2.1.

Transducer Design and Simulation

The most critical component of designing an ultrasonic transducer is the selection of the piezoelectric material, which converts the acoustic energy to an electrical signal. Possessing good piezoelectric properties, $\mathrm{Pb}(\mathrm{Zr},\mathrm{Ti}){\mathrm{O}}_3$ is the most popular polycrystalline ferroelectric ceramic material for fabricating ultrasonic transducers.^9,10 Besides polycrystalline ceramics, single crystals, such as ${\mathrm{LiNbO}}_3$ (Refs. 11 and 12) and $\mathrm{Pb}({\mathrm{Mg}}_{1/3}{\mathrm{Nb}}_{2/3}){\mathrm{O}}_3-{\mathrm{PbTiO}}_3$ (PMN-PT),^13–15 are also used for ultrasonic transducers due to their outstanding dielectric and electromechanical properties. Different from the polycrystalline ceramics, single crystals do not suffer from grain size and porosity issues when fabricating very thin piezoelectric layers and thus are more desirable for high-frequency transducers without scaling limitations.¹⁶ Both ${\mathrm{LiNbO}}_3$ and PMN-PT have a satisfactory electromechanical coupling coefficient $k_t$ (e.g., ${\mathrm{LiNbO}}_3$ $\sim 0.39$, PMN-PT $\sim 0.58$), which is favorable for high-sensitivity transducers. However, PMN-PT exhibits a much higher clamped dielectric permittivity ${\epsilon }^s/{\epsilon }_0$ than ${\mathrm{LiNbO}}_3$ (e.g., ${\mathrm{LiNbO}}_3$ $\sim 39$, PMN-PT $\sim 800$), which is selected for designing smaller-aperture transducers. Requiring a small aperture and high sensitivity, currently PMN-PT should be the first choice for PAOM imaging system.

In general, the piezoelectric element of an ultrasonic transducer operates at its thickness mode and the resonant frequency is determined by the thickness of the piezoelectric layer given by¹⁷

The lowest resonance frequency is achieved when $n=1$, where $n$ is an odd integer, and $c_p$ is the acoustic velocity in the piezoelectric material. The PA signal is broadband and most PAM systems use transducers with frequencies varying from 20 to 60 MHz.¹⁸ In PAOM, a 20 MHz ultrasonic transducer provides sufficient axial resolution (75 μm theoretically) to differentiate the retinal blood vessels from the retinal pigment epithelium (RPE) layer.¹⁹ A further increase in frequency may sacrifice the detection sensitivity due to the higher attenuation at higher frequencies. We first designed three unfocused transducers with different center frequencies (20, 30, and 40 MHz) to demonstrate this image quality change. To match the standard $50\mathrm{\Omega }$ impedance of the receiver electronics, the aperture size and thickness of matching layers of the transducers were determined by using PiezoCAD (KLM model-based simulation software, Sonic Concepts, Woodinville, WA). The thickness of the polarized PMN-PT material for 20, 30, and 40 MHz transducer were also optimized by PiezoCAD and lapped to be 100, 65, and 40 μm, respectively. The general fabrication technique described by Cannata et al.¹² was modified and used in this study. A first matching layer was laid over to the negative electrode (front site) of PMN-PT material to facilitate the efficient acoustic energy transfer between the PMN-PT element with high acoustic impedance and the water medium. The matching layer had a thickness of $\lambda /4$ ($\lambda$ is the wavelength in the material of the matching layer) and was made of a mixture of three parts 2–3 μm silver particles (Adrich Chem. Co., Milwaukee, Wisconsin) and 1.25 parts Insulcast 501 epoxy (American Safety Technologies, Roseland, New Jersey). A conductive backing material (E-SOLDER 3022, Von Roll Isola Inc., New Haven, Connecticut) was cured separately over the back of the PMN-PT element to damp the energy radiation from the back site of the element and provide structural support. Before assembling the functional element into the steel needle housing, an electrical connector was fixed into the backing material. Finally, parylene was vapor deposited on the whole transducer as a second matching layer and protective coating. A picture and the 3-D cutting drawing of the transducer are shown in Figs. 1(a) and 1(b), respectively.

Fig. 1

(a) Actual image of the needle transducer. (b) Three-dimensional cutting drawing of the transducer design.

Normally, focusing increases the sensitivity around the focal zone of the high-frequency ultrasonic transducer in pulse-echo imaging. According to acoustic reciprocity, the transmitting pressure response at an imaging point from a transducer is identical to that received by the transducer from a point source. In the far-field, the beam profile of the focused ultrasonic transducer diverges and may provide a larger FOV. To study the focusing effect on PAOM imaging, we designed and fabricated a 30 MHz focused transducer to make a comparison study with the 30 MHz unfocused transducer. Both of the transducers have a square shape with an aperture size of 0.5 mm. To make the focused transducer, a spherical ball of 1 mm radius was used to mechanically press on its functional element to achieve an $f$ number (F#) of 2 with the manufacturing process detailed below. A mold was customized to fit the size of the steel needle housing and align concentrically the needle housing with the spherical ball. The mold and transducer were incubated at temperatures $>90°\mathrm{C}$ during the press-focusing process and then cooled down to room temperature before the focusing was achieved. Since the incubated temperature may exceed the depoling temperature (60 to 95°C) for binary PMN-PT crystal, the focused transducer was poled in a DC electric field of $30\text{kV}/\mathrm{cm}$ for 5 min at room temperature to induce the piezoelectric effect. The sound fields of the two transducers were simulated using Field II with the same design parameters to visualize the effect of focusing on the FOV.^20,21 The simulation data were then compared to the in vivo imaging results.

2.2.

Experimental System

Figure 2 shows a schematic of the experimental system. The illuminating light source was a frequency-doubled Q-switched Nd:YAG laser (SPOT-10-100-532, Elforlight Ltd, Daventry, UK: 532 nm; $10\mu \text{J}/\mathrm{pulse}$; 2 ns pulse duration; maximal pulse repetition rate: 30 kHz). The output laser light was first attenuated with a neutral density (ND) filter before being coupled into a single-mode optical fiber. The light reflected from the surface of the ND filter is detected by a photodiode to provide the trigger signal for PAOM data acquisition. The output laser light from the fiber was collimated into a beam diameter of 6 mm, combined with the probing light of a spectra-domain optical coherence tomography by a hot mirror (Edmund Inc., Barrington), reflected by a mirror, scanned by an X-Y galvanometer, and finally delivered to the eye through a lens system consisting of an achromatic relay lens ($f=75\mathrm{mm}$) and an ocular lens ($f=19\mathrm{mm}$). When imaging a sample other than the retina in vivo, e.g., black tape, the ocular lens was removed and the relay lens was replaced with an objective lens ($f=30\mathrm{mm}$). The galvanometer scanner was controlled by an analogue-output board, whose sample clock was used to trigger the laser.

Fig. 2

Schematic of laser scanning photoacoustic microscopy system. ND, neutral density filter; PD, photodiode; I, iris; SMF: single-mode fiber; HM, hot mirror; UT, ultrasonic transducer; SLD, superluminescent diode; PC, polarization controller; L1 to L4, lens; FP, fiberport; M, mirror.

The induced PA waves from the sample were detected by an ultrasonic transducer, which was placed in contact with the eyelid coupled by ultrasound gel. The detected PA signals were first amplified by 66 dB and then digitized by a high-speed digitizer (CompuScope14200, Gage Applied Technologies, Lachine, Canada) at a sampling rate of $200\mathrm{MS}/\mathrm{s}$. In our current study, the laser pulse energy was measured to be 40 nJ after the ocular lens, which is lower than the ANSI safety limit. The OCT was used mainly for imaging guidance to reduce visible light exposure to the eye. During alignment, only the OCT light was on. After the retinal region of interest was identified and the OCT image was optimized, the OCT light was turned off and the PAOM imaging mode was activated.

2.3.

Imaging Subjects

We characterized the frequency response of the transducers and their performances in imaging in vivo samples. The transducer’s frequency response was measured by pulse-echo test in a deionized water bath at room temperature. Similar to the in vivo imaging condition, an X-cut quartz was placed 6 mm away from the transducer as the reflecting target. The transducers were excited by a 70 Vpp broadband negative pulse from a pulser/receiver. The center frequency ($f_c$) and fractional bandwidth (BW) are determined by the following equations:

For in vivo experiments, the eyes of albino rats with normal retina were imaged (Sprague Dawley rats, body weight: 300 to 450 g, Charles Rivers). The animals were anesthetized by intraperitoneal injection of a cocktail containing Ketamine ($54\mathrm{mg}/\mathrm{kg}$ body weight) and Xylazine ($6\mathrm{mg}/\mathrm{kg}$ body weight). Before imaging, the pupils were dilated with 10% Phenylephrine solution. After anesthetization, the animals were restrained in a mounting tube, which was fixed on a five-axis platform. Raster scans with the fast axis along the horizontal direction were performed. The scan angle spanned $\sim 20\mathrm{deg}$ for imaging the rat retina. All experiments were performed in compliance with the guidelines of the University of Southern California’s Institutional Animal Care and Use Committee.

3.1.

Transducer Performance

The pulse-echo measurement results and frequency response of the three unfocused transducers are displayed in Figs. 3(a) to 3(c). For the 20 MHz transducer, the peak-to-peak amplitude of the echo signal was 648 mV without gain at a reflection distance of 6 mm and a time delay of 8 μs. The measured $f_c$ and $-6\mathrm{dB}$ BW were 17 MHz and 37%. The 30 MHz transducer had an echo signal amplitude of 564 mVpp, $f_c$ of 30.5 MHz, and $-6\mathrm{dB}$ BW of 53%, respectively. The 40 MHz transducer had an echo signal amplitude of 430 mVpp, $f_c$ of 39 MHz, and $-6\mathrm{dB}$ BW of 39%, respectively. The measured axial resolutions of the 20, 30, and 40 MHz transducers were 78, 46, and 37 μm, respectively. For these three unfocused transducers, the measured center frequencies were in good agreement with the PiezoCAD simulation, while the 20 MHz transducer had the highest sensitivity as expected.

Fig. 3

(a) Frequency response of 20 MHz transducer at 6 mm ($f_c$: 17 MHz, BW: 37%). (b) Frequency response of 30 MHz transducer at 6 mm ($f_c$: 30.5 MHz, BW: 53%). (c) Frequency response of 40 MHz transducer at 6 mm ($f_c$: 39 MHz, BW: 39%). (d) Frequency response of 30 MHz focused transducer ($\mathrm{F}\#=2$) at 6 mm ($f_c$: 30 MHz, BW: 60%).

The frequency response of the 30 MHz focused transducer is shown in Fig. 3(d). The measured echo signal amplitude, center frequency, and bandwidth were 199 mVpp, 30 MHz, and 60%, respectively. Compared with the 30 MHz unfocused transducer, the 30 MHz focused transducer exhibited a broader bandwidth and a better axial resolution of 37 μm. However, it had a relatively irregular spectrum, which can be attributed to the slight degradation of piezoelectric properties caused by the mechanical pressing. According to the previous study by Knapik et al.,²² the mechanical losses would increase and electromechanical coupling in the devices would decrease after the press-focusing process, but the reported sensitivity did not appear to be significantly affected at the focal point. Nevertheless, note that the pulse-echo measurement was conducted at a distance (6 mm) far beyond its focal point (1 mm), where the transducer’s beam had diverged significantly into the far-field. Therefore, in contrast to the unfocused transducer, the focused transducers’ loss of sensitivity in the far-field was mainly caused by the focusing effect.

3.2.

Effect of Transducer Center Frequency on PAOM Image

The maximum-amplitude projection (MAP) PAOM images of the same albino rat’s eye acquired with the three unfocused transducers are shown in Figs. 4(a) to 4(c) (all of the three images are displayed in the same dynamic range of 25 dB). The major retinal blood vessels were clearly displayed with high PA signals within the FOV of the three images. In the retinal blood vessels, the PA signals originated from the high optical absorption of hemoglobin in the red blood cells. Since there is no melanin in the RPE cells of an albino rat, the laser light penetrated through the RPE layer into the choroid without generating PA signals in the RPE layer. As a result, the densely packed capillary vessel network appeared in the PAOM images indicated by the red arrow. The 20 MHz transducer [Fig. 4(a)] provided the highest contrast for the retinal blood vessels, and only the 20 MHz transducer can capture the choroidal vessels in the whole FOV, while the choroid layer was only partially captured by the 30 MHz [Fig. 4(b)] and 40 MHz [Fig. 4(c)] transducers. None of the three transducers were able to resolve a single capillary in the choroid by using the divergent receiving field far away from the focal zone.

Fig. 4

Maximum-amplitude projection photoacoustic ophthalmoscopy (MAP-PAOM) images of albino rats captured by (a) 20 MHz, (b) 30 MHz, and (c) 40 MHz ultrasonic transducers. B-scan images of corresponding red dashed line in MAP images by (d) 20 MHz, (e) 30 MHz, and (f) 40 MHz ultrasonic transducers. Scale bar: 200 μm.

The CNR was calculated by using the following equation:^23,24

3.3.

Effect of Transducer Focusing on PAOM Image

The simulation results of both the unfocused and focused transducers are shown in Fig. 5. The unfocused transducer had a natural focal zone located at 1.5 mm, and the focused transducer ($\mathrm{F}\#=2$) had a focal length of 1 mm. The beam profile 6 mm away from the transducer surface, where the rat retina was placed during in vivo imaging, shows that the focused transducer has a wider $-6\mathrm{dB}$ beam width to form a larger FOV, but a reduced sensitivity. To quantitatively estimate the FOV of the transducer, we imaged a black tape target. The full width of half maximum of the MAP of the received PA signals was calculated to provide an approximation of the transducers’ FOV. The $-6\mathrm{dB}$ intensity MAP retinal images were also processed to quantitatively measure the FOV in actual retinal images by using the same evaluation criteria. Figures 6(a) and 6(e) show the rat retinal images captured by the 30 MHz unfocused and focused transducers, respectively. The FOVs of the two images (indicated by the yellow circle) were estimated to be 0.8 and 1.0 mm in diameter in the $-6\mathrm{dB}$ intensity images [Figs. 6(b) and 6(f)], demonstrating that focusing was capable of providing approximately a 20% increase of FOV, which also agreed with the results of imaging with the black tape target. The B-scan images of the green dashed line in Figs. 6(a) and 6(e) are shown in Figs. 6(c) and 6(g). The unfocused transducer could only resolve two retinal vessels due to its limited FOV, while the focused transducer could still clearly differentiate three retinal vessels even with a relatively high noise level at the edge of the FOV. A tiny side branch of retinal vessels indicated by the dashed circle was resolved by the focused transducer in the MAP image [Fig. 6(e)] and B-scan image [Fig. 6(h)]; however, it did not appear in both MAP [Fig. 6(a)] and B-scan images [Fig. 6(d)] captured by the unfocused transducer. The two sets of B-scan images also proved that focusing could effectively enlarge the FOV in PAOM imaging. Nevertheless, the detection sensitivity of the focused transducer obviously dropped to a CNR of 10.8 dB compared with that (CNR of 14.8 dB) of the image obtained by the unfocused transducer. Since both transducers have a center frequency of $\sim 30\mathrm{MHz}$, other than the FOV difference, the images provided similar anatomical information of the retina as was previously described. Consequently, the comparison study reasonably agrees with the Field II simulation results and shows that the unfocused and focused transducers exhibited their respective strengths in PAOM imaging.

Fig. 5

(a) Acoustic field of the 30 MHz unfocused transducer. (b) Acoustic field of the 30 MHz focused transducer. (c) Lateral pressure profile 6 mm away from the 30 MHz unfocused transducer. (d) Lateral pressure profile 6 mm away from the 30 MHz focused transducer.

Fig. 6

MAP-PAOM images of albino rat retina captured by 30 MHz (a) unfocused and (e) focused ($\mathrm{F}\#=2$) ultrasonic transducers. (b) and (f) $-6\mathrm{dB}$ intensity images of (a) and (e). (c) and (g) PAOM B-scan images marked by green dashed line and red dashed line in (a). (d) and (h) PAOM B-scan images marked by green dashed line and red dashed line in (e). Scale bar: 200 μm.