Training With Uncertain Annotations for Semantic Segmentation of Basal Cell Carcinoma From Full-Field OCT Images

Semantic segmentation of basal cell carcinoma (BCC) from full-field optical coherence tomography (FF-OCT) images of human skin has received considerable attention in medical imaging. However, it is challenging for dermatopathologists to annotate the training data due to OCT's lack of color specificity. Very often, they are uncertain about the correctness of the annotations they made. In practice, annotations fraught with uncertainty profoundly impact the effectiveness of model training and hence the performance of BCC segmentation. To address this issue, we propose an approach to model training with uncertain annotations. The proposed approach includes a data selection strategy to mitigate the uncertainty of training data, a class expansion to consider sebaceous gland and hair follicle as additional classes to enhance the performance of BCC segmentation, and a self-supervised pre-training procedure to improve the initial weights of the segmentation model parameters. Furthermore, we develop three post-processing techniques to reduce the impact of speckle noise and image discontinuities on BCC segmentation. The mean Dice score of BCC of our model reaches 0.503±0.003, which, to the best of our knowledge, is the best performance to date for semantic segmentation of BCC from FF-OCT images.

High-resolution full-field optical coherence tomography microscope for the evaluation of freshly excised skin specimens during Mohs surgery: A feasibility study

Histopathology for tumor margin assessment is time-consuming and expensive. High-resolution full-field optical coherence tomography (FF-OCT) images fresh tissues rapidly at cellular resolution and potentially facilitates evaluation. Here, we define FF-OCT features of normal and neoplastic skin lesions in fresh ex vivo tissues and assess its diagnostic accuracy for malignancies. For this, normal and neoplastic tissues were obtained from Mohs surgery, imaged using FF-OCT, and their features were described. Two expert OCT readers conducted a blinded analysis to evaluate their diagnostic accuracies, using histopathology as the ground truth. A convolutional neural network was built to distinguish and outline normal structures and tumors. Of the 113 tissues imaged, 95 (84%) had a tumor (75 basal cell carcinomas [BCCs] and 17 squamous cell carcinomas [SCCs]). The average reader diagnostic accuracy was 88.1%, with a sensitivity of 93.7%, and a specificity of 58.3%. The artificial intelligence (AI) model achieved a diagnostic accuracy of 87.6 ± 5.9%, sensitivity of 93.2 ± 2.1%, and specificity of 81.2 ± 9.2%. A mean intersection-over-union of 60.3 ± 10.1% was achieved when delineating the nodular BCC from normal structures. Limitation of the study was the small sample size for all tumors, especially SCCs. However, based on our preliminary results, we envision FF-OCT to rapidly image fresh tissues, facilitating surgical margin assessment. AI algorithms can aid in automated tumor detection, enabling widespread adoption of this technique.

Tertiary lymphoid structures accompanied by fibrillary matrix morphology impact anti-tumor immunity in basal cell carcinomas

Tertiary lymphoid structures (TLS) are specialized lymphoid formations that serve as local repertoire of T- and B-cells at sites of chronic inflammation, autoimmunity, and cancer. While presence of TLS has been associated with improved response to immune checkpoint blockade therapies and overall outcomes in several cancers, its prognostic value in basal cell carcinoma (BCC) has not been investigated. Herein, we determined the prognostic impact of TLS by relating its prevalence and maturation with outcome measures of anti-tumor immunity, namely tumor infiltrating lymphocytes (TILs) and tumor killing. In 30 distinct BCCs, we show the presence of TLS was significantly enriched in tumors harboring a nodular component and more mature primary TLS was associated with TIL counts. Moreover, assessment of the fibrillary matrix surrounding tumors showed discrete morphologies significantly associated with higher TIL counts, critically accounting for heterogeneity in TIL count distribution within TLS maturation stages. Specifically, increased length of fibers and lacunarity of the matrix with concomitant reduction in density and alignment of fibers were present surrounding tumors displaying high TIL counts. Given the interest in inducing TLS formation as a therapeutic intervention as well as its documented prognostic value, elucidating potential impediments to the ability of TLS in driving anti-tumor immunity within the tumor microenvironment warrants further investigation. These results begin to address and highlight the need to integrate stromal features which may present a hindrance to TLS formation and/or effective function as a mediator of immunotherapy response.

In vivo tumor immune microenvironment phenotypes correlate with inflammation and vasculature to predict immunotherapy response

Response to immunotherapies can be variable and unpredictable. Pathology-based phenotyping of tumors into ‘hot’ and ‘cold’ is static, relying solely on T-cell infiltration in single-time single-site biopsies, resulting in suboptimal treatment response prediction. Dynamic vascular events (tumor angiogenesis, leukocyte trafficking) within tumor immune microenvironment (TiME) also influence anti-tumor immunity and treatment response. Here, we report dynamic cellular-level TiME phenotyping in vivo that combines inflammation profiles with vascular features through non-invasive reflectance confocal microscopic imaging. In skin cancer patients, we demonstrate three main TiME phenotypes that correlate with gene and protein expression, and response to toll-like receptor agonist immune-therapy. Notably, phenotypes with high inflammation associate with immunostimulatory signatures and those with high vasculature with angiogenic and endothelial anergy signatures. Moreover, phenotypes with high inflammation and low vasculature demonstrate the best treatment response. This non-invasive in vivo phenotyping approach integrating dynamic vasculature with inflammation serves as a reliable predictor of response to topical immune-therapy in patients.

Standard assessment of immune infiltration of biopsies is not sufficient to accurately predict response to immunotherapy. Here, the authors show that reflectance confocal microscopy can be used to quantify dynamic vasculature and inflammatory features to better predict treatment response in skin cancers.

Classification of Basal Cell Carcinoma in Ex Vivo Confocal Microscopy Images from Freshly Excised Tissues Using a Deep Learning Algorithm

Ex vivo confocal microscopy (EVCM) generates digitally colored purple-pink images similar to H&E without time-consuming tissue processing. It can be used during Mohs surgery for rapid detection of basal cell carcinoma (BCC); however, reading EVCM images requires specialized training. An automated approach using a deep learning algorithm for BCC detection in EVCM images can aid in diagnosis. A total of 40 BCCs and 28 negative (not-BCC) samples were collected at Memorial Sloan Kettering Cancer Center to create three training datasets: (i) EVCM image dataset (663 images), (ii) H&E image dataset (516 images), and (iii) a combination of the two datasets. A total of seven BCCs and four negative samples were collected to create an EVCM test dataset (107 images). The model trained with the EVCM dataset achieved 92% diagnostic accuracy, similar to the H&E model (93%). The area under the receiver operator characteristic curve was 0.94, 0.95, and 0.94 for EVCM-, H&E-, and combination-trained models, respectively. We developed an algorithm for automatic BCC detection in EVCM images (comparable accuracy to dermatologists). This approach could be used to assist with BCC detection during Mohs surgery. Furthermore, we found that a model trained with only H&E images (which are more available than EVCM images) can accurately detect BCC in EVCM images.

Combined PARP1-targeted nuclear contrast and reflectance contrast enhances confocal microscopic detection of basal cell carcinoma

Reflectance confocal microscopy (RCM) with endogenous backscattered contrast can noninvasively image basal cell carcinomas (BCCs) in skin. However, BCCs present with high nuclear density and the relatively weak backscattering from nuclei impose a fundamental limit on contrast, detectability, and diagnostic accuracy. We investigated PARPi-FL, an exogenous nuclear poly (ADP-ribose) polymerase (PARP1)-targeted fluorescent contrast agent and fluorescence confocal microscopy (FCM) towards improving BCC diagnosis. Methods: We tested PARP1 expression in 95 BCC tissues using immunohistochemistry, followed by PARPi-FL staining in 32 fresh surgical BCC specimens. Diagnostic accuracy of PARPi-FL contrast was evaluated in 83 surgical specimens. Optimal parameters for trans-epidermal permeability of PARPi-FL through intact skin was tested ex vivo on 5 human skin specimens and in vivo in 3 adult Yorkshire pigs. Results: We found significantly higher PARP1 expression and PARPi-FL binding in BCCs, as compared to normal skin structures. Blinded reading of RCM-and-FCM images by two experts demonstrated a higher diagnostic accuracy for BCCs with combined fluorescence and reflectance contrast, as compared to RCM-alone. Optimal parameters (time and concentration) for PARPi-FL trans-epidermal permeation through intact skin were successfully determined. Conclusion: Combined fluorescence and reflectance contrast may improve noninvasive BCC diagnosis with confocal microscopy.

Optical imaging guided- 'precision' biopsy of skin tumors: a novel approach for targeted sampling and histopathologic correlation

Dermoscopy and reflectance confocal microscopy (RCM) are two noninvasive, optical imaging tools used to facilitate clinical diagnosis. A biopsy technique that produces exact correlation with optical imaging features is not previously reported. To evaluate the applications of a novel feature-focused 'precision biopsy' technique that correlates clinical-dermoscopy-RCM findings with histopathology. This was a prospective case-series performed during August 2017 and June 2019 at a tertiary care cancer. We included consecutive patients requiring a precise dermoscopy-RCM-histopathologic correlation. We performed prebiopsy dermoscopy and both wide probe and handheld RCM of suspicious lesions. Features of interest were isolated with the aid of paper rings and a 2 mm punch biopsy was performed in the dermoscopy- or RCM-highlighted area. Tissue was processed either en face or with vertical sections. One-to-one correlation with histopathology was obtained. Twenty-three patients with 24 lesions were included in the study. The mean age was 64.6 years (range 22-91 years); there were 16 (69.6%) males, 14 (58.3%) lesions biopsied were on head and neck region. We achieved tissue-conservation diagnosis in 100% (24/24), 13 (54.2%) were clinically equivocal lesions, six (25%) were selected for 'feature correlation' of structures on dermoscopy or RCM, and five (20.8%) for 'correlation of new/unknown' RCM features seen on follow-up. The precision biopsy technique described herein is a novel method that facilitates direct histopathological correlation of dermoscopy and RCM features. With the aids of optical imaging devices, accurate diagnosis may be achieved by minimally invasive tissue extraction.

Abstract 2814: Dynamic imaging of tumor-immune microenvironment (TiME) and microvasculature identifies ‘hot' and ‘cold' tumor phenotypes in vivo in patients

Investigating the dynamic crosstalk between the tumor-immune microenvironment (TiME) and microvasculature in vivo in patients can lead to important insights into the underlying biology, help identify tumor phenotypes and reveal attractive druggable targets.

Dynamic non-invasive label-free imaging of TiME and microvasculature in real-time directly in patients using reflectance confocal microscopy (RCM) was investigated on 60 skin cancer patients (basal cell carcinoma, BCC; squamous cell carcinoma, SCC), followed by automated and machine-learning based quantification of TiME and microvasculature features such as vascular density, leukocyte trafficking and immune cell density. Manual (two readers) and histopathological evaluation (dermatopathologist) of these features was also performed. Molecular correlation of imaging features and phenotypes was performed using anti-CD3/anti-CD20 IHC staining for tertiary lymphoid structures (TLS) and total lymphocyte density (n=33), flow cytometry for immune cells (n=3), and differential RNA expression (n=14). Correlation of RCM features and phenotypes at baseline (before treatment) with treatment response was also evaluated on 9 cancer lesions undergoing topical immunotherapy imiquimod. High agreement for feature presence on RCM and Histology, and manual and automated RCM features was observed. Unsupervised clustering on total TiME and microvasculature features on RCM using principal component analysis (PCA) indicates four distinct tumor phenotypes (PCA 1). The phenotype with high inflammation, high trafficking and higher density of vessels or the denoted ‘hot' phenotype correlated with higher activated CD8+ Granzyme B+ cells (67% of total CD8+cells). The clustering pattern on RCM was compared to TLS and lymphocyte density (PCA 2) and gene expression following CIBERSORT analysis (PCA 3). The clustering in RCM correlated better with gene expression (PCA 1 and 3, 100% agreement) than TLS and lymphocyte density (PCA 1 and 2, 86% agreement). The ‘hot' phenotype in RCM correlated with higher immune gene signatures and higher TLS/lymphocyte density. Increased plasma, CD8, activated CD4 memory and activated NK cells, M1 macrophages and monocytes, along with up-regulation of JAK-STAT, chemokine and cell adhesion signaling cascade were found in the ‘hot' RCM phenotype. Statistical modeling for correlating phenotypes with treatment outcomes was performed using principal component-linear discriminant analysis (PC-LDA). Two responders with tumor regression were predicted as ‘hot' phenotype while the non-responding patients (remaining 7) were classified as cold phenotype; suggesting that RCM 'hot' phenotype correlates with better treatment response. Thus, we demonstrate the potential utility of noninvasive RCM imaging in identifying ‘hot' and ‘cold' tumor phenotypes directly in patients.

Citation Format: Aditi Sahu, Melissa Gill, Miguel Cordova, Anthony Santella, Kivanc Kose, Teguru Tembo, Anabel Alfonso, Pratik Chandrani, Christi Fox, Salvador Gonzalez, Nicholas Kurtansky, Melissa Pulitzer, William Phillips, Madison Li, Kimeil King, Stephen Dusza, Shuaitong Liu, Ning Yang, Haaris Jilani, Paras Mehta, Ashfaq Marghoob, Allan Halpern, Anthony Rossi, Liang Deng, Chih-Shan Jason Chen, Milind Rajadhyaksha. Dynamic imaging of tumor-immune microenvironment (TiME) and microvasculature identifies ‘hot' and ‘cold' tumor phenotypes in vivo in patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2814.

Reflectance confocal microscopy-guided laser ablation of basal cell carcinomas: initial clinical experience

Laser ablation offers a procedure for precise, fast, and minimally invasive removal of superficial and early nodular basal cell carcinomas (BCCs). However, the lack of histopathological confirmation has been a limitation toward widespread use in the clinic. A reflectance confocal microscopy (RCM) imaging-guided approach offers cellular-level histopathology-like feedback directly on the patient, which may then guide and help improve the efficacy of the ablation procedure. Following an ex vivo benchtop study (reported in our earlier papers), we performed an initial study on 44 BCCs on 21 patients in vivo, using a pulsed erbium:ytterbium aluminum garnet laser and a contrast agent (aluminum chloride). In 10 lesions on six patients, the RCM imaging-guided detection of either presence of residual tumor or complete clearance was immediately confirmed with histopathology. Additionally, 34 BCCs on 15 patients were treated with RCM imaging-guided laser ablation, with immediate confirmation for clearance of tumor (no histopathology), followed by longer-term monitoring, currently in progress, with follow-up imaging (again, no histopathology) at 3, 6, and 18 months. Thus far, the imaging resolution appears to be sufficient and consistent for monitoring efficacy of ablation in the wound, both immediately postablation and subsequently during recovery. The efficacy results appear to be promising, with observed clearance in 19 cases of 22 cases with follow-ups ranging from 6 to 21 months. An additional 12 cases with 1 to 3 months of follow-ups has shown clearance of tumor but a longer follow-up time is required to establish conclusive results. Further instrumentation development will be necessary to cover larger areas with a more automatically controlled instrument for more uniform, faster, and deeper imaging of margins.

Traditional versus streamlined management of basal cell carcinoma (BCC): A cost analysis

BACKGROUND

Facing rising incidence of basal cell carcinoma (BCC) and increasing pressure to contain health care spending, physicians need to contemplate cost-effective paradigms for managing BCC.

OBJECTIVE

We sought to perform a cost analysis comparing the traditional BCC management scheme with a simplified detect-and-treat scheme that eliminates the biopsy before initiating definitive treatment.

METHODS

A decision analytic model was developed to compare the costs of traditional BCC management with the detect-and-treat scheme, under which qualifying lesions diagnosed clinically were either treated with shave removal or referred to Mohs micrographic surgery for on-site histologic check. Values for model parameters were based on literature and our institutional data analysis. Costs were based on 2014 Medicare fee schedule.

RESULTS

The average cost per lesion with detect-and-treat scheme was $449 for non-Mohs micrographic surgery-indicated lesions (vs $566 with traditional management, $117 in savings) and $819 for Mohs micrographic surgery-indicated lesions (vs $864 with traditional management, $45 in savings). The combined weighted average savings per case was $95 (15% of total average cost). Conclusions were similar under various plausible scenarios.

LIMITATIONS

Model parameter values may vary based on individual practices.

CONCLUSIONS

A simplified management strategy eliminating routine pretreatment biopsy can reduce BCC treatment cost without compromising quality of care.

Clinical and dermoscopic features of combined cutaneous squamous cell carcinoma (SCC)/neuroendocrine [Merkel cell] carcinoma (MCC)

BACKGROUND

Merkel cell carcinoma (MCC) is a neuroendocrine carcinoma, associated with Merkel cell polyomavirus. MCC admixed with squamous cell carcinoma (SCC) is unassociated with polyomavirus, and is genetically distinct.

OBJECTIVE

We sought to distinguish clinically and dermoscopically between MCC and SCC/MCC.

METHODS

We compared patient data for SCC/MCC (n = 26) and MCC (n = 20), and reviewed clinical and dermoscopic images (n = 9) of SCC/MCC.

RESULTS

Patients with SCC/MCC were older (median 76.5 vs 69 years) and more often male (77% vs 60%), and had more nonmelanoma skin cancer (85% vs 25%), malignant extracutaneous tumors (25% vs 5%), lymphoproliferative disorders (23% vs 10%), and immunodeficient/proinflammatory states (77% vs 35%). In all, 58% of SCC/MCC versus 10% of MCC were clinically diagnosed nonmelanoma skin cancer. Patients with SCC/MCC had more metastases (77% vs 40%), more treatment failures (53% vs 45%), shorter survival (41 vs 54 months), and more death from disease (50% vs 40%). SCC/MCC demonstrated marked scale (7/9), and telangiectasia (1/9). Dermoscopically, small dotted and short linear irregular peripheral vessels and central milky-red areas with large-diameter arborizing vessels were seen.

LIMITATIONS

The rarity of SCC/MCC limits available data.

CONCLUSIONS

SCC/MCC is aggressive, arising within elderly patients' chronically ultraviolet-exposed skin, often in the setting of immunosuppression or inflammation. Dermoscopically, polymorphous vessels in lesions suspicious for nonmelanoma skin cancer are suggestive.

Confocal microscopy-guided laser ablation for superficial and early nodular Basal cell carcinoma: a promising surgical alternative for superficial skin cancers

IMPORTANCE

Laser ablation is a rapid and minimally invasive approach for the treatment of superficial skin cancers, but efficacy and reliability vary owing to lack of histologic margin control. High-resolution reflectance confocal microscopy (RCM) may offer a means for examining margins directly on the patient.

OBSERVATIONS

We report successful elimination of superficial and early nodular basal cell carcinoma (BCC) in 2 cases using RCM imaging to guide Er:YAG laser ablation. Three-dimensional (3D) mapping is feasible with RCM to delineate the lateral border and thickness of the tumor. Thus, the surgeon may deliver laser fluence and passes with localized control-ie, by varying the ablation parameters in sublesional areas with specificity that is governed by the 3D topography of the BCC. We further demonstrate intraoperative detection of residual BCC after initial laser ablation and complete removal of remaining tumor by additional passes. Both RCM imaging and histologic sections confirm the final clearance of BCC.

CONCLUSIONS AND RELEVANCE

Confocal microscopy may enhance the efficacy and reliability of laser tumor ablation. This report represents a new translational application for RCM imaging, which, when combined with an ablative laser, may one day provide an efficient and cost-effective treatment for BCC.

Confocal microscopy to guide erbium:yttrium aluminum garnet laser ablation of basal cell carcinoma: an ex vivo feasibility study

For the removal of superficial and nodular basal cell carcinomas (BCCs), laser ablation provides certain advantages relative to other treatment modalities. However, efficacy and reliability tend to be variable because tissue is vaporized such that none is available for subsequent histopathological examination for residual BCC (and to confirm complete removal of tumor). Intra-operative reflectance confocal microscopy (RCM) may provide a means to detect residual tumor directly on the patient and guide ablation. However, optimization of ablation parameters will be necessary to control collateral thermal damage and preserve sufficient viability in the underlying layer of tissue, so as to subsequently allow labeling of nuclear morphology with a contrast agent and imaging of residual BCC. We report the results of a preliminary study of two key parameters (fluence, number of passes) vis-à-vis the feasibility of labeling and RCM imaging in human skin ex vivo, following ablation with an erbium:yttrium aluminum garnet laser.

Current progress of immunostains in Mohs micrographic surgery: a review

Mohs micrographic surgery is often considered the treatment of choice for a variety of skin malignancies. In recent years, the application of immunostaining techniques has facilitated the successful removal of a number of common and less common cutaneous malignancies, including basal cell carcinoma, squamous cell carcinoma, malignant melanoma, dermatofibrosarcoma protuberans, microcystic adnexal carcinoma, sebaceous carcinoma, atypical fibroxanthoma, extramammary Paget's disease, and even sarcomas. Immunostains highlight the tumor cells and allow the Mohs surgeons to pinpoint and eliminate the residual tumor at the surgical margin. It is especially helpful when a tumor presents with subtle or nonspecific histologic features or when a tumor is masked in a pocket of dense inflammation. However, the cost, the labor, and the time consumption are of concern to many of our peers, as are the diversity of antigens, which may overwhelm some. This article serves as a review of the literature on current uses of immunostaining in Mohs micrographic surgery and as a summary of their realistic applications in the dermatologic surgeon's practice. We conclude that immunohistochemical technique has played an important role in Mohs surgery advancement. With greater use and more cost-effective staining methods, we believe that the use of immunostains in a Mohs practice will become routine.