1. Introduction

Nonmelanoma skin cancer (NMSC) is the most common cancer diagnosed in the United States, with estimates as high as 5.4 million in 2012 [1], more than all other cancers combined. NMSCs are typically diagnosed on a shave or punch biopsy [2,3]. The excised tissue is then sent to a pathology lab for formalin-fixed paraffin-embedded (FFPE) processing, sectioned into 4µm sections, and then stained with hematoxylin and eosin (H&E). A pathologist then reads the H&E-stained slide to determine if the sample is cancerous and, if so, what the subtype is. Due to the low-risk nature of NMSC, these samples are not typically prioritized, and processing delays can add weeks to the time for diagnosis. Following a positive diagnosis, treatments for NMSC include curettage with electrodessication [4], standard surgical excision with postoperative transverse-sectioned histopathology [5], and surgical excision with intraoperative en face frozen section histology (Mohs surgery). For small lesions or those in cosmetically insensitive areas, a conventional excision can remove the tumor and a large margin of healthy tissue around it [6]. However, NMSC frequently occurs on sun-exposed tissues such as the face [7], where large margins can lead to complicated surgical reconstruction or unacceptable cosmetic outcomes. In these cases, Mohs surgery removes the tumor in stages using intraoperative frozen sections histopathology to guide excision is preferred [7]. However, while faster than FFPE processing, frozen section processing is slow, typically 30 minutes for smaller specimens and potentially hours for larger specimens that must be dissected into multiple sections. Consequently, clinics can only treat a limited number of patients per day and costs are high.

Advanced fluorescence imaging modalities such as microscopy with UV surface excitation [8,9], confocal microscopy [10] and two photon fluorescence microscope (TPFM) [11,12] can provide rapid, real-time histological imaging of fresh tissues. These emerging methods could eliminate the days to weeks delay for FFPE processing as well as accelerate margin evaluation during Mohs surgery. However, these methods have limited imaging depth in highly scattering skin tissues and typically can only image the tissue surface or a relatively shallow region of the subsurface. While multiple z-planes can be used to image curved surfaces, this process is much slower and cannot handle tissues that fold over or are at steep edges, as in the case of many beveled surgical excisions. Finally, uneven contact can lead to inconsistent image quality, complicating image interpretation or obscuring diagnostic regions. Compounding these challenges, rapid diagnosis necessarily requires rapid tissue preparation with minimal manual intervention to enable high clinical throughput.

The need to develop means to flatten surgical margins rapidly has motivated several previous studies. These have typically relied on applying high pressure to press the tissue against the cover glass to flatten the tissue surface. Various tissue holders have been designed to apply an even and appropriate amount of pressure to the tissue. Patel et. al utilized a solid polycarbonate piston coated with agarose gel to press the tissue against the cover glass [13]. While the agarose gel could reduce the uneven application of pressure on the tissue, the solid piston could not provide entirely even pressure around tissue of varying thickness, especially the edges. Previous TPFM surgical imaging studies [12,14] utilized flexible histology foam that deforms around the tissue, providing relatively even pressure across the tissue. While providing relatively tunable pressure by changing the amount of foam stacking between the top plate and cover glass, the histology foam solution is cumbersome and ultimately limited in how much pressure can be applied, particularly if tissue thickness is highly variable. Abeytunge. et al. designed a more advanced flexible water bladder-based holder [15]. The flexible surface of the bladder hugs the tissue and provides even downward pressure. An updated version of the holder [16] improved its usability with an elastic membrane controlled by a water-filled chamber. A piston mounted on the chamber increases or decreases the water chamber volume, which expands or sinks the elastic membrane. However, the pressure is difficult to tune due to the incompressible nature of water, while force is uniformly downwards, which limits utility on some irregularly shaped specimens and is still challenging to use with many bowel-shaped surgical excisions that become thinner at the edges.

A common limitation of existing specimen holder designs is that they are unsuitable for tall samples with small contact areas, such as bread-loafed shave biopsy samples, which comprise a significant fraction of dermatological biopsies [17]. Due to the high aspect ratio of these samples, the direct downward force from any of these methods will tip over or crumple the sample, resulting in uneven contact and incomplete evaluation of the sample. Unfortunately, evaluation of biopsies in the breadloaf orientation, which produces tall, thin samples is extremely widely used for NMSC diagnosis. Thus, a reliable and repeatable mounting solution for shave biopsies is crucial for adapting TPFM to the histology workflow.

This work presents an improved piston-based specimen holder that enables imaging of a wider range of sample sizes and shapes, including small biopsies as well as large surgical excisions. The new holder system is designed to provide even pressure distribution across the specimen consistently. Mounting pressure is applied by a stretchable membrane controlled by an air chamber that easily attaches/detaches from the cover glass holder. Due to the compressible nature of air, the mounting pressure can be precisely controlled by the volume of the air being added into the chamber. Support fixtures are also developed to facilitate the pressure distribution from the rubber membrane to image specimens with small contact areas relative to thickness, such as bisected shave skin biopsies. We demonstrate this method by imaging diagnostic biopsy specimens and bowel-shaped fresh Mohs excisions in the clinic immediately after excision, demonstrating the potential for rapid diagnosis and margin evaluation.

2. Material and method

2.1 Piston based holder design

The piston design is derived from the surgical foam-based holder in previous works [12,18] and previous published holder designs [15,16]. The self-contained balloon chamber replaces the solid back plate in our previous design to maintain ease of rapid assembly, as shown in Fig. 1(A). The cover glass (0.035 inch, D263) is glued to the sample holder that fits into a motorized stage (Thorlabs Inc. MLS203). A flexible rubber membrane (McMaster-Carr 85995K12, durometer 40A, 0.006” thickness) is pressed against a rubber gasket (McMaster-Carr 1170N104) by a retaining ring to create a tight seal for the balloon chamber (Fig. 1(B)). We selected flexible soft membranes that can conform to the tissue at lower air pressure, reducing the chances of edge lifting while fitting around tissue supports. A thin strip of paper wick is placed between the rubber membrane and the tissue. The thin paper wicks out fluid during compression, reducing the water gap between tissue and coverglass. A manually operated air valve and a conventional 100 ml graduated syringe is used to control the pressure applied to the piston (typically ∼10 PSI), as shown in Fig. 1 C. During the inflation process, the balloon apex contacts the tissue first, causing an initial pressure hot point at the top, resulting in tissue edge lifting from the coverglass. Therefore, we place the balloon membrane as close to the cover glass as possible to reduce the curve of the balloon when the apex first contacts the tissue, in turn reducing lifting.

 

Fig. 1. Design of the inflatable piston. A: Components of the holder in exploded view and in cross-section. B: Conceptual diagram of the piston compressing tissue to the cover glass corresponding to fabricated piston components. Crucially, the rubber membrane is separated from the tissue by a thin paper wick that prevents a water-tight seal from forming between the glass and rubber that would trap fluid. C: Photograph of the holder with a conventional 100 mL graduated syringe for inflating the piston.

Download Full Size | PDF

2.2 Biopsy tissue support

Due to the controllable nature of our piston holder, we implemented support structures to redistribute the pressure from the piston balloon. For shave biopsies, which are typically evaluated in breadloaf orientation with small contact area relative to tissue height, we developed a supporting disk with two cutout channels to hold bisected shave samples as shown in Fig. 2(A). Thin samples with a small contact area (<1mm thickness) could use the side of the channel as support to stand up during the mounting phase. The support structure redistributes the downward pressure force to lateral force for thin specimens, as shown in Case 1 in Fig. 2(B). For specimens with relatively large contact areas capable of standing on the cutting surfaces independently, tissue support structure could prevent tissue from being crushed, as shown in Case 2 in Fig. 2(B), creating two mini pistons shaped by the support. Membrane deformation is controlled with the volume added via the syringe and adjusted depending on the tissue shape. Higher air volume is added (35ml, ∼13 PSI) to achieve greater membrane deformation to achieve Case 1. Case 2 requires lower pressure with 30ml air added to achieve around 11 PSI for less membrane deformation. Furthermore, the larger channel width allows more deformation. Thus, counterintuitively thinner sample needs wider channel and taller support to achieve Case 1, and thicker samples need narrower closely matching width and height support to achieve Case 2.

 

Fig. 2. Tissue support redirects piston pressure for different specimen types. A) Photography of tissue support for shave biopsies. B) Illustration of tissue support for shave biopsies. For thin shave samples, tissue support redirects the membrane to push the sample down and laterally to prevent tissue tipping (Case 1). For thicker samples, the tissue support reshapes the rubber membrane into two small pistons to prevent tissue from being crashed (Case 2).

Download Full Size | PDF

2.3 Mohs tissue support

Tissue supports are also used to redistribute mounting pressure for Mohs specimens and reduce pressure hotspots. Downward pressure is redirected with tissue support to locations needed the most (tissue margin) by the tissue support, as shown in Fig. 5(A). By blocking the tissue movement towards the gap of the whole mount excision, the tissue block re-directs the compression force radially and maintains the “Pacman” opening for orientation.

2.4 Fresh skin specimen collection

Diagnostic shave biopsy specimens and Mohs excisions were obtained and imaged under protocols approved by the Research Subjects Review Board at the University of Rochester Medical Center. In all cases, tissues were rapidly imaged in a histology laboratory while waiting in the queue for processing to avoid delays in treatment. As TPFM is a nondestructive modality that can image tissue specimens without altering subsequent histology [12], the RSRB waived the need for consent to image specimens so long as there was no patient contact. However, consent to the image was obtained to image the specimens in Fig. 2 and 3 to enable the collection of demographic and treatment data as part of other studies.

2.5 Shave biopsy imaging

Biopsy specimens were collected during routine treatment procedures and bisected lengthwise to expose a cross-sectional view of the shave by an attending physician. Following bisection, they were labeled with a solution of 40x (0.4% concentration of stock) SYBR Green (SGr) and 100 µg/ml Sulforhodamine 101 (SR101) dissolved in 70% ethanol for 3 minutes. They were then mounted in the piston holder, imaged for ∼ 60 seconds, returned to specimen jars, and placed back into the queue for processing.

2.6 Mohs margin imaging

First-stage Mohs specimens were similarly obtained following routine excision. Relaxing excisions were performed by the attending surgeon to flatten the tissue for frozen sectioning. The specimens were then labeled with a solution of 40x (0.4% concentration of stock) SYBR Green (SGr) and 100 µg/ml Sulforhodamine 101 (SR101) dissolved in 20% ethanol for 3 minutes. While higher concentrations of ethanol significantly improve penetration of the stain, alcohol prevents freezing of water during cryosectioning and must be removed before further processing. To speed up this process, lower concentrations of alcohol were used and specimens were rinsed in saline following both staining and again after imaging to remove residual alcohol.

2.7 TPFM system

Imaging was performed using a TPFM system described previously [12]. Briefly, 1040 nm light from a YLMO-2W (Menlo Systems) was scanned using a 16 KHz resonant scanner system (B-scope, Thorlabs, Inc.) and a 20x 0.7NA air objective with a correction collar (Olympus UCPLFLN20X). Excited fluorescence was detected using high dynamic range silicon photomultipliers (SiPMs) [19,20] operated in strip mode with synchronous position sampling to enable ∼ 32 MP/s peak and 25MP/s average mosaic imaging rate [21].

3. Result

3.1 Shave biopsies

Imaging of shave biopsy specimens in the same orientation as FFPE (standing upright on the thin side) is challenging without embedding in a supporting material. Without support, the balloon or foam compression collapsed the samples. To solve this problem we created 3D printed supports to allow these specimens to stand on their bisected face (Fig. 3). These supports allowed for compression to be applied either only from the top of the sample, simply compressing it down on the glass or from the top and side where the sample was pinched against the support and the balloon, again allowing flattening. For thinner specimens which end up containing mostly epidermis, the latter of the two was particularly important to forming good contact. These thinner specimens benefited greatly from being placed in printed chips with larger slot widths which allowed the balloon to conform to this gap, pinning the sample against the edge while also applying downward flattening force. For thicker specimens the opposite was true because of their ability to better hold their shape. These benefited greatly from being placed in a slot which was close to their thickness, allowing the balloon to push downward from above while being kept upright by the surrounding walls of the support. By printing a variety of different inserts, we were able to accommodate many oddly shaped specimens with excellent tissue contact. Total of 53 diagnostic biopsy specimens including both Basal Cell Carcinoma (BCC), Squamous Cell Carcinoma (SCC), Actinic Keratosis, and Sebaceous Keratosis.

 

Fig. 3. Mounting and imaging of shave biopsy specimens under piston compression. A: Bisected biopsy of a ∼5 × 1 × 5 mm shave specimens viewed from below (direction of microscope objective). B: After applying pressure from the piston holder, the same specimens in A show uniform contact with the glass without collapsing or folding the thin specimen. C: TPFM image of a biopsy specimen under ballon compression showing extensive squamous cell carcinoma. D: Corresponding paraffin section showing similar pathology.

Download Full Size | PDF

3.2 Fresh Mohs specimen imaging

We next explored mounting of fresh, diagnostic first stage Mohs excisions using the piston holder and 3D printed supports. Similar to the use of histology foam for compression, we found that applying uniform pressure on freshly excised, bowl-shaped Mohs excisions resulted in the ends flaring upwards and frequently folding the epidermis over, making evaluation of the dermal-epidermal junction impossible. To understand this effect, we performed live TPFM imaging of specimens while applying increasing pressure to watch how individual specimens deformed under compression. We observed that the opposing tension applied by the stiffer epidermis was responsible for the upwards motion and folding of the specimen edges under compression (Fig. 4) and that using shallow circumferential relaxing incisions could break transmission of this tension. While one standard technique is to use circular cuts following the epidermis [22,23], we found that a simpler superficial triangular cut into the epidermis approximately following the shape of the specimen was relatively effective and much quicker as only straight line cuts were required.

 

Fig. 4. Mohs tissue relaxation cut scheme and tissue support structure. A: Tissue support and relaxation cut scheme for fresh Mohs specimen imaging. The wedge support redirects pressure away to the tissue edge instead of inwards to the tissue gap. B: The relaxation cuts, illustrated with red lines, break the tension (pink arrows) that pulls the tissue edge away from the cover glass. Note that relaxation cuts were only through the epidermis and do not penetrate down to the diagnostic layers.

Download Full Size | PDF

 

Fig. 5. Mounting and imaging of Mohs surgical excision using piston compression. A: Top (user) view of a ‘Pacman’ Mohs excision with 3D printed support material inserted into the notch to support tissue under compression. B: The same specimen after compression is viewed from below (towards the objective). Fluid can be seen pushed out of the specimen by the compressive force and wicked away from the image surface. C: TPFM image of whole mount Mohs specimen.

Download Full Size | PDF

After selecting a method for relaxing incisions that is comparable to standard practice and could be rapidly applied to specimens, we next tested imaging of freshly excised, first stage Mohs excisions. As before, a plastic wedge was 3D printed to allow the sample to filet out more evenly (Fig. 5(A)) and a paper wick was used to allow fluid to escape after compression (Fig. 5(B)). As a result, we found that with sufficient relaxing excisions, most whole-mount first stage Mohs specimens could be flattened with all or nearly all of the dermal-epidermal junction in contact with the glass (Fig. 5(C)).

3.3 Foam vs. balloon based holder

To compare the piston-based compression to more conventional histology foam for compression, a first stage Mohs excision with the above relaxing incisions was prepared. The specimen was mounted in the previous specimen holder using high-stiffness open cell foam and then compressed to the maximum extent possible without bottoming out the foam. After imaging under foam compression (Fig. 6(A)), the specimen was remounted in the piston-based holder and imaged a second time (Fig. 6(C)). After imaging the tissue was submitted for normal frozen sections processing and then scanned (Fig. 6(E)). Comparison of a region of stiff epidermis showed limited visualization under foam (Fig. 6(B)) while the piston showed more of the margin (Fig. 6(D)), similar to frozen sections (Fig. 6(F)). Red arrows in Fig. 6 D&F point out the same hair follicle on epidermis. The surrounding epidermis features are slightly different due to cryosection when prepared FSA in F. Red arrow in Fig. 6(B) points to the similar region in pointed in Fig. 6(D), but the fair follicle I s missing due to uneven compression from surgical foam.

 

Fig. 6. Comparison between diagnostic Mohs margin imaged under compression by open cell foam (A&B), inflatable piston (C&D), and the final frozen sections cut after imaging (E&F). A: TPFM image of fresh Mohs specimen using open histology foam compression technique from previous study [12]. High resolution image link: https://imstore.circ.rochester.edu/papers/holder/FoamHolder/zstackRgb.html. B: The boxed region in A shows missing epidermis coverage due to uneven compression. Red arrow points the missing hair follicle visible in D. C: TPFM image of the same specimen in A mounted using piston based holder. High resolution image link: https://imstore.circ.rochester.edu/papers/holder/BalloonHolder/zstackRgb.html. D: The boxed region in C at the same location as B shows good epidermis coverage with hair follicle visible pointed out with arrow. E: Diagnostic FSA slide image of the same specimen in A&C. Features in the diagnostic FSA are slightly different from the true margin due to sectioning. F: The boxed region in E shows full epidermis coverage, and same hair follicle pointed out in D.

Download Full Size | PDF

3.4 Analysis of true Mohs margins using TPFM

It is widely recognized that cryosectioning results in the loss of substantial tissue, with one study finding an average loss of tissue of 285 microns from the true surgical margin [24] due to the loss of tissue inherent in cryosectioning a flat section from a curved surface. As Mohs specimens may be only 1-2 mm thick, this loss has the potential to change the margin status in a significant fraction of cases. Utilizing our holder and supplemental devices, we were able to assess both the true margin by imaging the actual excision surface before FSA and then the pseudo-margin formed during cryosectioning by imaging the frozen tissue block. While in most cases, FSA and TPFM agreed, during testing of the holder, we observed an instance where the true margin contained little or no tumor (Fig. 7(A)) while the subsequent FSA observed substantial tumor on the most superficial section (Fig. 7(C)). To confirm that this was not an artifact of TPFM imaging, following an additional stage of Mohs surgery to remove the positive area found on FSA, we imaged the remaining block (Fig. 7(E)). TPFM imaging confirmed the presence of BCC on the pseudo-margin seen in FSA (Fig. 7(D)) could be seen on TPFM of the residual block (Fig. 7(F)), but could not be seen on the TPFM of the true margin (Fig. 7(B)). Thus, we conclude that cryosectioning into tissue revealed extensive tumor that was not present on the true margin.

 

Fig. 7. True margin imaging vs pseudo-margin imaging after cryosectioning. A: TPFM image of a fresh Mohs excision at true surgical margin without sectioning. B: The boxed region in A shows no BCC suggesting the true margin is negative. C: The boxed region in B shows no BCC near the hair follicle. D: H&E stained FSA slide image of the same tissue in A after cryosectioning. Thus, the imaging plane is deeper (closer to the epidermis) with a higher possibility of seeing basal cell carcinoma (BCC). E: The boxed region in D (at a similar location as boxed region A). Unlike in B, The FSA slide shows clusters of infiltrative BCC. F: The boxed region in E showing BCC cluster near the same hair follicle in C. G: TPFM image of thawed frozen tissue in C. H: The boxed region in E showing same clusters of infiltrative BCC in D suggesting TPFM could produce same diagnosable images as FSA. I: The boxed region in H showing similar BCC cluster near the same hair follicle in F.

Download Full Size | PDF

4. Discussion

The design we present here incorporates features previously described in the literature, including our previous foam-based holder [12,14] and pioneering work on piston-based mounting [15,16]. To these, we added an air piston system with a flexible rubber membrane and wick that easily attaches/detaches from the cover glass holder. The mounting pressure is tunable by varying air volume added in the chamber. The process of balloon inflation allows adjusting pressure to suit different tissue types such as thin bisected shave biopsies samples and whole mount Mohs excision. To this, we add mechanical supports that can be placed around delicate biopsy or surgical specimens to guide compression along a preferred axis, replicating the mechanical support typically provided by embedding in paraffin or cryosectioning media. The membrane plays an important role in optimizing the system. The flexibility of the membrane allows us to shape and redistribute pressure with tissue support as shown in Fig2. However ultra-flexible membranes are generally fragile and easily punctured by tissue supports. We chose the thinnest nature rubber membrane that could withstand repeat inflate/deflate cycles. We picked the thin(0.006”) super-stretchable (Durometer 40A) natural rubber from McMaster-Carr (85995K12) for its durability and flexibility.

To the best of our knowledge, this work represents the first investigation of the mounting of fresh, diagnostic Mohs specimens imaged en face and before cryosectioning. In contrast to previous work, which has used discard frozen tissue blocks that have been cryosectioned flat, this work explores the contact of fresh with the true surgical margins intact and unmodified. Unlike thawed tissue, these specimens generally have uneven surfaces and steep bevels at the edges, making tissue mounting much more challenging than for discarded samples.

The importance of flat mounting of samples in both histopathology and fluorescence imaging is underappreciated. In conventional histopathology, specimens are frequently breadloafed, which produces a relatively flat surface that can be embedded in a paraffin block. To the extent that the surface is uneven, microtome sectioning removes tissue, resulting in a flat tissue section. Conversely, for en face histology, as used in Mohs surgery, a combination of relaxing excisions and compression during embedding is used to mount the tissue as flat as possible in order to minimize the amount of tissue lost during sectioning [25]. This is critical as the tissue lost during sectioning represents the true surgical margin, with excessive loss of tissue potentially changing the margin status if sectioning reveals an underlying tumor that was not actually on the true margin. The specimen holder we have developed enables the replication of both the breadloaf workflow on fresh tissue as well as the en face orientation with no loss of tissue.

While the design enables rapid and straightforward imaging of shave biopsy samples, embedding large or bowl-shaped, en face Mohs sections remains time-consuming and operator-dependent. Because the piston pushes down uniformly over the surface, careful relaxing excisions are critical to allowing the downward force to flatten the specimen. Without relaxing excisions, we found that the central downward force tends to cause the edges to fold upwards, preventing visualization of the critical dermal-epidermal junction where both superficial basal cell carcinoma and squamous cell carcinoma can be present. While shaping the piston to be as flat as possible enables the application of pressure to the entire sample at nearly the same time, the application of pressure to the central region is still translated outwards to the edges unless relaxing excisions are present to interrupt that transmission. Intriguingly, we observed that specimens where the relaxing excisions enabled flat mounting under the piston were also much easier to flatten during subsequent frozen section processing and could be sectioned with less loss of tissue during cryosectioning, indicating that the same effects are present during conventional frozen section processing. The design of shaped membranes or other means of applying pressure differentially to the sample may enable more reliable flattening and reduce the dependence on precise relaxing excisions. Alternatively, improved techniques for relaxing excisions would improve fresh tissue mounting during fluorescence imaging and cryosectioning during Mohs surgery.

Finally, during testing, we observed an instance where the piston-mounted tissue had no obvious cancer. However, following cryosectioning, both H&E staining and follow-up TPFM imaging of the new (deeper) pseudo-margin revealed extensive BCC. In all likelihood this represents a failure of conventional FSA to accurately diagnose the margin status due to the intrinsic and unavoidable loss of tissue present when sectioning a non-flat surface. The ability of fluorescent techniques to rapidly image the true margin surface without this loss of tissue may represent an unappreciated advantage that can improve the accuracy margin evaluation and reduce the false positive rate present when using FSA.

5. Conclusion

TPFM combined with the novel tissue mounting method provides a reliable fluorescence histology imaging solution that is compatible with downstream traditional histology. Our design can image both tall, thin biopsy specimens and bowel-shaped Mohs first stage excisions, enabling rapid diagnosis with a potentially lower false positive rate than FSA.