2.1.

Laser Microbiopsy Overview

An overview of the laser microbiopsy approach is shown in Fig. 1. A pulsed laser beam is shaped through a lens system to generate a converging annular beam profile at the tissue surface. The annular beam shape ablates the tissue, and the tissue section in the center of the annulus is ejected. Cryogen spray cooling is used before and after the laser pulse to limit tissue damage and patient discomfort. These tissue sections, with volumes $<1\mu \mathrm{l}$ ($=1{\mathrm{mm}}^3$), are collected onto a glass slide where they can be analyzed using desired methods including, but not limited to, conventional H&E staining, virtual H&E or other microscopy methods, and molecular profiling.

Fig. 1

Overview of laser microbiopsy workflow. An annular beam is used to cut a sub-microliter tissue section. Cryogen spray cooling is used pre- and post-ablation to limit thermal damage. The tissue is ejected and collected directly onto a glass slide for imaging/analysis. Figure and portions of methods adapted from Refs. 26 and 27.

2.2.

Benchtop Laser Microbiopsy Implementation

We have designed and built a benchtop laser microbiopsy system consisting of the following components (Fig. 2): (1) A Ho:YAG laser (MOSES Pulse 120H, Lumenis, Yokneam, Israel) delivered through a $230\text{-}\mu \mathrm{m}$ core diameter multimode fiber (Moses 200 D/F/L, Lumenis, Yokneam, Israel). An SMA connector was added to the distal end of the fiber to allow connection to a fiber mount with $x\text{-}y\text{-}z$ positioning and tilt control for precise optical alignment. (2) A ZnSe aspheric collimation lens (ASPH-ZC-25-50, ISP Optics, Orlando, Florida) to collimate light from the fiber. (3) Two fused silica axicons ($\alpha =20\mathrm{deg}$, Thorlabs, Newton, New Jersey) to shape the collimated circular beam into an annular beam. The axicons are oriented with the points of the conical surface facing in the direction of light propagation as opposed to the typical orientation with the axicon points facing each other. This axicon orientation provides an annular beam diverging at an angle of 0.6 deg rather than perfect collimation, which results in the beam retaining an annular shape at the focal plane. (4) A ZnSe aspheric focusing lens (ASPH-ZC-25-25, ISP Optics, Orlando, Florida) is utilized to focus the annular shape. The optical components produce a converging annular shape with decreasing diameter towards the beam focus. A similar beam shaping approach has been reported for material processing using an annular beam.²⁸ The converging annular shape conveniently allows a variable annulus diameter at the tissue surface by adjusting the distance between the focusing lens and the tissue surface. Additionally, the beam converges into the tissue, generating a pressure gradient necessary to eject the harvested tissue upward where it is collected directly onto a glass slide. Cryogen spray cooling (R134A) is used before and after pulsed laser emission to limit thermal damage. A solenoid fuel injector valve (800-1257N, GP Sorensen, Long Island City, New York) is used to deliver the cryogen spray pre-cooling directed at the tissue surface from 4 cm away and a second valve is used to deliver post-cooling directed at the biopsy collected on the glass slide from 4 cm away.

Fig. 2

Diagram of optics used in this study. A Ho:YAG laser is shaped into an annular beam using a pair of axicons. The annular beam is focused on the tissue surface through an aspheric focusing lens. Cooling and harvesting slide not shown.

2.3.

Transmission Measurement

Transmission through the beam-shaping optics was measured. A power meter (PM10, Coherent, Santa Clara, California) was used to record average power before and after the beam-shaping optics, and percent transmission was calculated.

2.4.

Beam Profile Measurement

The annular beam profile produced by our benchtop system was explored computationally and experimentally. The optical components were imported into raytracing software (Zemax), and the resulting beam profile was measured in software with a virtual detector. The beam profile was measured experimentally by irradiating a $100\text{-}\mu \mathrm{m}$ layer of water in a rectangular capillary tube and imaging the emitted radiation using an IR camera (420-0044-04-00, FLIR, Wilsonville, Oregon) [Fig. 5(a)]. For these measurements, the laser pulse energy was set to the minimum level (0.2 J). Pulse energy was further reduced by coupling light between two optical fibers with intentionally poor coupling efficiency. Beam-shaping optics were oriented horizontally, and the rectangular capillary was placed a variable distance, $Z$, from the back surface of the focusing lens. The capillary was initially placed at $Z=20\mathrm{mm}$ and increased in increments of 0.25 mm to measure the beam profile at varying focal locations. A ZnSe lens (ASPH-ZC-25-50, ISP Optics, Orlando, Florida) was placed between the capillary and the IR camera to magnify the beam profile image. A video was recorded by the IR camera at a frame rate of 432.5 Hz and an integration time of 0.50 ms. The first frame after laser emission was recorded and used for beam profile analysis. Computational and experimental beam profiles were compared.

2.5.

Computational Modeling

The laser microbiopsy processes were modeled computationally to gain insight in beam-shaping optimization, interaction with annular beam and tissue, tissue ablation, and thermal damage. The simulation is based off of a blow-off model as described by Vogel and Venugopalan.²⁹ This model is conditional on the laser pulse being shorter than the thermal diffusion time providing thermal confinement. In the case of annular ablation, we calculate the thermal diffusion time using Eq. (1), where l is the width of the annular beam from inner to outer diameter, $\alpha$ is the thermal diffusivity, $k$ is the thermal conductivity, $\rho$ is the tissue density, and $C_p$ is the tissue specific heat capacity (Table 1)

Table 1

Parameters used in computational modeling.

Using an annular beam width of $100\mu \mathrm{m}$ gives a thermal diffusion time of 24 ms. The pulse duration of the laser is much shorter with a FWHM duration of $71\mu \mathrm{s}$.³³ Therefore, thermal confinement during the laser pulse is a justified assumption, and the blow-off model can be used.

The computational model consisted of the following steps. (1) The optical fiber and beam-shaping optics were modeled in raytracing software (Zemax), and the beam profile at varying distance from the beam focus was simulated. (2) A volume detector with the absorption and scattering properties of soft tissue (Table 1) was inserted at a distance such that the outer diameter of the annular beam was $800\mu \mathrm{m}$. (3) The simulated fluence in the tissue was computed with raytracing and exported to MATLAB [Fig. 3(a)]. 4) Regions with a radiant energy density greater than a threshold of $2.5\mathrm{J}/{\mathrm{mm}}^3$ were assumed to be ablated, and the tissue inside the ablation annulus was assumed to be ejected as a microbiopsy at the conclusion of the laser pulse [Fig. 3(b) and 3(c)]. The ablation threshold is derived from the energy required to vaporize water within the tissue.³⁴ Energy absorbed in the ablation region was removed from the model as this energy is assumed to be spent on ablation and kinetic energy of tissue expulsion. The energy absorbed in the regions of tissue that are not ablated contributes to tissue heating. The temperature at the end of the laser pulse at each depth and radius, is calculated using Eq. (2), where 37°C is the initial body temperature, $F(r,z)$ is the fluence at each location, ${\mu }_a$ is the absorption coefficient, $\rho$ is the tissue density, and $C_p$ is the tissue specific heat capacity (Table 1):²⁹

Fig. 3

Computational model steps. (a) Fluence through the tissue was modeled using Zemax. (b) Regions with fluence above the ablation threshold were removed [displayed as dark blue in (b) and (c)]. (c) The microbiopsy separated from the bulk tissue at the conclusion of the laser pulse and the post ablation temperature evolution was calculated. (d) The microbiopsy volume and initial temperature were exported to COMSOL for modeling of bioheat transfer and thermal damage.

(5) The simulated biopsy was severed at half of the maximum ablation depth to give a simulated microbiopsy volume within the range of experimental microbiopsy volumes. The volume and geometry of the tissue in the middle of the ablated annulus and above the severed depth were saved to an STL file and imported into COMSOL multiphysics (COMSOL) as the volume of the ejected microbiopsy [Fig. 3(d)]. The temperature at the end of the laser pulse [Eq. (2)] within the ejected microbiopsy volume was imported into COMSOL [Figs 3(c) and 3(d)]. (6) The heat transfer through the tissue as it cools was modeled in COMSOL. (7) The Arrhenius damage integral [Eq. (3)] and the percentage of cell death [Eq. (4)] were calculated in COMSOL for each location in the microbiopsy.^29,35,36

In these equations, $\mathrm{\Omega }$ is the Arrhenius thermal damage parameter, A is the frequency factor, $E_a$ is the activation energy per mole, $R$ is the universal gas constant, and $T$ is the tissue temperature in Kelvin. Values used for these parameters are given in Table 1. Importantly, the Arrhenius damage integral is used here as an overall measure of thermal damage to make comparisons for optimization but does not necessarily indicate whether the tissue will retain integrity for analysis by virtual H&E or other analysis methods. An Arrhenius thermal damage parameter of $\mathrm{\Omega }=1$ indicates 63% cell death and is used as a threshold for thermal damage.

This model allows for various parameters to be varied to better understand the laser microbiopsy process and aid in optimization to limit microbiopsy thermal damage. Important parameters that were explored included tissue scattering coefficient, convective heat transfer coefficient, and external temperature. The effects of cryogen spray cooling before (pre-cooling) and/or after (post-cooling) the laser pulse were explored. To model pre-cooling, the heat transfer through tissue was calculated with a surface temperature of $-40°\mathrm{C}$ and a convective heat transfer coefficient at the surface of $\mathrm{10,000}\mathrm{W}/{\mathrm{m}}^2\mathrm{K}$ for a duration of 200 ms. Detailed descriptions of cryogen spray cooling including computational models and experimental results can be found elsewhere.^31,37–39 To model post-cooling the convective heat transfer coefficient and external temperature were set to $\mathrm{10,000}\mathrm{W}/{\mathrm{m}}^2\mathrm{K}$ and $-40°\mathrm{C}$. To model cooling in room air, the convective heat transfer coefficient and external temperature were set to $10\mathrm{W}/{\mathrm{m}}^2\mathrm{K}$ and 23°C.

The effects of tissue scattering were also explored. The model was run with normal tissue reduced scattering coefficient (${{\mu }_s}^{\prime }=6{\mathrm{cm}}^{-1}$),⁴⁰ scattering reduced by half (${{\mu }_s}^{\prime }=3{\mathrm{cm}}^{-1}$), and without scattering (${{\mu }_s}^{\prime }=0$). Experimental results with reduced tissue scattering are not available for microbiopsy volume comparison. Therefore, biopsy volumes were assumed to separate from the bulk tissue at the maximum ablation depth for these models.

The full set of parameters used in the computational model is summarized in Table 1. The key assumptions made in this model are summarized in Fig. 4.

Fig. 4

Key assumptions made in the thermal damage model for each stage: pre-cooling, ablation, and post-cooling.

2.6.

Microbiopsy Harvest Experiments

Microbiopsy harvest experiments were completed on ex vivo porcine tissues (Animal Technologies, Tyler, Texas). The protocol involved the following: (1) the tissue surface was aligned to the desired position relative to the beam focus using an aiming beam. This position determines the diameter of the annular beam at the tissue surface. (2) A $100\text{-}\mu \mathrm{m}$ thick glass coverslip was positioned $\sim 5\mathrm{mm}$ above the tissue surface to collect the ejected microbiopsy. (3) A custom shutter consisting of a foil paddle connected to a rotational stepper motor (17md302s-00, Anaheim Automation, Anaheim, California) was placed with the foil paddle initially directly below the focusing lens, preventing initial pulses emitted by the laser from reaching the tissue. (4) Pre- and post-cooling solenoid valves were directed to the tissue surface and glass coverslip, respectively. (5) The laser pulse energy was set. (6) A train of pulses was manually initiated from the laser. The first pulse emitted by the laser was detected by an InGaAs photodetector (2034, New Focus, San Jose, California) connected to a digital delay generator. This event was used to control the shutter and spray cooling timing. (7) Pre-cooling began 600 ms before microbiopsy collection for a duration of 400 ms. (8) The shutter rotated open to allow a single laser pulse to perform the microbiopsy before returning to the closed position to block the following pulse. (9) Post-cooling began immediately after microbiopsy collection for a duration of 200 ms.

This procedure was tested on ex vivo porcine skin and kidney tissues. Optimal pulse energy and tissue location in the beam (beam diameter) were experimentally determined for each tissue type. A fast video camera (Fastcam Mini UX100, Photron, Tokyo, Japan) was used to record the skin ablation process at 50,000 fps.

2.7.

Microbiopsy Volume Measurements

Microbiopsies were collected from ex vivo porcine skin at two distances from the focusing lens (separated by 0.5 mm) providing a larger and smaller beam with estimated inner annulus diameters of 530 and $420\mu \mathrm{m}$, respectively. Beam diameters outside this range were either too large or small to reliably harvest microbiopsies. Pulse energies of 1 and 1.3 J incident on the tissue surface were used. Pulse energies $<1\mathrm{J}$ were insufficient to reliably harvest microbiopsies from porcine skin. Five microbiopsies were collected for each combination of beam diameter and pulse energy. Harvested microbiopsies were imaged using a custom-built OCT system.³⁶ The OCT system used a single mode optical fiber with a swept source laser (1310 nm, Axsun, Billerica, Massachusetts). The OCT system used a Mach–Zehnder fiber interferometer configuration collecting field of views up to $6.5\mathrm{mm}(X)\times 6.5\mathrm{mm}(Y)\times 3\mathrm{mm}$ (depth). A custom MATLAB program was used to detect the microbiopsy surface from recorded OCT data, display a 3D reconstruction of each microbiopsy, and calculate microbiopsy volumes.

2.8.

Virtual H&E

To generate H&E images, acridine orange (AO) was chosen as the nuclear stain with sulforhodamine 101 (SR101) as a counter stain as described by Cahill et al.¹⁷ The stain was prepared by mixing 8 mg AO (A6014, Sigma-Aldrich, St. Louis, Missouri) and 5 mg SR101 (HB0838, Hello Bio, Princeton, New Jersey) in 50 ml 1:1 water:ethanol. Tissues were stained for 2 min followed by a 20 s rinse in 1:1 water:ethanol. Tissues were then imaged with a confocal microscope (Ultima IV, Bruker, Billerica, Massachusetts) with a $20\times$ objective. Two excitation wavelengths were used: 488 nm for AO excitation and 561 nm for SR101 excitation. Two emission channels were used: $525\pm 25\mathrm{nm}$ for AO emission and $605\pm 35\mathrm{nm}$ for SR101 emission. The two channels were false colored and merged as described by Gareau¹⁵ with the SR101 channel taking the place of the reflectance channel. The R and G values for the AO or $H$ channel were increased to reduce blueness after feedback from a pathologist. The RGB values used for the H&E digital staining were $H=[0.5,0.4,1]$ and $E=[1,0.55,0.88]$. Virtual H&E imaging was first done on bulk tissues as a control. Microbiopsies and ablation craters were then imaged and compared to bulk tissues to evaluate damage.

3.1.

Transmission Measurements

Transmission through the beam-shaping optics was measured to be 66% of pulsed radiant energy released from the fiber tip. Therefore, a pulse energy setting at the laser of 2 J results in a pulse energy at the tissue surface of 1.3 J while a pulse energy setting of 1.5 J results in a pulse energy at the tissue surface of 1 J. All values reported in this study are the Ho:YAG radiant pulse energy incident on the tissue surface.

3.2.

Simulated and Experimental Beam Profiles

Simulated and experimental beam profile measurements show an annular shape with good intensity uniformity around the circumference of the annulus [Fig. 5(b)]. The full width at half maximum (FWHM) width from outer to inner radius was larger for experimental vs. simulated beams [Fig. 5(c)]. The experimental FWHM outer beam diameter decreased from $874\mu \mathrm{m}$ (at position $Z_o=20\mathrm{mm}$ from the back side of focusing lens) to $795\mu \mathrm{m}$ ($Z=20.25\mathrm{mm}$) to $721\mu \mathrm{m}$ ($Z=20.50\mathrm{mm}$) to $643\mu \mathrm{m}$ ($Z=20.75\mathrm{mm}$) as the beam was measured in increments of 0.25 mm away from the focusing lens.

Fig. 5

(a) Beam profiles were measured by irradiating a thin layer of water and imaging with an infrared camera. (b) and (c) Simulated and measured beam profiles show a uniform annular shape that decreases in diameter with increased distance from the beam-shaping optics.

3.3.

Computational Thermal Damage

Computational results displaying biopsy volume geometry, post-ablation temperature, cooling profile, and resulting tissue damage are compared for biopsies with and without pre- and post-cooling are shown in Fig. 6. Computational results indicate that cooling does not significantly affect ablation depth. Results indicate that pre- and post-cooling works best in combination to preserve the largest volume of tissue. The model predicts that without post-cooling the entire biopsy would be thermally damaged at a level $\mathrm{\Omega }>1$ (Fig. 6, left two columns). With post-cooling alone a small fraction of tissue at the surface up to $55\mu \mathrm{m}$ into the tissue remains viable. With a combination of pre- and post-cooling a larger volume up to $140\mu \mathrm{m}$ into the tissue remains viable ($\mathrm{\Omega }<1$). The mean temperature in the simulated harvested microbiopsies over time, including before, during, and after the laser pulse is shown in Fig. 7. The time axis is shown on a log scale to display the orders of magnitude difference in cooling time for cases with and without post-cooling. From Figs. 6 and 7, we observe that pre-cooling limits the maximum temperature reached within the microbiopsy, while post-cooling with cryogen spray leads to much a much faster cooling time on the order of 50 ms. When used together, pre- and post-cooling minimize the extent of thermal damage.

Fig. 6

Temperature and thermal damage calculations for biopsies under different cooling schemes. Temperature profiles are shown immediately post ablation and after 50 ms of cooling time. Thermal damage profiles are shown as percentage of cell death.

Fig. 7

Average temperature within biopsy volume over time for each cooling scheme. The time axis is shown in log scale.

The model also indicates that laser biopsy thermal damage is primarily due to light scattered from the annular beam into the center causing increased absorption, heating, and damage of the tissue. Results for normal scattering (${{\mu }_s}^{\prime }=6{\mathrm{cm}}^{-1}$), half scattering (${{\mu }_s}^{\prime }=3{\mathrm{cm}}^{-1}$), and zero scattering are compared with all cases including pre- and post-cooling (Fig. 8). By reducing scattering strength, depth of viable tissue ($\mathrm{\Omega }<1$) increased from $140\mu \mathrm{m}$ for normal scattering to $260\mu \mathrm{m}$ for half scattering, and $620\mu \mathrm{m}$ for zero scattering, suggesting that optical clearing methods that reduce tissue scattering could reduce microbiopsy damage. Optical clearing methods that could be used in vivo include mechanical optical clearing devices⁴¹ or chemicals such as glycerol which provide reversable optical clearing.⁴²

Fig. 8

Computed fluence (top row) and tissue damage percentage (bottom row) of microbiopsies collected with pre- and post-cooling with different tissue scattering properties. (a) Normal scattering, ${{\mu }_s}^{\prime }=6{\mathrm{cm}}^{-1}$, (b) half scattering, ${{\mu }_s}^{\prime }=3{\mathrm{cm}}^{-1}$, and (c) no scattering.

3.4.

Microbiopsy Harvest and Ablation Volume Measurements

A pulse energy of 1.3 J was required to provide consistent and repeatable biopsy harvest when cooling was used. Fast camera videos show the microbiopsies ejecting from the center of the annular ablation region and reaching the coverslip after 100 to $150\mu \mathrm{s}$ [Fig. 9(a)]. OCT images show that the dimensions of ablation craters and harvested biopsies are on the order of 200 to $500\mu \mathrm{m}$ [Figs. 9(b) and 9(d)]. Ejected microbiopsy volumes collected on the glass slides increased with increasing beam diameter and volumes ranged from 0.02 to $0.08{\mathrm{mm}}^3$ [Fig. 9(e)].

Fig. 9

Biopsy harvest results: (a) select frames recorded by a fast camera ($\mathrm{50,000}\text{frames}/\mathrm{s}$) of microbiopsy harvest. (b) OCT image of a harvested microbiopsy. (c) OCT image of a bulk tissue crater at the microbiopsy site. (d) 3D reconstruction of a microbiopsy volume. E) Mean microbiopsy volumes for two pulse energies and two beam diameters (estimated inner annulus diameters of 420 and $530\mu \mathrm{m}$ for small and large, respectively).

3.5.

Virtual H&E Images

An example of the H&E virtual staining process for bulk porcine skin is shown in Fig. 10. AO fluorescence highlights the cell nuclei [Fig. 10(a)] while SR101 fluorescence highlights the cytosol and collagen [Fig. 10(b)]. The two channels were merged [Fig. 10(c)] and false colored to the purple and pink colors of H&E stains [Fig. 10(d)]. Example virtual H&E images for bulk porcine skin and kidney tissue, ablation craters, and harvested biopsies, are shown in Figs. 11 and 12, respectively.

Fig. 10

Virtual H&E imaging of bulk skin sample. (a) AO fluorescence; (b) SR101 fluorescence; (c) merged image; and (d) false colored virtual H&E image.

Fig. 11

Examples of skin virtual H&E images, including (a) undamaged bulk skin, (b) microbiopsy crater in bulk skin, and (c) and (d) representative microbiopsies. $\text{Scale bars}=100\mu \mathrm{m}$.

Fig. 12

Example kidney virtual H&E images, including (a) undamaged bulk kidney, (b) microbiopsy crater in bulk kidney, and (c) and (d) representative microbiopsies. $\text{Scale bars}=100\mu \mathrm{m}$.

The virtual H&E image of the skin ablation crater [Fig. 11(b)] suggests that the entire epidermis and a portion of the dermis was removed. The crater shows increased signal in the SR101 channel at the left and right edges of the crater in the epidermis, indicating thermal damage. No clear changes were observed in the image along the base of the crater in the dermis. While our model did not focus on predicting or preventing thermal damage to the bulk tissue near the microbiopsy crater, one might expect to see significant thermal damage along the full length of the crater due to high predicted temperature increase. The absence of observable damage along the base of the crater indicates less light scattering than predicted into the center of the annulus and/or limited thermal exposure due to pre- and post-cooling. Some skin microbiopsies appear to be entirely epidermis [Fig. 11(c)], while others appear to contain dermis and epidermis [Fig. 11(d)]. The biopsy edge in the epidermis shows increased signal in the SR101 channel [Fig. 11(c)].

The virtual H&E image of the kidney crater [Fig. 12(b)] shows increased signal in the SR101 channel along the full length of the ablation crater. Due to the weak mechanical properties of kidney, many smaller tissue fragments were ejected in addition to the larger microbiopsy. Some kidney microbiopsies largely retained their shape [Fig. 12(c)], while others were fragmented [Fig. 12(d)]. Portions of the biopsies appeared to have an increase in signal in the SR101 channel such as at the edges [Fig. 12(c)].