Backward-Mode Photoacoustic Imaging Using Illumination Through a CMUT With Improved Transparency

Abstract

In this paper, we describe a capacitive micromachined ultrasonic transducer (CMUT) with improved transparency for photoacoustic imaging (PAI) with backside illumination. The CMUT was fabricated on a glass substrate with indium tin oxide (ITO) bottom electrodes. The plate was a 1.5-μm silicon layer formed over the glass cavities by anodic bonding, with a 1-μm silicon nitride passivation layer on top. The fabricated device shows approximately 30%–40% transmission in the wavelength range from 700 nm to 800 nm, and approximately 40%–60% transmission in the wavelength range from 800 nm to 900 nm, which correspond to the wavelength range commonly used for in vivo photoacoustic (PA) imaging. The center frequency of the CMUT was 3.62 MHz in air and 1.4 MHz in immersion. Two preliminary PAI experiments were performed to demonstrate the imaging capability of the fabricated device. The first imaging target was a 0.7-mm diameter pencil lead in vegetable oil as a line target with a sub-wavelength cross-section. A 2-mm diameter single CMUT element with an optical fiber bundle attached to its backside was linearly scanned to reconstruct a 2D cross-sectional PA image of the pencil lead. We investigated the spurious signals caused by the light absorption in the 1.5-μm silicon plate. For pencil lead as a strong absorber and also a strong reflector, the received echo signal due to the acoustic excitation generated by the absorption in silicon is approximately 30 dB lower than the received PA signal generated by the absorption in pencil lead at the wavelength of 830 nm. The second imaging target was a “loop-shape” polyethylene tube filled with indocyanine green (ICG) solution (50-μM) suspended using fishing lines in a tissue-mimicking material. We formed a 3D volumetric image of the phantom by scanning the transducer in x and y directions. The two experimental imaging results demonstrated that CMUTs with the proposed structure are promising for photoacoustic imaging with backside illumination.

I. Introduction

Photoacoustic imaging (PAI) is emerging as an attractive hybrid imaging modality that combines the physics of optical imaging and ultrasound imaging [1]. It is a complement to the traditional pulse-echo ultrasound imaging since it can provide optical contrast information. Compared to optical imaging techniques, PAI provides deeper penetration and excellent spatial resolution primarily determined by ultrasound wavelength [2]. Therefore, PAI is especially useful for biological tissues with inhomogeneous optical but relatively homogeneous acoustic properties. In addition to providing structural images, functional or physiological information can be obtained by imaging at different wavelengths and then performing spectroscopic analysis. Thus PAI can also be helpful in applications such as functional imaging, molecular imaging, and therapy monitoring [3]–[5].

In PAI, laser pulses with nanosecond-range pulse widths are used to illuminate the target tissue to cause a brief heating of the absorbing structures. Ultrasound is therefore generated due to the thermoelastic effect and then can be detected by a transducer. Ultrasound transducer technology for optimized PAI is still under development. The most commonly used strategy is based on the traditional piezoelectric transducers [6]. Capacitive micromachined ultrasonic transducer (CMUT) technology has the advantages of fabricating large arrays, integration with electronics, wide bandwidth, and broad selection of processing materials. CMUT is especially suitable for PAI due to the broadband nature of the photoacoustic signal (typically tens of MHz). CMUTs therefore have been investigated for PAI [2], [7]–[11]. Recently, clinical PAI results have been demonstrated by using CMUT arrays as a receiver [12]. Besides, optical ultrasonic sensors, such as the etalon-based all-optical vibration sensor [13], [14], and the optical micro-ring resonator (MRR) [15], [16], have also been widely investigated for photoacoustic signal detection.

An important challenge in PAI is the arrangement of the laser source and the ultrasound transducer. Various approaches have been demonstrated for different applications [17]. One of the commonly used arrangement is to have the light source illuminate the target from the single side or two sides at a right angle to the acoustic path [7]. Another implementation is to have the light source integrated as two fiber bundles along the length direction of a 1D transducer array [5], [18]. However, this method does not illuminate the surface area under the transducer array and results in a blind spot in front of the transducer. Also, it is difficult to use this approach with 2D arrays, which would result in larger area under the 2D transducer array not being illuminated. This approach also limits the footprint of the imaging probe as the fiber bundle and the closure required to place it next to the ultrasound array occupy extra space. The small footprint is especially important for intracavital probes. The light illumination has also been distributed using a spherical mirror [19]. Another approach is face-to-face arrangement which is also referred to as forwardmode [12], [20], [21], which is not practical for many clinical applications.

In many diagnostic imaging and image-guided therapeutic applications, it is desirable to have the light source integrated behind the ultrasound transducer for implementing the so-called backward-mode PAI. First of all, this approach would allow a more uniform illumination in the near field. Also, it could lead to a compact form factor, which is particularly important for catheter- or endoscope-based imaging. One approach to implement the backside illumination is to use a ring array or annular array and introduce an optical fiber bundle into the central lumen [10], [18], or place a forward-viewing ultrasound catheter in the central lumen of an annular light ring [9]. However, the former arrangement limits array geometry and illuminated area and the latter results in a larger catheter size. The ideal implementation is to have a transparent transducer which allows the laser illumination to go through with minimal self absorption. Some transparent piezoelectric materials such as lithium niobate (LNO) [22] and polyvinylidene fluoride (PVDF) [23] have been proposed for such purpose. A PVDF annular transducer array has been adapted for PAI by introducing illumination through a hole in the center of the array [24]. Also, transparent Fabry-Perot polymer film optical ultrasound sensors and an optical micro-ring resonator (MRR) have been investigated for implementation of backward-mode PAI [13], [25]. However, there are only limited reports on using CMUTs for such an operation. An earlier attempt was based on CMUTs fabricated on a silicon substrate [11], where transmission is limited and the wavelength of the light source has to be above 1000 nm to reduce the silicon substrate absorption. We have previously presented a fabrication process to make CMUTs on a glass substrate with indium tin oxide (ITO) bottom electrodes for improved transparency [26]. The transmission was improved to 30% to 70% in the wavelength range from 700 nm to 900 nm. Here in this work, we use the device in immersion and have laser illumination through the fabricated CMUT to demonstrate preliminary photoacoustic imaging results. We also present a detailed characterization of the effects of light absorption in the silicon plate.

In the next two sections, the device fabrication and characterization are presented. In Section IV, the experimental setup for imaging and the image reconstruction method are described. In Section V, the experimental imaging results are presented and the effect of the silicon plate absorption is studied. Section VI discusses the limitations of the current device and suggests further improvements.

II. Device Fabrication

The proposed implementation is illustrated in Fig. 1. The CMUTs are fabricated on a glass substrate with ITO bottom electrodes. A 1.5-μm silicon layer was formed by anodic bonding over the glass cavities with a 1-μm silicon nitride passivation layer on top. An insulation layer could be incorporated in the device structure to prevent an electrical short circuit in case the plate pulls in [27]. In this paper, we present a design for use in conventional mode receive-only operation and therefore an insulating layer was not included. The laser output was fixed at the back of the CMUT so that the light passes through the device and illuminates the target to generate photoacoustic signals, which can be detected by the CMUT.

Fig. 1.

A schematic diagram of the CMUT with improved transparency with backside illumination for photoacoustic imaging.

The fabrication is based on the three-mask process we have reported before [27], with some modifications (Fig. 2). The starting substrate was a standard 0.5-mm thick, 100-mm diameter borosilicate glass wafer (Borofloat®33, Schott AG, Jena, Germany) that has a high surface quality with an RMS roughness of 0.7 nm and a good flatness (warp < 10 μm and bow < 58 μm). The cavities were patterned using a 2-μm-thick photoresist and then etched down to 350-nm depth in 10:1 buffered oxide etchant (BOE) in two cycles for a total time of 15 minutes. The wafer was baked at 125°C between the wet etching cycles to prevent photoresist from peeling off. The BOE etching process has a lateral-to-vertical etch rate ratio of 10:1. Therefore a 3.5-μm undercut can be achieved during the cavity formation, which is beneficial for the later ITO lift-off step. The wafer was then transferred into a rf sputtering system without removing the photoresist. An ITO film with a thickness of 170 nm was sputtered over the wafer in ambient temperature and then lifted off in a heated N-Methyl-2-pyrrolidone (NMP@70°C) solution. Then the wafer was annealed at 450°C for 5 minutes in a rapid thermal annealing (RTA) system to improve the ITO conductivity and transparency (Fig. 2a) [26]. The SOI device layer was a 2±0.5-μm-thick, n-type single-crystal silicon with 0.001–0.005 Ω·cm resistivity. The SOI wafer and the processed glass surface were bonded together by anodic bonding at 450°C under 2.5-kN force and 700-V bias voltage in vacuum. The top plate was formed after the handle wafer and BOX layer removal (Fig. 2b).

Fig. 2.

Simplified fabrication process flow: (a) Annealed ITO bottom electrode in glass cavities; (b) Anodic bonding and handle/BOX layer removal; (c) Silicon etch; (d) PECVD silicon nitride deposition; (e) Silicon nitride etch, evaporation and liftoff of metal pads.

The silicon plate was etched at the bottom pad location to evacuate the gas generated during anodic bonding (Fig. 2c). The device was sealed using a 1-μm conformal PECVD silicon nitride (Fig. 2d). The sealing nitride was plasma-etched to reach the conductive silicon top plate and the bottom electrode to form electrical contacts. The device fabrication was completed after evaporating and lifting off a stacked layer of 20-nm chromium and 180-nm gold as the bond pads (Fig. 2e). The physical dimensions of the fabricated device are summarized in Table I.

TABLE I.

Physical parameters of the fabricated CMUT

Shape of the cell	Circular
Number of cells per element	483
Cell diameter, μm	82
Cell-to-cell distance, μm	4
Element radius, mm	1
Silicon nitride layer thickness, μm	1
Silicon layer thickness in plate, μm	1.5
Vacuum gap height, μm	0.18
Bottom ITO thickness, μm	0.17
Glass substrate thickness, μm	500

III. Device Characterization

A. Electrical Input Impedance in Air

The real and imaginary parts of the electrical input impedance of the fabricated CMUT was measured in air using a network analyzer (Model E5061B, Agilent Technologies, Inc., Santa Clara, CA) (Fig. 3). The open circuit resonance frequency of the CMUT element was measured as 3.62 MHz at 18-V dc voltage, which is approximately 75% of the pull-in voltage. The 1-kΩ baseline in the real part corresponds to the series resistance of the device, which is mainly contributed by the resistance of the patterned ITO bottom electrode. This resistance could be reduced by depositing a thicker ITO layer or using parallel connections to the pads. For the PAI experiment described in Section IV, we wire bonded two connections to two pads reaching the ITO bottom electrode to reduce the series resistance.

Fig. 3.

Electrical input impedance measurements (V_dc = 18 V) (a) Real part; (b) Imaginary part.

B. Optical Transmission

Fig. 4a shows the optical image of the fabricated CMUTs. On the left is a CMUT die with six CMUT elements placed on a “NC STATE UNIVERSITY” logo under a microscope with backside illumination. It can be seen that the device has a good transparency in visible light wavelength range. The metal bond pads indicate the location of each element. The right side of Fig. 4a is a close-up image of a single CMUT element. Optical transmission was measured using a spectrophotometer (Cary 60 UV-Vis, Agilent Technologies, Santa Clara, CA) in the wavelength range from 400 nm to 1000 nm. The light source was focused through a converging lens to achieve a focal area of a 150 μm × 150 μm square. On the CMUT die, we measured the region where the active CMUT cells are (R1) and also where there are no CMUT cells (R2). The measured results indicate the 1.5-μm silicon plate is the main hurdle for transmission (Fig. 4b) while the ITO bottom electrode has little effect on the optical transmission [26]. This result also justifies illumination through the whole die in the photoacoustic imaging experiment.

Fig. 4.

(a) Optical picture of a CMUT die on a “NC STATE UNIVERSITY” logo (left) and optical picture of a CMUT element (right). (b) Transmittance characteristics of the CMUT die.

C. Pulse-echo Measurement

Although the presented transducer is designed primarily as a receiver for photoacoustic imaging, a pulse-echo measurement was performed in vegetable oil to characterize the small-signal bandwidth in immersion and also to help quantify the effects of optical absorption in the silicon plate on the generation of spurious transmit signals, which will be discussed in Section V. The CMUT was placed 1.2 cm away from a plane reflector and was biased at 18-V dc voltage (75% V_pull–in). A 1-V_pp, 250-ns pulse was used to excite the device. The received echo signal and its Fourier transform are shown in Fig. 5. The center frequency of the CMUT is 1.4 MHz with a 6-dB fractional bandwidth of 105%.

Fig. 5.

Pulse-echo measurement: (a) echo signal; (b) Fourier transform of the echo signal.

IV. Experimental Photoacoustic Imaging

A. Experimental Setup for Photoacoustic Imaging

The experimental setup for backward-mode PAI is shown in Fig. 6. The CMUT die was mounted and wire bonded on a printed circuit board (PCB) that has a rectangular cutout in the center to allow light to pass through. A bias-T circuit was designed to separate dc and ac signals and switches were used to select individual CMUT elements for testing. The excitation laser source was a tunable optical parametric oscillator (OPO) pumped by a Q-switched Nd:YAG laser (model Phocus Mobile, Opotek Inc., Carlsbad, CA) with a wavelength range from 690 nm to 950 nm. The laser pulses had a pulse-width of 4.5 ns and a repetition rate of 20 Hz. The output of the OPO was coupled using a fiber bundle (1:1 circular input/output, NA 0.37 ± 0.02, 5-mm diameter, 2-m length), which can generate a homogeneous output since the fibers are sufficiently long and arbitrarily arranged. The output of the fiber bundle was fixed to the cutout using a 3D-printed holder, so that the relative position of the CMUT and the fiber bundle output did not change during the experiment. The holder was mounted on a 3-axis linear stage (model PRO165, Aerotech Inc., Pittsburgh, PA, USA) to enable mechanical scanning. The target and the coupling medium were placed in a glass container under the PCB. A dc power supply (model E3631A, Agilent Technologies, Santa Clara, CA) was used to provide dc bias voltage. The received PA signal was filtered and amplified by a receiver (model 5072PR, Olympus Scientific Solutions Americas Inc., Waltham, MA, USA) and then recorded by a PC-controlled digitizer (model NI PCI-5124, National Instruments, Austin, TX). Fig. 6a shows the connection of each component. The side view of the fiber bundle, PCB, and the holder is shown in Fig. 6b. The bottom view of the PCB with the CMUT chip is shown in Fig. 6c, with the inset graph indicating the relative location of the laser beam and the CMUT elements. We chose to use the #2 CMUT element on the die because the light illuminated through this entire device. For each experiment, laser energy per pulse was measured using a pyroelectric energy detector (model: QE25, Gentec Inc., Quebec City, Canada) both at the fiber output and through the CMUT. The irradiation fluence at the fiber bundle output was calculated using the fiber bundle aperture area. Then the fluence through the CMUT can be estimated using the spot size of the beam on the CMUT front surface computed using the NA of the fiber, refraction through glass, and the propagation distance.

Fig. 6.

(a) Diagram of the connections of experimental setup (inset graph shows the CMUT mounted on the PCB with a cutout to allow illumination). (b) Side view of optical fiber bundle fixed at the back of the PCB using the 3D-printed holder. (c) Bottom view of the PCB with the CMUT mounted and wire bonded on the front. Inset graph indicates the relative locations of the fiber bundle output and CMUT elements.

The first imaging target was a 0.7-mm diameter pencil lead made of graphite. The pencil lead was suspended 2 cm above the bottom surface of the glass container as shown in Fig. 6b. Vegetable oil was used as the medium instead of water as the transducer surface and bond wires were not electrically insulated. We filled the vegetable oil up to approximately 1.5 cm above the pencil lead. Then we lowered the holder until the CMUT surface touched the oil. The wavelength was chosen as 830 nm according to the target absorption spectrum. The irradiation fluence from the fiber bundle output was 12 mJ/cm² (6.2 mJ/cm² through the CMUT due to beam spreading and absorption in silicon). The CMUT was biased at 18-V dc voltage. We set the receiver gain at 20 dB and the cutoff frequency of the lowpass filter at 10 MHz. The transducer was scanned across the pencil lead and the received signal at each location was sampled at a rate of 200 MSa/s, digitized, and averaged 16 times to improve the SNR before recording.

The second target was designed to better mimic the condition of biological tissues. We looped a polyethylene tube with an inner diameter of 2.3 mm and an outer diameter of 3.6 mm and filled it with an indocyanine green (ICG) solution (50-μM), which is commonly used as a contrast enhancement agent in PAI. The tube was then suspended using fishing lines inside the glass container. Then we filled up the container with a mixed solution of 1% Agar (Reagent grade, Carolina Biological Supply Co., Burlington, NC) and 1% Intralipid (20% emulsion, Sigma-Aldrich Co., St. Louis, MO) diluted in deionized water to build the photoacoustic imaging phantom. After the phantom solidified, we added a 5-mm oil layer on top of the solid phantom for acoustic coupling. The CMUT was again biased at 18-V dc voltage. This time the received signals were amplified with 40-dB gain. The wavelength was chosen as 790 nm and the irradiation fluence from the fiber bundle output was chosen as 20 mJ/cm² (10.3 mJ/cm² through the CMUT due to beam spreading and absorption in silicon). By mechanically scanning in x and y directions, volumetric data was recorded at a sampling rate of 200 MSa/s and by averaging each scan line 16 times.

B. Image Reconstruction

Photoacoustic images were reconstructed using the standard delay-and-sum (DAS) beamforming [28] along with a coherence factor (CF) weighting [29]. Prior to image reconstruction, every A-scan S(t) was processed as in [30]:

$$S_{\mathit{processed}}\left(t\right)=S\left(t\right)-t\frac{\partial S\left(t\right)}{\partial t},$$ (1)

which can be written in its discrete form as

$$S_{\mathit{processed}}\left(i\right)=S\left(i\right)-i\left(S\left(i\right)-S\left(i-1\right)\right),$$ (2)

This preprocessing suppresses the low-frequency component in the signal. Then, the A-scans were filtered by a 0.15-MHz – 4.5-MHz bandpass filter to eliminate out-of-band noise. After that, DAS receive-only beamforming, which can be expressed as

$$I\left(\overrightarrow{r}\right)={\displaystyle \sum _{j=1}^N{S}_{\mathit{processed},j}}\left(|\overrightarrow{r}-{\overrightarrow{r}}_j|/c\right),$$ (3)

was applied to form the PA image. c is the speed of sound, I $\left(\overrightarrow{r}\right)$ is the reconstructed pixel value at position $\overrightarrow{r}$ before envelope detection, S_{processed, j}(t) is the processed A-scan from element j at time t, ${\overrightarrow{r}}_j$ is the position of element j, and N is the number of elements in the synthesized 2D array. Since the size of the CMUT element was 2λ (λ is the acoustic wavelength in immersion), a threshold value of 14° was chosen according to the radiation pattern to maximize the image SNR, and the contribution from an element was not taken into consideration if the angle from its normal to the pixel location was larger than the threshold. Finally, envelope detection was performed and the image was multiplied by the coherence factor map. Logarithmic compression was performed before displaying the PA image.

V. Photoacoustic Imaging Results

A. Pencil Lead in Oil

The experimental results of imaging of the pencil lead in oil are shown in Fig. 7. A sample A-scan at X=0 mm is shown in Fig. 7a. Four signals (S1, S2, S3, S4) are marked on the A-scan with the signal paths shown in Fig. 7b. As the 4.5-ns wide laser pulse shines through the CMUT, some of the optical energy is absorbed in the silicon plate and converted to heat. Thermoelastic expansion of the silicon plate caused by rapid heating and cooling will set the plate into vibration at its natural frequency in oil, which results in S1 on the A-scan. S2 is the received PA signal generated by the pencil lead. The tail signal after S2 is because of the substrate ringing of the device and the reverberations in the pencil lead. S3 is the pulse-echo signal transmitted due to S1 and reflected by the pencil lead, and therefore occurred at double the distance compared to S2. S4 is the PA signal generated by the pencil lead reflected by the silicon plate and then reflected back by the pencil lead. Therefore it appeared at three times the target depth.

Fig. 7.

Pencil lead cross-sectional PAI results: (a) A sample A-scan at X=0 mm and its Fourier transform after applying a Gaussian window on S2; (b) Signal paths of the four signals on the A-scan; (c) Reconstructed crosssectional image.

The B-scan image of the cross section of the pencil lead was reconstructed according to the algorithm described in Section IV(B), as shown in Fig. 7c with 40-dB dynamic range. The pencil lead was seen at the depth of approximately 12 mm. Substrate ringing and the reverberations in pencil lead can be observed after the main signal. At the distance of 24 mm, a weaker signal (34 dB lower than the pencil lead signal) was detected, which is due to the pulse-echo signal generated by the silicon plate absorption and corresponds to S3 on the A-scan.

In order to further evaluate the effect of the silicon plate absorption, we designed the following two experiments. In the first experiment we compared the photoacoustic signals generated by the light absorption in the pencil lead (referred to as PA_target in the following context) and the pulse-echo signals generated by photoacoustically induced vibration of the silicon plate (referred to as PA_cmut in the following context) for the same travel distance in oil across the 690 nm to 950 nm wavelength range. PA_target was first recorded as described for the imaging experiments. To record PA_cmut also at 12-mm travel distance, we placed the CMUT 6-mm away from the glass container bottom, which served as a plane reflector without generating interfering PA signals. The laser wavelength was scanned from 690 nm to 950 nm with a step of 10 nm. The received signals were normalized to 1-mJ/cm² fluence from the fiber bundle output. The results are plotted in (Fig. 8a) with curve fitting. The oscillatory behavior in the curve is due to the fluctuations of the laser power coming through the CMUT caused by optical interference at the thin film interfaces, which corresponds to the optical transmittance shown in Fig. 4b. It can be seen that for the same travel distance, PA_cmut is much smaller than PA_target (approximately 30 dB lower at wavelength of 830 nm).

Fig. 8.

(a) Received PA signal amplitude normalized to 1-mJ/cm² fluence from fiber bundle output with 12-mm total travel distance: one-way target signal (PA_target) and two-way silicon plate signal (PA_cmut). (b) Equivalent electrical excitation amplitude normalized to 1-mJ/cm² fluence from fiber bundle output.

In the second experiment, we compared PA_cmut to the pulseecho signal generated by the electrical excitation (PE_cmut) for the same travel distance. The aim of this experiment was to find an equivalent excitation voltage for the transducer that would generate a PE_cmut with an amplitude equal to that of the PA_cmut. We used a 250-ns, 1-V unipolar pulse to perform a regular pulse-echo measurement. The received echo signal amplitude was 1.5 mV_pp. Thus, the equivalent electrical excitation signal amplitude can be calculated for 1-mJ/cm² fluence from fiber bundle output (Fig. 8b). As an example at the wavelength of 830 nm, the pulse-echo signal generated by photoacoustically induced vibration of the silicon plate using 1-mJ/cm² fluence from the fiber bundle output is equivalent to that generated by the CMUT using 0.29-V electrical excitation.

One should also note that pencil lead is both a strong absorber and a strong reflector. Therefore the spurious transmit signal generated by photoacoustically induced vibration of the silicon plate (S2) was strongly reflected back and received by the CMUT. In PAI applications, target is usually a strong absorber but not a strong reflector, such as human soft tissue or blood. In such a case, this signal will not be reflected and detected by the receiver. However, transmitted acoustic signal can still cause clutter issues in a photoacoustic image, especially as SNR degrades for tissue at depth. Therefore, further improvement of the device transparency is desired.

B. Loop-Shape ICG Tube Phantom

The photograph of the phantom is shown in Fig. 9a, where a looped polyethylene tube filled with the ICG solution was embedded in the tissue mimicking material and suspended using fishing lines. 3D volumetric data was collected by scanning the CMUT with fiber bundle along x and y directions and then the 3D image reconstruction was performed according to the algorithm described in Section IV(B). The 3D image was rendered using a medical image viewing software (Osirix, Pixmeo SARL, Bernex, Switzerland) [31] and displayed by using maximum intensity projection (MIP) (Fig. 9b). The tube filled with ICG solution and the fishing line knots could be seen in the rendered 3D image.

Fig. 9.

ICG tube phantom (a) Photograph of the ICG-filled polyethylene tube. (b) 3D rendered image of the ICG-filled polyethylene tube.

VI. Discussion

We have presented a CMUT element that resonates at 3.62 MHz in air and operates at a center frequency of 1.4 MHz, with a 105% fractional bandwidth in immersion due to the mechanical loading of the medium. The device demonstrates 30% to 70% transmittance in the 700-nm to 900-nm wavelength range. We performed backward-mode PAI with a fiber bundle output fluence up to 20 mJ/cm². The imaging artifact introduced by the absorption in the silicon plate was not significant, and no damage or degradation in the performance of the transducer was observed. For nanosecond laser pulses at 1064-nm wavelength, the damage threshold for silicon and glass have been reported in the >5 J/cm² fluence range [32], [33], which is far greater than the ANSI limit of irradiance of 20 mJ/cm², commonly followed in medical photoacoustic imaging. The presented CMUT structure is particularly suitable for backward-mode PAI and the following two aspects can be further improved.

Regarding the optical aspect, it is desirable to fabricate CMUTs with greater optical transparency to deliver high laser power to the tissue and minimize the device absorption to reduce imaging artifacts. CMUT technology enables wide selection of processing materials. The transparency of the device could be further improved by replacing the silicon plate with a more transparent material, such as silicon nitride, ITO, or glass. Besides, an optical lens can be incorporated to uniformly distribute the input radiant energy on the backside of the transducer.

Regarding the imaging aspect, arrays can be implemented instead of a single transducer that is mechanically scanned, so that faster data acquisition can be achieved to enable realtime imaging. The presented fabrication process can be used to fabricate 1D and 2D CMUT arrays [34]–[36]. High-frequency CMUT arrays operating up to 60 MHz have already been implemented for applications that require high resolution [37]–[40]. It should also be noted that ultrasound propagates one way in PAI. Thus, overall attenuation is lower compared to pulse-echo imaging for the same imaging depth. Furthermore, the maximum imaging depth and SNR can benefit from closely integrated electronics, which is especially important for 2D arrays and high-frequency arrays. Photoacoustic imaging as deep as 5 cm has been demonstrated inside a tissue-mimicking phantom using a 2D CMUT array with closely integrated electonics that greatly helps mitigate effects of parasitics and improve noise performance [2].

VII. Conclusion

This paper presented a CMUT element with improved transparency for backward-mode photoacoustic imaging. The device has 30% to 70% transmittance in the wavelength range from 700 nm to 900 nm. The CMUT operates in immersion with a center frequency of 1.4 MHz and fractional bandwidth of 105%. Two backward-mode PAI experiments were performed by mechanically scanning the CMUT element with a fiber bundle attached on the backside. A graphite target was first imaged at a wavelength of 830 nm. The artifact introduced by light absorption in the silicon plate was 34 dB lower compared to the signal intensity of the target. Then a polyethylene tube filled with 50-μM ICG solution embedded in a tissue-mimicking material was imaged at 790-nm wavelength and a volumetric image was reconstructed.

We are currently working on designing and fabricating 1D and 2D CMUT arrays with ITO bottom electrodes as well as further improving the plate transparency to reduce the light absorption in the device. By doing that a faster data acquisition rate and a better image quality can be realized. In the long term, we will implement the through-illumination approach in intravascular/intracardiac imaging, endoscopy, laparoscopy, and other intracavital applications that require compact integration of optical and acoustic components for photoacoustic imaging.

Biographies

Xiao Zhang (S’13) received the B.S. degree from Xi’an Jiaotong University, Xi’an, China, in 2012, and the M.S. degree from North Carolina State University, Raleigh, NC, in 2015, both in electrical engineering. He is now pursuing a Ph.D. degree in the Department of Electrical and Computer Engineering at North Carolina State University, Raleigh, NC.

His current research focuses on design and fabrication of 1D and 2D capacitive micromachined ultrasonic transducer (CMUT) arrays and associated MEMS transmit/receive switches on glass substrates as well as their integration with frontend electronics for medical imaging and therapy applications. He also works on developing transparent CMUTs for medical and air-coupled applications.

Mr. Zhang received student travel awards in the 2015, 2016, and 2017 IEEE International Ultrasonics Symposia. He was a student paper competition finalist in the 2016 and 2017 IEEE International Ultrasonics Symposia. He won the first place in 2017 NCSU ECE research symposium and he was also a recipient of the UGSA conference award at NCSU in 2015. Mr. Zhang has authored more than 15 scientific publications.

Xun Wu (S’14) received his B.S. degree from Southeast University, Nanjing, China, in 2012, and his M.S. degree from North Carolina State University, Raleigh, NC, in 2016, both in electrical engineering. He is currently working toward his Ph.D. degree in electrical engineering at North Carolina State University, Raleigh, NC.

His research interests include portable and programmable medical imaging system design as well as ultrasound and photoacoustic image processing to monitor high-intensity focused ultrasound therapy.

Oluwafemi Joel Adelegan (S’16) received his B.S. degree from University of Lagos, Lagos, Nigeria, in 2008, and his M.S. degree from North Carolina Central University, Durham, NC, in 2014, both in Physics. He is currently pursuing a Ph.D. degree in Electrical Engineering at North Carolina State University, Raleigh, NC.

His current research focuses on design and fabrication of wideband and ultra-wideband high frequency 1D capacitive micromachined ultrasonic transducer (CMUT) arrays with their integration with frontend 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 the M.S. and Ph.D. degrees from Sabanci University, Istanbul, Turkey, in 2006 and 2011, respectively, all in electrical and electronics engineering. He received 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. 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, North Carolina 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. His research focuses on developing micromachined devices for biological and chemical sensing, ultrasound imaging, and therapy.

Ömer Oralkan (S’93-M’05-SM’10) received the B.S. degree from Bilkent University, Ankara, Turkey, in 1995, the M.S. degree from Clemson University, Clemson, SC, in 1997, and the 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 North Carolina 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 an Associate Editor for the IEEE Transactions on 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 160 scientific publications.