Bendable graded index lens-based laser-scanning optical-resolution photoacoustic microendoscopy

Abstract

Minimally invasive, three-dimensional (3D), flexible in-situ optical microendoscopy with cellular resolution has the potential to elucidate key biological mechanisms inaccessible with conventional approaches. Developing such microendoscopy is, however, still technically challenging. To this end, we recently proposed a bendable graded index (bGRIN) lens two-photon microendoscopy, with imaging quality resistant to lens bending. In this Letter, we improve on this bGRIN lens microendoscopy with another complementary modality—photoacoustic imaging. This distal scan-free bGRIN lens photoacoustic microendoscopy (bGRIN-PAM) has optical resolutions of ~2 μm (lateral) and ~60 μm (axial), with a lateral field of view of 100.6 μm and a maximum penetration depth of 600 μm. We have demonstrated microendoscopic imaging of mouse hairs using our bGRIN-PAM.

Imaging of tissue morphology and cellular activity in vivo with cellular resolution is of transformative importance for many biomedical applications, including disease diagnosis, drug screening, and precision targeting of surgical and interventional tools. For example, a high-throughput implantable microdevice was invented to simultaneously screen multiple anti-cancer drugs in situ [1]. While the microdevice has demonstrated great promise in clinical trials [2], it relies on surgical removal of the microdevice and the surrounding drug-exposed tissue for off-site histopathological imaging and analysis. Minimally invasive microendoscopy that may monitor the in-situ drug effects would significantly augment its clinical impact. Another example is the image-guided placement of interventional needles and catheters [3], where cellular-resolution microendoscopic imaging for robust tissue phenotyping would considerably improve targeting and enable new procedure capabilities with microscopic precision.

Optical microendoscopy is among the most promising technologies to enable the above-described capabilities, but its implementation remains technically challenging. One major obstacle arises from the miniaturization of flexible microendoscopic imaging probes. Typically, an optical fiber is used as the flexible lead-in optical waveguide for a distal scan assembly. This standard endoscopy design has been used in a variety of imaging modalities [4-6]. However, the distal scan assembly is bulky and thus impedes their further miniaturization. Over the last decade, a single multimode optical fiber-based distal scan-free microendoscopy has attracted growing interest [7-10]. This technology requires pre-calibration from the distal end, which degrades sharply as the fiber deforms and represents a bottleneck not overcome yet despite ongoing efforts [11-13]. Our group recently proposed and demonstrated the first bendable graded index (bGRIN) lens two-photon microendoscopy [14]. The bGRIN lens, fabricated using glass and a non-toxic silver ion exchange process, features a parabolic index profile which is superior to a step index profile for deformation-resistant imaging [15]. Its mechanical flexibility while maintaining the imaging quality is a sought-after innovation unavailable in all existing rigid GRIN endoscopy [16-18].

Here, we equip our side-view bGRIN lens with an additional optical modality—photoacoustic imaging, which offers complementary (absorption) information to previously demonstrated two-photon fluorescence imaging. Our bGRIN lens photoacoustic microendoscopy (bGRIN-PAM), which uses laser scanning at the proximal end (Fig. 1), is capable of 3D optical-resolution photoacoustic imaging. This novel bGRIN-PAM represents a significant advance to existing PAM, which typically applies distal scan and a conventional short and rigid GRIN lens to collimate or moderately focus the excitation laser [19-21].

Fig. 1.

Conceptual illustration of the proposed flexible, distal scan-free, optical-resolution 3D photoacoustic microendoscopy using a bGRIN lens.

Our bGRIN-PAM is demonstrated using the experimental setup schematically shown in Fig. 2. An externally triggered 1064nm pulsed laser (Cobolt Tor^™ XE, HÜBNER Photonics) fires ~2.2 ns pulses (maximum rate: 1 KHz). This wavelength is close to the design wavelength (1040 nm) of the bGRIN lenses, which are detailed in our previous work [14]. A tunable attenuator (VA5-PBS253, Thorlabs) is applied to regulate the laser power sent to the sample, and the remaining laser is disposed of by a beam dump. Lateral scans (x and y directions, inset of Fig. 2) of the laser beam are implemented by two orthogonal galvo mirrors (GVS002, Thorlabs), which is followed by beam expansion through two tube lenses. The expanded laser beam goes through a tunable lens (EL-10-30-TC-VIS-12D, Optotune) paired with an offset lens (LC4232, Thorlabs), which modifies the divergence of the laser beam for axial scan (z direction). A 20× objective (UPLFLN, Olympus) is used to focus the laser before it is coupled into the bGRIN lens. Photoacoustic signals are collected by a 6.35-mm-diamater focused electronic ultrasound sensor (IAG062, ndtXducer), which has a focal length of 12.7 mm, a peak responsivity at 6.25 MHz, and –6 dB bandwidth of 2.63 MHz. The output of the ultrasound sensor is amplified by a low-noise preamplifier (TOFD, Phoenix), which has a fixed 40 dB gain and a passband of 80 KHz–40 MHz. The amplified photoacoustic signal is sent to a data acquisition device (US-KEY, LeCoeur Electronique), which has an internal gain set to 30 dB. Synchronization of the laser pulses and data acquisition is completed using an I/O device (PCIe-6351 and BNC 2110, National Instruments), which also outputs the scan signals for the galvo mirrors. Injection current of the tunable lens is controlled through another port of the computer. It takes about 1 minute to acquire an image with 50 × 50 pixels (one ultrasound waveform for each pixel), which is currently limited by the speed of the ultrasound data acquisition device.

Fig. 2.

Experimental setup. Inset: coordinate definition.

We first characterized the field of view (FOV) and resolution of the bGRIN-PAM. A positive variable line grating (R1L3S6P, Thorlabs) was used as the testing target (Fig. 3(a)). Using the 10-μm-period subgrating, the lateral FOV was calculated to be around 100.6 μm (Fig. 3(b)). The axial FOV (i.e., maximum penetration depth, z direction) was measured by manually translating the target, which was determined to be around 600 μm. The lateral resolution was measured by the full width at half maximum (FWHM) of the line spread function (LSF) of the acquired edge spread function (ESF) of a sharp grating edge. To compute the LSF, a five-point moving average of the ESF was utilized. The computed lateral resolution was 2.0 μm and 2.1 μm for the x and y direction, respectively. The axial resolution of 59.2 μm, measured around half of the maximum penetration depth, was calculated from the signal amplitude vs. target translation curve (Fig. 3(d)). The much lower axial resolution might be due to the low effective NA and inherent aberration of our long bGRIN lens as well as the lack of a “confocal mechanism” to suppress out-of-focus signals. Note that this optical axial resolution is independent of the ultrasound bandwidth, which is related to the acoustic axial resolution defined in existing photoacoustic microscopy [22-24]. We also want to mention that a lower axial resolution can be desirable when imaging with an extended depth of field is needed [25].

Fig. 3.

Characterization of FOV and resolution. (a) A photo of the setup, and measurement of (b) lateral FOV, (c) lateral resolution, (d) axial resolution. ESF, edge spread function; LSF, line spread function.

We then verified the lateral optical resolution of bGRIN-PAM by imaging a negative 1951 USAF resolution test chart (R3L1S4N1, Thorlabs). Group 7 Elements 3 and 4 (Fig. 4(a)) were imaged. A typical signal waveform is shown by the red curve in Fig. 4(b), and its frequency spectrum calculated by fast Fourier transform (FFT) is illustrated by the red curve in Fig. 4(c). The FFT spectrum suggests the main frequency of ~2 MHz, with a sidelobe around 4.1 MHz. To improve the signal-to-noise ratio, we devised a digital filter with a passband of 1–5 MHz. Attenuation outside the passband was set to be 60 dB. The filtered waveform is shown as the blue curve in Fig. 4(b). After filtering, a delay of ~0.475 μs was observed due to the phase shift imposed by the filter (~2π at 2.1 MHz). Since we use the signal amplitude (not the time of flight) to reconstruct the image, this delay does not influence the image. The FFT spectrum after filtering is shown as the blue curve in Fig. 4(c), which demonstrates a clear rolling off for frequencies outside the passband. The image before and after the digital filtering is displayed in Figs. 4(d) and 4(e), respectively. The digital filter improved the contrast considerably, especially at low-signal regions (lower left corner of the image).

Fig. 4.

Demonstration of the optical-resolution bGRIN-PAM using an USAF chart. (a) Brightfield image of the tested Group 7 Elements 3 and 4, (b) a typical photoacoustic waveform and (c) its FFT spectrum, and the reconstructed image (d) before and (e) after digital filtering.

Next, we verified the photoacoustic imaging while the bGRIN lens was subject to bending, which was demonstrated using the 5-μm-period subgrating of the same target in Fig. 3(a). The optical resolution of our bGRIN-PAM enabled visualization of the grating when the lens was straight (Fig. 5(a)). To facilitate the bending test, the lens was bent in a plane parallel to the target surface (and also the prism surface, see inset of Fig. 2). Cantilever-like deflection with a 5-mm distal displacement, which was about 5 times that (1 mm) of over 50% interventional needles inserted into live tissues [26], was introduced to the bGRIN lens. When the lens was bent either leftward (Fig. 5(b)) or rightward (Fig. 5(c)), the grating was entirely resolved. More in-depth analysis of bending effects on imaging can be found in our previous work on bGRIN lens microendoscopy [14].

Fig. 5.

Photoacoustic imaging of a 5-μm-period grating when the bGRIN lens is either (a) straight or bent, (b) leftward, and (c) rightward. The dark regions annotated by the arrows were due to the peeling off of the chrome coating.

Finally, we tested our bGRIN-PAM using mouse ear samples. A brightfield image of the tested sample, which was glued to a glass slide, is shown in Fig. 6(a). Because the laser wavelength of our current system is not optimized for label-free blood vessel imaging, only hairs of the mouse ear sample were imaged. A typical photoacoustic waveform is shown in Fig. 6(b). There is also a time delay of ~0.5 μs in the signal after filtering (2π phase shift at 2 MHz), similar to that in Fig. 4(b). The FFT spectrum of the waveform is shown in Fig. 6(c), suggesting a main frequency range of 2–4 MHz, different from the spectrum of the USAF target (Fig. 4(c)). In addition, there is a sidelobe around 5.8 MHz, likely originating from its proximity to the responsivity peak of the ultrasound sensor, which is effectively suppressed by the digital filter (blue curve of Fig. 6(c)).

Fig. 6.

Photoacoustic imaging of hairs on a mouse ear sample. (a) Brightfield image of the tested sample, (b) a typical photoacoustic waveform and (c) its FFT spectrum, (d) overlay of the brightfield and photoacoustic images, and (e) photoacoustic sectioning of the sample.

Photoacoustic imaging of the region designated by the dashed box in Fig. 6(a) was carried out, with the close-up view displayed in the upper left panel of Fig. 6(d). Because of the limited FOV, we performed four scans and then digitally stitched them to represent a larger area. Adjacent scans had partial overlap to facilitate the stitching. The stitched bGRIN-PAM image and its overlay with the brightfield image are shown in the lower left and right panels of Fig. 6(d), respectively. The two images generally show excellent spatial overlap. It is worth mentioning that the ear sample was not perfectly flat and had hairs pointing outward, thus the brightfield image only shows a 2D project view. However, the photoacoustic images were acquired from a single plane. Therefore, perfect spatial overlap in the overlay image would not be expected. The 3D imaging capability of our bGRIN-PAM is demonstrated by photoacoustic sectioning of the ear sample. Lateral sections at varying depths (35 μm interval, z direction) are shown in Fig. 6(e). Gradual evolution of the morphology was observed as the imaging depth decreased, verifying its optical sectioning capability.

While we demonstrated microendoscopic imaging using our bGRIN-PAM, the current system has some limitations. First, because the focused ultrasound sensor is placed at an inclined angle behind the bGRIN lens which has a short working distance (maximum ~600 μm), the spherical wavefront of the generated photoacoustic signal may be partially disturbed by the lens itself, which may lead to reduced signal efficiency. This arrangement may also contribute to the unsymmetric brightness of the image (e.g., the brighter upper right corner of images in Figs. 4(d) and 4(e)). Second, while our minimally invasive bGRIN lens can be inserted in deep tissues for imaging, the current bulky ultrasound sensor is unsuitable. A flexible miniature ultrasound sensor integrable with the bGRIN lens may address these two limitations. Third, while the current 1064 nm wavelength offers deeper tissue penetration, a shorter wavelength may better visualize vasculature and even nuclei, which are important elements for photoacoustic histology [27, 28]. Future advancements could also explore nanoparticle-based contrast enhancement [29] to improve vascular visualization while maintaining a low background at the current wavelength. A system optimized for vasculature imaging (beyond the fur imaging here) will fulfill the full potential of our bGRIN-PAM.

In summary, we have developed bGRIN-PAM, which features flexibility of the imaging probe and capability of optical-resolution, distal scan-free microendoscopic photoacoustic imaging. The FOV and resolution were measured using a line grating. The optical resolution was verified by a 1951 USAF chart. A digital bandpass filter was devised to improve the signal-to-background ratio. Feasibility of imaging during lens bending was also illustrated. Finally, photoacoustic sectioning of hairy mouse ears was demonstrated. Our bGRIN-PAM will potentially facilitate the precise placement of the aforementioned drug-screening microdevice and interventional needles into deep-seated tissues.

Data availability.

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.