Technical Note: Effect of transducer position and ground plane configuration on image quality in MR-guided focused ultrasound therapies

Abstract

Purpose:

Evaluate the effect of a focused ultrasound transducer position and ground plane configuration on magnetic resonance image quality.

Methods:

The effect of transducer position with respect to the MRI B0 field and the radiofrequency receive coils was evaluated in a breast-specific MRgFUS system with an integrated RF phased array coil. Image signal to noise ratio was evaluated at different transducer locations. The effect of ultrasound transducer ground plane configuration was evaluated using a replica transducer with twelve ground plane configurations. All evaluations were performed at 3 Tesla.

Results:

Both transducer position and ground plane configuration were found to have a considerable effect on overall image SNR. A 67% increase in SNR was achieved by positioning the transducer face perpendicular to the B0 field. A 25% increase in SNR was achieved by segmenting the replica transducer ground plane from one continuous plane to 9 individual segments.

Conclusions:

Advances in focused ultrasound hardware allow for integrated radiofrequency MRI coils as well as adjustable transducer positioning. The placement of the ultrasound transducer with respect to both the magnetic field and RF coils can have a considerable effect on image SNR and the resulting MR images that are used for MR guided focused ultrasound treatment planning, monitoring and assessment.

Introduction

In magnetic resonance guided focused ultrasound (MRgFUS) treatments, image quality is generally lower than diagnostic images. This is due to the requirements of the MRgFUS hardware, including the presence of acoustic coupling fluid (1) and the ultrasound transducer and associated equipment needed for patient positioning and transducer manipulation. High image signal-to-noise ratio (SNR) is achieved by placing the radiofrequency (RF) coils close to the region of interest. But in MRgFUS treatments, the geometry of the ultrasound transducer and the acoustic coupling medium often make it difficult if not impossible to place RF coils near the region of interest. In such cases the large volume body RF coil is used for transmit and signal reception, greatly reducing the image quality that can be obtained during MRgFUS treatments (2). In some cases, such as transcranial applications, the treatment planning or assessment may be performed in an anatomy-specific diagnostic coil (3) while treatment monitoring is done using the body coil. While using anatomy-specific coils can increase the SNR during treatment planning and assessment, using different coils during treatment monitoring can potentially introduce registration errors that need to be addressed to ultimately correlate planning, monitoring and assessment images.

To increase SNR in MRgFUS systems and ultimately allow treatment planning and assessment to be performed in the same orientation as treatment, novel RF coils are being developed that allow for increasingly customizable configurations (4) and the direct integration of RF coils into MRgFUS system design. These novel designs and use of dedicated RF coils that are placed in very close proximity to the transducer and to the anatomy of interest will increase the options to increase SNR in MRgFUS treatments. The most basic coil integration is achieved through placing an RF coil around the therapeutic ultrasound transducer. This has been implemented successfully in both preclinical (5) and clinical applications (6) that have a smaller acoustic aperture. A more complex phased-array radiofrequency coil has been integrated into a breast-specific MRgFUS system design, demonstrating a significant SNR increase when compared to the body coil (7).

Typical therapeutic ultrasound transducers are constructed using a conductive ground plane between the target and the transducer element. Often this transducer ground plane is large and continuous, even for multi-element phased array transducers. The presence of this ground plane can impact MR image quality, even when RF coils are integrated in MRgFUS systems. The RF and gradient field eddy current effects caused by the ultrasound transducer ground plane can significantly decrease the available SNR and cause MR image artifacts. These effects can be exacerbated by MR sequences that require the use of large amplitude and long duration gradients (8), particularly when a high slew rate is employed (9), including MR acoustic radiation force imaging (MR-ARFI) and diffusion weighted imaging.

Despite the large number of studies published considering MRgFUS applications, only a few studies have considered how SNR and imaging artifacts are not only impacted by the design of the RF coil, but also by the therapeutic transducer ground plane design and transducer position with respect to the RF coil. An open source preclinical MRgFUS system considered the tradeoffs between the acoustic coupling of the ultrasound transducer, coil positioning for SNR measurements and transducer placement (5). In the design of an endocavitary transducer with an integrated RF coil surrounding the transducer, quartering the transducer ground plane drastically reduced the magnitude of the eddy currents as assessed through reduction in the image artifact. SNR was increased by 40%, although this improvement was sequence dependent (6). Another study investigated reducing gradient induced eddy currents in a transducer ground plane of a transcranial therapeutic ultrasound system (10). The investigators evaluated six different ground plane configurations and found more than one transducer ground plane design that significantly reduced both image artifacts and improved overall SNR when compared to the commercially available seven-segment transducer ground plane configuration (10).

This work investigates the effects of the ultrasound transducer on MRI image quality in MRgFUS treatments through evaluating both transducer ground plane design and transducer position with respect to the B0 field for an MRgFUS system. While past work has considered gradient eddy currents (10), this work evaluates the combined contributions of RF and gradient eddy currents by analyzing the effect that transducer location and ground plane design have on treatment volume SNR. Evaluation was performed with a breast-specific MRgFUS system that has an integrated phased-array RF coil and transducer that has multiple positioning degrees of freedom. The results from this work provide considerations and strategies for transducer design and integration, and treatment-planning techniques that can reduce RF and gradient eddy currents and provide improved SNR for MRgFUS therapies.

Methods

The effect of the transducer ground plane on MR image quality in MRgFUS treatments was evaluated in a breast-specific MRgFUS system as shown in Figure 1. All experiments were performed on a Siemens 3T Prisma^FIT MRI scanner. This system features a 256-element phased-array ultrasound transducer (14.4 × 9.8 cm aperture, f = 940 kHz, radius of curvature = 10 cm) with a solid, conductive ground plane that is acoustically coupled to the breast with lateral mounting on a 1.0 L treatment cylinder. The system also has an integrated custom 8-channel RF coil array fitted around the treatment cylinder (14 cm diameter, 11 cm height) (11). The coil consists of a 6-channel ladder geometry (9,10) phased array (11) wrapped around the cylinder (elements 5 × 9 cm), a 17 cm diameter loop encircling the top of the treatment cylinder, and a single loop (13 × 15 cm) surrounding the transducer opening. This customized RF coil provides increased SNR for both improved image quality and parallel imaging capabilities (12). This design provides 6-degrees of freedom for positioning the system allowing for movement of the transducer with respect to both the breast and the B0 field. The coil array rotates with the transducer resulting in changing coil sensitivity profiles as a function of position for those coil that are wrapped around the treatment cylinder.

Figure 1:

The breast-specific MRgFUS system used in this work. (a) Top-view of the treatment cylinder, designed to be placed on the table of a clinical MRI scanner. The integrated RF coil and laterally-mounted ultrasound phased-array transducer are indicated. (b) 3D model of the ultrasound transducer. 256 4-mm diameter elements are semi-randomly placed on the transducer face. A solid conductive ground plane is used in this design. Aperture is 14.5 × 8 cm with a 10 cm radius of curvature. (c) The 8-channel RF MRI receive coil integrated into the treatment cylinder shown in (a). The coil consists of a 6-channel ladder array with additional loops placed at the chest wall and around the transducer opening. (d) Top and (e) side schematic views of the breast MRgFUS system. Three rotation angles (Ψ, α, θ) and three translation distances (X_gantry, Z_gantry, X_slide) are indicated as well as the θ=0° and 90° locations and direction of B0. Ψ is measured from the horizontal.

Evaluation of transducer position

The variation in RF coil performance with transducer and RF coil position was evaluated by measuring the SNR at eight different transducer positions, summarized in Table 1. A homogeneous Cu₂SO₄ cylindrical phantom (1.955 g/L deionized water, 12 cm diameter, 11 cm height), designed to match the RF coil load of a human breast, was placed in the treatment cylinder and SNR was measured in a slice positioned 6.5 cm from the top of the treatment cylinder using a 2D gradient echo sequence (coronal orientation, TR/TE = 500/10ms, flip angle = 90°, FOV=256 × 256 mm, Resolution = 1 × 1 × 5 mm, 5 averages). The signal from each coil was optimally combined using the method of Roemer et al. (13). SNR and noise correlation were calculated for each treatment configuration (14). SNR measurements were averaged over the cross-sectional area of the phantom and in an 11×11 voxel region at the approximate geometric focus of the transducer position, simulating a tumor target region.

Table 1:

Transducer positions and orientations evaluated for the effect of transducer orientation on SNR. X_gantry, Z_gantry and α, as defined in Figure 1(d), were fixed. σ_T was computed for a 3D MRTI sequence that has a 77.5% SNR reduction compared to the 2D GRE sequence.

Transducer position	θ (rotation)	xslide (mm)	ψ (tilt)	SNR over entire phantom (mean±1std)	SNR over target region (mean±1std)	σT (°C) over target region
1	0°	15	0°	749±539	414±25	1.08
2	45°	15	0°	808±400	537±17	0.83
3	90°	15	0°	880±331	692±5	0.65
4	90°	15	30°	833±293	687±5	0.65
5	0°	50	0°	629±432	489±12	0.91
6	45°	50	0°	695±333	560±16	0.80
7	90°	50	0°	677±293	592±15	0.75
8	90°	50	30°	712±304	564±9	0.79

Related MR temperature measurement precision can be derived from the SNR measurement, as the standard deviation of the image phase proportional to the inverse of the SNR (15). Temperature measurement precision can be calculated using

$${\sigma }_T=\frac{\surd 2/SNR}{\gamma {\alpha }_{PRF}{B}_{0}TE}$$

where SNR is the measured relative SNR, γ is the gyromagnetic ratio, B₀ is the magnet field strength (T), TE is the echo time (s), and α_PRF is the PRF water proton shift constant. SNR for all evaluated transducer positions was measured using a 2D GRE sequence, using parameters typically not utilized for MR temperature imaging. However, the relative SNR differences measured for the different transducer positions with the 2D GRE sequence are the same as would be obtained with any other sequence. To gain a temperature measurement precision estimate, a MR temperature measurement sequence (3D gradient echo with segmented echo planar imaging readout, TR/TE = 15/11 ms, flip angle = 14°, bandwidth = 752 Hz/pixel, echo train length = 7, resolution = 1.5 × 1.5 × 2 mm, FOV = 192 × 168 × 20 mm) was run at transducer position 3 in Table 1 and the SNR was computed to determine the relative SNR difference.

Evaluation of transducer ground plane design

The effects of the transducer ground plane design on SNR were evaluated with a replica of the MRgFUS system, shown in Figure 2. A single rectangular RF coil (13 × 15 cm), representing the transducer loop of the breast tank was used to image the homogeneous phantom described above. A replica 3D printed plastic transducer (ABS, Hatchbox; Monoprice Maker Ultimate 3D printer) was laterally mounted in the X_slide = 50mm position, close to the transducer coil loop, where it would cause the greatest RF eddy currents. The θ and ψ positions were fixed at 90° and 0°, respectively. The transducer ground plane was simulated by placing copper tape on the face of the replica transducer, replicating the copper ground plane that exists with the transducer used in the breast MRgFUS system. The resonant frequency shift seen by the RF coil with the replica transducer was equivalent to the actual transducer. To create different transducer ground plane configurations, the conducting plane was broken into multiple rows and columns of copper patch elements by removing thin copper strips (~1 mm wide) in the short and long axis directions, as shown in Figure 2b. Twelve different configurations were tested, as seen in Figure 2c. To minimize gradient eddy currents and maximize RF eddy currents, the mock MRgFUS setup was positioned with the RF loop axis perpendicular to the B0 field of the scanner. For each ground plane configuration, the RF coil loop was tuned and matched to 123 MHz with an insertion loss better than −35 dB. Active and preamplifier detuning were better than −35 dB and −20 dB, respectively. SNR was evaluated using a 2D gradient echo sequence, as previously described.

Figure 2:

Mock-up of the breast MRgFUS system used to evaluate effect of transducer ground plane configuration on image quality. (a) Schematic of components in mock-up, including a Cu₂SO₄ phantom, a 3D-printed transducer facsimile and a single 13 × 15 cm rectangular coil. The red arrow indicates the direction of the B0 field. (b) Different transducer ground planes were simulated using copper tape placed on the transducer face. A 3×2 configuration is shown. (c) All evaluated ground plane configurations with number of ground plane segments ranging from 0 to 9.

Results

Evaluation of transducer orientation

Coronal SNR images of the Cu₂SO₄ phantom at the plane of the ultrasound focus with the transducer at the eight positions evaluated as outlined in Table 1 are shown in Figure 3. The mean ± one standard deviation of the SNR over the entire phantom and in the tumor target region (shown as the red boxes in Figure 3) are presented in Table 1. MR temperature measurement precision (σ_T) was calculated assuming γ = 42.58 MHz/T, B0 = 3 T, α_PRF = −0.01 ppm/°C and TE = 11 ms. The relative SNR difference between the 2D GRE SNR sequence and 3D MR temperature measurement sequence was measured to be −77.5% and the temperature precision measurement over the target region listed in Table 1 was scaled by that factor. The SNR as a function of transducer orientation with respect to the B0 field (θ) and distance from the treatment cylinder (x_slide) is shown in Figure 4. In general, SNR increases with increased θ rotation, up to θ =90°, or when the ultrasound propagation is perpendicular to the B0 field. The highest SNR in the simulated tumor region occurs when θ=90°. This orientation has the most homogeneous SNR throughout the entire phantom as well. The effect of the distance of the transducer from the treatment cylinder also significantly affects SNR when analyzing the tumor target region. Tilting the transducer (ψ) has less effect on the total SNR, resulting in a 0.7% or 4.7% reduction when the transducer is moved further (X_slide = 15mm) and closer (X_slide = 50mm) to the RF coil, respectively. When comparing transducer position 1 and 3 in Table 1, the SNR increases in the tumor target region by 67%.

Figure 3:

Coronal SNR maps demonstrating the effect of transducer position with respect to the B0 field. The B0 field direction is indicated by the red arrow on the left. The top and bottom row shows the cases with the transducer further and closer to the treatment cylinder, respectively. The rotation (θ) and tilt (ψ) of the transducer is also indicated. The transducer face in each image is denoted by the white text. SNR values were evaluated over the entire cylinder, and in the tumor target region, an 11×11 voxel region of interest indicated by the red square, located at the geometric focus of the ultrasound transducer.

Figure 4:

Spatial mean SNR as a function of θ rotation and X_slide position over (a) the entire phantom and (b) the tumor target region, indicated by a red box at each transducer position in Figure 3. Error bars are one standard deviation over the spatial mean.

Coronal SNR maps of four of the studied ground plane conditions are shown in Figure 5a. The spatial mean SNR values as a function of number of ground plane segments are plotted in Figure 5b. As the number of ground plane segments was increased, RF eddy currents were reduced resulting in an increased SNR trend. These results indicate that segmenting the transducer ground plane could potentially increase the SNR by 25% at the center of the phantom, when compared to using a continuous transducer ground plane.

Figure 5:

Effect of transducer ground plane configuration on SNR. (a) Coronal normalized SNR maps for 4 of the 12 configurations evaluated in Figure 2c in the replica MRgFUS setup. The direction of the B0 field is denoted by the red arrow. (b) The normalized SNR as a function of total number of ground planes.

Discussion

The breast-specific MRgFUS system used in this work has been engineered to allow treatment from many transducer positions relative to the target anatomy. While some MRgFUS systems place the ultrasound transducer in a relatively fixed orientation, the continuing improvements to the technology including RF coil integration and transducer-positioning flexibility will allow for improved image quality and increased treatment options (4,16,17). This work has presented results that demonstrate SNR variability should be considered both in the design phases and treatment planning using these advancing technologies.

Given the phased array RF coil design and flexible transducer positioning of the breast-specific MRgFUS system, this work demonstrates that SNR variation is a function of both transducer ground plane design, transducer positioning relative to the RF coils, and system orientation with respect to the B0 field. While transducer ground plane issues may be addressed with a design that increases the number of segments of the conductive ground plane, the expected SNR variation with system positioning can be an input that is used in treatment planning to increase image quality. While transducer positioning is largely driven by the location of the tumor in the breast, considerations including ultrasound incidence angle and the potential of acoustic phase aberration are also evaluated. The large number of positioning degrees of freedom provides several transducer location options for each target position. Therefore, in addition to the positioning considerations listed above, SNR effects may also be evaluated when positioning the transducer. The presented results indicate increased SNR homogeneity is seen when the transducer is placed perpendicular to the B0 field. If the transducer is oriented parallel to the B0 field, where gradient eddy currents (18) and transducer loop RF coil signal sensitivity are the worst, the entire volume will have reduced SNR, but particularly at the tumor target location. This diminished image quality would likely be in the near field of any ultrasound sonication. This could adversely affect treatment planning, monitoring and assessment results.

This work also confirms what other investigators have demonstrated (10), that use of a single continuous transducer ground plane in the MRgFUS transducer can result in reduced image SNR. This loss can be dramatically reduced by splitting the ground plane into more segments. The ground plane configuration should be evaluated at the beginning of the ultrasound transducer design process for a given application. Segmenting the ground plane shouldn’t have an effect on the FUS transducer performance, assuming that the segmentation boundaries are thin gaps in the ground plane, that the segments are connected to each other somewhere remotely and that the additional wire connections to the individual elements do not obstruct the ultrasound beam. The reduction of RF eddy current effects in the imaging volume will increase image SNR and enhance both anatomic and temperature imaging during MRgFUS treatments.

Due to the requirements of MRgFUS hardware and limitations of available RF imaging coils, MR images during MRgFUS often suffer from low SNR and eddy current effects of the conductive transducer ground plane. While advancing technologies are integrating RF coils into MRgFUS designs and are considering transducer design techniques that reduce eddy currents, variations in SNR still exist that can impact all aspects of an MRgFUS treatment including planning, temperature monitoring and assessment. However, the overall effect of reduced SNR on treatment outcome is not clear, particularly in clinical cases. Reduced image SNR will result in lower temperature precision and potentially more difficulty in identifying and monitoring both the desired target and critical surrounding areas. While any increase in SNR would benefit MRgFUS treatments during all phases, the ultimate impact on treatment success and patient response requires further study. Use of relative SNR maps, based on different treatment configurations for any given transducer conductive ground plane design, will allow the use of SNR information to be incorporated into the patient treatment planning process by identifying treatment configurations that provide the best SNR for a given treatment volume.