2.1.

Excised Tissue Collection

Skin excisions from Mohs surgeries were obtained from the Mohs Surgery Unit in the Dermatology Service at Memorial Sloan-Kettering Cancer Center (MSKCC) under Human Tissue Utilization Committee (HTUC) approval. For each excision, the Mohs technician usually prepares three to four frozen sections, with each section being approximately 5 to $6\mathrm{\mu }\mathrm{m}$ thick. After the Mohs surgeon has obtained and evaluated the frozen sections for histology, the remnants of each excision are usually discarded. All discarded excisions were collected at the end of the day, after the completion of Mohs surgeries. Thus, our research did not compromise the standard Mohs surgical and pathology procedures, and did not affect routine patient care. Fresh excisions of normal skin were also collected. Such excisions of redundant normal skin, also referred to as “dog ears” because of their shape, is sometimes performed to optimize surgical wound closure and maximize cosmetic-reconstructive outcome.

2.2.

Excised Tissue Preparation

The collected excision specimens were frozen in the embedding medium that is used in preparing histology sections. Each excision was thawed, removed from the embedding medium, washed with lukewarm water for several minutes, placed in a vial of potassium phosphate monobasic-sodium hydroxide buffer solution (PBS) and then refrigerated. The excisions were living, as confirmed by viability staining with fluorescein diacetate and fluorescence-imaging of the subsequent fluorescein-labeled cytoplasm (unpublished data).

The confocal imaging of each excision was performed in two stages: before acetowhitening (control) and after. For the control, we imaged the excision from the PBS solution. The excision was then acetowhitened by soaking in acetic acid for a set time and again imaged. A total of 125 excisions were imaged in this study. The tested concentrations of acetic acid were 10, 5, 3, 2, and 1%, with soaking times of 5, 3, 2, and $1\min$ , and $30\mathrm{s}$ . For each combination of concentration and time, 5 excisions were imaged to test repeatability, resulting in the total of 125 excisions.

The skin excisions were typically of size 10 to $20\mathrm{mm}$ and thickness $2\mathrm{mm}$ , such that the tissue volume was no more than $1\mathrm{ml}$ (considering that 60 to 70% of the content of tissue is water). The excisions were soaked in 15 ml of acetic acid. Thus, for the acetowhitening, the ratio of the volume of acetic acid to the volume of tissue was 15:1. The low concentrations of acetic acid that were used do not affect the compliance of the tissue and do not affect the subsequent sectioning and staining for frozen histology. This was verified in the initial studies.^{4, 5} (Low-concentration acetic acid, whenever used for acetowhitening on skin or mucosa, can be rinsed out of the tissue prior to preparation of frozen histology. Note, however, that at much higher concentrations such as 50%, acetic acid fixes tissue. High-concentration acetic acid is, in fact, used as a fixative for several tissues. In bone tissue, for example, acetic acid facilitates the removal of calcified material, which is required prior to fixation.)

2.3.

Confocal Microscope Instrumentation

The imaging was performed with a modified laboratory breadboard version of a commercial reflectance confocal microscope (VivaScope 2000, Lucid Inc., Rochester, New York) that is designed specifically for imaging fresh tissue excisions ex vivo. The design and instrumentation details were reported earlier.^{4, 5} The illumination is with a near-IR diode laser at wavelength of $830\mathrm{nm}$ and low power of about 1 to $5\mathrm{mW}$ on the tissue. The objective lens is a $30\times$ , 0.9 numerical aperture (NA) water immersion (StableView, Lucid Inc., Rochester, New York) that was custom-made to image through a 1-mm-thick glass window. Standard 1-mm-thick microscope cover slides were used as the glass window (part number 12-550A, Fisher Scientific, Pittsburgh, Pennsylvania). The objective lens provides a field of view of $430\mathrm{\mu }\mathrm{m}$ . The nominal diffraction-limited lateral resolution is $0.6\mathrm{\mu }\mathrm{m}$ and optical section thickness is $2\mathrm{\mu }\mathrm{m}$ . The detection pinhole diameter is $150\mathrm{\mu }\mathrm{m}$ , which corresponds to five times the nominal lateral resolution. The illumination laser beam is linearly polarized ( $p$ -state) and detection is through a cross-polarizer or analyzer ( $s$ -state). By rotation of a quarter-wave plate behind the objective lens, the illumination can be switched between brightfield and cross-polarized. Images are captured with a frame grabber (Imaq 1408, National Instruments Inc., Austin, Texas).

As in previous studies, ^{4, 5, 23, 24} water as an immersion medium was often substituted with water-based gels such as Aquasonic 100 ultrasound gel and ${\mathrm{Suave}}^{\circledR }$ hair gel. These gels have refractive indices of 1.36 and 1.34, respectively, that are reasonably close to that (1.33) of water. Furthermore, the gels have a higher degree of viscosity and can therefore be placed on the inverted objective lens. By comparison to water, the use of these gels (particularly, ${\mathrm{Suave}}^{\circledR }$ ) appeared not to degrade the image quality of the superficial tissue layer. The immersion gel is placed between the objective lens and the coverglass window, while the skin excision is kept hydrated with PBS. Once the excision was placed in the fixture, properly positioned and oriented, and the illumination power adjusted, confocal images and mosaics were obtained.

The surface of the excision was imaged. Because the frozen excision undergoes thawing, rinsing, acetowhitening, and mounting in the fixture, distortions in the imaged surface may occur due to the mechanical compliance of tissue. Thus, a close but not exact correspondence was expected between the confocal mosaics of the imaged surface and the sections of frozen histology that were prepared by the Mohs technician.

2.4.

Mechanical Tissue Fixture

Fresh skin excisions from Mohs surgery are of complex and awkward 3-D shape and size, and are living and hydrated, and therefore not easy to mount for microscopy. To enable confocal mosaicing, to observe large fields of view covering large areas of the excision, we designed a mechanical fixture to hold the surgical excision and precisely control its flattening, tilt, sag, and stability (Fig. 1 ). The design of the fixture parallels the functionality of commercially available cryostats (for example, the CM1510 from Leica Microsystems Inc.) that are used for positioning and orienting Mohs surgical excisions prior to preparing frozen histology sections.

Fig. 1

Mechanical fixture to mount surgical skin excisions and enable mosaicing over large areas of tissue. The fixture encases the inverted objective lens.

The tissue fixture (Fig. 1) accommodates the inverted objective lens configuration of the confocal reflectance microscope that uses the custom $30\times$ , $0.9\mathrm{NA}$ water immersion objective lens, which images through a 1-mm-thick cover slide. Creating mosaics of a large area requires the fixture to gently flatten and accurately position and orient the lower surface of the excision that is to be imaged. This requires gentle and uniform pressure on the excision, with variable tilt and adjustable objective lens-to-excision distance. (In close analogy, proper positioning and orientation of the skin excision in the cryostat is a crucial task for Mohs technicians when preparing frozen histology sections.) Once the excision is placed on the glass window, a piston is used to apply pressure on the top to flatten the lower surface. The 1-mm-thick coverglass slide is stiff enough to prevent the lower surface (i.e., object plane to be imaged) from sagging under the pressure of the piston. (Standard $170\text{-}\mathrm{\mu }\mathrm{m}$ -thick microscope coverslips that are normally used with standard objective lenses do not work as well. The coverslips tend to sag or break such that capturing a 2-D sequence of images to create a mosaic over a large contiguous area is not possible.) The tilt and orientation of the window relative to the objective lens is adjusted via spring-loaded thumbscrews. (This is similar to the kinematic mount that is commonly used in optics.)

To ensure uniform pressure and uniform flattening of the lower surface, a mechanically compliant material is necessary between the piston and the irregular corrugated shape of the excision. We use cold-form agarose gel disks, prepared with concentration of 3% and thickness 2 to $3\mathrm{mm}$ . Agarose powder of $0.9\mathrm{g}$ is dissolved in 30 ml of hot water, placed in circular Delrin-plastic molds of $1.1875\mathrm{in.}$ diam and cooled to room temperature. Under the piston-induced pressure, the gel conforms to the shape of the excision, which then becomes embedded within the disk. (This is analogous to the embedding of skin excisions in the cryostat for preparing Mohs histology.) Agarose gel concentrations of 1 to 5% with disk thickness of 1 to $5\mathrm{mm}$ were tested. With concentrations of 1 and 2%, the disks were too fragile and broke under the pressure applied by the piston. With concentrations of 4 and 5%, the disks were too stiff and did not conform to the excision. Agarose gel disk concentration of 3% conformed to the excision without breaking. Disks with a thickness of $1\mathrm{mm}$ broke while those with thicknesses of 2 to $5\mathrm{mm}$ did not. This fixture encompasses the objective lens and is mounted onto the 3-D $xyz$ translation stage of the reflectance confocal microscope with user-controlled height adjustment. The system is robust and enables accurate and repeatable imaging and mosaicing.

2.5.

Confocal Mosaicing

Skin excisions from Mohs surgery are typically 10 to $20\mathrm{mm}$ in extent, whereas the confocal field of view with the current $30\times$ objective lens is $0.43\mathrm{mm}$ . (More recently available lenses with lower magnification such as Nikon $20\times ∕0.75\mathrm{NA}$ , Olympus $20\times ∕0.95\mathrm{NA}$ , and Lucid $10\times ∕0.8\mathrm{NA}$ provide increased fields of view of 0.75 to $1.00\mathrm{mm}$ .) To observe large areas of an excision, a rastered 2-D sequence of images is captured and stitched in software to create a mosaic that displays a much larger field of view. The rastered sequence of images is captured by translating the tissue fixture with stepper-motors-driven linear $XY$ stages (Hayden Inc., Stamford, Connecticut). Individual images are captured using Vivascope software and mosaics are subsequently created with MATLAB software (Version 7.0.1, MathWorks, Natick, Massachusetts). The software routine consists of the following steps: (1) normalization of images with a reference background image, to correct for illumination curvature and vignetting, and (2) cropping of images at the four edges by 5 to 15%; the amount of cropping depends on the size of the excision, which necessitates a trade-off between amount of overlap and mosaicing speed; (3) merging or concatenation of the images into a single mosaic; (4) removal of residual shading bands at the edges of each image with a reverse polynomial fit; (5) Fourier-transform-based removal of residual dark stitching lines; and (6) scaling down to resolution and pixelation that is equivalent to that in $2\times \mathrm{magnified}$ views of histology. (The mosaicing software and technical support is freely available to anyone who may be interested.) Currently, up to $36\times 36$ images are stitched together to display a field of view of up to $12\times 12\mathrm{mm}$ , which simulates light microscope magnification of $2\times$ . Mohs surgeons routinely use an objective lens with $2\times$ magnification to quickly examine large areas of frozen histology sections during surgery. A continuous step-and-capture routine requires about $5\min$ for $36\times 36$ frames.

When routinely examining frozen histology sections with a standard brightfield microscope, using a $2\times$ , $0.08\mathrm{NA}$ objective lens and white-light illumination (wavelength $\lambda \sim 0.5\mathrm{\mu }\mathrm{m}$ ), the Mohs surgeon observes a field of view of $\sim 10\mathrm{mm}$ with a resolution of $\sim 4\mathrm{\mu }\mathrm{m}$ . The observed field of view consists of $\sim$ 2500×2500 pixels, assuming 1 pixel per optical lateral resolution element. Thus, the confocal mosaics must be displayed with at least $2500\times 2500$ pixels to ensure the display resolution does not degrade the optical lateral resolution. We observe mosaics on a 21-in. PC monitor (ViewSonic Corporation, Professional Series $\mathrm{P}220f$ , Walnut, California) using visualization software (PhotoImpact, MicroTek, or IPLab Spectrum version 3.0, Scanalytics Inc.). This software enables digital zooming to increase the display magnification to 4 or $10\times$ , which mimics the Mohs surgeon’s occasional ad hoc use of 4 or $10\times$ objective lenses to examine histology with higher magnifications.

Each image consists of $640\times 480$ pixels, is 8-bit gray scale, and requires $\sim 1∕4\mathrm{MB}$ of memory, such that a mosaic of $36\times 36$ images at full resolution consists of more than $\mathrm{23,000}\times \mathrm{17,000}$ pixels and requires $\sim 300\mathrm{MB}$ of memory. The mosaic is scaled down using bilinear interpolation, to make the final display equivalent to a histology-like $2\times$ view. This display consists of $\sim 2500\times 2500$ pixels and requires $\sim 4\mathrm{MB}$ .

2.6.

Preparation of Histology

The frozen histology sections of each excision, which were prepared by routine hematoxylin and eosin (H&E) staining during Mohs surgery, were available for comparison to the confocal mosaics. The confocal images, mosaics, and submosaics were visually correlated to the histopathology at standard microscope magnifications (2 to $40\times$ ). As explained earlier, a close but not exact correlation to histology is expected.

3.1.

Acetowhitening: Optimum Concentration and Soaking Time

Figure 2 shows that a range of acetic acid concentrations at varying soaking times is effective for sufficiently brightening nuclei in acetowhitened excisions. Compared to controls, concentrations of 10 and 5% require only $30\mathrm{s}$ , whereas 3, 2, and 1% concentrations require 1, 2, and $5\min$ , respectively. Rapid brightening within $1\min$ is possible for concentrations of 3% and higher. The brightening of nuclei, as observed only on the exposed surface of the excision, is independent of the size of the excision for the acetic acid-to-skin excision volume ratio of 15:1.

Fig. 2

Soaking time required at varying acetic acid concentrations. Concentrations of 10 and 5% require only $30\mathrm{s}$ , whereas 3, 2, and 1% concentrations require 1, 2, and $5\min$ , respectively. Rapid brightening within $1\min$ is possible for concentrations of 3% and higher. The brightening of nuclei on the surface is independent of the size of the excision for the acetic acid-to-skin excision volume ratio of 15:1. This observation of rapid acetowhitening is limited to only the exposed surface of the excision that was imaged.

3.2.

Confocal Mosaics: Visual Comparison to Histology

Confocal mosaics of Mohs excisions show good visual correlation to frozen histology. The observed morphology of BCCs in terms of shape, size, and location of nests compares well to that in the corresponding histology. The atypical morphology of nuclei such as polymorphism (varying shapes and sizes), crowding (increased density), and palisading is clearly visualized in the mosaics.

Figure 3 shows a confocal mosaic of a micronodular BCC and the corresponding histology. Both the mosaic and the histology show nests of micronodular BCCs within the deeper reticular dermis [ $*$ in Figs. 3A, 3B]. The epidermal margin of the excision along with the dermo-epidermal junction can be differentiated (arrows). The acetowhitening clearly shows enhanced BCC-to-dermis contrast.

Fig. 3

(A) Mosaic shows $15\times 15$ frames stitched together in software to show an equivalent $2\times \mathrm{magnified}$ view of the specimen. Within the dermis the clusters of micronodular BCCs are differentiated in the deep dermis because of brightening by the acetowhitening technique $(*)$ . The epidermis is seen along the superficial peripheral edge (arrows). Acetowhitening: 2% for $2\min$ . Scale bar $1\mathrm{mm}$ . (B) Corresponding frozen H&E stained histology showing correlation to the confocal mosaic. Scale bar $1\mathrm{mm}$ .

Figure 4 shows a submosaic of a superficial BCC at higher magnification of $\sim 4\times$ . Bright nuclear detail is more clearly visible in both the epidermis (arrows) and in the BCC within the underlying deeper dermis $(*)$ . The individual layers of the epidermis, namely, stratum corneum, granular, spinous, and basal cell layers, are seen.

Fig. 4

(A) Confocal submosaic of a superficial BCC shows $8\times 6$ frames stitched together to show an equivalent $\sim 4\times$ magnified view. Bright nuclei are more clearly seen in epidermis along the peripheral edge (arrows) and the BCC $(*)$ in the underlying deep dermis. Acetowhitening: 3% for $1\min$ . Scale bar $500\mathrm{\mu }\mathrm{m}$ . (B) Corresponding frozen H&E stained histology. Scale bar $500\mathrm{\mu }\mathrm{m}$ .

Figure 5 shows a micronodular BCC in a smaller submosaic at higher magnification of $10\times$ . Small nests of BCCs are seen in the deeper dermis $(*)$ with bright nuclei. Again, the epidermis is seen with bright nuclei throughout (arrows). Within the dermis, nuclear detail in hair follicles near the dermo-epidermal junction is differentiated [“h” in Figs. 5A, 5B] along with individual inflammatory cells and collagen throughout the papillary and reticular dermis. The thin collagen fibers in the superficial papillary dermis and the thicker collagen bundles in the deeper reticular dermis are seen in the mosaics, similar to that seen in the corresponding histology.

Fig. 5

(A) Confocal submosaic of micronodular BCC shown in $4\times 4$ frames stitched together to show an equivalent $10\times$ magnified view. The submosaic shows small nest of BCCs with bright nuclei $(*)$ , the epidermis along the edge (arrows), and bright nuclear detail in hair follicles (h). Bright nuclei of inflammatory cells are also seen infiltrating the dermis. Acetowhitening: 3% for $1\min$ . Scale bar $250\mathrm{\mu }\mathrm{m}$ . (B) Corresponding frozen H&E stained histology. Scale bar $250\mathrm{\mu }\mathrm{m}$ .

Tumor nests consisting of BCC cells mixed with inflammatory infiltrative cells are easily visualized in the confocal mosaics. Large nodular and small micronodular BCCs are easily detected in the mosaics and correlated to the frozen histology. However, infiltrative or sclerosing BCCs are difficult to detect because they consist of very small nests or thin strands of nuclei that tend to remain hidden within the bright collagen in the dermis.

3.3.

Comparison of Cross-polarized versus Brightfield Contrast

A comparison of cross-polarized versus brightfield mosaics is shown in Figs. 6 and 7 . The cross-polarized images show better overall contrast. Images obtained by cross-polarization suppress the light backscatter from the cytoplasm and dermis,⁴ thereby increasing the contrast of nuclear detail relative to the dermis. Nuclei in the epidermis [arrows in Figs. 6A, 6B] and in the nests of BCCs [ $*$ in Figs. 6A, 6B, 7A, 7B] appear with enhanced contrast. Collagen within the superficial papillary dermis darkens consistently in cross-polarized images due to its homogeneous morphology. However, collagen does not consistently darken in the deeper reticular dermis due to the increased heterogeneity in terms of inherently thicker, denser, and more randomized arrangement of fibers and bundles. The enhancement of BCC-to-dermis contrast is thus consistent within the papillary dermis but not within the reticular dermis. The enhancement of contrast is better appreciated in individual images and submosaics at high magnification, rather than in the large mosaics at lower magnification. Cross-polarized mosaics do not necessarily improve the detectability of BCCs in the deeper dermis, compared to brightfield mosaics.

Fig. 6

Comparison of (A) brightfield versus (B) cross-polarized submosaics. Under cross-polarization, the brightness of the cytoplasm and the collagen is suppressed. Nuclei in the epidermis and the nests of BCCs appear with enhanced contrast. Scale bar $100\mathrm{\mu }\mathrm{m}$ .

Fig. 7

Comparison of (A) brightfield versus (B) cross-polarized submosaics. Under cross-polarization, the brightness of the collagen is suppressed. Nuclei in the nests of BCCs appear with enhanced contrast. Scale bar $250\mathrm{\mu }\mathrm{m}$ .