2.1.

Development of Confocal Mosaicking Microscopy Imaging Protocol in Laboratory Settings

Rajadhyaksha et al.⁹ described the use of 5% acetic acid and confocal cross-polarized microscopy to image thick skin excisions containing BCC and SCC obtained from Mohs surgery. Acetic acid was used as a contrast agent to “aceto-whiten” or brighten nuclei in reflectance mode.^9,11,12,15 However, the surrounding normal dermis also appears bright. Consequently, large and densely nucleated tumors, such as superficial and nodular BCCs (Fig. 1), were readily identified, whereas small and sparsely nucleated tumors, such as micronodular and thin strands or cords of infiltrative and sclerosing BCC tumors, were lost in the background (Fig. 2). Later studies that were conducted on larger sample sizes using RCM-acetic acid approach also reported similar limitations.²⁷

Fig. 1

Acetic acid stained reflectance confocal mosaic (a) shows nodular BCCs (red arrows) that can be readily identified at $2\times$ magnification and compares well to (b) the corresponding H&E frozen histopathology. Nests of tumor (red arrows) within the dermis, the epidermal margin (green arrows) and hair follicles (yellow arrows) are seen. Scale bars: $\mathrm{A}=1\mathrm{mm}$. Figure reproduced with permission, courtesy of Wiley.¹⁵

Fig. 2

Acetic acid stained reflectance confocal mosaic (a) shows nodular and infiltrative BCCs and (b) corresponding H&E stained frozen histology. Tiny strands of infiltrative BCC (red boxed area) are difficult to identify and differentiate from thin bright strands of collagen fibres (yellow dotted circles) at low magnification in contrast to the nodular BCC (red arrow) that appears distinct (a). (c) Digital zooming to $30\times$ magnifications is often required to identify the tiny strands of infiltrative BCC. The epidermal margin (green arrows) is seen. Scale bars: $\mathrm{A}=1\mathrm{mm}$; $\mathrm{C}=150\mu \mathrm{m}$. Figure reproduced with permission, courtesy of Wiley.¹⁵

Therefore, RCM-acetic acid methodology was, subsequently, replaced by FCM imaging again using specific nuclear contrast agents but in fluorescence, such as acridine orange, methylene blue, and toluidine blue;^13,14 methylene blue and toluidine blue primarily stain nucleic acid (DNA and RNA) and are frequently used as conventional histology stains.¹³ Acridine orange interacts with the DNA and RNA of the nuclei, without staining collagen tissue.¹⁴ Thus, when stained tissue is imaged in fluorescence mode, the nuclei appear bright with minimal to weak interfering signal from collagen, imparting good contrast for the detection of even small tumor nests or cords^13,14 (Fig. 3).

Fig. 3

Low tumor volume of infiltrative BCC can be readily detected on a wide-field fluorescent confocal mosaic. Acridine orange stained fluorescence confocal mosaic (a) shows a small focus of residual infiltrative BCC (dashed red box) as bright strands that can be identified within the reticular dermis even at this low magnification, corresponding to $2\times$ magnification on (b) H&E-stained frozen histopathology. However, it can sometimes be challenging to differentiate these particularly thin tumor strands (four to eight cells thick; dashed red box) from the surrounding inflammatory infiltrate (green dotted circles) and may require digital zooming to higher magnification ($4\times$ to $10\times$). (c) Submosaic shows digital zooming of the area under the red dashed box from mosaic “A” necessary to appreciate pleomorphic nuclei and enlarged nuclear cytoplasmic ratio of the tumor strands (red arrowheads) and differentiate them from bright smaller monomorphic inflammatory cells (green dotted circles). Epidermis (green arrow) and hair follicle (yellow arrow) are seen. On a separate note, the dark bands that are seen in this particular mosaic (but not the others shown in this paper) are due to incomplete correction for the illumination fall-off (vignetting) that occurs across each image and incomplete stitching along the edges of images. This was merely due to the developmental state of the stitching algorithm and software at the time. (Further development led to a second algorithm that produces much more seamless appearing mosaics, as seen in the other figures.) Scale bars: A, $\mathrm{C}=1\mathrm{mm}$.

Of these contrast agents, acridine orange has been, to date, more widely tested in skin, both for optimal use and staining^14,33 and, subsequently, for use in Mohs surgery.^27–31 Gareau et al.¹⁴ demonstrated the FCM with acridine orange (FCM-acridine orange) method to be able to detect both large (and densely nucleated) as well as small (and sparsely nucleated) residual BCC tumors in Mohs surgical margins. Bini et al.³³ determined the optimal staining approach, in terms of concentration and immersion time, for maximizing contrast and detectability. Fifty discarded and thawed Mohs skin excisions were stained with a variable concentration (0.3 to 1.0 mM) of acridine orange for various periods of time (5 to 120 s). A concentration of 0.6 mM and immersion for 20 s was discovered to be optimal for detection of both large and small tumors and was subsequently validated on 50 more discarded samples.¹⁴

A modified research breadboard version of a Vivascope 2000 was used for imaging in fluorescence mode, with illumination of 488 nm, a water immersion $30\times$, and 0.9 numerical aperture objective lens.¹⁴ Specialized devices and methods were developed for mounting and flattening Mohs surgical skin excision.^12,14 This system provided optical sectioning up to $\sim 1\mu \mathrm{m}$ and high resolution images of $\sim 0.25\mu \mathrm{m}$. Individual images were acquired and up to $36\times 36$ images stitched and processed to create mosaics that displayed field-of-view (FoV) of up to $12\mathrm{mm}\times 12\mathrm{mm}$, which is equivalent to a $2\times$ view in standard light microscopy [Figs. 4(a) and 4(b)].¹⁴ Further zooms at high magnifications of $10\times$ to $30\times$ were feasible when necessary [Figs. 5(a) and 5(b)].¹⁴ The procedure for creating a single mosaic required $\sim 9\min$. All subtypes of BCC (superficial, nodular, micronodular, infiltrative, and sclerosing) were imaged. Comparison of fluorescence mosaics to histology showed that this method detected both large and small tumors at low magnification [$2\times$; Figs. 4(a) and 4(b)]. Although thin strands of infiltrative BCC could be readily identified in the collagenous background (in contrast to imaging with acetic acid) at low magnification ($2\times$), it can sometimes be difficult to detect them in the background of bright nucleated structures, especially inflammatory cells, and required investigation at higher magnification ($4\times$ to $10\times$; Fig. 3). Similar challenge is also encountered on histology. Not only did Gareau et al.¹⁴ demonstrate that acridine orange in fluorescence is a better contrast agent than acetic acid in reflectance for the detection of small and thin strands of tumors, but they also found that acridine orange does not interfere with subsequent histopathological evaluation. Thus, this study proved the feasibility of using FCM-acridine orange for the rapid detection of both small and large BCC tumors.

Fig. 4

Acridine orange stained fluorescence confocal mosaic of a micronodular BCC (a) that equates well to $2\times$ view on standard H&E stained histology section (b). Small and tiny nodules or nests of tumor (red dotted area) can be identified at this magnification. Epidermis (green arrow), hair follicle (yellow arrow), sebaceous glands (green asterisk), dermal collagen (yellow asterisk) are seen. Scale bar: A = 2mm. Figure reproduced from Gareau et al. “Confocal mosaicing microscopy in Mohs skin excisions: feasibility of rapid surgical pathology.”¹⁴

Fig. 5

Acridine orange stained fluorescence confocal submosaic (a) obtained by digitally zooming ($6\times$ magnification) in the mosaic (Fig. 4(a), dashed boxed area) to appreciate morphological features of micronodular BCC tumor (red dotted area) such as nuclear pleomorphism, increased nuclear density, and clefting. Epidermis (green arrow), hair follicle (yellow arrow), sebaceous glands (green asterisk), dermal collagen (yellow asterisk) are seen. This corresponds well with H&E stained frozen histopathology (b). A = 0.5 mm. Figure reproduced from Gareau et al. “Confocal mosaicing microscopy in Mohs skin excisions: feasibility of rapid surgical pathology.”¹⁴

To validate their results, the same group later performed a semiquantitative preclinical study using the same experimental set-up on discarded BCC Mohs excisions.³⁴ For this study, 45 confocal mosaics were blindly evaluated for the presence (or absence) of BCC tumors by two clinicians: a senior Mohs surgeon with several years’ experience and expertise in interpreting confocal images and a novice Mohs fellow with limited few months of experience. BCCs (all subtypes) were detected with an overall high sensitivity and specificity of 96.6% and 89.2%, respectively. The negative predictive values (NPV) were 98.4% for the senior (expert) Mohs surgeon and 91.9% for the novice. A relatively lower NPV reported by the novice reader was the result of lack of experience in evaluating confocal images. Several interesting outcomes were drawn from this study.³⁴ Detection of superficial BCC was most challenging in the absence of adjacent epidermal margins, which is probably related to the incomplete or uneven flattening of the epidermis in the tissue mounting device. Flattening of the edges of Mohs surgical excisions is critically necessary when preparing frozen sections to allow viewing of the complete epidermal margins (Fig. 6).³⁴

Fig. 6

(a) Acridine orange stained fluorescence confocal submosaic shows a small bright focus (dotted circle) of superficial BCC along the epidermal margin (green arrow) raising suspicion for residual tumor and (b) corresponding H&E-stained frozen histopathology at $4\times$ magnification. Digital zooming to higher magnification ($30\times$) reveals the subtle but characteristic features of (c) superficial BCC on confocal image as well as on (d) H&E-stained frozen histopathology. This figure demonstrates that small foci of superficial BCC may be sometimes missed on confocal imaging either due to their subtle morphological features or uneven flattening of the tissue. Scale bar: $\mathrm{A}=1\mathrm{mm}$. Figure reproduced with permission, courtesy of Wiley.³⁴

Identification of infiltrative BCC was relatively easy when present as multiple tumor clusters, however, remained relatively challenging when there were only few clusters, as reported previously (Fig. 3).¹⁴

This study also demonstrated the existence of a learning curve for reading FCM images as well as the bias of previous history of training and experience on reading accuracy. The expert Mohs surgeon was highly sensitive (98.9%), but historical experiences also led to somewhat conservative and cautious reading for absence of residual tumor, resulting in relative lower specificity (83.3%). By comparison, the novice was not as accurate in detecting the presence of tumor and was relatively less sensitive (94.4%) but the shorter historical experience led to relatively bolder reading for absence of tumor, resulting in relatively higher specificity (95.0%). Some technical challenges were also highlighted such as the need for improving mosaicking speed while maintaining image quality, as well as for building inexpensive and user-friendly equipment for increased acceptance of this technology in clinical realms. (While the current speed of mosaicking is certainly faster by more than $\sim 2\times$ than the time taken in current surgical practice to prepare frozen sections, this time does not include the time taken for sample preparation, which includes acridine orange-staining followed by mounting and flattening of tissue. Moreover, multiple excisions (2 or more) are often taken during a single Mohs surgical procedure. For these reasons, further improvement in mosaicking speed will help toward wider acceptance and adoption of the CMM approach.

2.2.

Confocal Mosaicking Microscopy Imaging from Bench to Bedside

The laboratory work, along with the advent of a second generation commercialized smaller and faster microscope (VivaScope^R 2500, Caliber Imaging and Diagnostics, Rochester, New York), paved the path for FCM-acridine orange testing in actual clinical and Mohs surgical settings directly on fresh excised tissue (instead of discarded frozen-thawed tissues that was used in the laboratory development work).

Bennassar et al.²⁸ were the first group to advance FCM to testing and use at the bedside. Fresh tissues from shave biopsies were obtained from two whitish facial papules in two different patients with equivocal dermoscopic diagnosis. The biopsies were stained with acridine orange and imaged with FCM. A diagnosis of intradermal nevi (IDN) and infiltrating BCC, respectively, was reached from the FCM images within 170 s of tissue excision. Immediate point-of-care clinical management was feasible in both cases, based on the results of FCM imaging alone. In case of IDN, electrodesiccation of the wound base was performed without causing much harm to the tissue. For BCC, Mohs surgery was performed in the same setting, resulting in total tumor clearance. FCM diagnosis correlated well with their corresponding histopathology. This report, for the first time, demonstrated the feasibility of performing FCM-acridine orange mosaicking rapidly and reliably in an actual clinical setting for the evaluation of fresh ex vivo tissues.

Later, Bennàssar et al.²⁹ advanced the implementation and testing of FCM imaging in clinics by defining FCM criteria for all subtypes of BCC. This was a quintessential step to improve the reproducibility of BCC diagnosis and its differentiation from surrounding dermis and adnexal structures (especially small hair follicles) that often pose diagnostic dilemma. Sixty-nine tumor samples from 66 patients were obtained as 1-mm strips from the first Mohs stage excision without disrupting the tissue margins for ulterior frozen section assessment. FCM mosaics were acquired and evaluated by the attending Mohs surgeon, experienced in analyzing both frozen sections and FCM images in the operating room in real time. Tissues were subsequently submitted for histopathological evaluation.

Broadly, the tumors were classified as either superficial BCC (sBCC), nodular (nBCC), or micronodular-infiltrative BCC (iBCC) (Fig. 7 and Table 1). Eight morphological-fluorescent criteria were defined for BCC, including presence of fluorescence, clear demarcation, nuclear crowding, clefting, nuclear pleomorphism, palisading, increased nuclear cytoplasmic (N/C) ratio, and the presence of stromal reaction (Fig. 7 and Table 1). Of these, hyperfluorescent aggregates with nuclear crowding (91%), palisading (96%), and nuclear pleomorphism (100%) were identified in all subtypes of BCCs. However, certain criteria were more pronounced than others, which aided in subtyping of BCC. For example, sBCCs appeared poorly defined with hyperfluorescent structures (tumoral nuclei) sparsely distributed along the basal layer of the epidermis with surrounding clefting [Fig. 7 and Table 1]. Stromal reaction and inflammation was less prominent in sBCCs. On the contrary, nBCCs appeared as well-demarcated nodules with crowded hyperfluorescent pleomorphic nuclei, peripheral palisading, and clefting [Fig. 7 and Table 1]. Unlike sBCCs, a significant stromal “starry sky” pattern was associated with nBCCs. Likewise, iBCCs were described as nests and strands of hyperfluorescent pleomorphic nuclei infiltrating dermis with variable nuclear crowding, palisading, and clefting [Fig. 7 and Table 1]. Strong stromal and inflammatory reaction was the most prominent feature of iBCCs [Fig. 7 and Table 1]. FCM criteria, such as clefting, palisading, and “starry sky” stroma, had high specificity for BCC detection and were useful in differentiating BCC nests from the surrounding adnexal structures. The authors stressed the importance of “starry sky” stroma that should initiate a thorough search of BCC foci in the absence of tumor nests. The outcome of the FCM mosaics analysis was later compared with the results of H&E slides. The comparisons were independently evaluated by a dermatopathologist and demonstrated high overall concordance ($\kappa =0.9$) in classifying BCC subtypes.²⁹ In order to further validate the aforementioned FCM criteria, the same FCM images were also independently evaluated by two dermatologists. These criteria were validated with $\kappa$ values greater than 0.7 for most criteria in both independent observers.

Fig. 7

Fluorescent confocal microscopy criteria for various subtypes of BCCs. (a, c, e) Submosaics show characteristic morphological features of BCCs: presence of fluorescence, nuclear pleomorphism, increased N/C ratio, and palisading in (a) superficial BCC, (c) nodular BCC, and (e) infiltrative BCC. The tumor islands (red arrows) are well demarcated in (c) nodular BCC and (e) infiltrative BCC than the (a) superficial BCC, whereas clefting (red arrowheads) appears more prominent in superficial BCC (a) and nodular BCC (c). (e) “Starry sky” (yellow asterisk) appearance of the tumor stroma is pronounced in infiltrative BCC. (b, d, f) And corresponding H&E-stained frozen histopathology ($10\times$). Scale bars: $\mathrm{A}=200\mu \mathrm{m}$; C & $\mathrm{E}=600\mu \mathrm{m}$.

Table 1

FCM criteria for BCC and their subtypes.29

Recently, Bennàssar et al.³⁰ conducted another study, in the Mohs surgical setting, where—for the first time—true surgically excised fresh tissue margins from 80 BCC lesions (with or without tumors) were prospectively imaged with FCM. The goals of this study were (a) to evaluate the sensitivity and specificity of ex vivo FCM imaging in detecting residual BCC in Mohs tissue excisions and (b) to calculate the time invested up to the diagnosis for both FCM and frozen sections. The attending Mohs surgeon evaluated mosaics in the operating room in real time for the presence or absence of BCC, especially at the lateral margins. The reading of the FCM mosaics also mimicked the usual frozen section slide-reading process during Mohs surgery, as described previously. Later, frozen pathology was independently evaluated by a dermatopathologist in a blinded manner.

The results of both FCM and frozen section were compared and an overall high sensitivity and specificity of 88% and 99%, respectively, of detecting BCC was found. Their result³⁰ is on par with and validates the results of the initial studies.³⁴ Micronodular/infiltrating BCC was the main histological subtype in this study. The diagnosis of BCC was missed in 10 cases, which were later confirmed as infiltrating BCC with thin strands (two to three cell layers) of tumor cells on frozen section. This misdiagnosis was attributed to either the lack of sufficient training for the Mohs readers or to the differences in the level of tissue section examined: surface imaging of true margin by FCM but deeper subsurface frozen sections. Only one false-positive margin was reported in this study, making FCM a highly specific tool for detection of tumor margins in fresh tissues. However, the main limitation of this study was low prevalence of positive margins for BCC, making the results of high positive predictive value and NPV of 98% and 97%, respectively, less reliable. This study also proved that FCM markedly reduces the time of tissue preparation and evaluation by almost two-thirds, in comparison to traditional method of frozen sections.

Following the initial reports from Barcelona, Longo et al.,³¹ published a case report of recurrent BCC that was subjected to FCM-acridine imaging for margin control using Mohs surgery. The lesion was excised under local anesthesia without surgical debulking and then the excised tissue was sliced into several smaller samples to evaluate margins. The time required to image all samples was about 9 min. The infiltrative type of BCC was clearly identified in the deeper margin and correlated well with corresponding frozen section diagnosis. In this particular case, the patient refused further surgery and was subsequently treated with intensity modulated radiation therapy. This study highlights the potential role of FCM mosaicking in the management of complex and recurrent tumors.

2.3.

Translation of Confocal Mosaicking Microscopy Guided Mohs Technology Platform to Other Perioperative Settings

2.4.

Current Limitations of Confocal Mosaicking Microscopy and Future Advancement Toward Clinical Translation

All of the studies to date provide compelling evidence of FCM’s translational potential to expedite ex vivo evaluation of fresh tissue for the detection of residual BCC in Mohs surgical margins and potentially other clinical applications. Nonetheless, these studies also highlighted limitations that will need to be overcome to integrate FCM into routine clinical workflow. Here, we list some of these limitations and their potential solutions.

2.4.3.

Learning Curve

Another reason for the lack of readers is the reluctance of trained clinicians to learn to interpret grayscale (black-and-white) confocal mosaics instead of more familiar purple and pink appearing images of histopathology. To address this barrier, an advanced multimodal-digital staining approach is being developed to recreate images that appear similar to conventional purple-and-pink appearance of H&E images.^48,49 This approach simultaneously images tissue in both fluorescence (FCM) and reflectance (RCM) modes. Images acquired in fluorescence contrast highlight only nuclei and can thus be digitally colored purple (this simulates hematoxylin staining component of H&E), whereas those acquired in reflectance contrast highlight only cellular cytoplasm and dermal morphology and can thus be digitally colored pink (this simulates eosin staining component of H&E Figs. 8 and 9).

Fig. 8

DSCMs of normal skin tissue closely mimicking H&E stained tissue histology. Mosaic of normal skin tissue in (a) fluorescence mode, (b) in reflectance mode, and (c) digital H&E (DHE) image created by overlapping fluorescence and reflectance channels. Nucleated structures, such as epidermis (red arrows) and hair follicle (yellow arrow), appear bright in fluorescence mode (a) and dull-grey on reflectance mode (b). Adipocytes (blue asterixis) appear darker and well defined in the fluorescence mode than in reflectance mode. By contrast, dermal collagen (green asterixis) appears bright in both the modes. (d) The H&E image ($2\times$) compares well with the DHE image (c). Scale bars: A, $\mathrm{B}=2\mathrm{mm}$.

Fig. 9

DSCMs of BCCs closely mimicking H&E stained tissue histology. Submosaics shows (a, c) infiltrative BCC and nodular BCC in fluorescence mode, (b, d) in reflectance mode, and (e, g) digital H&E (DHE) images created by overlapping fluorescence and reflectance channels, respectively. (a, e) Nucleated structures, such as hair follicle (yellow arrows) and tumor cords and nodules (red arrows), appear bright in fluorescence mode and (b, f) are not visible (yellow and red asterixis) on reflectance mode. By contrast, dermal collagen (green arrows) appears more pronounced in the reflectance mode bright in (b, f; green asterixis). Corresponding H&E images of (f) infiltrative and (h) nodular BCC at $10\times$ magnification compares well with the DHE images (d and g, respectively). Scale bars: A, $\mathrm{B}=0.3\mathrm{mm}$; C, $\mathrm{D}=0.15\mathrm{mm}$.

Overlaying the two on each other produces purple and pink colored confocal mosaics. Gareau et al.⁴⁹ developed and tested this approach on 35 BCC excisions, stained with acridine orange, from Mohs surgery in a preclinical study and showed good correlation with the H&E stained frozen pathology. Recently, the same group conducted a retrospective study to assess the potential of digitally stained confocal mosaics (DSCMs) for screening SCC and BCC in Mohs excisions.⁵⁰ For this, 133 DSCMs from 64 Mohs discarded tissue excisions were evaluated by three Mohs surgeons in a blinded manner, before and after brief training. They demonstrated average respective sensitivities and specificities of detecting NMSC of 90% and 79% before training and 99% and 93% after training. There was also a significant increase in the interobserver agreement from 72% pretraining to 98% ($\kappa =0.97$; $P=0.04$) post-training. They also noted a marked reduction of training time to 5 min as compared to traditional grayscale confocal microscopy that requires a training period minimum of 30 min with at least 2 months of confocal imaging experience.^34,48 However, the appearance of the mosaics remains somewhat variable in quality, due to artifacts from acridine orange staining in the presence of tissue heterogeneity, mainly saturation in nuclei and/or residue in the dermis. This approach still needs optimization and larger scale validation studies before it could actually be implemented into regular clinical workflow.⁴⁹