Implementation of a novel 288-element dual-frequency array for acoustic angiography: in vitro & in vivo characterization

Abstract

Acoustic angiography is a superharmonic contrast-enhanced ultrasound imaging method that produces high-resolution, three-dimensional maps of the microvasculature. Previous acoustic angiography studies have used two-element, annular, mechanically actuated transducers (called “wobblers”) to image microvasculature in preclinical tumor models with high contrast-to-tissue ratio and resolution, but these early wobbler transducers could not achieve the depth and sensitivity required for clinical acoustic angiography. In this work, we present a system for performing acoustic angiography with a novel dual-frequency transducer – a coaxially stacked dual-frequency array. We evaluate the dual-frequency array system both in vitro and in vivo and demonstrate improvements in sensitivity and imaging depth up to 13.1 dB and 10 mm, respectively, compared to previous wobbler probes.

I. Introduction

Ultrasound is a widely used modality for soft tissue imaging due to its portability, safety, high temporal resolution, and low cost compared to computed tomography and magnetic resonance imaging. With the addition of microbubble contrast agents (MCAs) as a blood pool marker, ultrasound can also be used to image blood flow and perfusion [1]. Contrast-specific imaging schemes, such as pulse inversion [2] and amplitude modulation [3], utilize the nonlinear response of MCAs by receiving at the second harmonic frequency and are commonly used to reduce background tissue signal in contrast-enhanced ultrasound imaging. Subharmonic [4], [5] and superharmonic [6], [7] techniques are also used to enhance nonlinear microbubble signal over background tissue.

Bouakaz and colleagues [6] first introduced superharmonic imaging, in which ultrasound images are formed from higher harmonic echoes (≥3x the fundamental frequency) rather than the fundamental or second harmonic response. The authors used a custom phased array with interleaved transmit and receive elements, centered at 0.9 and 2.8 MHz, respectively, to show that greater contrast-to-tissue ratio (CTR) can be achieved with superharmonic over second harmonic imaging [6]. Kruse and Ferrara [7] extended this concept and demonstrated high resolution superharmonic M-mode imaging using 2.25 MHz transmit and 15 MHz receive with two confocal pistons.

Building on these seminal studies, Gessner and colleagues developed acoustic angiography, a three-dimensional (3D) microvascular ultrasound imaging technique, using custom dual-frequency (DF) wobbler transducers with transmit at 2 – 4 MHz and receive at 30 MHz [8], [9]. The authors used this DF strategy to isolate intravascular microbubble signals from signals of the surrounding tissue with high sensitivity and resolution [8], [9]. Implementation of the technique in 3D produces volumetric maps of the microvasculature, which can be quantitatively analyzed for characteristics of vessel morphology, coining the term “acoustic angiography” [8], [9]. Quantitative analysis of tortuosity and vascular density on acoustic angiography images has been used to differentiate tumors from healthy tissue [10]–[12], monitor response to therapy [13]–[16], and assess vascular development [17], [18] in various preclinical models.

However, previous acoustic angiography studies were necessarily focused on the preclinical development of the technique. The DF wobblers were designed specifically for small animal imaging in rodents, with shallow focal depth and depth of penetration limited by the high receive frequency [9]. While preclinical use of acoustic angiography has proven highly useful for tumor differentiation and monitoring treatment, the first clinical application of acoustic angiography in human breast found the technique lacking in several aspects. Most notably, the limited imaging depth excluded many patients from the study, the sensitivity of the transducer was limited at a clinical MCA dose, and respiratory motion significantly degraded image quality during data acquisition [19]. For the progression of this microvascular imaging technique toward clinical translation, improvements in DF transducer technology are required to overcome these issues.

Most clinical ultrasound imaging is performed with array transducers, which have been called “the most important advance in transducer technology” for the benefits they provide in terms of beam focusing, steering, and rapid acquisition sequences [20]. For dual-frequency imaging and therapeutic applications, including acoustic angiography, it follows that the next generation of transducers must be DF arrays [21]. In this work, we present the latest developments in acoustic angiography through the implementation of a coaxially stacked DF array (DFA) transducer, which has been developed by our group [22]. We begin by characterizing this new acoustic angiography system in vitro and go on to demonstrate its in vivo imaging capabilities.

II. Materials and Methods

A. Ultrasound System Description

Fig. 1 provides a schematic of the acoustic system used in this work. The transducer used here is a vertically stacked DF array, as described in [22]. It consists of a 32-element low frequency (LF) linear array stacked behind a 256-element high frequency (HF) linear array for LF transmit and HF receive. The nominal center frequencies of the LF and HF stacks are 2 MHz and 18 MHz, respectively. The design of this DFA allows for confocal imaging through electronic focusing in the lateral (azimuthal) and axial directions. The DFA is operated by two programmable ultrasound machines that share a system clock for phase-accurate synchronization during imaging (Vantage 256, Vantage 256 high-frequency configuration, and Multi-System Synchronization Module, Verasonics, Kirkland, WA, USA). The system is programmed for focused line-by-line imaging with 128 ray lines with transmit at 2 MHz on the LF array and receive with 62.5 MHz sampling rate and a 10 – 30 MHz bandpass filter on the HF array. Hereafter, this form of imaging will be referred to as “dual-frequency mode” or “DF-mode.” Conventional high-frequency tissue imaging (transmit and receive at 15.625 MHz) will be referred to as “B-mode.”

Fig. 1.

Schematic of the acoustic system depicting the overall system configuration (left) and internal dual-frequency (DF) array geometry (right).

For imaging in DF-mode, the transmit aperture consisted of all 32 elements of the LF array, and all 256 elements of the HF array were used on receive. For all transmissions, the full LF aperture was excited with a single-cycle pulse with a frequency of 2 MHz. Each DF-mode frame consists of 128 ray lines with their origins located at the center of every other HF element. Data were beamformed with dynamic receive beamforming on the native Verasonics beamformer after tuning to account for the unique geometry of the DFA, including offsets in the delay profiles for azimuthal focusing on both the LF and HF stacks. These offsets were empirically determined using a needle hydrophone (HNA-0400, Onda Corporation, Sunnyvale, CA, USA). For B-mode, 200 focused ray lines and F-number of 1.5 were used. In both modes, uncompressed, envelope-detected, beamformed radiofrequency data were saved for offline analysis. In all experiments, frame rate was selected to allow comparison between the DFA system and the DF wobbler systems described previously [8], [11].

To acquire 3D image stacks, the DFA was attached to a linear translation stage (Velmex, Inc., Bloomfield, NY, USA) controlled by a custom program (LabVIEW, National Instruments, Austin, TX, USA). Translation of the stage was triggered by the imaging system, and image stacks were collected with 0.1 mm between slices. The number of slices per scan was determined by tumor size and ranged from 50 to 200 slices, corresponding to elevation distances of 5 to 20 mm.

B. Acoustic Characterization

The beam alignment and pressure output of the DFA were measured in degassed water with a calibrated needle hydrophone (HNA-0400, Onda Corporation, Sunnyvale, CA, USA) and a digital three-axis motion stage (Newport Corporation, Irvine, CA, USA). Acoustic signals were digitized (CSE1222, DynamicSignals LLC, Lockport, IL, USA) and recorded with a custom acquisition program (LabVIEW, National Instruments, Austin, TX, USA). Pressure maps in the elevational-axial dimension were collected with plane wave transmissions to demonstrate the effect of the elevation lens on the array. Pressure maps were acquired with 250-μm axial and 100-μm lateral or elevational grid spacing and linearly interpolated to a 50-μm grid for display.

For frequency bandwidth measurements, a calibrated high-frequency needle hydrophone (NH0040, Precision Acoustics, Dorchester, UK) was used to acquire single-cycle transmit waveforms from both LF and HF stacks. For each stack, a single transmit beam axially focused at 10 mm and laterally centered was generated, and 150 waveforms were recorded at a pulse repetition frequency of 10 Hz. Waveforms were recorded and digitized as described above. All data analysis presented in this work was performed in MATLAB (Mathworks, Natick, MA, USA). To obtain the frequency responses, the 150 signals from each stack were averaged, and a fast Fourier transform was performed. The center frequency of each stack was measured as the maximum of the magnitude of the Fourier transform, and the −6 dB cutoff frequencies were measured relative to the center frequency.

C. Contrast Agent Formulation

The contrast agent used in this work was an in-house formulation of lipid-shelled, perfluorocarbon-filled microbubbles, as previously described [11], [23]. Briefly, a 9:1 molar mixture of distearoylphosphatidyl-choline (18:0 PC, Avanti Polar Lipids, Alabaster, AL, USA) and PEGylated dipalmitoylphosphatidyl-ethanolamine (16:0 PEG2000 PE, Avanti Polar Lipids, Alabaster, AL, USA) was prepared in phosphate buffered saline (PBS) containing 15% (v/v) propylene glycol and 5% (v/v) glycerol. After preparation, the lipid solution was aliquoted into cleaned 3 mL vials, and the air headspace in each vial was replaced with decafluorobutane gas (Fluoromed, Round Rock, TX, USA). Contrast was activated with a mechanical agitator (Vialmix, Lantheus Medical Imaging, North Billerica, MA, USA) to form polydisperse microbubbles. The average size and undiluted concentration of these MCAs as measured by single particle optical sizing (Accusizer 780A, Particle Sizing Systems, Santa Barbara, CA, USA) were 0.96 μm and 3.1 × 10¹⁰ microbubbles/mL, respectively.

D. In Vitro Resolution

To assess the resolution of the imaging system, an in vitro experiment was performed. A beaker containing 575 mL of distilled water was placed on a stir plate (Thermolyne Cimarec, Barnstead International, Dubuque, IO, USA) and constantly mixed with a magnetic stir bar on the lowest setting to prevent contrast microbubbles from floating during data acquisition. A solution containing approximately 1.87×10⁵ microbubbles was injected into the bath to create a suspension of spatially separated bubbles. The bubble solution was imaged in DF-mode with a frame rate of 4 fps and focal depth between 5 – 30 mm in 5 mm increments. The maximum driving voltage (28 V) was applied at all depths to maximize pressure and signal-to-noise ratio. At each depth, 100 frames were collected.

In vitro resolution was analyzed as follows. For each focal depth, each frame was cropped to 1 mm on either side of the focal depth (e.g., for 5 mm focal depth, images were cropped to 4 – 6 mm axial range). For any single bubbles in this region, lateral and axial profiles were extracted using the “improfile” function in MATLAB. The profiles were normalized, and the resolution was computed as the full width at half-maximum (FWHM) for each profile. The final resolution metrics were the average FWHMs in each direction. At least 100 bubbles were measured for each focal depth.

E. In Vitro Sensitivity

To measure the sensitivity of the imaging system to contrast, a regenerated cellulose tube with 200 μm inner diameter (Spectrum Laboratories, Rancho Dominguez, CA, USA) was suspended in a water bath orthogonal to the axis of propagation, such that the length of the tube was in the imaging plane. The depth of the tube from the transducer surface was varied with the focal depth of the imaging scheme from 7.5 – 27 mm, and alignment was confirmed with B-mode imaging. The peak rarefactional pressure was held constant at 490 kPa for all depths (MI = 0.35), and a frame rate of 4 fps was employed to allow the tube to fully perfuse between frames. MCAs were diluted in PBS to 1 × 10⁷ microbubbles/mL and infused through the microtube at a volume flow rate of 60 μL/min (31.8 mm/s) with a syringe pump (Pump 11 Elite, Harvard Apparatus, Holliston, MA, USA). Three trials were performed for each depth with a fresh MCA dilution prepared for each trial, and 50 frames were collected while contrast was flowing. For comparison, data was also collected with a DF wobbler probe as previously described [8], [9]. Briefly, a single-cycle cosine-windowed sine wave was transmitted at 4 MHz and MI = 0.35 with the LF element of the wobbler, while a preclinical HF scanner controlled receive at 30 MHz center frequency with a 15 MHz high-pass filter on the receive line (Vevo 770, FUJIFILM VisualSonics, Inc., Toronto, ON, Canada). Compressed image data were saved for analysis.

In vitro sensitivity was quantified by contrast-to-noise ratio (CNR) as defined below. For each depth, a maximum intensity projection (MIP) was created from the image stack. Each MIP was used to draw regions of interest (ROIs) for “contrast” within the tube and “noise” in the surrounding water. Next, the mean envelope amplitude inside each ROI was computed for each frame. CNR was defined as:

$$CNR=20{\log }_{10}\left(\frac{E_{\mathit{contrast}}}{E_{\mathit{noise}}}\right),$$ (1)

where E_contrast and E_noise were the mean envelope amplitudes inside the contrast and noise ROIs, respectively. To obtain the final CNR metric, the CNR values were averaged over the 150 frames collected over three trials for each depth.

F. In Vivo Imaging

All animal studies performed in this work were approved by the University of North Carolina Institutional Animal Care and Use Committee. Subcutaneous fibrosarcoma tumors in female rats (Fischer 344, Charles River Laboratories, Wilmington, MA, USA) were imaged. This tumor model has been described previously [24] and was chosen for its well-vascularized tumors characterized by increased vascular density and tortuosity [10], [25]. Briefly, fibrosarcoma tissue (1 mm³) from donor rats was implanted subcutaneously in the right flank. Tumor allografts grew for approximately two weeks before imaging, and animals were humanely euthanized when tumors reached 2 cm in the largest dimension.

For imaging, rats were anesthetized with vaporized isoflurane (induced at 5%, maintained at 2 – 2.5%), and the skin surrounding the tumor was shaved. A catheter was inserted in the tail vein for administration of MCAs, which was performed with an infusion pump (Pump 11 Elite, Harvard Apparatus, Holliston, MA, USA) with a 15 μL/min volume flow rate. MCAs were diluted in sterile saline to a concentration of 1.5 × 10¹⁰ microbubbles/mL before infusion, resulting in an effective circulating concentration in the range of 0.5 – 1.5 × 10⁸ microbubbles/mL. A water bath and ultrasound gel were used to couple the transducer to the animal’s skin. For in vivo imaging, the focal depth of the DFA was set equal to the center of the tumor, frame rate was set to 4 fps, and MI = 0.48 was used. Tumors were imaged with both the DFA and a DF wobbler for comparison, as described in the previous section. On the DFA system, four frames were averaged at each position, compared to five frames on the wobbler system.

For image analysis, DFA images collected on the Verasonics Vantage were formed by log-compressing envelope-detected data and were displayed with 44 dB dynamic range. However, the Vevo 770 used in this work could not provide uncompressed data; instead, it exported 8-bit images with unknown compression. Therefore, the dynamic range for display of DF wobbler images was chosen such that image contrast was comparable to the corresponding DFA image when displayed at 44 dB. Images are displayed as MIPs unless otherwise specified. Because the wobbler and array had to be physically moved to acquire images of the same tumor, a true slice-by-slice comparison could not be achieved. MIPs were therefore used to compare approximately the same volume with each device.

III. Results

A. Acoustic Characterization

The results of hydrophone measurements are shown in Fig. 2. Fig. 2A provides the frequency bandwidths measured on transmit for the LF and HF stacks. The center frequency of the LF array is measured at 1.95 MHz with −6 dB cutoffs at 1.32 and 2.55 MHz. For the HF array, a center frequency of 17.1 MHz is measured, and the −6 dB bandwidth reaches from 13.4 to 22.6 MHz. The −20 dB lower and upper cutoffs of the main lobe of the HF bandwidth are 10.8 and 28.3 MHz, respectively.

Fig. 2.

Acoustic characterization: A) transmit frequency bandwidths of the 2 MHz (red) and 18 MHz (black) stacks; B) maximum peak rarefactional pressure vs. focal depth for constant applied voltage (28 V) on the 2 MHz stack; C) lateral-axial pressure maps demonstrating effective electronic focusing in azimuth and alignment of both beams at 10 mm focal depth; and D) elevational-axial pressure maps showing the beam thickness in elevation. On all pressure maps, the −6 dB beamwidth is denoted by the white contour.

Maximum rarefactional pressure output is generated by the LF stack for a focal depth of 10 mm, producing 674 kPa and corresponding to MI = 0.48 (Fig. 2B). Based on these measurements, imaging in DF-mode can be performed in the range of MI = 0.3 – 0.5 for depths up to 3 cm. The LF and HF beams produced by electronic focusing at 10 mm depth are shown in Fig. 2C. Here, the white contour denotes the −6 dB beamwidth for each array, confirming alignment of the beams for coaxial dual-frequency imaging. Furthermore, the LF and HF beams are shown in elevation in Fig. 2D. The HF beam is narrowest from 6 – 12 mm axially, due to the elevational lens on the DFA; over this range, the average elevational beamwidth is 0.7 mm. At all depths, the LF beamwidth is larger than 3 mm.

B. In Vitro Resolution

Fig. 3 depicts the axial and lateral resolution of the DFA measured in vitro. The theory surrounding superharmonic imaging is complex, but in general, the received pulse is determined by the nonlinear response of the microbubble shell, which is especially high frequency when the shell is disrupted. Here, the system is sensitive to signals up to 30 MHz, based on the transducer bandwidth and receive sampling and filtering used. Measurements show an average axial resolution of 68.4 μm. Near the HF focus of the elevation lens of the array at 10 mm, lateral resolution is measured as 104.6 μm. Table I provides all resolution values measured in this experiment.

Fig. 3.

In vitro measurement of axial and lateral resolution vs. focal depth. Data presented as mean full-width at half-maximum ± standard error.

TABLE I.

In Vitro Resolution

Depth (mm)	Number of Bubbles	Axial Resolution (μm)	Lateral Resolution (μm)
5	160	65.5 ± 1.0	96.3 ± 1.7
10	121	69.5 ± 1.6	104.6 ± 2.3
15	127	72.3 ± 1.3	105.0 ± 2.1
20	117	68.6 ± 0.9	111.7 ± 2.5
25	116	67.3 ± 0.7	124.9 ± 2.9
30	115	67.5 ± 0.8	151.2 ± 3.1

Data presented as mean full-width at half-maximum ± standard error.

C. In Vitro Sensitivity

Results of the in vitro sensitivity experiment are given in Fig. 4. Example MIPs collected with the DF wobbler system and the DFA system at a focal depth of 20 mm are shown in Fig. 4A and 4B, respectively. The ROIs used for CNR calculation are provided for reference. It should be noted that the post-excitation peaks [26], [27], observed as an extension of signal outside the tube in the axial dimension, were not included in the contrast ROI. While these signals are present in this water bath experiment, they are not typically observed in attenuating media, such as tissue. Fig. 4C gives CNR as a function of focal depth for the two devices. Here, we observe roughly constant CNR for all depths with the DFA, illustrating the benefit of using electronic focusing to match target depth and maintain constant pressure. The DFA exhibits maximum CNR (38.8 dB) at 10 mm, near the elevation focus of the HF array. At 15 mm, we find that the DFA system achieves 13.1 dB greater CNR compared to the wobbler system at the same MI. As expected, the CNR achieved with the wobbler probe follows a Gaussian shape, with maximum CNR at the focal depth of 15 mm. The limited field of view of the DF wobbler did not allow for calculation of CNR at a depth of 27 mm. All measured CNR values are provided in Table II.

Fig. 4.

In vitro contrast-to-noise ratio measurement: example maximum intensity projections at 20 mm depth for the dual-frequency A) wobbler and B) array systems with contrast (red) and noise (yellow) regions of interest; C) mean contrast-to-noise ratio for increasing focal depth.

TABLE II.

In Vitro Sensitivity

Depth (mm)	Wobbler CNR (dB)	Array CNR (dB)
7.5	0.4 ± 0.1	33.4 ± 0.5
10	5.2 ± 0.4	38.8 ± 0.4
12	14.2 ± 0.3	37.7 ± 0.8
15	21.7 ± 0.3	34.8 ± 0.6
17	20.9 ± 0.5	35.9 ± 0.9
20	19.7 ± 0.4	35.1 ± 0.9
24	12.5 ± 2.6	33.2 ± 1.0
27	-----------	34.6 ± 0.6

Data presented as mean ± standard deviation. CNR = contrast-to-noise ratio.

D. In Vivo Imaging

Fig. 5 provides B-mode slices (A – B) and DF-mode MIPs (C – D) of a large rat fibrosarcoma tumor, demonstrating the ability of the DFA to perform high-contrast acoustic angiography in vivo. This dataset consisted of 220 slices and was acquired in less than 5 minutes. In comparison to the DF wobbler system, the DFA system exhibits comparable resolution with higher sensitivity, shown by greater MCA signal and reduced contamination from artifact (Fig. 6). In Fig. 6B – 6C, when imaging with the wobbler, we observe bright artifacts from the animal’s skin (indicated by arrows) that do not appear when imaging with the DFA, as shown in Fig. 6E – 6F. The DFA also exhibits 4 – 5 mm greater depth of field (i.e., the axial range over which the transducer is sensitive to microbubble signals) than the DF wobbler (Fig. 6C and 6F) using a single-focal-zone imaging scheme. Finally, the DFA system is able to resolve vessels in vivo on the order of 100 – 200 μm, as illustrated in Fig. 7.

Fig. 5.

In vivo images of a 20 mm rat fibrosarcoma tumor acquired with the dual-frequency array system: A – B) high-frequency B-mode, C – D) dual-frequency acoustic angiography. B-mode images are single slices from the 3D dataset, while dual-frequency images are presented as maximum intensity projections in either the sagittal (C) or transverse (D) plane. Scale bar = 4 mm.

Fig. 6.

Comparison of in vivo images of three different rat fibrosarcoma tumors imaged with a dual-frequency wobbler (A – C) or the dual-frequency array (D – F). Images are presented as maximum intensity projections in the transverse plane. Red arrows indicate skin artifacts. Scale bar = 3 mm.

Fig. 7.

In vivo images of an 11 mm rat fibrosarcoma tumor acquired with the dual-frequency array system: A) Maximum intensity projection through a 15 mm scan of the tumor and B – D) single slices taken from positions 3.5 mm (B), 6.9 mm (C), and 14.6 mm (D) in the dataset, illustrating resolution of 100 – 200 μm vessels. Images are displayed in the transverse plane. Scale bar = 3 mm.

IV. Discussion

In this study, we have presented a novel imaging system consisting of a coaxially stacked dual-frequency array and programmable research scanners for performing acoustic angiography. Overall, the results demonstrate that this system can be used for high-sensitivity, high-resolution, real-time DF-mode imaging up to 27 mm in depth.

The device described in this work differs significantly from the interleaved or co-linear dual-frequency array transducers implemented in previous works [6], [28], [29]. Here, the integrated stacked design of the DFA allows for truly coaxial imaging. Conversely, co-linear array designs have a fixed elevation focus dependent on the intersection of the outer beams [28], [29]. While interleaved arrays can have more ergonomic designs than co-linear probes, they can suffer from the presence of grating lobes [6]. The DFA used here does not suffer from these limitations, but the design of appropriate stacking layers and aperture sizes for such a device can be technically challenging. While other coaxial DF transducer designs have been reported for intravascular and intracavitary applications of superharmonic imaging [30]–[32], the transducer implemented in this study represents the first coaxially integrated transcutaneous DF array used for acoustic angiography. Detailed reviews on dual-frequency transducer technology for superharmonic imaging can be found elsewhere [21], [22], [33].

The beamwidth of the DFA in elevation is 0.7 mm for the HF stack and >3 mm for the LF stack. In comparison, the elevation beamwidth of the wobbler probe used in this work is 0.36 mm for the HF element and 0.46 mm for the LF element (similar to the device described in [9]). Conventionally, the acoustic point spread function (PSF) is equal to the convolution of the transmit and receive PSFs. For dual-frequency imaging, however, the system PSF is more complex and is affected by the highly nonlinear bubble response, in addition to the transmit and receive beams. While the HF receive PSF contributes to the resolution of superharmonic imaging, other factors, including the transmit beam shape and amplitude and the microbubble response, also affect the overall system PSF. In this work, the wobbler probe likely has a smaller overall PSF in elevation and therefore better resolution in this dimension compared to the array. These effects must be considered in the design of future DF arrays.

While an improvement in sensitivity was expected when transitioning to the DFA system, the difference in superharmonic artifacts within the in vivo images (usually from skin or bone) was unexpected. It should be noted that the positioning of the coupling gel and water bath was not altered when switching between the DF array and wobbler devices. There are two central differences between the array and wobbler systems that may be responsible for this phenomenon. First, there is a great difference in the frequencies used for acoustic angiography with a wobbler compared to the array. With the DF wobbler transducers described in the past, the fundamental (i.e., transmit) frequency was 4 MHz, while the center frequency of the receive bandwidth was the 6th or 7th harmonic (25 – 30 MHz) [9]–[11]. However, with the DFA presented here, the fundamental frequency was 2 MHz, and the center frequency of the receive bandwidth was the 8^th – 9^th harmonic. Second, on receive, the wobbler is a single-element system, while the DFA system receives with 256 elements. The beamforming used on each system is therefore very different, and the high number of elements on the DFA may make this system less susceptible to clutter. Taken together, these two disparities may be responsible for the higher prevalence of artifacts within wobbler images.

The system described in this work can resolve vessels on the order of 100 μm in diameter at depths less than 3 cm. While this array represents a step toward the clinical translation of acoustic angiography, DF transducers that can image at more clinically relevant depths will be necessary. The development of such devices will likely require a further reduction of the transmit and receive frequencies to accommodate the greater attenuation of high frequency signals encountered when imaging at depth, in addition to improving the design of the transducer geometry for deeper imaging. Further work is needed to elucidate the best frequency combinations for performing acoustic angiography in relevant clinical scenarios, such as the evaluation of suspicious breast lesions. Translation to clinical applications warrants further investigation to optimize DF transducer designs, including finite-element modeling and experimental validation.

The potential impact of microvascular ultrasound imaging, initially enabled by techniques like acoustic angiography, likely motivated the development of ultrasound localization microscopy (ULM), a super-resolution ultrasound technique inspired by optical photo-activation localization microscopy [34]–[36]. With ULM, vessels can be resolved well below the diffraction limit, on the order of 10 μm [36], [37]. However, the time required for image acquisition and processing can be quite lengthy for 3D applications when compared to acoustic angiography, in which volumes containing vessels on the order of 100 μm can be reconstructed in less than 5 minutes.

Our group has previously demonstrated that superharmonic imaging with a DF transducer can be used for high-sensitivity bubble detection for ULM [27]. Using a prototype co-linear DF probe, the authors were able to image 20 μm vessels in a rodent kidney [27], [28]. While we have previously shown that acoustic angiography can be used to differentiate tumors from healthy tissue, this analysis was based on vessels with diameters greater than approximately 150 μm [10], [11]. The diagnostic power of this technique may be improved with the ability to analyze even smaller vessels for characteristics of malignancy. In the future, this combination of superharmonic imaging and ULM will be implemented on a DFA, such as the one described herein, for tumor imaging. This will allow real-time visualization of larger microvasculature with acoustic angiography, followed by detailed analysis of high-resolution vascular features with ULM. However, for mechanically scanned techniques like superharmonic ULM and acoustic angiography, tissue motion between elevation slices can lead to errors in microvascular reconstruction. We have previously demonstrated that B-mode images interleaved with DF images can be used for speckle tracking and motion compensation [27]. In future work, this approach can be implemented on the DFA system described in this work to improve visualization of microvasculature in high-motion targets, such as the human breast [19].

V. Conclusion

In this work, we have presented the first use of a novel system for acoustic angiography using a coaxially stacked dual-frequency array. This system allows acoustic angiography to be performed at depths up to 3 cm, an improvement of 1 cm over previous dual-frequency wobblers. Additionally, a reduction in transmit and receive frequencies allows an improvement in sensitivity of at least 13.1 dB in vitro, and we have shown that the DFA can resolve vessels as small as 100 μm in vivo. Overall, the integrated array and system presented herein represents the next step in dual-frequency technology to facilitate the clinical translation of acoustic angiography. Future work will focus on further improvements in imaging depth and sensitivity with dual-frequency transducers specifically designed for imaging in clinical applications.

Biography

Isabel G. Newsome (Student Member, IEEE) received the B.S. degree in physics from Clarkson University, Potsdam, NY, USA, in 2016. She is currently pursuing the Ph.D. degree with the Joint Department of Biomedical Engineering, The University of North Carolina at Chapel Hill and North Carolina State University, Chapel Hill, NC, USA, where she focuses on advancements in superharmonic ultrasound imaging technology and the clinical translation of acoustic angiography.

She is currently a Trainee at the UNC’s Integrative Vascular Biology Program.

Mrs. Newsome was a recipient of the Ruth L. Kirschstein Predoctoral Individual National Research Service Award from the National Cancer Institute.

Thomas M. Kierski (Student Member, IEEE) received the B.S. degree in biomedical engineering from The University of North Carolina at Chapel Hill, Chapel Hill, NC, USA, in 2017. He is currently pursuing the Ph.D. degree with the Joint Department of Biomedical Engineering, The University of North Carolina at Chapel Hill and the North Carolina State University, where he is researching superharmonic imaging and novel approaches to ultrasound localization microscopy.

Guofeng Pang (Member, IEEE) received the Ph.D. degree in physics from Queen’s University, Kingston, Ontario, Canada in 2003.

Starting in 2004, he worked as a Research Scientist at the Sunnybrook Health Sciences Center, Toronto, ON, Canada, where he focused on research and development of ultrasound linear arrays for biomedical applications. In 2006, he joined FUJIFILM VisualSonics, Inc., Toronto, ON, Canada. At VisualSonics, he has worked on research, development, and fabrication of ultra-high frequency ultrasound array transducers for pre-clinical and clinical micro-ultrasound imaging as a Production Coordinator and Transducer Developer.

Currently, Dr. Pang is a Senior Transducer Developer, where his research involves photoacoustic imaging, dual-frequency imaging, functional imaging, and creating novel ultrasound transducer arrays and medical devices.

Jianhua Yin received the Ph.D. degree in physics from Nanjing University, Nanjing, China, in 1990.

His previous research involved surface acoustic wave devices and their applications in electronics, Nanjing University, ultrasonic nondestructive evaluation, Tokai University, Shimizu, Japan, and domain study and characterization of piezoelectric crystals, Pennsylvania State University, State College, PA, USA. He is a Research Physicist with Sunnybrook Research Institute, Toronto, ON, Canada, where he is currently working on finite-element modeling, ultrasound transducers, and arrays design and fabrication.

Jing Yang (Student Member, IEEE) received the B.A.Sc. degree in engineering science from University of Toronto, Toronto, ON, Canada in 2016. She is currently pursuing the Ph.D. degree with the Department of Medical Biophysics, University of Toronto. Her research focuses on superharmonic ultrasound imaging with ultrafast imaging techniques.

Emmanuel Cherin (Member, IEEE) received the Ph.D. degree in physical acoustics from University Pierre etMarie Curie (Paris VI), Paris, France, in 1998.

He joined Imaging Research Group, Sunnybrook Health Sciences Centre, Toronto, ON, Canada, as a Postdoctoral Fellow. Since 2001, he has been a Research Associate with Sunnybrook Research Institute, Toronto, where he is involved in research projects related to high-frequency ultrasound imaging, photoacoustic imaging, ultrasound contrast agents, and transducer characterization.

F. Stuart Foster (Fellow, IEEE) is currently a Senior Scientist with Sunnybrook Research Institute, Toronto, ON, Canada, and a Professor with the Department of Medical Biophysics, University of Toronto, Toronto. His current research centers on the development of high-frequency clinical and preclinical imaging systems, array technology, intravascular imaging, photoacoustics, and molecular imaging.

Dr. Foster is a fellow of the American Institute of Ultrasound in Medicine, the Royal Society of Canada, and the Canadian Academy of Engineering. He was elected as a foreign member of the U.S. National Academy of Engineering in 2017. He was a recipient of the Eadie Medal for his major contributions to engineering in Canada, the Queen’s Golden Jubilee Medal, the Manning Award of Distinction for Canadian Innovation, the Ontario Premier’s Discovery Award, the 2010 Rayleigh Award, and the 2020 Biomedical Engineering Award from the IEEE. He is also the Founder and the Former Chairman of Fujifilm VisualSonics Inc., a company dedicated to preclinical and clinical micro-ultrasound. He co-founded Mouse Imaging Centre (MICe) now at the Toronto Centre for Phenogenomics. He has served on the Board of Directors for the National Cancer Institute of Canada and as the Chairman of its Committee on Research (ACOR). He serves on numerous advisory bodies and is currently an Associate Editor of Ultrasound in Medicine and Biology.

Claudia A. Carnevale received the B.S. degree in physics from The University of Guelph, Guelph, ON, CANADA, in 2009. She completed post-graduate studies in advanced lasers at Niagara College and joined FUJIFILM VisualSonics, Inc. in 2011.

Mrs. Carnevale is currently the Manager of Transducer Development at VisualSonics., where her team conducts research, development, and fabrication of ultra-high frequency ultrasound array transducers for pre-clinical and clinical micro-ultrasound imaging. Her team is developing products incorporating dual-frequency, photoacoustic, and phased array transducer technology.

Christine E. M. Démoré (Member, IEEE) received the B.Sc. degree in engineering physics and the Ph.D. degree in physics from Queen’s University, Kingston, ON, Canada, in 2000 and 2006, respectively.

From 2007 to 2015, she was based at Institute for Medical Science and Technology, University of Dundee, Dundee, U.K., and the School of Engineering, University of Glasgow, Glasgow, U.K., in 2016. She is currently a Scientist with Sunnybrook Research Institute, Toronto, ON, Canada, and an Assistant Professor with the Department of Medical Biophysics, University of Toronto, Toronto. Her research includes creating novel ultrasound transducer arrays for biomedical applications, with a specialization in probes for micro-ultrasound imaging. She is also exploring acoustic and photoacoustic contrast-enhanced imaging with new transducer arrays and novel contrast agents.

Dr. Démoré is an Elected Member of the IEEE UFFC-S Administrative Committee. She was a recipient of the Royal Society of Edinburgh/Caledonian Research Fund Biomedical Personal Research Fellowship and the IEEE Ultrasonics Young Investigator Award in 2015 for her development of ultrasonic arrays for particle manipulation.

Paul A. Dayton (Senior Member, IEEE) received the B.S. degree in physics from Villanova University, Villanova, PA, USA, in 1995 and the M.E. degree in electrical engineering and the Ph.D. degree in biomedical engineering from the University of Virginia, Charlottesville, VA, USA, in 1998 and 2001, respectively.

He pursued postdoctoral research and was later research faculty at the University of California at Davis, Davis, CA, USA. Much of his training was under the mentorship of Dr. Ferrara, where his initial studies involved high-speed optical and acoustical analysis of individual contrast agent microbubbles. In 2007, he moved to the Joint Department of Biomedical Engineering, The University of North Carolina at Chapel Hill, Chapel Hill, NC, USA, and North Carolina University, Raleigh, NC, USA, where he is currently a Professor and the Interim Department Chair. He is also the Associate Director of Education for the Biomedical Research Imaging Center. His research interests involve ultrasound contrast imaging, ultrasound-mediated therapies, and medical devices.

Dr. Dayton is a member of the Technical Program Committee of IEEE UFFC and editorial boards of the IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control and Molecular Imaging. He was the recipient of the 2020 IEEE UFFC Carl Hellmuth Hertz Award.