A Biopsy-needle Compatible Varifocal Multiphoton Rigid Probe for Depth-resolved Optical Biopsy

Abstract

In this work, we report a biopsy-needle compatible rigid probe, capable of performing 3-dimensional two-photon optical biopsy. The probe has a small outer diameter of 1.75 mm and fits inside a gauge-14 biopsy needle to reach internal organs. A carefully designed focus scanning mechanism has been implemented in the rigid probe, which, along with a rapid 2D MEMS scanner, enables 3D imaging. Fast image acquisition up to 10 frames per second is possible, dramatically reducing motion artifacts during in vivo imaging. Equipped with a high numerical aperture (NA) micro-objective, the miniature rigid probe offers a high two-photon resolution (0.833 × 6.11 μm, lateral x axial), a lateral field of view of 120 μm, and an axial focus tuning range of 200 μm. In addition to imaging of mouse internal organs and subcutaneous tumor in vivo, first-of-its-kind depth-resolved two-photon optical biopsy of an internal organ has been successfully demonstrated on mouse kidney in vivo and in situ.

1. Introduction

Optical biopsy refers to methods of using optical imaging or spectroscopy to evaluate tissue histopathology in vivo and in situ without tissue excision [1–5]. Among various potential technologies for optical biopsy such as confocal laser scanning microscopy (CLSM) [6–8], optical coherence tomography (OCT) [9–14], and multiphoton microscopy (MPM), MPM is of particular interest because of its capability to perform 3D imaging of unstained tissue at subcellular resolution [15–19]. The near-IR excitation wavelength of MPM allows for deeper tissue penetration, and its confined nonlinear interaction helps reduce out of focus photodamage [18, 20]. Moreover, both two-photon fluorescence (2PF) from endogenous fluorophores (such as reduced nicotinamide adenine dinucleotide (NADH) and oxidized flavin adenine dinucleotide (FAD)) and second-harmonic generation (SHG) signals from collagen fibers can be acquired simultaneously with single-wavelength excitation [17, 21–24], enabling label-free and simultaneous multi-contrast imaging. MPM has been approved for skin imaging in European Union [25], and it provides considerable label-free contrast for skin cancer detection [26, 27]. However, conventional bench-top multiphoton microscopes are bulky and not suitable for internal organ imaging. Recently, the development of miniature fiber-optic multiphoton endomicroscopes enables label-free, in vivo, high-resolution and functional histological assessment of internal organs that was previously impossible [28–35]. An alternative approach for MPM to reach internal organs was to use a rigid probe made of a gradient index (GRIN) relay lens to deliver the excitation light to and collect the fluorescence (or SHG) from tissue [36–38]. Compared with the fiber-optic flexible endomicroscope, the rigid probe is more desirable in laparoscopic applications or in interfacing with a biopsy probe (i.e., by going through the cannula of a biopsy needle). In addition, the rigid probe can be made smaller since the beam scanning mechanism can be packed outside the probe at its proximal end, offering more flexibility in design and additional functionalities. Nonetheless, all prior efforts enabling MPM imaging of internal organs involved surgical elevation of target organs to reduce motion artifact and most lacked a built-in mechanism for depth scanning [35, 37, 39, 40], which fell short of the in situ requirement for optical biopsy.

To perform optical biopsy, several issues and challenges need to be addressed for the rigid probe. Firstly, the probe should be compatible with existing biopsy protocols, i.e., the probe needs to be protected and durable, but at the same time it should have an appropriate length and small enough diameter to fit inside the cannula of a biopsy needle. This requires the objective lens and the beam delivery optics to be small in diameter. Secondly, the probe should be compact and flexible for clinical use, which calls for fiber-optical delivery of the excitation beam and signal photons, and a mechanism to manage/minimize temporal broadening of the femtosecond excitation pulses in optical fibers. Performance-wise, the probe should offer a histopathological resolution with an adequate imaging signal-to-noise ratio (SNR). For MPM, it requires a tight spatiotemporal confinement of the excitation beam at the tip of the probe [41] and a superior collection efficiency. Thus, a miniature high-NA objective is needed. Function-wise, a built-in focus (or depth) scanning mechanism is desired to perform 3D imaging. Physically moving the probe for depth scanning is sub-optimal for in vivo applications, since it requires an additional precision actuation and translation mechanism [31, 34, 42], and the longitudinal probe translation would demand extra caution in avoidance of potential damages to the target tissue (e.g., tear or perforation) and even the probe itself. To realize focus scanning for MPM, a reasonably high NA for the excitation beam over the entire focus tuning range should be maintained, which poses an additional challenge in optics design for the rigid probe. Finally, the probe should provide a sufficient imaging speed in order to minimize motion artifacts introduced by physiological movements (breathing and heartbeat), and thus a fast beam scanner is highly desirable.

In this work, we developed a multiphoton handheld rigid probe for optical biopsy that addressed these challenges. The handheld rigid probe consisted of two functional parts: a handheld compact scanning box (3D) and a compound GRIN objective. The compound GRIN objective was 15 cm long (including a micro objective) with an outer diameter of 1.75 mm (including the housing hypodermic tube), and could fit within a 14-gauge biopsy needle. The scanning box included a MEMS mirror for 2D raster beam scanning up to 10 frames per second (FPS) and a piezoelectric stage for focus scanning. Our optics design ensured a continuously tunable working distance from 100 μm to 300 μm in tissue and yielded a measured 120 μm field of view with a 0.833 × 6.11 μm resolution (lateral x axial). A single-mode fiber was used in our rigid probe imaging system for delivery of femtosecond pulses, whereas a multi-mode fiber with a large core diameter was used at the proximal end of the rigid probe to deliver the nonlinear optical signal (collected by the compound GRIN objective) to a separate detection module. The use of optical fibers for both excitation pulse delivery and nonlinear signal collection offers the flexibility for positioning the rigid handheld probe during imaging, which is particularly desirable for in vivo imaging. Depth-resolved two-photon autofluorescence (2PAF) imaging of internal organs and subcutaneous tumor on mouse models in vivo were performed with the biopsy needle compatible rigid probe. In addition, enabled by its high frame rate and 3D imaging capability, a first-of-its-kind depth-resolved two-photon optical biopsy of internal organ was successfully demonstrated. The optical biopsy was performed on mouse kidney in vivo and in situ without surgical elevation of the organ, and with the rigid probe going through a biopsy needle.

2. Material and methods

2.1. Miniature handheld rigid probe

As mentioned in the previous section, the design of our handheld rigid probe takes into account multiple target operational and performance features, including (1) compatibility with a biopsy needle, (2) an excellent SNR and resolution for label-free imaging, (3) capability of 3D focus scanning, (4) a high imaging framerate, and (5) a compact and flexible system for in vivo or clinical use. Figure 1a shows the detailed schematic of the handheld probe which consists of a scanning box and a compound GRIN objective. Lateral beam scanning was performed by a MEMS scanner, which was capable of raster scanning at 10 FPS with a frame size of 234*234 pixels. Depth scanning is generally more challenging and several elegant methods have been demonstrated, including the use of an electrically tunable lens (ETL) in front of the objective lens [43]. However, a typical ETL can only provide 10 to 25 diopters of focusing power tuning range, which is not suitable for this application. In our design, depth scanning was realized by a piezoelectric stage (Figure 1a). By mechanically translating the input focus at the proximal end of the compound GRIN objective, the distal focus on the sample could be varied with minimal influence on the excitation NA.

Figure 1.

Handheld rigid probe. (a) Handheld probe design schematic. EFC: Excitation Fiber Connector and Collimator; CFC: Collection Fiber Connector; DM: Dichroic Mirror; L1-L4: Lens. (b) Photo of the handheld rigid probe. (c) Photo of the rigid probe inside a gauge-14 biopsy needle

2.1.1. Scanning box

The scanning box (Figure 1a) featured a MEMS scanner (A1S1.2, Mirrorcle Technologies, Inc.) for 2D lateral (X-Y) scanning, a piezoelectric micro translation stage (PP-18, Micronix) for focus (Z) scanning, a dichroic mirror for separation of signal beam from excitation beam and their corresponding relay optics. The excitation beam (denoted by red) from the PM-SMF was first collimated by a collimator (F230FC-780, Thorlabs), and then directed to the MEMS scanner for lateral 2D scanning. The MEMS scanner had a 1.7 mm aperture diameter, a resonance frequency of 1.3 kHz for both axes, and a maximum beam steering angle of 5 degrees. After the MEMS scanner, the beam was expanded by a pair of achromatic lenses (6.25 mm diameter, Edmund), as denoted by L1 (f=7.5 mm) and L2 (f=10 mm), and then refocused into the proximal (input) end of the compound GRIN objective by another achromatic lens L3 (f=10 mm). Depth scanning was realized by moving L3 along the optical axis with the piezoelectric translational stage. The nonlinear optical signal from the tissue was collected by the compound GRIN lens, collimated by L3, then deflected by a dichroic mirror (FF660-Di02–8.5*10.5, Semrock). The signal from the dichroic mirror was re-focused by a singlet lens L4 (6 mm diameter, f=6 mm, Edmund) to the multi-mode fiber for detection. Given the fact that all the components in the compound GRIN lens were not optimized for achromatic operation, the detection optics were expected to suffer from severe chromatic focal shift, thus a multi-mode fiber with a large core (800 μm in diameter) was adopted to minimize the impact of chromatic focal shift on fluorescence collection efficiency.

The chassis of the scanning box was designed in house and fabricated from aluminum with computer numerical control (CNC) machining. All the lenses and optical-mechanical components were either screwed or glued in their dedicated mounting space machined on the chassis. The MEMS scanner was held in place with a 3D printed mount with two screws for minor adjustments. Apart from the MEMS scanner, all the alignments of optics were passive owing to the high precision CNC machining process (offered by Proto Labs, Inc.).

2.1.2. Compound GRIN objective

Compatibility with a needle biopsy procedure is a major consideration in designing the compound GRIN objective. The objective should have an appropriate size and diameter to fit inside the cannula of a biopsy needle, and at the same time provide a superior resolution and sufficient field of view. We chose the dimension of the objective to be 15 cm long and 1.75 mm in diameter to fit inside the cannula of a gauge-14 biopsy needle (BARD® MAGNUM® MN1416). To ensure a good imaging quality, we have designed the objective as a compound lens consisting of a long doublet relay rod lens and a short micro objective (Figure 2a). The micro objective (GT-MO-080-018-810, GRINTech) has a high NA (0.8) at the sample and a 120 μm field of view. The doublet relay rod lens (GT-ERLS-100-275-075-10-20-NC, GRINTech) serves to deliver the excitation beam and the fluorescence photons over a long distance (>10 cm) between the scanning box and the micro objective. Due to the size constraint, we have chosen GRIN rod lenses of a 1 mm diameter to construct the doublet relay rod lens. In order to reduce the chance of beam clipping when the excitation beam is scanned off axis, a GRIN rod lens of 0.1NA was chosen for the proximal end of the doublet relay lens. To match the input NA of the micro objective (0.19), a second GRIN rod lens of a 0.2NA was added between the first rod lens and the micro objective. As a result, the doublet relay lens consisted of a long 0.1NA (2.75 pitch, 122 mm) GRIN rod lens at the proximal end, and a short 0.2NA (0.75 pitch, 18 mm) GRIN rod lens at the distal end. This design also came with the advantage of reducing the total GRIN lens pitch number (as compared with directly using a single 0.2NA GRIN rod lens), and thus reduced optical aberration in the excitation beam path. The GRIN relay doublet and the micro objective were housed inside a customized hypodermic tube of a 1.75 mm outer diameter, with a ~0.5 mm air gap in between the doublet and the micro objective.

Figure 2.

Ray tracing of the compound GRIN objective. (a) Compound GRIN objective ray tracing model. (b) Distal focus (or imaging depth D1) within the sample vs. input (proximal) beam focus (D2). The compound objective had an ~7X magnification, corresponding to an ~50X axial magnification. (c) Excitation NA vs. distal focus (D1). The excitation NA dropped 10% over the entire depth scanning range. (d) RMS spot size/airy radius vs. distal focus (D1). A diffraction limited focused spot size can be achieved when the imaging depth is scanned between 97 μm and 270 μm.

Ray tracing simulation (Figure 2) was performed with ZEMAX to evaluate and fine-tune the performance of the compound GRIN objective. As mentioned before, focus scanning (D1) of the imaging beam within the sample was realized by translating the input beam focus (D2) to the compound objective. At a given D2, the corresponding D1 was numerically determined and plotted in Figure 2b. At the designed two-photon excitation wavelength (750 nm), moving D2 from −5 mm to 5 mm led to a change of D1 (depth scanning) from 302 μm to 97 μm. The compound GRIN objective had a transverse magnification of ~7X, which translated to an ~50X axial magnification, matching nicely the ~20 μm/mm slope in Figure 2b. The excitation NA was also plotted against depth scanning of D1 in Figure 2c, where the NA dropped about only 10 % (i.e., from 0.66 to 0.59) over the entire depth scanning range. The root-mean-square (RMS) spot radius on the focal plane determined by ray tracing was plotted against D1 as well and was compared with the airy radius (Figure 2d). The results showed that the diffraction-limited beam focusing (with the RMS spot radius smaller than the airy radius) could be achieved when the beam focus (D1) was tuned between 97 μm and 270 μm. Outside the diffraction-limited focusing range, the aberrations from the compound GRIN objective will become non-negligible and contribute to degradation of the imaging resolution.

It is noted that our compound GRIN objective design could be directly applied to confocal reflectance endomicroscopy. For confocal fluorescence endomicroscopy in the short wavelength regime, additional chromatic correction might be required to further minimize the chromatic focal shift of the compound GRIN objective.

2.2. Rigid probe imaging system

To make the handheld rigid probe flexible for in vivo and potential future clinical use, we have adopted fiber-optic delivery and collection to separate the compact 3D scanning handheld probe from the benchtop optics system (Figure 3). A single-mode fiber (SMF) was used to deliver the excitation beam to the handheld probe, and a multi-mode fiber was used to transmit fluorescence or SHG signals collected by the probe from the sample to the photon detection unit within the benchtop optics system.

Figure 3.

Rigid probe imaging system. M: Mirror; CL: Coupling Lens; PM-SMF: Polarization Maintaining Single-mode Fiber; PMT: Photomultiplier Tube; EF: Emission Filter; L: Lens; FC: Fiber Connector.

Fiber delivery of femtosecond laser pulses from a Ti:Sapphire laser has always been a challenge. Material dispersion and nonlinear effects in optical fibers require careful management, which would otherwise lead to severe temporal broadening of the laser pulses and thus dramatically reduce the fluorescence yield for MPM [41]. Here we briefly describe the implementation of a dual-fiber based femtosecond pulse delivery scheme [44]. As shown in Figure 3, a Ti:Sapphire laser (120 fs, Chameleon Ultra II, Coherent, with a built-in pre-chirping unit) was first coupled into a 20 cm long polarization maintaining single-mode fiber (PM-SMF, PM780-HP, Thorlabs) for spectral broadening. The beam from the PM-SMF was then collimated and chirped by a grating pair (1200 lines/mm, Wasatch Photonics) to introduce anomalous dispersion. The negatively chirped pulse was re-coupled into another 1-meter-long PM-SMF of the same type to deliver the femtosecond pulse to the probe. With the anomalous dispersion provided by the grating pair balanced by the sum of the normal dispersion from the system (both fiber and probe), the pulse delivery system could deliver ~100 fs pulses up to 2nJ of pulse energy at the end of the second fiber.

The nonlinear signal from the rigid endoscopic probe was collected by a multi-mode fiber with an 800 μm core diameter (FT800EMT, Thorlabs), and then delivered to the signal detection module (Figure 3). In the detection module, the signal from the multi-mode fiber was first collimated, then fed through an emission filter (FF02–694/SP-25, Semrock), and finally directed to a photomultiplier tube (H10771–40P, Hamamatsu) for detection. The fiber launch, the grating pair and the detection unit were mounted on an 18*12 inch breadboard with all-fiber-optic connections for easy integration with a portable femtosecond laser in the future.

2.3. Animal preparation for in vivo imaging

2.3.1. In vivo imaging protocol

For kidney and small intestinal imaging, we adopted the same imaging protocol as previously published [35]. For kidney imaging, a small dorsal incision was created on an isoflurane sedated mouse, through which the left kidney was exteriorized and lifted for imaging. As for the case of small intestine imaging, a small abdominal incision was created on another isoflurane sedated mouse. From the incision, a loop of small intestine was lifted and cut open to expose the mucosa for imaging. For kidney optical biopsy in a laparoscopic fashion, similarly, an isoflurane sedated mouse was fitted with a nose cone on a feedback-controlled heating pad. With the mouse lying sideways, the position of its kidney was identified, and a tiny dorsal incision (~2 mm) was created on top of the kidney. The rigid probe was first mounted sideways on a translational stage, with the probe going through the cannula of a gauge-14 biopsy needle. The cannula was then fixed, acting as a guide allowing the probe to slide inside. Aimed at the incision, the tip of the probe was adjusted carefully to go through the incision, until reaching the surface of the kidney. In vivo subcutaneous tumor imaging was performed with the tumor secured by a spring-loaded clamp. Images were first acquired with the rigid probe held carefully against the intact skin atop the tumor, then the tumor was cut open for further imaging. In cases of severe hemorrhage, a small piece of No. 0 cover glass was fitted on top of the tissue to control the amount of bleed. For all the in vivo imaging experiments, 30–50 mW of optical power on tissue at 750 nm was used.

2.3.2. Subcutaneous tumor model preparation

The A431 (human epidermoid carcinoma) cells were first cultured in Dulbecco’s modified eagle medium (DMEM, Gibco) supplemented with 10% (v/v) of fetal bovine serum (Sigma-Aldrich) at 37°C with 5% $C{O}_2$ and a humidified atmosphere. Five NCr nude mice (male, 6–8 weeks, Taconic biosciences) were injected subcutaneously with ~$3\times {10}^6$A431 cells in the right flank. 7–10 days after tumor induction (tumor reaching 5–6 mm in diameter), the mice were isoflurane-sedated and prepared for imaging.

All animal housing and experimentation procedures were performed in accordance with the standards of humane animal care described in the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Protocols were approved by the institutional animal care and use committee of the Johns Hopkins University.

3. Results and discussion

3.1. System performance characterization

Several key parameters such as the temporal pulse width, optical throughput, field of view (FOV), two-photon imaging resolution (lateral and axial) and maximum frame rate were experimentally investigated to assure the handheld rigid probe was performing as designed. We first tested our fiber-based pulse delivery scheme. The excitation beam from the tip of the probe was collimated by an aspherical lens with a short focal length, and then directed to a home-built interferometric autocorrelator for pulse width measurement. With the grating pair separation carefully tuned in the dual-fiber based grating pair dispersion management setup, the shortest achievable temporal pulse width of the excitation beam was 65 fs at 750 nm, assuming a squared hyperbolic secant pulse shape (Figure 4a). Note that the resultant pulse width was shorter than the input pulse width (120 fs, FWHM) from the Ti:Sapphire laser. This was due to the nonlinear effects in the fibers, and a resultant overall spectral broadening present in the pulse delivery system [44]. At the design wavelength of 750 nm, the optical throughput of the pulse delivery system (from the laser output to the rigid probe input) was measured to be 28%, and the optical throughput of the handheld rigid probe was 45%, yielding an overall system optical throughput of 12.6%.

Figure 4.

System characteristics. (a) Autocorrelation trace of the excitation beam. The FWHM of the autocorrelation trace was ~100 fs, corresponding to a 65 fs temporal pulse width (assuming a sech² pulse shape). (b) Wide-field reflectance image of a USAF resolution target showing the group 6 element 1 pattern (64 lp/mm, scale bar: 20 μm). (c) Two-photon signal intensity envelope measured across the entire FOV by a fluorescence slide, showing an FWHM of 101.6 μm and 85.6 μm along the x- and y-axis, respectively (with the Focus (D1) tuned to 200 μm). (d) The lateral point-spread-function (PSF) of the two-photon fluorescence signal measured across a 200 nm yellow-green fluorescent bead (with the Focus (D1) tuned to 200 μm). Experimental data was fit with a Gaussian profile and the FWHM of the lateral PSF was found to be 833 nm. (e) The axial PSF of the two-photon fluorescence signal measured by the ‘knife edge resolution test’ with the surface of a thick fluorescence slide (with the Focus (D1) tuned to 200 μm). The axial PSF data was derived from the step response curve (dashed line) by taking its derivative and a fit with a Lorentzian profile yielded a 6.11 μm axial PSF (see text for detailed descriptions). (f) The lateral resolution (PSF) was measured along the entire focus tuning range (every 20 μm between 100 to 300 μm focus tuning). The resolution remained submicron between 100 to 280 μm, with best resolution (833 nm) measured with the focus around 200–220 μm.

The FOV of the rigid probe was validated by imaging a USAF 1951 resolution target with the imaging system configured in a wide-field reflection imaging mode (excitation filter removed and PMT replaced by a photo diode). Figure 4b shows a representative wide-filed microscopy image of the group 6 element 1 pattern (64 line pairs/mm) of the resolution target. The FOV was estimated to be ~120 μm in diameter which agreed well with the specs of the micro objective. Another important parameter is the FWHM of the two-photon intensity envelope across the entire FOV. This was measured by imaging a green fluorescence reference slide (#2273, Ted Pella, Inc.) with the focus (D1) tuned to 200 μm. The measured intensity envelopes along the X and Y directions were plotted in Figure 4c, with their FWHM identified as 101.6 μm and 85.6 μm, respectively. The asymmetry in the intensity profile was noted, which was caused by a slight misalignment (<1̊) between the compound GRIN objective and the scanner box.

The lateral two-photon imaging resolution was measured with 200 nm sub-resolution yellow-green fluorescence beads (#09834, Polysciences, Inc.) with the focus (D1) tuned to 200 μm. Figure 4d shows the point spread function (PSF) measured by the fluorescent beads and its Gaussian fitting. The FWHM of the Gaussian fitted curve was calculated to be 833 nm. The axial resolution was measured by a ‘knife edge resolution test’ with the same green fluorescence reference slide. By moving the surface of the thick fluorescence slide towards the focal plane of the rigid probe axially with a precision translation stage, the increase of fluorescence intensity in the center of the FOV (averages of 50 pixels) was recorded as a function of the axial displacement of the slide. This curve (dashed line in Figure 4e) is the axial step response curve, as it is effectively the convolution of a unit step function (approximated by the surface transition of the fluorescence slide) with the axial PSF of the system. The axial PSF (axial impulse response) was then derived by taking the derivative of the axial step response curve and it was then plotted in Figure 4e. A Lorentzian fit was applied to the axial PSF and the FWHM was found to be 6.11 μm (with the focus (D1) tuned to 200 μm). This method was proven to be more reliable and consistent compared with directly acquiring the XZ images of the fluorescence beads, as the slide was much more resilient to photobleaching.

Simulations showed that diffraction-limited focusing could be achieved when the beam focus was tuned between 97 μm and 270 μm, this was also experimentally validated by measuring the lateral resolution along the entire focus tuning range. It was done by embedding the fluorescence beads into 1% agarose ($~{10}^6$ particles/mm³) and taking 3D volumetric images of the embedded beads along the focus tuning range (100 μm to 300 μm, 20 μm interval). The measured resolution as a function of the beam focus (D1) was plotted in Figure 4f, and the trend of the curve matched nicely with the simulated spot size shown in Figure 2d. In Figure 4f, the lateral resolution remained submicron when the focus was tuned between 100 to 280 μm, with best resolution measured with the focus around 200–220 μm (833 nm). Note that all the resolution data in Figures 4d-f were averages of three different measurements, and all the data in Figures 4c-e were acquired with the focus (D1) tuned to 200 μm for consistency.

The maximum frame rate that yielded a reasonable image size supported by our MEMS scanner was also investigated. For high frame rate scanning applications, the MEMS should be driven in a hybrid raster scanning mode, where one axis is given a sinusoidal voltage at a frequency close to its resonance for line scan, and the other axis is given a ramp signal for it to work in a quasi-static mode for frame scan. The resonance frequency of our MEMS scanner was 1.3 kHz for both axes, and when the X-axis was driven at 1.424 kHz, imaging at 10 FPS could be achieved with each frame consisting of 234*234 pixels. The off-resonance 1.424 kHz driving frequency was chosen empirically not only to avoid coupling between the X and Y axes, but also to achieve a good compromise between scanning range and speed. If switched to a MEMS scanner with a higher resonance frequency (e.g., 3.8kHz, Model A3I8.2, Mirrorcle Technologies), the framerate could be easily pushed to ~30 FPS with this scanning mode (given a sufficient fluorescence photon rate). The near-resonant scanning of X-axis also introduced nonlinear distortion in the raw images, where the center portions of the raw images were stretched. An additional numerical interpolation step was introduced in post processing, where the images were mapped back to the equidistance sampling space with linear interpolation, correcting the nonlinear distortion introduced by sinusoidal scanning.

Finally, the imaging performance of the rigid probe was qualitatively compared with a conventional table-top two-photon microscope (modified from Olympus FLUOVIEW confocal microscope, equipped with a XLUMPlanFL 20X objective). A convallaria majalis (lily of the valley) fluorescent microscopy slide (#1027337197, Carl Zeiss Microscopy LLC) was imaged by the rigid probe first, then the same regions of interest were identified under the table-top microscope with a much larger FOV. And finally, images were acquired with 6X zoom-in (~130 μm FOV) with the microscope. Figures 5a-c shows the unaveraged raw images acquired by the microscope, with 512*512 pixels, 4 μs pixel dwell time, 4 mW average power on the sample, and the images were cropped to match the image size of the rigid probe. Figures 5d-f shows the corresponding two-photon fluorescence images (unaveraged raw frames) acquired by the rigid probe, with 500*500 pixels, 1.33 μs pixel dwell time and 13 mW average power on the sample. Apart from the smaller FOV, the rigid probe provided very similar imaging contrast as compared with the microscope, with slightly blurred edges due to lower spatial resolution and some field distortion. The FOV of the rigid probe was majorly limited by the small diameter of the micro objective and relay lens. For laparoscopic and endoscopic applications with a much larger working channel, larger objectives and relay optics could be fit into our design, potentially allowing larger FOVs and/or better imaging resolutions.

Figure 5.

Comparison of the rigid probe with a conventional bench-top two-photon fluorescence microscope. (a-c) Conventional two-photon microscopy images of a convallaria majalis (lily of the valley) fluorescent slide (512*512 pixels, unaveraged raw frames with 4 μs pixel dwell time, 4 mW average power on the sample, ImageJ ‘green’ color map, scale bar: 20 μm. Images were cropped to match the image size of the rigid probe). (d-f) Two-photon fluorescence images of the convallaria majalis slide acquired by the rigid probe (500*500 pixels, unaveraged raw frames with 1.33 μs pixel dwell time, 13 mW average power used, ImageJ ‘green’ color map, scale bar: 20 μm).

3.2. In vivo depth-resolved two-photon imaging of mouse internal organs

To demonstrate the feasibility of the probe for optical biopsy in vivo, we performed depth-resolved two-photon auto-fluorescence (2PAF) imaging of internal organs on live mice. Here we chose kidney and small intestine as the organs of interest, and for initial performance testing, surgeries were performed to elevate these organs from the body to suppress motion during imaging (laparoscopic imaging was also performed later without surgeries for organ elevation). A single PMT was used to detect all the autofluorescence signals ranging from 400 nm to 660 nm, including those from intrinsic fluorophores such as NAD(P)H and FAD. Figure 6a shows some representative images of mouse kidney cortex in vivo where the renal tubule (RT) and dark nucleus (N) are evident on the surface of the cortex. As the imaging depth was tuned 24 μm below the surface, the empty lumen (L) of the renal tubule shows up. Figure 6b shows representative images at subcellular level acquired from villi of mouse small intestinal mucosa. On the surface of the mucosa, one can appreciate the tiling pattern of the columnar epithelial (CE) cells, with occasional mucus-secreting goblet (G) cells identified as dark patches surrounded by epithelial cells [45]. Crypts (C) of the small intestinal mucosa are also visible in the same field of view. As the depth is tuned 20 μm below the surface towards the basal side, columnar epithelial cells in another orientation (with elongated columnar shape) emerge near one of the crypts.

Figure 6.

2PAF images of internal organs in vivo. (a) Representative in vivo depth-resolved 2PAF images of mouse kidney cortex. RT: Renal Tubules; N: Nucleus; RCI: Renal Cortical Interstitium; L: Lumen. (500*500 pixels, 5-frames-averaged with 10 μs effective pixel dwell time, ImageJ ‘green’ color map, scale bar: 20 μm) (b) Representative in vivo depth-resolved 2PAF images of mouse small intestinal mucosa. C: Crypt; G: Goblet Cell; CE: Columnar Epithelial Cell. (500*500 pixels, 10-frames-averaged with 20 μs effective pixel dwell time, ImageJ ‘green’ color map, scale bar: 20 μm)

3.3. In vivo two-photon imaging of subcutaneous tumor

We further demonstrated the probe’s capability of label-free imaging of cancerous tissue by performing 2PAF imaging of A431 (human epidermoid carcinoma) subcutaneous tumor in vivo on a mouse model by gently placing the probe in direct contact with the intact skin and the exposed tumor core. Figure 7 shows the representative images collected by the rigid probe of the subcutaneous tumor as well as normal skin, with Figure 7a showing the keratinocytes of normal intact skin [46], Figure 7b showing the cells of intact skin directly on top of the subcutaneous tumor, and Figures 7c-d showing the cells of the exposed tumor core. In the tumor core, cellular features such as large nuclei, pronounced variation in cell size and shape, and disorganized cellular arrangement can be directly visualized, which are hallmarks of cancer cells [47, 48].

Figure 7.

2PAF images of normal mouse skin and A431 subcutaneous tumor. (a) 2PAF image of keratinocytes of normal mouse skin. (b) 2PAF image of the skin cells directly atop the subcutaneous tumor, where cells closely resemble the normal keratinocytes. (c-d) 2PAF images of A431 subcutaneous tumor core, showing the distinct cellular features of tumor cells (Images a-d: 500*500 pixels, 10-frames-averaged with 13.3 μs effective pixel dwell time, ImageJ ‘green’ color map, scale bar: 20 μm)

3.4. Optical biopsy of internal organ

The in vivo 2PAF imaging of internal organs described previously involved surgical elevation of the organ from the body for minimizing motion artifacts, which did not satisfy the in situ requirement of optical biopsy. Here we report 2PAF optical biopsy of mouse kidney in vivo and in situ with the rigid probe, where no surgeries were performed on the organ. Figure 8a shows the experimental setup, where imaging was performed laparoscopically on the kidney of an anesthetized mouse with the rigid probe delivered through a biopsy needle cannula. Figures 8b-d show the representative 2PAF images acquired at different depths of the kidney cortex. Microscopic structures of the kidney cortex such as renal tubules (RT), lumen (L) of the renal tubule and renal cortical interstitium (RCI) could again be visualized by the rigid probe at an imaging speed of 10 FPS, but with a reduced image quality when compared with Figure 6a. This was due to a much shorter pixel dwell time and downsampling (~2X reduction of scan line density) in the 10-FPS scanning mode. Video clips of real-time imaging at a fixed and variable depth with the rigid probe are also provided (see Visualization 1 and Visualization 2).

Figure 8.

Optical biopsy of mouse kidney. (a) Photo of the rigid probe based optical biopsy setup. (b)-(d) Representative depth-resolved 2PAF images of mouse kidney cortex acquired during optical biopsy (Visualization 1 and Visualization 2). RCI: Renal Cortical Interstitium; RT: Renal Tubules; L: Lumen. (234*234 pixels, unaveraged raw frames with 2 μs pixel dwell time, ImageJ ‘green’ color map, scale bar: 20 μm)

4. Conclusion

In conclusion, we developed a biopsy-needle compatible varifocal multiphoton handheld rigid probe for depth-resolved optical biopsy of unlabeled biological tissues in vivo and in situ. The probe could perform 2PAF imaging near histopathological resolution with a good SNR owing to careful temporal pulse management for the excitation pulses, the use of a high NA objective and the excellent nonlinear light collection efficiency. The probe was compatible with the needle biopsy protocol, and its small diameter (1.75 mm) and sufficient length (15 cm) fit inside the cannula of a 14-gauge biopsy needle. The probe was capable of 3D imaging, with a 120 μm field of view and a 200 μm focus scanning range. Imaging could be performed in real time at 10 FPS with an integrated MEMS scanner. Depth-resolved 2PAF imaging of healthy mouse internal organs and subcutaneous tumor in vivo were demonstrated. With our rigid probe, microscopic structures of kidney cortex and small intestinal mucosa could be clearly visualized in a label-free fashion, and tumor cellular features such as large nuclei, pronounced variation in cell size and shape, as well as disorganized cellular arrangement could be clearly resolved. In addition, depth-resolved optical biopsy of internal organ was demonstrated on mouse kidney for the first time with the organ untouched, and the rigid probe going through a biopsy needle. In the future, more detection channels can be easily added to the rigid probe imaging system, potentially allowing applications such as high-resolution metabolic functional imaging by simultaneous NAD(P)H and FAD detection [35], as well as evaluation of breast cancer malignancy by SHG imaging during needle biopsy [49], and preterm birth risk assessment [32].