. (B) A 2 mm sagittal T2-W slices obtained along midvaginal planes, spanning from left to right. The bias correction filters remove the strong image intensity gradients present close to the endovaginal array and make reading the images close to the endovaginal array easier. Since the bias correction filter normalizes the mean signal intensity in different regions of the image, it does increase the noise levels further way from the endovaginal array, where only the weaker body and spine arrays are active.

4 |. DISCUSSION

In this preclinical study, we showed that a relatively large FOV could be covered by an endovaginal array that included both forward-looking and sideways-looking elements, and that approximately four times SNR enhancement was obtained in-vivo at distances of 40 mm from the array.

The main technological difference between this endovaginal array and other endovaginal, cervix,^20,21,41 endourethral,²⁴ or endorectal coil arrays,^42,43 is the novel coil structure used in the forward-looking coil at the tip of the array, which enables acquiring high SNR images above the tip (i.e., in the forward-looking direction). The metallic-backed forward-looking coil was specifically optimized to allow high-sensitivity imaging of the cervix and posterior parametrium, which are regions that cannot be accessed with most coil designs, due to the narrow diameter of the cervical canal, which prohibits advancement of large (>10 mm) diameter coils upwards into the cervix and parametrium. Here, we utilized a metallic backing placed on a spiral coil, with the exact shape optimized with the aid of EM simulation and subsequently validated in vitro and in vivo, to provide high SNR at distances of up to 40 mm above the tip of the coil. We have not found this design used in any commercial receiver coil, although it is employed in MRI body-transmit coils.

The four channel sideways-looking array is similar in design (e.g., a loop coil) to the endorectal coil, but has (i) imaging in all radial directions instead of only in one direction, (ii) has higher SNR due to the fact that it has four coil elements instead of one, as well as integral preamplifiers on each coil, which is a modern method used to increase SNR and reduce sensitivity to intercoil coupling.

When applied clinically, this endovaginal array will be combined with existing BT devices to most precisely image the remnant cervical tumor left after external beam radiation. Contouring the gross tumor volume (GTV) at the time of BT is part of the new International Congress on Radiation Units and Measures (ICRU) report 89⁴⁴recommendations. Identifying the GTV, however, is very challenging. A higher-sensitivity array that images the vagina, cervix, para-, and endometrium during the BT implant may improve treatment planning based on a better estimation of the residual tumor utilizing high-resolution multiparametric MRI.

We expect that the clinical advantages of the endovaginal array will be found in the combination of an interventional tool with a sensitive imaging array that should be able to image multiple MRI contrasts in approximately half the currently required imaging time. For intervention, the endovaginal array shape and exterior replicate the capabilities of the standard Syed-Neblett obturator; it anchors the vaginal anatomy and cervix location, and its exterior provides directional holes for both the central and interstitial needles that are inserted into tumors found in the vaginal wall, the cervix, and parametrium. Additionally, the obturator provides access in its center for the tandem coil which is needed to anchor the parametrium location to the obturator, and allows introducing radioactive sources into the center of the parametrium.

We chose the Syed-Neblett obturator as the obturator to emulate since transperineal interstitial BT applicators, such as the Syed-Neblett and Martinez Universal Perineal Interstitial Template (MUPIT), are the most common applicators for the treatment of cervical cancer with (i) vaginal and extensive parametrial extension and (ii) bulky (>5 mm) primary/recurrent vaginal malignancies. The Syed and MUPIT applicators both employ the use of a vaginal obturator, both of which can be substituted with the combined endovaginal MRI-array-coil obturator device.

In totality, this “active obturator” should allow performing imaging in exactly the same pelvic setup, as is used during the subsequent intervention. That means there is no motion of the vaginal wall or cervix between the imaging phase and the subsequent treatment. This is important, since performing nonrigid registration of the deformable anatomy in this region is not easy and is frequently associated with large errors.⁴⁵ As a result, the BT catheters can be inserted under MRI guidance at far higher precision. Furthermore, the obturator enclosure is intended to be used after the MRI imaging phase, for example, during CT or X-ray scanning and during the BT dose delivery phase, which is why the imaging insert can be removed, without moving the outer obturator enclosure, so that the pelvic anatomy is not displaced during the removal process.

The active obturator’s enclosure (Figure 2B), which is sterilizable and disposable, is intended to be inserted into the patient’s vaginal canal before the imaging phase of the MRI-guided BT procedure. This section includes a 14-mm diameter central hole for insertion of a tandem. The internal section (Figure 2A), which includes the coil, which is sterilizable and reusable, is then inserted into the cover and locked in place with screws on the proximal end of the obturator. This design is specifically arranged to allow removal of the internal section once imaging is complete. This leaves the enclosure in place during the subsequent interventional portions of the BT procedure, which include placement of BT catheters into the tumor, an optional phase of patient CT imaging with the implanted catheters in place, and the final insertion of radiation sources into the catheters.

This coil was developed for eventual human use and passed the heating tests required for interventional MRI devices, and biomedical-device electrical safety standards.

This study had limitations. We have not used the array in adult females. Imaging of humans will require IRB approval, which will require additional testing. Specifically (a) the inner nondisposable (the coil array and cables) section must endure Ethylene Oxide (ETO) sterilization while the (b) outer disposable enclosure must endure conventional steam sterilization. We have not completed the sterilization validation tests, which are required for IRB approval. Once the array is cleared for use in human subjects, we will be able to evaluate its clinical advantages, which we hypothesize will be found in combination with multiparametric MRI sequences that will strongly benefit from its higher SNR such as DWI.

We saw some phantom image artifacts at distances of up to 20 mm outside the endovaginal array enclosure (Figure 5C). The artifacts were larger in large flipangle acquisitions, such as FSE, and probably result from phase cancellations, which vary with the array orientation relative to B₀. In swine imaging, however, there were relatively few image artifacts (Figure 8). The image intensity gradient close to the obturator surface, however, was large. The analog gain-equalization circuit should reduce this issue in clinical scenarios. When it will not suffice, we may be able to apply regulatory-approved digital image filters,³⁷ which can typically be performed in less than 1 min for all the images, as demonstrated in Figure 9.

Inhomogeneous SNR distribution of the sideways-looking array was observed in phantoms in the coronal plane at large flip angles (Figure 5B), which may be due to intercoil inductive coupling. Although this issue was not observed in swine (Figure 8), a future sideways-array could include an interior metallic layer, which may reduce this effect.

Due to the need to perform ETO resterilization of the internal section, which requires heating at approximately 60°C and applying elevated gas pressures, we moved the decoupling diodes from each coil’s circuit board to (Figure 2C) the receiver box. It has been suggested that adding a series fuse to each coil’s circuitry would reduce heating risks during a coil open-circuit malfunction, and we will study this in the future.

Performance of the BT procedure in a 3T scanner forms a possible alternative to use of the endovaginal array. In two diagnostic comparisons performed between 1.5 T MR imaging with the endorectal coil versus 3T imaging without the endorectal coil,^22,23,42,46 the endorectal coil provided superior spatial resolution, and higher sensitivity to disease. In one comparison,⁴² the conclusion was “Intraindividual comparison shows that image quality and delineation of prostate cancer at 1.5 T with the use of an endorectal coil in a pelvic phased array is superior to the higher field strength of 3 T with a torso phased-array coil alone.” Additionally, at 3 T, there are larger air-soft tissue susceptibility artifacts that affect vaginal wall and cervix imaging, which may require insertion of a dielectric gel/fluid into the vagina.⁴⁷ Lastly, there is a strong preference for use of 1.5 T systems for interventional applications, due to (i) a larger fear of heating of interventional devices due to the shorter wavelength and stronger RF (35 KW vs. 15 KW peak) at 3 T and (ii) the far stronger RF inhomogeneity seen at 3 T (e.g., “the dielectric effect”), which results in large differences in signal amplitude between regions, particularly, in abdominal imaging.

It is important to note that all prior comparisons of 1.5 T imaging with an endorectal coil to 3 T imaging without an endorectal coil were conducted for diagnostic imaging, whereas this endovaginal obturator is primarily intended for interventional procedures, performed after most of the tumor has been eradicated with external beam radiation. In these interventions, the objective is to find small residual tumors with multiparametric imaging, and precisely target them with a large dose of catheter-based radiation. This requires precise tumor localization, with minimal changes in tumor location between the diagnostic and therapy stages. To address this, the endovaginal array’s inner section was placed within an external (obturator) enclosure that remains in the patient after imaging, and can be used during therapy delivery, so that minimal changes in patient anatomy and pose occur. The substantial SNR gain from the endovaginal array supports much faster multiparametric imaging.

Additionally, we plan to use the active obturator to image between successive radiation fractions, which may allow changing the radiation dose spatial distribution after detecting changes in residual tumor topology or physiology.

Forward-looking coils of a metal-backed design can also be placed at the tips of deflectable vascular catheters, to provide “flashlight” enhancement toward regions in front of the advancing catheter,²⁵ thus, reducing the risk of vessel-wall perforation or vascular-occlusion puncture without advance warning. They can also be employed in endoscopic coils to assist in MR-guided navigation through complex anatomy, such as within the gastrointestinal tract, while the sideways-looking elements of the array can be used to scan the walls.⁴⁸

While the first impact of this endovaginal array will be in the field of radiation oncology, we do believe there are applications in radiology such as in the early detection of tumors in the cervix and vaginal wall and their biopsy. However, we have not explored these in this article.

We provided photos of the biocompatible version of the device (Supporting information Figure S1). Once the array is cleared for use in human subjects, we will be able to evaluate its clinical advantages, which we hypothesize will be found in combination with multiparametric MRI sequences that will strongly benefit from its higher SNR such as DWI.

5 |. CONCLUSIONS

An endovaginal MRI receiver array coil, including forward-looking and sideways-looking components, was developed for improved SNR of the parametrium, endometrium, cervix, and vaginal-wall imaging at 1.5 T. The forward-looking element utilized metal surfaces within the coil to extend field penetration in front of the coil. The active obturator may be used once IRB approval is obtained for preprocedural imaging and radiation-source placement. Future work will focus on clinical evaluation in cervical-cancer patients during MR-guided BT procedures.