1. Introduction

The combination of pulsed illumination and acoustic detection enables optoacoustic (photoacoustic) imaging to visualize the optical absorption in biological tissues at depths exceeding the optical scattering mean free path [1,2]. At such depths, the illumination wavelength determines the contrast of the optoacoustic image, whereas the acoustic bandwidth, central frequency, and geometry determine the resolution. As the optoacoustic signals are relatively weak compared to medical ultrasound, the detector’s sensitivity is a crucial parameter that affects the system performance, often requiring bulky piezoelectric detectors for high-resolution deep tissue imaging or when high imaging speeds are needed.

The two most common configurations for optoacoustic imaging are tomography and microscopy. Optoacoustic tomography uses array transducers with central frequencies and bandwidths generally below 10 MHz, allowing imaging penetration of several centimeters with typical resolutions of hundreds of micrometers [3,4]. In contrast, optoacoustic microscopy uses single-element focused ultrasound detectors with central frequencies of 50 MHz and above, achieving resolutions below 50 µm. Because of the attenuation of high-frequency ultrasound in the tissue, optoacoustic imaging of internal organs with high resolution may only be performed as parts of intraoperative or minimally invasive procedures [5–7], requiring adaptation and often extreme miniaturization of the optoacoustic apparatus.

One of the most challenging medical applications for optoacoustic technology is intravascular photoacoustic (IVPA) imaging of the coronary arteries [8,9] due to the limited dimensions of the IVPA probes. IVPA catheters consist of an optical fiber used to deliver the optical pulses to the tissue and an intravascular ultrasound (IVUS) transducer used for optoacoustic signal detection. IVPA imaging can assess the composition of atherosclerotic plaques by analyzing images produced at several illumination wavelengths [8,9] and detect macrophages associated with plaque [10].

Recent works on IVPA have focused on lipid imaging due to the importance of lipid content in assessing atherosclerotic plaque vulnerability [6,11]. Although lipid imaging is also possible with near-infrared spectroscopy (NIRS), the resolution of NIRS is insufficient due to strong light scattering in the tissue [12]; moreover, NIRS can only produce two-dimensional (2D) images that lack depth information. In contrast, IVPA produces 3D images with resolutions comparable to those of IVUS, enabling the measurement of the volume of the atherosclerotic lipids [13–16].

The main challenge in IVPA imaging is the size-dependent sensitivity of conventional piezoelectric detectors used in IVUS catheters. While typical focused piezoelectric transducers used in optoacoustic microscopy have diameters of several millimeters and achieve noise-equivalent pressures (NEP) on the scale of 1 mPa/$\sqrt {\textrm{H}{\textrm{z}^{}}} $ [17], the typical NEP of IVUS, whose dimensions are below one millimeter, is 450 mPa/$\sqrt {\textrm{Hz}} $ [2,18]. This lack of sensitivity has limited the applicability of IVPA, requiring flushing the artery’s lumen with heavy water [19] to improve the signal-to-noise ratio (SNR) and enable the high imaging rates required in coronary interventions.

One promising approach for overcoming the sensitivity limitation of IVPA catheters is using resonator-based optical ultrasound detectors. While sub-millimeter optical resonators have achieved sensitivities of at least an order of magnitude higher than in IVUS [2], the lack of acoustic focusing limits their ability to achieve IVUS-like imaging performance [20,21]. Indeed, previous attempts at miniaturized optoacoustic probes using optical resonators have relied on optical focusing of the illumination source to achieve high lateral resolutions, limiting the imaging performance to superficial tissue layers where light diffusion is not dominant [21,22].

In this work, we developed a miniaturized optical ultrasound detector that combines high sensitivity–an order of magnitude higher than in IVUS­–with acoustic focusing, facilitating the development of all-optical IVPA catheters. The demonstrated detector uses an acoustic lens with a diameter of 0.8 mm bonded to a π-shifted waveguide Bragg grating [20,23] (π-WBG) fabricated using a silicon-photonics platform. In addition to high sensitivity, the detector demonstrated a lateral resolution of at worst 50 µm and a bandwidth of at least 84 MHz. In comparison, conventional piezoelectric IVUS catheters typically achieve a lateral resolution of 200 µm and a bandwidth of 20 MHz [24]. The detector’s capability for high-fidelity optoacoustic imaging is demonstrated in the raster-scan configuration on an optical phantom with complex geometry. Additionally, we demonstrate an acoustically absorbing layer composed of tungsten microparticles dispersed in a polymer matrix as an aperture stop; the layer removes image artifacts resulting from out-of-focus acoustic signals, improving the detector’s overall performance.

2. System overview

Figure 1 shows a schematic illustration of the focused detector and optical readout system. The acoustics lens is on top of the device, focusing the ultrasound wave on the sensing element at the bottom. The acoustic lens has a numerically optimized concave aspheric profile with a diameter of 0.8 mm and a depth of 0.1 mm, with a designed focal distance of 1 mm, corresponding to a numerical aperture of 0.4. The lens was fabricated on top of a smooth quartz glass substrate with a thickness of 0.87 mm and a diameter of 25.4 mm using laser micromachining. The area outside the lens diameter is covered with a 0.5 mm thick acoustically absorbing layer to block acoustic paths that bypass the lens. The absorbing layer consists of a composite material that we developed to provide sufficient acoustic attenuation and is composed of tungsten microparticles dispersed in a polydimethylsiloxane (PDMS) polymer matrix. While the PDMS bulk acts as an acoustically absorbing medium, the tungsten microparticles act as acoustic scatterers. The scattering significantly improves the absorption efficiency of the PDMS by increasing the travel distance of the acoustic waves inside the material, increasing the material’s effective thickness.

 

Fig. 1. Schematic illustration of the proposed device, the optical readout system, and the experimental setup. (a) Cross-section of the device. The π-WBG rests between a layer of silica (light blue) and a layer of BCB (green) and is coupled to two PM fibers (purple). The acoustic lens is on top of the glass slab (blue) and is covered with an acoustically absorbing layer (black and gold). (b) Schematic illustration of the experimental setup and the phase-monitoring optical readout scheme.

Download Full Size | PDF

The sensing element, a silicon π-WBG resonator, is embedded between two cladding layers – a silica under-cladding and benzocyclobutene (BCB) polymer over-cladding. The BCB layer serves to both enhance the optical response of the resonator to acoustic pressure [25] and bond the resonator to the glass substrate. Two polarization-maintaining (PM) optical fibers couple the resonator to the readout system via vertical grating couplers [26]. The π-WBG resonator was fabricated in a silicon strip-waveguide using silicon-on-insulator (SOI) technology [20]; the 270 µm long π-WBG had a Q-factor of 6 × 10⁴. The resonator’s effective sensing length was approximately 30 µm, i.e., considerably shorter than the physical length of the π-WBG, due to light localization around the phase shift [20]. The π-WBG resonator was fabricated at IMEC (Leuven, Belgium) via the Europractice Multi-Project Wafer program using Si-Photonics Passives technology; the technical details related to the resonator are available in the Supplementary Materials (Sec. 1).

Ultrasound detection is performed by monitoring the pressure-induced modulation in the resonance wavelength of the π-WBG. In our setup, a continuous wave (CW) laser is tuned to the peak transmission of the resonance, where its phase response is linear, and the phase at the output of the π-WBG was continuously monitored by an interferometric setup [27] (Fig. 1(b)), leading to a waveform that represents the resonance wavelength modulation. The advantage of phase monitoring is that it effectively eliminates laser phase noise via balanced detection and provides signal amplification by interfering with a strong reference. In this scheme, the interrogation signal is split equally into two channels, such that one channel passes through the Bragg resonator and is modulated by the ultrasound waves, while the second channel bypasses the resonator. Eventually, when the channels are combined, the second channel’s output acts as the strong reference signal, mitigating signal attenuation due to losses in the resonator and the fiber-to-chip grating couplers of the silicon photonics chip. The technical details of the interferometric system are described in the Supplementary Materials (Sec. 2).

3. Device fabrication

3.1 Acoustic lens

The acoustic lens was designed using Zemax Optic Studio (Supplementary Materials, Sec. 3) and fabricated using laser micromachining. Quartz glass windows with 25 mm diameter and a nominal thickness of 1 mm (Suprasil 3001, Heraeus, Germany) were used as the lens material. While one side of the substrate had an acoustic lens in the center, the other side had a 50 µm deep recess to accommodate the silicon detector (Fig. 2(a)). All glass machining tasks were done by laser ablation with a Ti:Sapphire femtosecond laser with a central wavelength of 800 nm, a pulse duration of <30 fs, and a pulse repetition rate of 1 kHz (Femtopower Compact Pro, Femtolasers Produktions GmbH, Austria). Debris generated in the process was constantly removed with a compressed air jet and an extraction system. The laser output was circularly polarized using a lambda quarter-wave plate to prevent laterally direction-dependent ablation results. For the acoustic lens machining, focusing was done with an achromatic lens with a 30 mm focal length resulting in a focus diameter of 7 µm measured at 1/e² of the maximum intensity. An achromatic lens with a 140 mm focal length and a focus diameter of 31 µm has been used for machining the recesses.

 

Fig. 2. The detector’s fabrication process. (a) Schematic description of the bonding and the substrate etching steps. (b) A waveguide array fabricated on top of an SOI die (left), an acoustic lens within a quartz substrate (center), and an etched waveguide array bonded to the acoustic lens (right). (c) Fiber bonding setup. The detector is placed under a microscope and between two rotating fiber holders, each connected to a 5-degree of freedom manipulator (x, y, z, pitch, yaw). (d) The assembled detector mounted on the scanning system (3D stage) inside a water tank.

Download Full Size | PDF

A z-stage was used for setting the vertical focus position, while lateral positioning was done by moving the glass sample with an x/y-stage system (M-511.DG z-stage, two crossed M-531.DG x/y-stages, C-843 controller, Physik Instrumente, Germany). The z-stage also held a six-axis tactile sensor (TP200B, Renishaw, United Kingdom). For fabricating the acoustic lenses, the sensor was equipped with a ruby ball stylus with a 300 µm diameter for depth measurements and edge detection. For the recess fabrication, a ruby ball stylus with a 3 mm diameter was used (A-5000-7800 and A-5000-4160, Renishaw, United Kingdom). The smaller ruby ball had a curvature significantly above the highest lens curvature, allowing depth measurements inside the lens. The surface roughness (Ra) of all micromachined surfaces was less than 5 µm

The machining setup was computer-controlled, with complete 3D-path control, and was synchronized with the laser. The laser can be switched on and off on the fly with position accuracy, preventing variations in ablation depths at switch-on and switch-off points. The machined samples were cleaned in distilled water in an ultrasonic bath for 15 minutes. The remaining surface water was removed with filtered, dry compressed air.

3.2 Absorbing layer

A thin (∼0.5 mm) layer of acoustically absorptive, scattering, and reflective material was created by mixing tungsten powder (2-4 µm microparticles, Holland Moran Ltd., Israel) into PDMS (Sylgard 184, Dow Corning, Midland, MI, USA). The weight ratio of PDMS to tungsten was 1:0.75. First, the powder was mixed with the polymer base (part A), and only after that the curing agent (part B) was added; the A:B weight ratio was 10:1. The mixture was then spread on a glass slide and cured on a 110 °C hotplate for 20 minutes. After the curing was finished, the PDMS layer was peeled off the glass substrate and coated with a 300 nm layer of gold to protect the layer from light absorption. Finally, a 1 mm circular hole was punched through the PDMS layer, creating an aperture for the acoustic lens.

3.3 Integration

Figure 2 summarizes the main steps of the device fabrication and assembly process. First, the SOI chip was bonded to the backside of the glass slab with the acoustic lens on top of it. Next, the backside silicon substrate was removed using plasma etching (Figs. 2(a),(b)), and the acoustic lens was coated with a thin gold layer and the acoustically absorbing layer. Finally, two PM fibers were bonded to the device’s backside, and the assembled detector was mounted on the 3-axis mechanical scan system and connected to the optical readout setup (Figs. 2(c),(d)).

To bond the SOI die to the backside of the quartz substrate, a BCB polymer was used (Cyclotene 3022-57, Dow Corning, Midland, MI, USA). BCB was chosen for three reasons: 1. BCB has a refractive index similar to that of SiO₂ – 1.54 at the wavelength of 1536 nm [28], making BCB a suitable cladding material. 2. BCB is transparent in the 1500-1600 nm optical window. 3. BCB’s strong photo-elastic response enhances the sensor’s sensitivity and reduces the adverse effects of surface acoustic waves [25]. Moreover, BCB is thermally stable at temperatures above 300 °C [29], which is crucial for the plasma etching of the silicon substrate.

The SOI die was bonded to the lens by placing a small drop of BCB on top of the die, attaching the die to the alignment groove, and curing the BCB polymer (Fig. 2(a)); we did not use any adhesion promoters. The BCB curing process consisted of two stages: 1. the device was kept on a hotplate at 120 °C for 10 minutes. 2. The temperature was slowly ramped to 230 °C, and the device was kept on the hotplate for another 30 minutes. During the curing process, the device was in a nitrogen-filled chamber. To avoid significant non-uniformity in the layer’s thickness, we applied uniform pressure on the silicon chip during the adhesion process; however, the post-curing planarity of the resulting layer was not assessed.

After the bonding, the 700 µm silicon substrate was etched at the Technion’s Micro & Nano Fabrication Unit (MNFU) using the inductively coupled plasma (ICP) tool (Versaline, Plasma-Therm, St. Petersburg, FL, USA) with an etching recipe based on the repetitive loops of C₄F₈/Ar and SF₆/Ar gases. The 2 MHz ICP RF power was 1200 W (of 2500 W possible) to keep a reasonable etch rate on the one hand and to avoid overheating on the other hand. During the etch process, the device’s quartz substrate was attached with a thermal paste to a sapphire carrier (to eliminate the vacuum gap) that was kept at 10 °C. Since thermal management was critical in this process, the etch was performed in small portions with intermediate cooling.

The transparent lens was coated with a 300 nm layer of gold using electron-beam physical vapor deposition (EB PVD) at the MNFU (BAK-501A, Evatec AG, Trübbach, Switzerland) to prevent thermal interference from light absorption in silicon during our imaging experiments. Although we did not use any adhesion layers since the gold layer proved stable enough, adding a chromium or titanium adhesion layer for increased robustness is possible. We note that since silicon is transparent at wavelengths relevant for lipid imaging, i.e., 1200 nm and 1700 nm, the gold coating is unnecessary when working at these wavelengths.

The PM fibers were polished at an angle of 42° and bonded to the grating couplers with NOA61 optical adhesive (Thorlabs Inc, Newton, NJ, USA). The polish angle was optimized to minimize the coupling losses near the resonance wavelength of the π-WBG. The polished fiber facets were subsequently coated with a 200 nm layer of gold using EB PVD. The gold layer acted as a mirror preventing the light transmission through the facet when the device was submerged in water. We did not use adhesion layers, such as titanium or chromium, since they would absorb a part of optical energy and thus reduce the fiber-to-waveguide coupling efficiency. In the final assembly step, the acoustically absorbing layer was placed on top of the gold-coated quartz substrate and glued with a few drops of clear glue around the layer’s perimeter.

4. Results

4.1 Acoustic characterization

To characterize the acoustic performance of the focused detector, we used an optoacoustic point-like source generated at the tip of metal-coated multi-mode (MM) fiber (Fig. 3(a)); the fiber had a core diameter of 50 µm, and the coating thickness was 80 nm. The fiber was coupled to a 1 ns pulsed laser operating at the wavelength of 1064 nm (Waveguard-D, Optogama, Vilnius, Lithuania), thus generating high-frequency ultrasound signals due to the light absorption in the metallic coating [30–32]. The laser’s repetition rate was 1 kHz, and the output power was 120 µJ per pulse. The fiber coating consisted of a thin layer of chromium (bottom layer) and gold (top layer); both layers had a thickness of 40 nm and were done using EB PVD. The fiber was coupled to the laser via a simple focusing lens. The laser’s output power was reduced with an absorptive neutral-density filter (ND 2.0) to prevent damage to the metal coating. The signals were acquired using InfiniiVision DSOX4154A oscilloscope (Keysight Technologies, Santa Rosa, CA, USA) with a bandwidth of 1.5 GHz.

 

Fig. 3. Results of the acoustic lens characterization experiments. (a) The experimental setup. The optoacoustic source was created by coupling a pulsed laser to a MM fiber with a 50 µm core; the fiber tip was coated with a thin layer of gold and placed in the detector’s focal plane. (b) 1D peak-to-peak plot of the waveforms acquired during the axial scan, shown as a function of the distance between the lens and the acoustic source. (c) 2D peak-to-peak image of the acoustic signals acquired for the lateral resolution characterization. (d) 1D peak-to-peak plot of the signals obtained from a high-resolution line scan through the center of the optoacoustic source (shown in (c)).

Download Full Size | PDF

The focal length of the acoustic lens was determined experimentally by centering the metal-coated fiber in front of the detector and scanning the fiber axially over a span of 2 mm with a step size of 50 µm. Figure 3(b) shows the peak-to-peak amplitude of the acoustic waveforms acquired at each step. According to the results, the lens’s focal length is 1.8 mm relative to the detector’s surface. As shown in Fig. 3(b), the depth-of-field, approximated as the plot’s full width at half-maximum (FWHM), is 0.9 mm.

The spatial resolution characterization was performed by placing the fiber at the focal point of the acoustic lens and scanning the fiber in two dimensions with the detector to generate a 2D image of the fiber tip. The scan resolution was 10 µm, and each acquired signal was averaged 128 times. Figure 3(c) shows the 2D image of the scanned object obtained by taking peak-to-peak of the acoustic pressure waveforms acquired at each point. A one-dimensional (1D) slice of the 2D image is shown in Fig. 3(d), revealing the FWHM of 49 µm, in agreement with the diameter of the optoacoustic source. The data used in Fig. 3(c) was acquired with a separate one-dimensional (1D) scan with a resolution of 1 µm.

Figure 4(a) shows the acoustic waveform acquired from a 50 µm optoacoustic source, revealing an initial short bipolar signal accompanied by low-frequency signals, which may be attributed to reverberations inside the device’s multi-layer backside. Figure 4(b) shows the Fourier transform of the waveform, obtained by applying a narrow (green curve) and a wide (blue curve) window on the time-domain signal, shown in Fig. 4(a). According to the spectrum, the leading bipolar signal (narrow window) has a central frequency of 44.5 MHz and an FWHM bandwidth of 84 MHz. The FWHM bandwidth was calculated using the green curve shown in Fig. 4(b), with the bottom and top half-maximum frequencies of the leading bipolar signal being 11 MHz and 95.1 MHz, respectively. According to the spectrum obtained using the wide window, even though the reverberations had a lower magnitude in the time domain, they dominated the acoustic response at frequencies below 25 MHz. Figure 4(c) shows several of the acoustic waveforms acquired during the axial scan of the 50 µm fiber, and Fig. 4(d) shows the corresponding Fourier transforms of the waveforms. Figures 4(c) and 4(d) demonstrate the decrease of the signal’s amplitude and the distortion of the acoustic waveform as the optoacoustic source shifts on the acoustic axis. According to the results shown in Fig. 4(c), as the source moves out of the lens’s focal point, the amplitude of the primary bipolar signal decreases relative to the secondary, low-frequency signal. The waveform distortion corresponds to the loss of frequencies above 25 MHz due to defocus, as shown in Fig. 4(d). We note that frequencies above 75 MHz are relatively stable in the region between 1.60 and 2.15 mm, i.e., there is no severe high-frequency loss inside the 0.55 mm span around the focal point.

 

Fig. 4. Acoustic waveforms acquired from the 50 µm optoacoustic source and the corresponding Fourier transforms. (a) The acoustic pressure waveform obtained when the optoacoustic source was at the lens’s focal point. The two windows (blue and green) were used for the Fourier transform. (b) The Fourier transform of the waveforms shown in (a). The transform was performed with the narrow and the wide windows. (c) Acoustic waveforms obtained at different distances between the optoacoustic source and the detector. The distances are shown above each separate waveform at the top of the figure. (d) The Fourier transforms of the waveforms shown in (c).

Download Full Size | PDF

To assess the NEP of the detector, we measured the signal from the 50 µm optoacoustic source with a calibrated needle hydrophone (1 mm Needle Hydrophone, Precision Acoustics, Dorchester, UK) and converted the acoustic signal obtained by the optical detector from voltage to pressure. The hydrophone was submerged in water and placed at the same distance from the OA source as the acoustic lens, i.e., 1 mm. Since the hydrophone’s response was limited to acoustic frequencies below 20 MHz, with its central frequency around 10 MHz, the leading bi-polar signal from the optical sensor was low-passed at 20 MHz as part of the calibration process, revealing a NEP of 37 mPa/ Hz^1/2. Since an ideal optoacoustic point-source response depends linearly on the frequency, the measured NEP represents a good estimation up to the central frequency in Fig. 4(b), 44.5 MHz. At higher frequencies, we may attribute the drop in the signal to a possibly worse NEP, to a reduction in the signal due to the finite size of the source, and the limited bandwidth of the optoacoustic source.

To quantify the performance of the acoustically absorbing layer, we performed a 1D scan of the 50 µm fiber tip over a larger span of 4 mm, i.e., beyond the span of the lens aperture, for two implementations of the focused detector, with and without the absorbing layer. Figures 5(a)-(c) describe the case of the bare detector, i.e., where no absorbing layer was used; the figures show a schematic illustration of the fiber position relative to the detector (Fig. 5(a)), the corresponding measured sinogram (Fig. 5(b)), and the simulated sinogram obtained for the same configuration. Two features are clearly visible in both the measured and the simulated results: a narrow central lobe, which corresponds to sources covered by the lens aperture, and two side lobes, which correspond to sources outside the lens aperture, i.e., the stray acoustic waves that bypass the lens. Figures 5(d)-(f) show the results obtained with the covered detector: Fig. 5(d) schematically shows the effect of the absorbing layer and Figs. 5(e) and 5(f) show the measured and simulated sinograms, respectively. The measured and simulated results show that the covered detector produces a localized sinogram with no stray signals. To further demonstrate the effect of the absorbing layer, the peak-to-peak values of the sinograms shown in Figs. 5(b) and 5(e) were calculated over the time axis (Fig. 6), describing the magnitude of the measured acoustic signal as a function of its offset from the principal axis of the lens. As shown in the figure, the absorbing layer effectively removes the stray acoustic signals without affecting the focal width of the lens.

 

Fig. 5. Effect of the acoustically absorbing layer on the out-of-focus signals. (a, d) Illustrations of the measurement setup for the bare and covered detector. (b, e) Measured sinograms obtained with the bare and the covered detector. (c, f) Simulated sinograms obtained with the bare and the covered detector.

Download Full Size | PDF

 

Fig. 6. The peak-to-peak amplitude of the acoustic signals shown in Figs. 5(b) and 5(e), taken along the time axis, as a function of the offset from the lens’ principal axis.

Download Full Size | PDF

4.2 Imaging

The optoacoustic imaging capabilities of the developed detector were demonstrated by imaging optical phantoms arranged in different shapes. The phantoms were created by embedding black nylon monofilament surgical sutures with diameters of 20 µm (10-0 Dermalon, Cyanamid of Great Britain, Davis & Geck, Gosport, Hampshire, UK) and 40 µm (8-0 Nylon, Sharpoint, Surgical Specialties Corp., Reading, PA, USA) in agar mixture (2 grams of agar to 110 ml water). To facilitate homogeneous illumination, we added a small amount of cow’s milk (3% fat) to the agar mixture; the milk contains light scattering lipids and thus facilitates light diffusion in the agar phantoms, leading to more uniform light distribution. The sutures were embedded near the agar’s surface and illuminated by a pulsed OPO laser (SpitLight DPSS EVO I OPO, InnoLas Laser GmbH, Krailling, Germany) through the agar bulk. The laser’s wavelength range was 730 nm, the repetition rate was 100 Hz, and the output power was 35 mJ per pulse. The signal acquisition was performed using a PCI express digitizer with a bandwidth of 200 MHz (M4I.4480-X8, Spectrum Instrumentation GmbH, Grosshansdorf, Germany). The sutures were positioned in the focal plane of the detector and illuminated with 7 ns light pulses with a wavelength of 730 nm and pulse energy of 30 mJ. An in-plane scan was performed with the focused detector to acquire the images, recording the acoustic waveform at each position.

Figure 7 shows several imaged sutures with a diameter of 40 µm and their corresponding maximum-amplitude-projection (MAP) optoacoustic images for the case in which the detector did not have the acoustically absorbing layer. As the figure shows, when the target was simple and composed of only a single segment (Fig. 7(a)), the suture structure could be clearly recognized in the image due to the relatively low magnitude of the artifacts. However, as the complexity of the target increased (Figs. 7(b) and 7(c)), the artifacts became more pronounced, considerably degrading the image quality. The resulting image distortions may be understood from the response to a point source of the bare detector, i.e., the detector not covered by the absorbing layer (Fig. 5(a)); such response reveals that the stray signals, although weak, are spread out over broad regions. Thus, in the case of a complex object, stray signals originating from different positions can constructively interfere, leading to amplification of the artifacts; this effect is clearly seen in Fig. 7(c), where there is constructive interference at the center of the ring-shaped suture.

 

Fig. 7. Optoacoustic images obtained without the absorbing layer. Top row: optical images of surgical sutures embedded in agar. Bottom row: corresponding optoacoustic images obtained by the acoustic lens. The red boxes in the top row mark the areas used for optoacoustic images shown in the bottom row.

Download Full Size | PDF

Figure 8(a) shows the optoacoustic MAP image obtained from a 20 µm suture obtained with a focused detector covered with the acoustically absorbing layer, demonstrating artifact-free imaging despite the complexity of the imaged object. Figures 8(b)-(d) show several depth-resolved slices of the image reconstruction at distinct positions in which the sutures appear as points or lines. Here, by depth-resolved, we mean that the signals obtained from different depths can be distinguished by their time delays, such that this information can be used for 3D reconstruction. Both the MAP and depth-resolved slices did not exhibit any visible artifact due to stray waves, demonstrating the capability of the coated device for imaging complex structures.

 

Fig. 8. Optoacoustic image of a double-loop-shaped suture obtained with the covered lens: (a) the image obtained from the optical phantom. (b-d) sinograms corresponding to the acoustic signals obtained along the dashed lines.

Download Full Size | PDF

5. Discussion

In this work, we developed a miniaturized all-optical focused ultrasound detector for optoacoustic imaging. The detector is based on a silicon-photonics optical resonator and an acoustic micro-lens with a diameter of 0.8 mm, surrounded by a 0.5 mm thick layer of acoustically absorbing composite material. The combination of acoustic focusing and monolithic design in a sub-millimeter effective size makes our sensor potentially useful for optoacoustic and ultrasound intravascular imaging. Although acoustic focusing may be achieved with unfocused detectors by using the synthetic aperture focusing technique (SAFT), using SAFT may impose geometrical constraints that are incompatible with the stringent restrictions of intravascular imaging. For example, in Ref. [33], an additional rotating mirror was needed to perform the acoustic scan required for SAFT, limiting the level of miniaturization that may be achieved.

The acoustic lens and the waveguide cladding are optically transparent, making the thin silicon layer of the device the only optically absorbing element in the device. Nonetheless, for the optoacoustic excitation wavelengths used in this work, 730 nm and 1064 nm, the 220 nm thick silicon layer absorbed a non-negligible amount of the impinging light, requiring the use of the gold coating to reflect the background light back to the medium. For wavelengths typically used for lipid imaging⁶, i.e., 1200 and 1700 nm, all the device layers should be transparent, making the gold coating redundant. The potential transparency of the detector at longer wavelengths may further facilitate its integration in an optoacoustic probe as it allows collinear illumination directly through the lens. In contrast, conventional optoacoustic imaging catheters employ opaque piezoelectric transducers, imposing geometrical restrictions that increase the device’s size.

Acoustic characterization of the detector was performed using an optoacoustic point-like source created on the tip of an optical fiber. The detector successfully performed optoacoustic imaging of black surgical sutures embedded in an optically scattering medium. In both the characterization and imaging experiments, the addition of the acoustically blocking layer to the sensor proved essential for the proper performance of the device. In the characterization experiments, the absorbing layer reduced the out-of-focus signals below the noise level without affecting the signals in the focal region. In the imaging experiments, the images produced with the uncoated sensors were dominated by out-of-focus artifacts, whereas the coated sensors produced high-contrast artifact-free images. Since the out-of-focus signals result from acoustic waves that bypass the lens and travel directly through the substrate that surrounds it, it is expected that their effect would be less significant in a fully miniaturized device that is cut to the width of the lens.

In terms of performance, our sensor possesses several advantages over piezoelectric-based IVUS, which may facilitate new imaging applications. First, while IVUS transducers are relatively narrowband, our sensor achieves a bandwidth of 84 MHz, with the upper half-maximum frequency reaching up to 95 MHz, potentially enabling imaging at several acoustic bands or high-resolution applications. Second, the achieved focal width of the sensor is at most 50 µm, limited by the 50 µm diameter of the optoacoustic source used in the experiments; in comparison, the typical beam width of IVUS transducers is in the range of 100-300 µm [34]. Third, the NEP of the sensor is 37 mPa/Hz^1/2, an order of magnitude lower than in single-element piezoelectric transducers of comparable size [35].

The detector’s depth of field is approximately 0.9 mm. However, as the acoustic source shifts away from the focal point, the acoustic waveforms begin to lose higher frequencies. According to the results, frequencies above 75 MHz rapidly deteriorate when the optoacoustic source is closer than 1.60 mm or farther than 2.15 mm from the detector, i.e., outside the 0.55 mm window centered at the focal point. Since the decrease in high frequencies likely corresponds to the increased focal width, the detector exhibits depth-dependent resolution, possibly reducing the depth of field for smaller optoacoustic sources and limiting the penetration depth for high-resolution imaging. Further investigations are required in order to quantify the resolution deterioration due to defocus.

The improved performance of the acoustic sensor may enable new applications in minimally invasive imaging. In particular, the device’s high sensitivity may enable optoacoustic imaging of lipids using low excitation energies, reducing the overall cost, improving safety, and potentially increasing imaging rates. Specifically, by reducing the pulse energy, higher repetition rates may be used, thus enabling higher imaging rates without exceeding the tissue damage threshold, a major limitation of current IVPA setups. The addition of ultrasound-generation capabilities to the device may enable pulse-echo ultrasound, as commonly done in all-optical ultrasound devices. If the gold layer is replaced by a high optical absorption coating, illuminating the device from its backside will lead to the optoacoustic generation of a focused ultrasound beam, as previously demonstrated in the literature [31,32,36–38]. The optically absorbing layer may be designed to be opaque at some wavelength bands but transparent in others. For example, plasmonic nanostructures may enable the device to operate simultaneously as an all-optical ultrasound transducer and an optoacoustic imaging detector. In such a configuration, two pulse lasers with different wavelengths would illuminate the backside of the sensor, where one pulse would be transmitted through the lens to excite tissue chromophores and generate an optoacoustic signal, and the second pulse would be absorbed on the lens surface to produce the focused acoustic beam.

Future intravascular applications would require adjustments of the detector geometry to conform to the constraints imposed by the clinical procedure. For example, in our current implementation, the silicon-photonics chip was fiber-coupled from both ends of the waveguide, leading to a transmission-mode setup that would limit intravascular applications, in which the detector is accessible from only one side. This limitation may be directly addressed by using a curved waveguide, e.g., with a U-shaped geometry [39], to lead the output of the optical resonator back to the direction of its input, enabling bonding of both input and output fibers on the same side of the chip. Alternatively, the interrogation system may be modified to measure the ultrasound-induced modulation in the resonator’s reflection rather than transmission, enabling the use of only a single fiber for both input and output. Additional geometrical modifications may be required in the acoustic design, namely, further miniaturization of the lens, which may reduce the lateral resolution and increase the depth of field. Additionally, using a smaller lens could also increase the width of the acoustic beam focused on the resonator, thus leading to weaker signals and worse NEP. In such a case, an improvement in the NEP may be required, which may be achieved by replacing the BCB over-cladding layer with other polymers with a higher elasto-optic response, such as PDMS [23].