2.1.

Compact sLSM Probe Design

Figure 1 shows a schematic of the compact sLSM probe. In the illumination path, a laser (B073H21NJP, Sunshine electronics; wavelength = 638 nm; output power = 1.0 W) was used as the light source. In our previous work investigating the optimal wavelength for sLSM, we found that a wavelength around 600 nm provides a good balance between the imaging depth and resolution. Therefore, the wavelength of 638 nm was used in the compact sLSM probe.¹⁶ Light from the laser was coupled into a multimode fiber (02-511, Edmund optics; fiber diameter = 2.0 mm; core diameter = 1.96 mm; NA = 0.5). In between the laser and fiber, a rotating diffuser (DG10-1500, Thorlabs; grit = 1500; rotation speed = 600 rpm) was used to reduce the speckle noise. An illumination slit (2-2.5-3+HS+M-0.5, National Aperture; $\text{width}=2.5\mu \mathrm{m}$; length = 3.0 mm) was used to shape the spatially incoherent light from the fiber into a narrow line, which was imaged on the tissue to generate the light sheet illumination.¹⁷ The light passing through the slit was collimated by an achromatic doublet (49-926, Edmund optics; $f=15\mathrm{mm}$). The collimated light passed through a three-dimensional (3D)-printed rectangular aperture and focused by another achromatic doublet (49-927, Edmund optics; $f=22.5\mathrm{mm}$). The illumination beam was folded by a dielectric mirror (87-367, Edmund optics; $\text{reflectivity}>99\%$) and transmitted through a plastic imaging window [material = polymethyl methacrylate (PMMA); $\text{thickness}=250\mu \mathrm{m}$; flat] to form a light sheet on the tissue. An immersion medium (GenTeal Tears Gel, Alcon; refractive index = 1.34) was used to fill the space between the plastic imaging window and the tissue.

Fig. 1

Schematic of the compact sLSM probe.

In the detection path, scattered light from the tissue was collected by a custom objective lens ($f=14.9\mathrm{mm}$; NA = 0.25) and focused by a tube lens (66-896, Edmund optics; multi-element micro video lens; $f=35\mathrm{mm}$; F/2) onto a monochromatic CMOS sensor (daA3840-45um, Basler; $\mathrm{3,840}\times \mathrm{2,160}\text{pixels}$; $\text{pixel size}=2.0\mu \mathrm{m}$; imaging speed = 45 fps). The distal end of the bi-concave lens was in contact with the immersion medium. The detection optics had a magnification of 2.35, resulting in the pixel size of $0.85\mu \mathrm{m}$ and the FOV of $3.26\mathrm{mm}\times 1.84\mathrm{mm}$ on the tissue $xy$-plane.

Mechanical holders were custom designed and fabricated using a 3D printer (Form 3+, Formlabs) and biocompatible material (biomed black resin, Formlabs). Most of the optical elements were passively aligned using tight fit with the mechanical holders to ensure concentricity, whereas some optical components were actively aligned.

2.2.

Custom Objective Lens Design

Previously, several custom objective lenses were successfully developed for microscopic imaging of tissues. These objective lenses achieved a high NA (0.5 to 1.0) while either reducing the overall lens size for endoscopy applications^19–21 or increasing the FOV for large-area imaging.²² However, these lenses needed to use multiple elements ($>5$) in order to achieve a high NA. In sLSM, the use of a single wavelength and small NA ($<0.3$) could make it possible to use fewer elements for the custom objective lens. In the compact sLSM probe, we designed a custom objective lens composed of two elements, an aspheric convex lens (material = PMMA) and an aspheric bi-concave lens (material = PMMA) as shown in Fig. 2(a). The custom objective lens was designed using OpticsStudio (Zemax).

Fig. 2

(a) Schematic of the custom objective lens, (b) wavefront error performance as a function of field, and (c) wavefront error performance as a function of defocus.

The aspheric convex lens had an aspheric surface on the proximal side (farther from the tissue) and a nearly flat surface on the distal side (closer to the tissue). The lens stop was located on the proximal side. The bi-concave lens had an aspheric surface on the proximal side and a spherical surface on the distal side. The two aspheric lenses were fabricated by diamond turning. The convex lens had the clear aperture of 7.4 mm in diameter and center thickness of 5.0 mm. The bi-concave lens had the clear aperture of 4.6 mm in diameter and center thickness of 2.9 mm. The distance between the two lens elements was designed as 3.0 mm. During the alignment procedure, the distance between the two lenses was actively adjusted to minimize the full-width-half-maximum (FWHM) of the measured point-spread function (PSF). A stainless-steel washer (thickness = 0.05 mm, outer diameter = 4.7 mm, inner diameter = 3.2 mm) was placed on the distal flange of the bi-convex lens to block stray light. During the optimization process, both the working distance and the diameter of the bi-concave lens were regulated to prevent mechanical interference between the lens and tissue when the objective lens is angled at 45 deg relative to the tissue surface.

The effective focal length and NA of the objective lens were 14.9 mm and 0.25, respectively. The theoretical resolution at 638 nm was $1.32\mu \mathrm{m}$. The lens was designed to have a diffraction-limited FOV of $\pm 1.0\mathrm{mm}$ in the tissue space, as shown in the root-mean-square (RMS) wavefront error plot as a function of field [Fig. 2(b)]. In our previous sLSM work imaging anal tissues with water-immersion objective lenses and immersion medium with the index of 1.34, sLSM was able to visualize nuclear features over the 100 to $200\mu \mathrm{m}$ thickness of anal epithelial tissues.²³ We evaluated the RMS wavefront error of the custom objective lens when imaging epithelial tissues with an index higher than 1.34 at the detection angle of 45 deg. For an assumed tissue index of 1.357,²⁴ the custom objective lens is expected to have a diffraction-limited performance over the imaging depth of $200\mu \mathrm{m}$. For a tissue index of 1.393,²⁵ the custom objective lens is expected to have a diffraction-limited performance over the depth of $100\mu \mathrm{m}$, sufficient for evaluating nuclear morphologic differences between normal/low-grade lesions and high-grade lesions. The field curvature of the lens was controlled to have diffraction-limited performance within the defocus range of $\pm 2.5\mu \mathrm{m}$, corresponding to the theoretical thickness of the light sheet illumination, as shown in Fig. 2(c). A Monte-Carlo tolerance analysis was conducted on the objective lens based on the lens manufacturing tolerances.²⁶ The tolerance analysis showed that more than 72% of the objective lenses would provide diffraction-limited performance.

2.3.

Light Sheet Thickness Simulation

For sLSM imaging, it is critical to maintain a thin light sheet thickness over a large depth range (along the $y$-axis in Fig. 1). This can be achieved by using a small NA for illumination. However, with such a small illumination NA, a highly scattering object in the tissue can cast shadow on cellular structures underneath it. This shadow effect manifests as dark vertical stripes in sLSM images.²⁷ Previously, we demonstrated that a rectangular aperture in the illumination path with an aspect ratio greater than 3 can effectively reduce the shadow artifacts while maintaining a thin light sheet thickness.¹⁷ In the compact sLSM probe, the aperture length along the $x$-axis and width along the $y$-axis were set to 9.0 and 2.9 mm, respectively, to provide the illumination NA of 0.2 along the $x$-axis and 0.06 along the $z$-axis (after 90 deg reflection by the fold mirror). This aperture arrangement resulted in hourglass-shape light sheet illumination with a theoretical thickness of $5.0\mu \mathrm{m}$ and depth of focus (DOF) of $206.2\mu \mathrm{m}$.

The choice of lenses in the illumination optics was initially based on first-order geometric optics. Therefore, the illumination performance degradation caused by aberrations was not initially considered. Given that the PMMA window and immersion medium would introduce aberrations, it was necessary to ensure the focusing light sheet meets the target thickness requirement. Due to the larger NA along the $x$-axis and smaller NA along the $z$-axis, aberrations were mainly expected along the $x$-axis, as shown in the simulated Huygens PSF [Fig. 3(a)]. The light sheet illumination, determined by the convolution between the PSF and image of the slit (longer along the $x$-axis), therefore maintains a narrow thickness, as shown in Fig. 3(a). We evaluated the light sheet thickness for different fields (along the $x$-axis) and different defocus locations (along the $y$-axis) to ensure the light sheet thickness was consistently small across the entire imaging FOV and depth range. The light sheet thickness for different fields ($x=0$ and $\pm 1.0\mathrm{mm}$) and defocus locations ($y=0$ and $\pm 103.1\mu \mathrm{m}$) was simulated using the extended diffraction image analysis function in OpticsStudio, which considered both diffraction and aberration effects of the illumination optics. As shown in Fig. 3(b), the theoretical light sheet thickness was maintained small, $\sim 5.0\mu \mathrm{m}$, even though aberrations were present along the $x$-axis.

Fig. 3

(a) Simulated Huygens PSFs and light sheet illumination for the on-axis field at the focus and at $103.1\text{-}\mu \mathrm{m}$ defocus and (b) light sheet thickness for different fields and defocus locations.

2.4.

Custom Objective Lens Performance Test

Optical performance of the objective lens was evaluated by imaging a UASF resolution target and a custom multi-pinhole target. The UASF resolution target image was used to identify the smallest resolvable line period. The image of the custom multi-pinhole target was used to measure the FWHM of the PSF across the entire FOV. The custom multi-pinhole target had pinholes periodically spaced along horizontal and vertical directions ($\text{pinhole diameter}=0.67\mu \mathrm{m}$; $\text{period}=20\mu \mathrm{m}$). The multi-pinhole target was fabricated by etching the pinhole pattern on a glass substrate (thickness = 1.5 mm) deposited with reflective chrome coating. We conducted a simulation of the effect of the pinhole size by convolving the theoretical PSF of the objective lens with the pinhole geometry and calculating the FWHM of the convolved profile. The simulation results showed that only a small amount of measurement error (5%) was expected when measuring the expected resolution of $1.32\mu \mathrm{m}$. The period of the pinhole pattern, $20\mu \mathrm{m}$, was set significantly larger than the pinhole diameter so that the image of each pinhole does not affect the image of neighboring pinholes.

In order to evaluate the performance of the objective lens alone, the detection optics were assembled and mounted vertically facing downward. The USAF resolution target or the multi-pinhole target was placed horizontally under the objective lens and was trans-illuminated from below the target. Light transmitted through the target was collected by the objective lens and imaged by the tube lens onto the CMOS sensor.

2.5.

Probe Performance Test

The axial and lateral resolution of the sLSM probe was measured simultaneously by imaging a custom resolution target with a periodic line pattern (line $\text{width}=0.5\mu \mathrm{m}$; $\text{period}=30\mu \mathrm{m}$) we previously developed.¹⁶ The line pattern was made on a chrome-coated glass substrate (thickness = 1 mm) by photolithography and etching. The target was placed horizontally with the reflective lines parallel to the $u$-axis and facing upward. Only a small portion of each line was illuminated by the light sheet, with the illumination width determined by the thickness of the light sheet. The lines reflected the illumination light toward the detection optics, producing an image that showed multiple partial lines. In the sLSM image of the line pattern, the FWHM of the line height along the $y$-axis was measured as the axial resolution of sLSM. The FWHM of the line width was measured as the line-spread function (LSF) or the lateral resolution of the detection optics. Therefore, both the lateral and axial resolution was measured from a single sLSM image of the line pattern at a given depth. The resolution target was translated along the $v$-axis with a step size of $10\mu \mathrm{m}$ to measure the resolution at different imaging depths. The exposure time of the CMOS sensor was adaptively adjusted to avoid the signal saturation. Index-matching gel (GenTeal Tears Gel, Alcon) was placed between the resolution target and the probe as an immersion medium.

Tissue imaging performance was tested by imaging formalin-fixed human anal tissues ex vivo. The sLSM images acquired by the compact probe were compared with the images previously acquired with the bench sLSM device¹⁷ using off-the-shelf objective lenses (N10XW-PF, Nikon; $f=20\mathrm{mm}$; $\mathrm{NA}=0.3$; water immersion) and H&E-stained histologic images. The epithelial side of the tissue was placed upward. The index-matching gel was placed between the tissue and probe. The protocol for imaging formalin-fixed human anal tissues ex vivo was reviewed and approved by the Stanford University and University of Arizona Internal Review Boards (IRBs). Informed consent was waived by the IRBs for this study imaging discarded fixed tissues. Cellular features visualized in sLSM images were analyzed in comparison with the features shown in the corresponding histologic images by an expert pathologist (EY).

3.1.

Compact sLSM Probe

A photograph of the compact sLSM probe is shown in Fig. 4. Overall dimensions of the prototype were 4 cm in height and width and 10 cm in length. The illumination power on the tissue was 0.14 mW. The exposure time was set at $\sim 0.5\mathrm{ms}$, resulting in a real-time imaging speed with a frame rate of 45 fps, which is the highest frame rate the CMOS sensor supported. The material cost of the compact sLSM probe was low, <$2000. The unit price of the custom objective lens was relatively high, $1200, due to the cost associated with diamond turning and anti-reflection coating. The objective lens cost can be reduced significantly by using injection molding in the future.

Fig. 4

Photo of the sLSM probe.

3.2.

Custom Objective Lens Performance

A representative image of the USAF resolution target is shown in Fig. 5(a). The smallest line pattern resolvable by the custom objective lens was group 9, element 1 ($\text{linewidth}=0.97\mu \mathrm{m}$) along both the horizontal and vertical directions, as shown in the inset of Fig. 5(a). However, the USAF image exhibited obvious side lobes surrounding the line pattern, indicating the presence of aberrations. A representative image of the custom multi-pinhole target is shown in Fig. 5(b). The measured PSF FWHM of the custom objective lens as a function of field [Fig. 5(c)] shows that the custom objective lens provides the resolution of 1.65 to $1.97\mu \mathrm{m}$ across the FOV of $\pm 1.0\mathrm{mm}$.

Fig. 5

(a) USAF target image, (b) custom multi-pinhole target image, and (c) measured resolution as a function of field of the custom objective lens.

3.3.

Probe Resolution Measurement

Figure 6(a) shows a representative image of the custom line pattern target captured by the compact sLSM probe. The measured LSF FWHM, also the measured lateral resolution, was $1.90\pm 0.07\mu \mathrm{m}$ over the theoretical illumination DOF of $206.2\mu \mathrm{m}$. The axial resolution at the center of the imaging depth range was measured to be $5.2\mu \mathrm{m}$. Within the theoretical illumination DOF of $206.2\mu \mathrm{m}$, the light sheet thickness, also the axial resolution, was smaller than $5.6\mu \mathrm{m}$.

Fig. 6

(a) Custom resolution target image and (b) measured axial resolution as a function of depth.

3.4.

Tissue Imaging Performance

Figure 7 shows representative sLSM and histologic images of fixed normal anal squamous mucosa. The image acquired by the compact sLSM probe [Fig. 7(a)] visualizes tissue details over a width of 2 mm. The epithelium and stroma are clearly delineated. A magnified view of the same image [Fig. 7(b)] visualizes a honeycomb pattern of squamous epithelial cells with hypo-reflective nuclei (arrows) and hyper-reflective cytoplasm and cell membranes, in a similar manner to the image obtained with the bench sLSM device [Fig. 7(c)] and with a similar nuclear size and spacing to the histologic image [Fig. 7(d)].

Fig. 7

sLSM images of normal anal squamous mucosa obtained with (a, b) the compact sLSM probe and (c) bench sLSM device, and (d) histologic image of the same tissue. Arrows, nuclei.

Figure 8 shows representative sLSM and histologic images of fixed low-grade squamous intraepithelial lesion (LSIL). The image obtained with the compact sLSM probe [Fig. 8(a)] clearly distinguishes the layer of parakeratosis from the underlying epithelium, which is thickened compared to normal squamous epithelium. The higher-magnification view of the compact sLSM probe image [Fig. 8(b)] shows that the honeycomb pattern of the squamous epithelial cells is still retained but the cellular features have more variability in size and shape than normal squamous epithelium. A similar trend is shown in the image obtained with the bench sLSM device [Fig. 8(c)] and the histologic image [Fig. 8(d)].

Fig. 8

sLSM images of LSIL obtained with (a, b) the compact sLSM probe and (c) bench sLSM device, and (d) histologic image of the same tissue.

Figure 9 shows representative sLSM and histologic images of fixed high-grade squamous intraepithelial lesion (HSIL). A large-area sLSM image obtained with the compact sLSM probe [Fig. 9(a)] shows the lack of clear distinction between the epithelium and stroma. The increased epithelial thickness and stronger scattering in HSIL might have degraded the resolution at the base of the epithelium more in HSIL than in normal squamous epithelium, making the epithelium-to-stroma interface less distinctive. A higher-magnification view [Fig. 9(b)] and its inset visualize dense and irregular arrangement of small hypo-reflective nuclei (arrows) with a varying size and shape. The trend of crowded nuclei is also shown in the image obtained with the bench sLSM device [Fig. 9(c)] and the histologic image [Fig. 9(d)].

Fig. 9

sLSM images of HSIL obtained with (a, b) the compact sLSM probe and (c) bench sLSM device, and (d) histologic image of the same tissue. Arrows, nuclei.