A Handheld Imaging Probe for Acoustic Angiography with an Ultra-Wideband Capacitive Micromachined Ultrasonic Transducer (CMUT) Array

Abstract

This paper presents an imaging probe with a 256-element ultra-wideband (UWB) 1D CMUT array designed for acoustic angiography (AA). This array was fabricated on a borosilicate glass wafer with a reduced bottom electrode and an additional central plate mass to achieve the broad bandwidth. A custom 256-channel handheld probe was designed and implemented with integrated low-noise amplifiers and supporting power circuitry. This probe was used to characterize the UWB CMUT, which has a functional 3-dB frequency band from 3.5 to 23.5 MHz. A mechanical index (MI) of 0.33 was achieved at 3.5 MHz at a depth of 11 mm. These promising measurements are then combined to demonstrate AA. Use of alternate amplitude modulation (aAM) combined with a frequency analysis of the measured transmit signal demonstrates the suitability of the UWB CMUT for AA. This is achieved through measuring only a low level of unwanted high-frequency harmonics in both the transmit signal and the reconstructed image in the areas other than the contrast bubbles.

I. Introduction

ONE of the hallmark characteristics of cancer is angiogenesis, which leads to an increase in tortuosity and density of microvasculature [1]–[3]. Computed tomography and magnetic resonance angiography allow imaging of this microvasculature, but these modalities either use ionizing radiation or are costly and time intensive [4]–[7]. They also suffer from low resolution on the order of approximately 1 mm. Ultrasound is a safe, portable, and low-cost alternate imaging modality to these angiography methods. To achieve the necessary specificity and resolution for angiography, contrast agents, also known as microbubbles, are injected into the bloodstream and subsequently imaged through a method called acoustic angiography (AA) [4], [8], [9].

Conventional B-mode images are formed by transmitting a series of pressure waves and receiving and reconstructing the reflected signals at this fundamental transmit frequency [10], [11]. Conversely, in AA a focused low-frequency ultrasound beam is used to excite the microbubbles, which then oscillate at a broad frequency band as shown in Fig. 1 by the dotted green line [4], [9], [12]. By receiving only the superharmonics (3^rd and higher harmonics), the unwanted reflections from the surrounding tissue can be eliminated. This increases the image contrast for the microvasculature by imaging only the microbubbles [9], [13]. An additional benefit of the high-frequency receive band is higher image resolution that enables viewing microvasculature.

Fig. 1.

Diagram depicting key features of acoustic angiography (AA) including the low-frequency transmit band, wide-band microbubble response, and high-frequency receive band. The AA image is formed where the microbubble response and high frequency receive overlap. Image inspired by [4], [14].

An ideal imaging probe for AA would be able to transmit at a range of low frequencies and also receive in a wide band of high frequencies. The probe would also need to be able to achieve a sufficiently high mechanical index (MI) (>0.2) to activate the microbubbles [4], [8], [15], [16].

Previously used probes have had design limitations. Initial probes had a single element for transmit and a separate single confocal element for receive [8], [9]. Mechanical scanning with this probe is time consuming; the probe has a limited transmit and receive band, and a limited depth of field. To increase the frame rate, dual-frequency probes were aligned such that the image planes of two parallel 1D arrays overlapped at a set focal depth [17], [18]. This limits the field of view and does not allow for focusing at variable depths. Another design introduced interleaved elements in a 1D array [19], but this introduced imaging artifacts such as grating lobes into the image. A more recent design has improved the performance with stacked multi-frequency piezoelectric arrays [12], [20]. These allow for beamsteering and multiple focal depths with a higher achievable frame rate. Their target transmit and receive frequencies must be decided before the probe has been created, and this does not allow for optimal frequency-selection flexibility when performing experiments. A wider receive band would also allow for a higher number of received microbubble harmonics.

Capacitive micromachined ultrasonic transducers (CMUTs) are known for their wide bandwidth in immersion, which results from the low mechanical impedance of the vibrating plate compared to the loading medium impedance when the CMUT is in contact with tissue [21]. CMUTs additionally have design flexibility in the frequency of operation.

Early work on CMUT imaging arrays targeted the design and fabrication of arrays and their characterization at the individual element level [22], [23]. Later these arrays were used with supporting discrete electronics with a limited number of active parallel channels with offline image reconstruction to demonstrate the image quality that can be achieved with CMUT arrays. However, real-time imaging suitable for in-vivo translation requires use of an increased number of parallel data acquisition channels and a probe package in a small form factor with a suitable interface to tissue. First such probe with a 192-element CMUT array was an extension of traditional piezoelectric transducer probe construction with no active electronics in the probe [24]. A side-by-side comparison of medical images between PZT and CMUT medical imaging arrays was included in this work showing the potential of CMUTs in terms of improved axial resolution and broadband operation. Other similar probes were reported in the literature [25]–[27]. A 1D imaging probe with a 12-MHz 192-element CMUT array included analog frontend electronics based on discrete components [28]. In 2009, the first commercial CMUT-based imaging probe was announced by Hitachi Corporation [29], [30]. When one considers only the close integration of CMUT arrays with supporting electronic circuits, the monolithic integration predates all the work cited above. A BiCMOS process had been used with only minor modifications to fabricate CMUTs side-by-side with electronic circuits on the same substrate [31]. The alternative approach of fabricating the electronic circuits first using a standard foundry process and then building CMUTs on top of finished electronics by further processing has also been demonstrated [32]–[34].

Several CMUT designs have been demonstrated for acoustic angiography. The first is a CMUT array with interleaved high- and low-frequency cells [35], which is similar in high-level concept to the interleaved piezoelectric elements. The second design is a single-element dual-frequency CMUT, which utilizes both the conventional and collapse mode [36]. This element remained biased for collapse mode while not transmitting and thus uses this mode to receive high frequency superharmonics. A conventional transmit signal is then generated at a low frequency by releasing the CMUT plate.

The third CMUT design for AA is shown in Fig. 2 and includes a reduced bottom electrode and added central mass on a thin plate [37]. This ultra-wideband (UWB) CMUT has been fabricated in arrays of 256 elements, and applications with these arrays are explored in this paper.

Fig. 2.

Diagram depicting a UWB CMUT cell [37]. The top plate is partially etched on the side in addition to the reduced bottom electrode. Both features contribute to the increased bandwidth of the device.

After a brief discussion of the UWB CMUT fabrication, we present the design and construction of a 256-channel handheld probe designed for use with these UWB CMUT arrays. Next is the characterization of the described probe and subsequent demonstration and analysis of AA from the 3^rd to the 7^th harmonics. A discussion of future directions and further improvements concludes this paper.

II. Array Design and Fabrication

A UWB CMUT array was used for implementing the imaging probe reported in this paper. This broadband response is achieved by three key features: the thin vibrating plate, reduced bottom electrode, and added top plate mass shown in Fig. 2 [37]. In a standard CMUT design, the plate mass and spring constant are closely linked, and there is a tradeoff between a higher pressure, narrower bandwidth device and a lower pressure, wider bandwidth device [38]. With the added mass on the CMUT plate, the spring constant and the plate mass can be decoupled to result in a broad bandwidth and high pressure device [39]. An additional method of design and analysis includes considering the added mass as a way to tune the higher order vibrational mode frequencies [37], [40], [41].

A second key feature of the UWB cell design is the reduced bottom electrode. The resulting electrostatic force acts only on the central area of the plate where it is the most effective. This increases the bandwidth by decreasing the less efficient force on the edges of the plate as well as decreasing the in-cell parasitic capacitance [42].

The UWB CMUT fabrication process follows the general process flow outlined for CMUTs on borosilicate-wafers in [43]. The total process requires seven masks and is outlined in more detail in [37]. Multiple designs have been fabricated that have a band out to 25 MHz [37]. Feature details for the UWB CMUT used in this paper are described in Table I.

TABLE I.

CMUT dimensional parameters

Parameter		Unit
Array
Element pitch	97	μm
Element width	99	μm
Element length	2988	μm
Array outer width	3.0	mm
Array outer length	24.9	mm
Number of cells per element	237	–
Resonant frequency in air	9.25	MHz
CMUT cell
Cell diameter	34	μm
Distance between cells	3.5	μm
Total plate thickness	0.9	μm
Mass on plate thickness	0.4	μm
Mass on plate diameter	28	μm
Insulation layer thickness	200	nm
Vacuum gap height	250	nm
Bottom electrode diameter	14	μm
Contact pad thickness (Cr/Au)	200	nm
Simulated pull-in voltage	75	V
Measured pull-in voltage	80	V

III. Methods

A handheld probe that enables all 256-elements of the UWB CMUT array has significant focusing and mobility advantages over a bench-top system that only activates 32 elements as previously described with the UWB CMUT [44]. The following sections describe the probe construction and system modifications made, which enable later AA with the UWB CMUT.

A. Probe Design and Construction

The 256-channel handheld probe adds bias for all 256 elements and amplifies the received signal. Packaging considerations are also described in detail.

1). System overview:

A block diagram of the handheld probe is given in Fig. 3, where channel paths are blue and power paths are red.

Fig. 3.

Block diagram of the handheld UWB CMUT probe. Signal paths are shown in blue, and power paths are shown in red. Charge pumps with built-in LDOs (LTC3260, Analog Devices, Wilmington, MA, USA), LNAs with TX/RX switches (MAX4805A, Maxim Integrated, Champaign, IL, USA), and biasing circuitry are depicted.

Each printed circuit board (PCB) contains 128 channels, and the two boards slot into a head PCB on which the individual UWB CMUT array is wire bonded. This allows the head piece and thus the CMUT to be exchanged without any other probe modification. The channel PCBs are described further in this section, and the head-piece packaging is described in further detail in Section III–A2.

The electronic circuitry connects the CMUT to the programmable PC-based imaging backend system (Vantage256, Verasonics, Inc., Redmond, WA, USA), which is used to both transmit and receive electrical signals.

The component electronic portions of the channel PCB design included:

• Optionally enabling/disabling individual channels through 0-Ω resistors.

• Applying bias to activate the CMUT through a bias-T circuit.

• Amplifying the received signals through low-noise amplifiers (LNAs) that also had integrated transmit/receive switches (MAX4805A).

• Supplying the necessary power for the LNAs through charge pumps with integrated low-dropout regulators (LDOs) (LTC3260), which require only a single +7-V power rail (V_IN).

• Connecting to the programmable imaging system’s signal channels and voltage rails through a custom ultrasound cable.

The selected LNAs (MAX4805A) have octal packaging and allow for a +/−100-V sinusoidal burst in transmit [45]. The LNA has a 44-MHz 3-dB bandwidth and a 8.7-dB gain. As previously mentioned, there are also integrated transmit switches to isolate the receive path. These amplifiers were chosen in part because they were also used in previous CMUT-probe work [28], [46], [47], and because they include the requisite bandwidth, channel count, and amplification.

The ideal handheld probe would not require additional power supplies beyond what the programmable imaging system can supply. This system can supply three rails: V_PP, which can provide a low-power +100 V; V_NN, which can provide a low-power −100 V; and V_IN, which can be set to +9 V with a 1.5-A current draw. Due to the resistance of the cabling, a 2 V drop is apparent from the system to the probe. This results in an approximate +7-V rail available to the probe. To enable use with only these inputs, the PCB provides power conversion using charge pumps with built-in low-dropout regulators (LDOs) (LTC3260). The system requires one charge pump to supply +/−2 V for every 16 signal channels, or 2 octal-packaged LNA subgroup. It also requires one charge pump to provide +/−5 V for every 128 channels, which resulted in only one +/−5 V charge pump necessary per board.

Fig. 4 shows a breakout view of the modular probe design. Each 128-channel board has 6 layers and consisted of LNA and bias-T connections on one side with charge pumps with built-in LDOs and power conversion circuitry on the other side. The probe shell was 3D printed and lined with copper tape for shielding, and the internal probe construction was surrounded with insulating tape. The external case shielding was connected to the board shielding via a screw through the tape.

Fig. 4.

Photographs of (a) probe casing with copper shielding and (b) yellow insulation tape and front and back of completed probe PCBs with connectors attached. The side with LNAs is shown on the left, and the side with power conversion circuitry is shown on the right.

The two boards were each connected to three outputs from a custom cable: two connectors each carried 64 channels, and the third connector carried power rails. On the opposite end (proximal) of each cable, the standard zero-insertion-force (ZIF) connector is connected to the mating connectors on the programmable imaging system where high-voltage power rails and signal channels were available from the system. This cable and ZIF connector were salvaged from a commercial imaging probe (Model L12–5, ATL Inc., Bothell, WA). The option to connect external power supplies to the ZIF connector through drilled holes in the casing was also available. An optional jumper cable connected the power rails of the two boards to allow for a single ZIF socket to control power selection. Since the bias can be provided by one cable and shared to both PCBs, future adaptations of the custom cable are possible where the bias voltage is controlled via wireless communication to a custom PCB in the ZIF-side connector. This wireless communication will enable easier integration of the bias control with external systems such as an Arduino microcontroller for flexible experimental design.

Fig. 5 shows the initially planned, 3D model of the handheld probe as well as a picture of the fully assembled completed probe. This probe measures 4 by 7.9 by 12.3 cm at the widest points when fully assembled, measurements shown in Fig. 5(b).

Fig. 5.

Photographs showing (a) planned 3D model and (b) completed and assembled handheld probe that measures 4 by 7.9 by 12.3 cm at the widest points.

2). Probe head design:

The decision was made to package the CMUT in a small oil well that was covered with a thin, flexible membrane. Vegetable oil was chosen to provide a good acoustic match to the water while insulating the interconnects of the CMUT and PCB. This quick-turnaround design allows for initial testing in phantom experiments using standard acoustic coupling measures such as ultrasound coupling gel or water. The thickness of the oil layer was 3.3 mm [48].

B. Imaging

There are several important considerations when using CMUTs to demonstrate AA. This section explains the motivation for the two imaging methods used in this paper, conventional and alternate amplitude modulation (aAM). This is followed by a discussion of the experimental phantoms as well as the calculations used to analyze the resulting images.

1). Alternate amplitude modulation (aAM):

As discussed in Section I, traditional AA has utilized separate transducers for transmitting and receiving. This is disadvantageous because it requires knowledge of the desired transmit and receive bands before probe construction, and it can limit the imaging flexibility and frame rate.

If a transmit pulse contains superharmonics, they can add spurious noise to the receive signal. Considering Fig. 1, this would be equivalent to some of the transmit signal in red existing in the AA image band indicated in teal. CMUTs also have naturally occurring nonlinear components that occur in transmission, and so reducing any transmitted harmonics that fall in the receive band is of interest [49]–[52].

There are a variety of methods that can be used to suppress unwanted harmonics in transmit. Some of these methods use an arbitrary waveform generator to generate precise and controlled transmit waveforms that have been calculated to avoid unwanted harmonics in bands of interest [51], [53]. The programmable imaging system used in this study does not have this level of control and precision in transmit.

Other methods modify the CMUT design [54], [55] or add a resistor-inductor matching circuit to filter the transmit signal [56]. These methods require either a completely different CMUT design or electronics that would limit the inherent flexibility of the UWB design.

Another common way to reduce unwanted harmonics involves transmitting multiple times for each image line and summing the resulting beams that are then used to create an image. Three common methods in this category are pulse inversion, amplitude modulation, and the combination of these two [57]. Amplitude modulation requires that the transducer’s harmonic components in transmit be proportionally the same with a varying AC voltage; however, the harmonic generation of a CMUT’s transmit signal varies with the AC amplitude, and so this method does not perform as well with CMUT transmission [58]. Pulse inversion preserves the even harmonics, and thus does not cancel out all unwanted harmonic signals [58].

One multi-pulse-and-sum method that has been designed specifically for CMUT activation is known as bias voltage modulation [59]. The method keeps a constant AC amplitude and alters the biasing DC voltage instead. It has been shown to work in transmission with a CMUT; however it requires a DC bias below 60% of the collapse voltage and an AC voltage that also remains even proportionally smaller. These requirements conflict with the need to achieve a high MI to strongly activate the microbubbles in AA. Additionally, this requires a system that can swiftly change the bias voltage multiple times per beam. This adds system complexity or severely limits the frame rate, which induces motion artifacts.

A final multi-pulse-and-sum method is known as alternate amplitude modulation (aAM), which has been shown to work well with CMUTs in transmission and reception [58]. In [58], spectrum- and image-analysis were used to show aAM is a good choice for suppressing linear reflections as well as both even and odd harmonics with a CMUT. An overview of the transmit pattern is shown in Fig. 6. The even-, then all, then odd-numbered elements are consecutively enabled for transmission, and all elements are enabled in receive. The received A-scans from the partially-enabled transmits are subtracted from the all-enabled A-scans. The resulting data is beamformed, and an image is formed. Since the reduced amplitude is generated by disabling half of the elements, many harmonics generated by the CMUT are cancelled by subtraction. Additionally, this method is more readily integrated with the programmable imaging system used in this study.

Fig. 6.

Diagram explaining alternate amplitude modulation (aAM). (a) All even elements are transmitted first, followed by (b) all elements enabled in a second transmit event and (c) a third transmit event with all odd elements enabled. All elements are enabled in receive for each event. The received A-scans for (a) odd-only and (c) even-only are subtracted from the received A-scans for (b) the all-on event. All linear response will be canceled, leaving only the harmonic response. The resulting data are used to create an aAM beamformed image.

This aAM method was implemented for this study by modifying a conventional linear scan. A linear scan was chosen first because of the wider aperture of the 256-element array and secondly due to the necessity of generating the required MI in multiple locations across the image. Two methods worth noting were implemented to reduce motion artifacts with aAM.

The first method is the half-all-half order of transmission [Fig. 6 (a), (b), and then (c)] instead of the all-half-half order [Fig. 6 (b), (a), and then (c)] previously reported [58]. This transmit order averages out the effects of motion.

The second implemented method involved the order the image data was obtained. A standard linear imaging scheme consists of creating the image by transmitting once for each image line, often with a subset of elements that are closest to that target transmit location. To modify this linear imaging method for aAM, two broad methods can be used. The first method involves generating the entire image with even-elements [Fig. 6 (a)]; followed by generating the entire image with all-elements [Fig. 6 (b)]; and lastly by generating the entire image with odd-elements [Fig. 6 (c)]. The second method consists of transmitting the even-, all-, odd-elements pattern for each beam formed in the image before transmitting the next beam used to form the image. The second method is an improvement over the first because it reduces the artifacts due to motion for each image line.

A third method of reducing motion artifacts for the entire image that was not implemented here involves modifying the order in which the beams are formed. The code used here stepped the beams from one end of the transducer to the other; however, alternating or pseudo-randomly selecting the beams can reduce motion artifacts as well. This would be more important for imaging targets that fill a larger portion of the imaging display. Since our interest in this paper was only on a smaller, more localized region, the third method was not necessary to reduce motion artifacts.

The imaging experiments performed in the presented study interleaved conventional linear imaging as well as conventional linear imaging with aAM applied. Additionally AA is performed, which effectively resulted in the same imaging with a higher-frequency receive filter applied. This allowed for comparable images that had the same target position.

2). Phantom imaging:

The imaging experiments presented in this paper utilized tissue-mimicking phantoms with embedded synthetic microvessels to provide acoustically interesting targets. The gelatin-based phantoms contained 4.5% graphite to acoustically mimic scattering and attenuation (0.3-dB/cm/MHz) of tissue. These phantoms were formed in containers as shown in Fig. 7 where there are two tubes passing through the phantom at the same depth. For each experiment, one tube is filled with deionized water, and the other tube is used to flow a solution of microbubble contrast using a syringe pump (Model 780210, KC Scientific, Holliston, MA, USA). The contrast agent used in this work was an in-house formulation consisting of lipid-shelled microbubbles with decafluorobutane core gas, as has been described previously [60]. The microbubbles had a mean diameter of 1 μm and were diluted in deionized water to an approximate concentration between 2 × 10⁸ - 2 × 10⁹ #/mL for imaging experiments.

Fig. 7.

(a) Diagram explaining phantom setup with flowing microbubbles in one tube and deionized water in the other. (b) Picture of probe imaging a phantom.

Two types of phantoms were used in the imaging experiments, and their key differences are outlined in Table II. Experiment A has a phantom with shallower-depth, larger polyethylene tubes that contained a higher concentration of microbubbles. Experiment B has a phantom with deeper, smaller cellulose tubes with a lower concentration of microbubbles.

TABLE II.

Imaging experiment parameters

Parameter	Experiment A	Experiment B	Units
Transmit Signal
Center frequency (f0)	3.5	3.5	MHz
Number of cycles	1.5	1.5	–
Leading polarity	positive	positive	–
VAC peak to peak	64	64	V
VDC bias	−43	−50	V
Focal depth	11	25	mm
Number of transmit elements	256	256	–
Receive parameters
Number receive beams	256	256	–
Receive band - conventional	2.44 to 4.46	2.44 to 4.46	MHz
Receive band - AA	8.54 to 27.71	11.17 to 27.53	MHz
Receive band - AA	3 f0 to 7 f0	4 f0 to 7 f0	–
Phantom setup
Nominal tube depth	10	20	mm
Tube material	polyethylene	cellulose	–
Tube inner diameter	0.38	0.20	mm
Microbubble tube location	left	right	–
Microbubble (μB) concentration	1:10	1:100	μB:water
Flow rate	120	30	μL/min

3). Contrast-to-water-tube ratio (CWR):

When performing AA, a common metric of interest is the contrast-to-tissue ratio (CTR), which measures the ability to clearly and reliably distinguish between the target and the background tissue [8], [15], [61], [62]. In this study, we are interested in examining the effects of the transducer-emitted harmonics on the quality of the AA image. To do this, we will be comparing the microbubble-filled tube to the water-filled tube as this provides a clearer indication of the presence of any unwanted higher-order harmonics. To this end, we have modified the CTR definition to a contrast-to-water-tube ratio (CWR), which is the ratio of the peak microbubble signal to the peak signal from the water and tube instead of the background tissue. This is defined as

$$CWR=20\cdot {\log }_{10}\left(\frac{V_{Microbubble}}{V_{WaterTube}}\right),$$ (1)

where V is the peak value of the received beamformed data in the area of interest. A high CWR is desired.

IV. Results

The following section describes the resulting experimental data using the materials and processes described in Section III. These results include MI, bandwidth, and imaging data. This section also includes signal-to-noise ratio (SNR) advantages gained through the addition of in-probe LNAs.

A. Probe Design Validation through Signal-to-Noise Ratio Comparison

One metric of interest in this study was the effect of the addition of LNAs to the probe. To understand this, a set of the 6-layer probe PCBs were populated with only bias-T elements to superimpose the AC signals and DC bias. This resulted in effective probe boards without in-probe receive amplification. Pulse-echo measurements were then obtained using both versions of the PCBs: the version with the LNAs and power conversion circuitry and the version without either. The transmit signal in both cases was a 25-V_pp, 3.5-MHz, 1.5-cycle, positive polarity sinusoidal plane wave, and the applied bias was −19 V_DC. The packaged UWB CMUT was used in air, and the reflection from the oil-membrane-air interface was used as the signal of interest.

The following process was repeated for both the with-LNA and without-LNA boards, and the difference was taken between the two calculated values in dB. The individual pulse-echo A-scans for 31 selected channels were averaged 4 times, and the signal-to-noise ratio (SNR) was calculated as

$$SNR=20\cdot {\log }_{10}\left(\frac{V_{RMS,signal}}{V_{RMS,noise}}\right).$$ (2)

The values from the with-LNA and without-LNA boards were subtracted. This resulted in a 12 +/− 2.4 dB advantage in SNR from the added in-probe LNAs.

B. CMUT Characterization: Bandwidth and MI at Target Depth

The 256-element UWB CMUT array with the same parameters described in Table I was used for all following measurements. This device has a 9.25-MHz center frequency in air.

The mechanical index (MI) is a metric of interest and is a measure of pressure at a given frequency. An MI between 0.2 and 0.8 is often adequate for AA [4], [8], [15]. It is calculated as

$$MI=\frac{P_n}{\sqrt{f}},$$ (3)

where P_n is the negative peak pressure in MPa and f is the center frequency of the pulse in MHz.

In addition to measuring the MI, two other important data for successful AA with our UWB CMUT are the corresponding transmit harmonics and the bandwidth of the device (Fig. 8). To measure these, a calibrated hydrophone (HGL-0200, Onda Corp., Sunnyvale, CA, USA) was placed at the focal point in water by using a linear stage to scan. The peak waveforms were captured and analyzed.

Fig. 8.

(a) Measured MI levels corresponding to pressures for experiments A and B with parameters given in Table II and (b) corresponding frequency spectrum of these measurements with −25-dB demarcated by a horizontal dashed line. (c) Small-signal frequency response characterization of the UWB CMUT used in both phantom measurements. Corresponding depth and bias voltages listed in Table II were applied with a 25-ns, 5-V unipolar pulse to simulate the small-signal AC operation. Possible low-frequency transmit frequencies are indicated in red on (b) and (c), and possible receive bands are highlighted in blue

This was performed for the MI measurements [Fig. 8(a)] using the matching experimental parameters for later imaging, tabulated in Table II. The MI corresponding to the experiment A parameters (1-cm depth) was 0.33, and the MI for the experiment B parameters (2-cm depth) was 0.21.

The Fourier transforms of the corresponding MI measurements were calculated [Fig. 8(b)]. Potential transmit (low) frequencies are highlighted in red, and potential receive (high) frequencies are highlighted in blue. Ideally, the energy of the transmit pulse in AA would be high in the (red) transmit band and low in the (blue) receive band.

The corresponding receive bandwidth measurements were achieved by the same method, at the same position, with the same DC-bias voltages respectively. However, the transmit signal was a 5-V_pp, 25-ns unipolar pulse to measure the small-signal bandwidth. The resulting bandwidths were then corrected for the transmit pulse shape and hydrophone response and are shown by the solid line. Attenuation-corrected values can be seen as a dotted line in Fig. 8(c). These values were calculated based on the attenuation in the 3.3-mm oil well as this is the dominant attenuation. The resulting frequency bands are displayed in TABLE III for data both before and after attenuation adjustments. As discussed in Section II, the higher order vibrational modes have been tuned by the added plate mass to bring them into a desirable frequency range for receive. The second 3-dB delineation for this higher harmonic is thus included in the table.

TABLE III.

Bandwidth measurements

		3-dB low (MHz)	3-dB high, 1st (MHz)	3-dB high, 2nd (MHz)
Exper. A	Not adj.	3.5	7.4	–
	Adj.	3.5	12.4	23.2
Exper. B	Not adj.	4.2	11.6	–
	Adj.	4.5	12.0	24.5

C. Imaging: aAM Advantages

We recorded a sample phantom image to display the advantages of aAM when compared to conventional linear imaging, shown in Fig. 9. Our transmit signal was a 102-V_pp, 1.5-cycle sinusoidal burst with a positive polarity centered at 3.5 MHz. This was superimposed with a −49-V_DC bias, and the handheld probe had PCBs with only bias-T circuitry and without the LNA and power circuitry for this image. These PCBs were used since the receive bandwidth was from 2.44–4.46 MHz, and thus buffering the higher frequencies along the cable was not necessary.

Fig. 9.

Images of a phantom containing a tube with microbubbles, tube with water, and several pieces of pencil lead with (a) linear imaging and (b) linear imaging with aAM applied. All targets are clearly visible in (a) the linear image, and only the nonlinear targets (the microbubbles) are visible in (b) the aAM image.

The microbubbles in this phantom were at a concentration of 1:100 and flowing in a phantom with characteristics similar to the phantom in experiment A described in Table II. The main difference between this phantom and that in experiment A was the addition of graphite pencil lead in the phantom.

The probe imaged a phantom with the cross section of two tubes, one with water and the other with microbubbles. There were also pencil lead inserted into the phantom in varying locations to add more reflectors. Despite the numerous reflectors, the aAM image [Fig. 9(b)] clearly displays the location of the only nonlinear reflector.

D. Imaging: Experiment A

To demonstrate AA with the UWB CMUT array in the handheld probe, the phantom in experiment A described in Section III–B2 was imaged with the parameters in Table II (Fig. 10). Corresponding MI and bandwidth measurements for this experiment are shown in Fig. 8.

Fig. 10.

Ultrasound images in experiment A where the microbubble tube is on the left (yellow) and the tube with only water is on the right (blue). Images are displayed with a low-frequency receive band using (a) standard linear and (b) aAM linear imaging as well as AA (high-frequency receive band) images with (c) standard linear and (d) aAM linear imaging. There was a slight position shift between measurements a/b and c/d. Images are normalized to their own frame and square root compressed, displaying on a scale from 0 to 1. CWR results are in the lower-right corner of each frame.

The high concentration of microbubbles and relatively shallower 10-mm tube placement resulted in strongly received microbubble signal. The receive bandwidth for the AA images [Fig. 10(c) and (d)] spanned from the 3^rd to 7^th harmonics. This was further confirmed by testing for detectable signal at these frequencies. The PE20, polyethylene tubes were reflective and can be seen clearly in the conventional linear image [Fig. 10(a)] and faintly in the AA image without aAM [Fig. 10(c)] along with reflections from the probe-head membrane.

CWR values as described in Section III–B3 are calculated and displayed in the bottom-right corner of each image in Fig. 10. These give the strongest contrast in the low-frequency receive, aAM image [Fig. 10(b)]; however, this image has a larger target spot size (0.3 mm) than the AA images (0.2 mm) [Fig. 10(c) and (d)] due to the lower resolution resulting from the lower frequency band.

E. Imaging: Experiment B

Similarly, this section describes phantom measurements that are outlined in Section III–B2 with parameters described in experiment B of Table II. The resulting images and calculated CWR values can be seen in Fig. 11.

Fig. 11.

Ultrasound images of the phantom in experiment B where the microbubble tube is on the right (yellow), and the tube with only water is on the left (blue). Images are displayed with a low-frequency receive band using (a) standard linear and (b) aAM linear imaging as well as AA (high-frequency receive band) images with (c) standard linear and (d) aAM linear imaging. Images are normalized to their own frame and square root compressed, displaying on a scale from 0 to 1. CWR results are in the lower-right corner of each frame.

This phantom was more acoustically challenging with a lower concentration of microbubbles (1:100) and deeper (20-mm), smaller targets. This increase in depth corresponds to a summarily lower MI due to acoustic attenuation; however microbubble response could still be detected to the 7^th harmonic. Cellulose tubes with higher acoustic transparency were used in this phantom to validate that peak measured signal was indeed primarily from the microbubble and not tube reflection. As in Fig. 10, the lowest and thus worst CWR was seen in the conventional linear image [Fig. 11(a)], followed by AA without aAM [Fig. 11(c)]. The highest CWR was seen in the low-band receive aAM image; however, this image also had a larger spot size (0.45 mm) due to the lower frequency in receive when compared to the AA images (0.12 mm).

V. Discussion

A highly-portable handheld ultrasound probe was presented to enable AA with the unique UWB CMUT design. Other probes have been created in research settings for CMUT arrays [28], including row-column arrays [63] and probes for dual ultrasound and photoacoustic imaging [46]. The probe presented here enables a full 256-channel arrangement with integrated amplifiers instead of 192 or 128 channels, and it contains power conversion for these LNAs in the probe handle to enable a more mobile design. This probe-side power conversion leaves room in the ZIF socket for additional circuitry such as potential future bias control via wireless communication. Additionally, the individual 256-channel pathing eliminates the need for multiplexing for accessing a 256-element CMUT array, which allows for higher-frame-rate, full-aperture, real-time imaging than would otherwise be possible.

The unique packaging design of the probe is fully modular to allow for disassembly and reassembly for design troubleshooting and testing multiple CMUTs with the same probe body. An area for future improvement in this probe design is implementing a water-tight seal for the areas past the probe head while retaining the design’s modularity. Additionally, incorporating a lens will allow for more protection and elevation focusing.

The power draw necessary for the LNAs from the programmable imaging system results in elevated temperatures for an LDO in the ZIF socket. Future designs will take advantage of switching power supply chips such as LT1076 to alleviate heating.

The promise of these UWB CMUTs was first demonstrated through MI measurements that show a sufficiently strong MI with low high-order harmonics, which could introduce unwanted spurious signal in the resulting AA image. Next, small-signal bandwidth characterization demonstrated a wide band with the potential to transmit at the fundamental frequency (in this paper, 3.5 MHz) and receive up through several higher-order harmonics (in this paper, through the 7^th harmonic). Other combinations of low- and higher-order transmit and receive frequencies are possible that can be further investigated based on these bandwidth characterizations.

After initial characterization methods demonstrated the potential for AA with this UWB CMUT probe, AA was then demonstrated via imaging. Shallower, higher-concentration and deeper, lower-concentration microbubble contrast agents were imaged, and CWR data was calculated to demonstrate the minimal effect of CMUT transmit harmonics. An alternate to conventional imaging, aAM, was demonstrated as a potential method to further reduce the effect of transducer-generated harmonics that might be present in transmit.

When aAM was applied to conventional imaging, it demonstrated the highest microbubble specificity of all tested imaging methods at the expense of imaging resolution. This increase in specificity ranged from 3 dB to 20 dB, depending on the compared-to imaging method and experimental conditions. This aAM-in-conventional-mode method can be conveniently enabled and disabled, and may be useful in clinical settings to more quickly and easily target the microvasculature region of interest before continuing with AA.

Additionally, a smaller advantage in CWR was present when comparing AA with and without aAM. A difference of less than 2 dB demonstrates that minimal higher-order transmit harmonics were present in the AA image and further strengthens the support for CMUTs as transducers for future AA. This reduced need for aAM in AA will allow for a higher resulting frame rate since aAM requires three transmit-receive events for every image line instead of the more conventional one transmit-receive event.

Further work demonstrating use of this probe at clinically relevant concentrations (<1:1000) will be an important future step. Additional work may also include comparing variations of UWB CMUT designs to optimize their performance in this setting as well as in vivo experiments with this probe.

VI. Conclusion

In this paper, we presented a 256-channel handheld US imaging probe. The probe enabled AA from the 4^th to the 7^th harmonics. An MI of 0.33 was recorded with this probe employing a UWB CMUT array, which also showed an effective receive frequency band from 3.5 to 23.5 MHz. A multi-pulse-and-sum imaging method, aAM, was demonstrated with this probe as a method of further reducing any CMUT transmit harmonics. MI frequency-content analysis and the resulting CWR showed that aAM may not be strictly necessary to achieve good AA results with this design in future work. These results demonstrate UWB CMUTs as promising candidates for future in vivo AA experiments.

Biographies

Jean L. Sanders (S’14) received her B.S., M.S., and Ph.D. degrees from NC State University, Raleigh, NC, USA in 2014, 2015, and 2020 respectively, all in electrical engineering. She is currently working as a postdoctoral scholar in engineering education on diversity, equity, and inclusion at Arizona State University.

Ali Önder Biliroğlu received his B.S. and M.S. degrees from Yildiz Technical University, Istanbul, Turkey, in 2005 and 2009 respectively, both in Electronics and Communication Engineering. He worked for TUBITAK (Technological and Scientific Research Council of Turkey) from 2007 to 2018 as senior researcher focused on both the hardware and software development of traction systems used in Electrical Vehicles and Railroad Vehicles.

He is currently pursuing his career as a Research Scholar in Electrical and Computer Engineering at NC State University. His current studies focus on the development of electronic systems used in CMUT based ultrasound applications both in hardware and software.

Isabel Newsome (S’17) received the B.S. degree in physics from Clarkson University, Potsdam, NY, USA, in 2016. She completed the joint Ph.D. degree in biomedical engineering from the University of North Carolina (UNC), Chapel Hill, NC, USA and North Carolina State University, Raleigh, NC, USA, in 2021.

From 2017 to 2021, she was a trainee in the Integrative Vascular Biology program at UNC. She was the recipient of a Ruth L. Kirschstein Predoctoral National Research Service Award from the National Cancer Institute from 2019 to 2021. Her interests include ultrasound contrast agents, nonlinear contrast imaging, and preclinical research. She is currently based in North Carolina and travels as a Scientific Applications Specialist for FUJIFILM VisualSonics, Inc., Toronto, ON, Canada.

Oluwafemi J. Adelegan (S’16) received his B.S. degree from the University of Lagos, Lagos, Nigeria, in 2008, and his M.S. degree from North Carolina Central University, Durham, NC, USA, in 2014, both in Physics. He completed his Ph.D. degree in Electrical Engineering at the NC State University, Raleigh, NC, where he is currently a postdoctoral scholar. His current research focuses on design and fabrication of wideband and ultra-wideband high frequency 1D and 2D capacitive micromachined ultrasonic transducer arrays, and their integration with front-end electronics for medical imaging and therapy applications.

Feysel Yalçın Yamaner (S’99-M’11) received his B.S. degree from Ege University, Izmir, Turkey, in 2004 and his M.S. and Ph.D. degrees from Sabanci University, Istanbul, Turkey, in 2006 and 2011, respectively, all in electrical and electronics engineering. He received the Dr. Gursel Sonmez Research Award in recognition of his outstanding research during his Ph.D. study. He was a Visiting Researcher with the VLSI Design and Education Center (VDEC) in 2006, and he was a visiting scholar in the Micromachined Sensors and Transducers Laboratory, Georgia Institute of Technology, Atlanta, GA, USA, in 2008. He was a Research Associate with the Laboratory of Therapeutic Applications of Ultrasound, French National Institute of Health and Medical Research, from 2011 to 2012. He was with the Department of Electrical and Computer Engineering, NC State University, Raleigh, NC, USA, as a Research Associate from 2012 to 2014. In 2014, he joined the School of Electrical and Electronics Engineering, Istanbul Medipol University as an Assistant Professor. Currently, he is a Research Faculty in the Department of Electrical and Computer Engineering at NC State University. His research focuses on developing micromachined devices for biological and chemical sensing, ultrasound imaging, and therapy.

Paul A. Dayton (S’98-M’01) received his B.S. degree in physics from Villanova University, Villanova, PA, USA, in 1995, 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 is a member of the Technical Program Committee for IEEE UFFC, and a member of the editorial boards for the journals IEEE ULTRASONICS, FERROELECTRICS, AND FREQUENCY CONTROL and Molecular Imaging, and Bubble Science, Engineering, and Technology. He pursued post-doctoral 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. K. Ferrara; his initial studies involved high-speed optical and acoustical analysis of individual contrast agent microbubbles. In 2007, he joined the Joint Department of Biomedical Engineering, The University of North Carolina and North Carolina State University, Chapel Hill, NC, USA, where he is currently a Professor and the Interim Department Chair. He is currently an Associate Director of Education for the Biomedical Imaging Research Center. His research interests involve contrast ultrasound imaging, ultrasound-mediated therapies, and medical devices.

Ömer Oralkan (S’93-M’05-SM’10) received his B.S. degree from Bilkent University, Ankara, Turkey, in 1995, his M.S. degree from Clemson University, Clemson, SC, in 1997, and his Ph.D. degree from Stanford University, Stanford, CA, in 2004, all in electrical engineering.

He was a Research Associate (2004–2007) and then a Senior Research Associate (2007–2011) in the E. L. Ginzton Laboratory at Stanford University. In 2012, he joined NC State University, Raleigh, where he is now a Professor of Electrical and Computer Engineering. His current research focuses on developing devices and systems for ultrasound imaging, photoacoustic imaging, image-guided therapy, biological and chemical sensing, and ultrasound neural stimulation.

Dr. Oralkan is the Editor-in-Chief of the IEEE Open Journal of Ultrasonics, Ferroelectrics and Frequency Control and serves on the Technical Program Committee of the IEEE International Ultrasonics Symposium. He received the 2016 William F. Lane Outstanding Teacher Award at NC State, 2013 DARPA Young Faculty Award, and 2002 Outstanding Paper Award of the IEEE Ultrasonics, Ferroelectrics, and Frequency Control Society. Dr. Oralkan has authored more than 200 scientific publications.