Hidden Variables in Deep Learning Digital Pathology and Their Potential to Cause Batch Effects: Prediction Model Study.

BACKGROUND: An increasing number of studies within digital pathology show the potential of artificial intelligence (AI) to diagnose cancer using histological whole slide images, which requires large and diverse data sets. While diversification may result in more generalizable AI-based systems, it can also introduce hidden variables. If neural networks are able to distinguish/learn hidden variables, these variables can introduce batch effects that compromise the accuracy of classification systems. OBJECTIVE: The objective of the study was to analyze the learnability of an exemplary selection of hidden variables (patient age, slide preparation date, slide origin, and scanner type) that are commonly found in whole slide image data sets in digital pathology and could create batch effects. METHODS: We trained four separate convolutional neural networks (CNNs) to learn four variables using a data set of digitized whole slide melanoma images from five different institutes. For robustness, each CNN training and evaluation run was repeated multiple times, and a variable was only considered learnable if the lower bound of the 95% confidence interval of its mean balanced accuracy was above 50.0%. RESULTS: A mean balanced accuracy above 50.0% was achieved for all four tasks, even when considering the lower bound of the 95% confidence interval. Performance between tasks showed wide variation, ranging from 56.1% (slide preparation date) to 100% (slide origin). CONCLUSIONS: Because all of the analyzed hidden variables are learnable, they have the potential to create batch effects in dermatopathology data sets, which negatively affect AI-based classification systems. Practitioners should be aware of these and similar pitfalls when developing and evaluating such systems and address these and potentially other batch effect variables in their data sets through sufficient data set stratification.

Overdiagnosis of melanoma - causes, consequences and solutions.

Malignant melanoma is the skin tumor that causes most deaths in Germany. At an early stage, melanoma is well treatable, so early detection is essential. However, the skin cancer screening program in Germany has been criticized because although melanomas have been diagnosed more frequently since introduction of the program, the mortality from malignant melanoma has not decreased. This indicates that the observed increase in melanoma diagnoses be due to overdiagnosis, i.e. to the detection of lesions that would never have created serious health problems for the patients. One of the reasons is the challenging distinction between some benign and malignant lesions. In addition, there may be lesions that are biologically equivocal, and other lesions that are classified as malignant according to current criteria, but that grow so slowly that they would never have posed a threat to patient's life. So far, these "indolent" melanomas cannot be identified reliably due to a lack of biomarkers. Moreover, the likelihood that an in-situ melanoma will progress to an invasive tumor still cannot be determined with any certainty. When benign lesions are diagnosed as melanoma, the consequences are unnecessary psychological and physical stress for the affected patients and incurred therapy costs. Vice versa, underdiagnoses in the sense of overlooked melanomas can adversely affect patients' prognoses and may necessitate more intense therapies. Novel diagnostic options could reduce the number of over- and underdiagnoses and contribute to more objective diagnoses in borderline cases. One strategy that has yielded promising results in pilot studies is the use of artificial intelligence-based diagnostic tools. However, these applications still await translation into clinical and pathological routine.

[Plexiform fibrohistiocytic tumor]. / Plexiformer fibrohistiozytärer Tumor.

Plexiform fibrohistiocytic tumors are rare, low-to-moderate malignant soft tissue tumors that occur primarily in children and adolescents and are located on the upper extremity. The diagnosis must be made histologically. We report on a young woman who presented a growing, painless lesion on the cubital fossa. Histopathology as well as the standard of treatment are discussed.

A 3-dimensional histology computer model of malignant melanoma and its implications for digital pathology.

BACKGROUND: Historically, cancer diagnoses have been made by pathologists using two-dimensional histological slides. However, with the advent of digital pathology and artificial intelligence, slides are being digitised, providing new opportunities to integrate their information. Since nature is 3-dimensional (3D), it seems intuitive to digitally reassemble the 3D structure for diagnosis. OBJECTIVE: To develop the first human-3D-melanoma-histology-model with full data and code availability. Further, to evaluate the 3D-simulation together with experienced pathologists in the field and discuss the implications of digital 3D-models for the future of digital pathology. METHODS: A malignant melanoma of the skin was digitised via 3 µm cuts by a slide scanner; an open-source software was then leveraged to construct the 3D model. A total of nine pathologists from four different countries with at least 10 years of experience in the histologic diagnosis of melanoma tested the model and discussed their experiences as well as implications for future pathology. RESULTS: We successfully constructed and tested the first 3D-model of human melanoma. Based on testing, 88.9% of pathologists believe that the technology is likely to enter routine pathology within the next 10 years; advantages include a better reflectance of anatomy, 3D assessment of symmetry and the opportunity to simultaneously evaluate different tissue levels at the same time; limitations include the high consumption of tissue and a yet inferior resolution due to computational limitations. CONCLUSIONS: 3D-histology-models are promising for digital pathology of cancer and melanoma specifically, however, there are yet limitations which need to be carefully addressed.

Genetic and Methylation Analysis of CTNNB1 in Benign and Malignant Melanocytic Lesions.

Melanocytic neoplasms have been genetically characterized in detail during the last decade. Recurrent CTNNB1 exon 3 mutations have been recognized in the distinct group of melanocytic tumors showing deep penetrating nevus-like morphology. In addition, they have been identified in 1-2% of advanced melanoma. Performing a detailed genetic analysis of difficult-to-classify nevi and melanomas with CTNNB1 mutations, we found that benign tumors (nevi) show characteristic morphological, genetic and epigenetic traits, which distinguish them from other nevi and melanoma. Malignant CTNNB1-mutant tumors (melanomas) demonstrated a different genetic profile, instead grouping clearly with other non-CTNNB1 melanomas in methylation assays. To further evaluate the role of CTNNB1 mutations in melanoma, we assessed a large cohort of clinically sequenced melanomas, identifying 38 tumors with CTNNB1 exon 3 mutations, including recurrent S45 (n = 13, 34%), G34 (n = 5, 13%), and S27 (n = 5, 13%) mutations. Locations and histological subtype of CTNNB1-mutated melanoma varied; none were reported as showing deep penetrating nevus-like morphology. The most frequent concurrent activating mutations were BRAF V600 (n = 21, 55%) and NRAS Q61 (n = 13, 34%). In our cohort, four of seven (58%) and one of nine (11%) patients treated with targeted therapy (BRAF and MEK Inhibitors) or immune-checkpoint therapy, respectively, showed disease control (partial response or stable disease). In summary, CTNNB1 mutations are associated with a unique melanocytic tumor type in benign tumors (nevi), which can be applied in a diagnostic setting. In advanced disease, no clear characteristics distinguishing CTNNB1-mutant from other melanomas were observed; however, studies of larger, optimally prospective, cohorts are warranted.

Does the peritumoral stroma of basal cell carcinoma recapitulate the follicular connective tissue sheath?

BACKGROUND: There are compelling embryologic and anatomic relationships within adnexal tumors. However, these are mostly perceived within the epithelial component while the stromal component of the tumors is frequently overlooked. In postnatal skin, nestin is almost exclusively expressed in the perifollicular mesenchyme. This study examines the expression of this neuroepithelial stem cell protein in trichoblastoma/trichoepithelioma and in basal cell carcinoma (BCC), which is increasingly being viewed as follicular in nature. METHODS: We employed standard immunohistochemical methods with three different antibodies to examine the expression of nestin in 25 BCCs and compared the staining pattern with that of 7 trichoblastomas and 11 trichoepitheliomas. RESULTS: Nestin is expressed in the peritumoral stroma of all tumors examined and is limited to the immediate layer of mesenchymal cells surrounding the tumor epithelium. In BCC, nestin-immunoreactive cells are found as a sheath of thin, spindled fibroblasts, while reactive cells are plump in trichoepitheliomas/trichoblastomas. CONCLUSIONS: We postulate that the peritumoral stroma of BCC imitates the perifollicular connective tissue sheath, while in contrast that of trichoepithelioma/trichoblastoma is similar to the papillary and immediate peripapillary follicular mesenchyme. Further functional and animal experimental studies are needed to test this hypothesis.

Deep learning approach to predict sentinel lymph node status directly from routine histology of primary melanoma tumours.

AIM: Sentinel lymph node status is a central prognostic factor for melanomas. However, the surgical excision involves some risks for affected patients. In this study, we therefore aimed to develop a digital biomarker that can predict lymph node metastasis non-invasively from digitised H&E slides of primary melanoma tumours. METHODS: A total of 415 H&E slides from primary melanoma tumours with known sentinel node (SN) status from three German university hospitals and one private pathological practice were digitised (150 SN positive/265 SN negative). Two hundred ninety-one slides were used to train artificial neural networks (ANNs). The remaining 124 slides were used to test the ability of the ANNs to predict sentinel status. ANNs were trained and/or tested on data sets that were matched or not matched between SN-positive and SN-negative cases for patient age, ulceration, and tumour thickness, factors that are known to correlate with lymph node status. RESULTS: The best accuracy was achieved by an ANN that was trained and tested on unmatched cases (61.8% ± 0.2%) area under the receiver operating characteristic (AUROC). In contrast, ANNs that were trained and/or tested on matched cases achieved (55.0% ± 3.5%) AUROC or less. CONCLUSION: Our results indicate that the image classifier can predict lymph node status to some, albeit so far not clinically relevant, extent. It may do so by mostly detecting equivalents of factors on histological slides that are already known to correlate with lymph node status. Our results provide a basis for future research with larger data cohorts.

Combining CNN-based histologic whole slide image analysis and patient data to improve skin cancer classification.

BACKGROUND: Clinicians and pathologists traditionally use patient data in addition to clinical examination to support their diagnoses. OBJECTIVES: We investigated whether a combination of histologic whole slides image (WSI) analysis based on convolutional neural networks (CNNs) and commonly available patient data (age, sex and anatomical site of the lesion) in a binary melanoma/nevus classification task could increase the performance compared with CNNs alone. METHODS: We used 431 WSIs from two different laboratories and analysed the performance of classifiers that used the image or patient data individually or three common fusion techniques. Furthermore, we tested a naive combination of patient data and an image classifier: for cases interpreted as 'uncertain' (CNN output score <0.7), the decision of the CNN was replaced by the decision of the patient data classifier. RESULTS: The CNN on its own achieved the best performance (mean ± standard deviation of five individual runs) with AUROC of 92.30% ± 0.23% and balanced accuracy of 83.17% ± 0.38%. While the classification performance was not significantly improved in general by any of the tested fusions, naive strategy of replacing the image classifier with the patient data classifier on slides with low output scores improved balanced accuracy to 86.72% ± 0.36%. CONCLUSION: In most cases, the CNN on its own was so accurate that patient data integration did not provide any benefit. However, incorporating patient data for lesions that were classified by the CNN with low 'confidence' improved balanced accuracy.

Skin cancer classification via convolutional neural networks: systematic review of studies involving human experts.

BACKGROUND: Multiple studies have compared the performance of artificial intelligence (AI)-based models for automated skin cancer classification to human experts, thus setting the cornerstone for a successful translation of AI-based tools into clinicopathological practice. OBJECTIVE: The objective of the study was to systematically analyse the current state of research on reader studies involving melanoma and to assess their potential clinical relevance by evaluating three main aspects: test set characteristics (holdout/out-of-distribution data set, composition), test setting (experimental/clinical, inclusion of metadata) and representativeness of participating clinicians. METHODS: PubMed, Medline and ScienceDirect were screened for peer-reviewed studies published between 2017 and 2021 and dealing with AI-based skin cancer classification involving melanoma. The search terms skin cancer classification, deep learning, convolutional neural network (CNN), melanoma (detection), digital biomarkers, histopathology and whole slide imaging were combined. Based on the search results, only studies that considered direct comparison of AI results with clinicians and had a diagnostic classification as their main objective were included. RESULTS: A total of 19 reader studies fulfilled the inclusion criteria. Of these, 11 CNN-based approaches addressed the classification of dermoscopic images; 6 concentrated on the classification of clinical images, whereas 2 dermatopathological studies utilised digitised histopathological whole slide images. CONCLUSIONS: All 19 included studies demonstrated superior or at least equivalent performance of CNN-based classifiers compared with clinicians. However, almost all studies were conducted in highly artificial settings based exclusively on single images of the suspicious lesions. Moreover, test sets mainly consisted of holdout images and did not represent the full range of patient populations and melanoma subtypes encountered in clinical practice.

Skin mesenchymal stem cells: prospects for clinical dermatology.

Stem cell-based therapies are expected to have a great impact on the medicine of the 21st century. The focus of dermatologic stem cell research is on the epidermis and the hair follicle. In contrast, the characterization of stem cells in the mesenchymal compartments of the skin has largely escaped the attention of the dermatologic community. This is surprising because the dermis may represent a larger reservoir for adult stem cells than the epidermis and the hair follicle together. In 2001, mesenchymal stem cells residing within the dermis were first isolated. They have the capacity to differentiate into adipocytes, smooth muscle cells, osteocytes, chondrocytes, and even neurons and glia as well as hematopoietic cells of myeloid and erythroid lineage. The perifollicular connective tissue sheath and the papilla crystallize as the likely anatomic niche for these multipotent dermal cells. These previously unidentified mesenchymal stem cells have the potential to function as an easily accessible, autologous source for future stem cell transplantation. Potential therapeutic applications include the treatment of acute and steroid-refractory graft-versus-host disease, systemic lupus erythematosus resistant to currently available therapies, or idiopathic pulmonary fibrosis. The neuronal differentiation potential of cutaneous mesenchymal stem cells may also be exploited in the treatment of neurodegenerative disorders. The most immediate impact can be expected in the field of wound healing. In line with the cancer stem cell hypothesis, the potential contributions to dermatopathology include a conceptual understanding of mesenchymal skin-based neoplasms as evolving from a genetically altered dermal stem cell clone.

Spatiotemporal expression pattern of neuroepithelial stem cell marker nestin suggests a role in dermal homeostasis, neovasculogenesis, and tumor stroma development: a study on embryonic and adult human skin.

BACKGROUND: Whereas keratinocytic bulge stem cells are well characterized, comparably little is known about cutaneous mesenchymal stem cells. The follicular connective tissue sheath is proposed as a niche for dermal stem cells. OBJECTIVE: Because the neuroepithelial stem cell marker nestin represents a marker for mesenchymal stem cells in various tissues, our aim was to characterize its spatiotemporal expression pattern in the skin with special reference to the follicular mesenchyme. METHODS: We studied immunohistochemically nestin expression over the course of human cutaneous embryogenesis, in postnatal skin, in scalp wounds, and in the peritumoral stroma of basal cell carcinomas and compared its expression with that of other known mesenchymal markers. RESULTS: Nestin is expressed throughout the entire early embryonic dermis but confined later during development to the follicular connective tissue sheath, where it can also be found in postnatal human hair follicles. Its expression is up-regulated in scalp wounds and the nestin-positive cells seem to originate from the follicular mesenchyme. Nestin is also expressed in a thin layer of fibroblasts in the immediate vicinity of basal cell carcinomas. LIMITATIONS: The examination for nestin expression of scalp wounds is considered preliminary, because we examined scalp wounds representing re-excisions of previously diagnosed neoplasms from which we had no exact time table available as to when the original excision took place. CONCLUSION: We propose that nestin functions as a stem cell marker of the follicular mesenchyme and has a major regulatory role in dermal homeostasis, cutaneous neovasculogenesis, and tumor stroma development.

Trichoadenoma of Nikolowski is a distinct neoplasm within the spectrum of follicular tumors.

BACKGROUND: Trichoadenoma is a rare benign follicular tumor first described by Nikolowski 50 years ago. Both trichoadenoma and desmoplastic trichoepithelioma are composed of cords of epithelial cells and cornifying cysts embedded in sclerotic stroma. In trichoadenoma the cystic component predominates, while desmoplastic trichoepithelioma is a mostly solid neoplasm. Therefore trichoadenoma was suggested to represent a cystic variant of desmoplastic trichoepithelioma. OBJECTIVE: The aim of this study was to investigate whether the morphologic overlap between trichoadenoma and desmoplastic trichoepithelioma translates into a similar immunohistochemical profile. METHODS: We studied 19 trichoadenomas and 21 desmoplastic trichoepitheliomas for cytokeratin 20, Ber-EP4, and androgen receptor expression. RESULTS: Eighteen of 19 trichoadenomas and all desmoplastic trichoepitheliomas demonstrated the presence of Merkel cells as detected by a monoclonal antibody against cytokeratin 20. In contrast, while all desmoplastic trichepitheliomas were positive for Ber-EP4, only 4 of 19 trichoadenomas showed any kind of reactivity for this marker. None of the trichoadenomas or desmoplastic trichoepitheliomas expressed androgen receptor. LIMITATIONS: This study is limited by the moderate number of these rare tumors available for immunohistochemical analysis. CONCLUSION: Our data demonstrate that trichoadenoma typically retains cytokeratin 20-positive Merkel cells but lacks Ber-EP4 and androgen receptor expression. Trichoadenoma is a distinct follicular tumor related but not identical to desmoplastic trichoepithelioma.

Cutaneous mesenchymal stem cells: status of current knowledge, implications for dermatopathology.

Stem cell biology is currently making its impact on medicine, which will probably increase over the next decades. It not only influences our therapeutic thinking caused by the enormous plasticity of stem cells but also affects diagnostic and conceptual aspects of dermatopathology. Although our knowledge of the keratinocytic stem cells located within the follicular bulge has exploded exponentially since their discovery in 1990, the concept of cutaneous mesenchymal stem cells (MSCs) is new. Described initially in 2001 in mice, MSCs later were also found in the human dermis. The connective tissue sheath and the papilla of the hair follicle probably represent the anatomical niche for cutaneous MSCs. In line with the cancer stem cell hypothesis, mutations of these cells may be the underlying basis of mesenchymal skin neoplasms, such as dermatofibrosarcoma protuberans. Furthermore, research on cutaneous MSCs may impact our thinking on the interaction of the epithelial component of skin neoplasms with their surrounding stroma. We are only in the early stages to recognize the importance of the potential contributions of cutaneous MSC research to dermatopathology, but it is not inconceivable to assume that they could be tremendous, paralleling the early discovery of the follicular bulge as a stem cell niche.

Sox9, more than a marker of the outer root sheath: spatiotemporal expression pattern during human cutaneous embryogenesis.

BACKGROUND: The sex-determining gene Sox9 was recently unexpectedly found to have an essential role in outer root sheath differentiation. It was also characterized as a general marker of basal cell carcinoma. Herein, we describe its spatiotemporal expression pattern outside the hair follicle during human cutaneous embryogenesis. METHODS: We examined immunohistochemically samples from embryonic and fetal human skin for the expression of SOX9 using standard techniques. For comparison reasons, we also included scalp skin from adults. RESULTS: SOX9 is expressed in the developing nail organ, eccrine glands, blood vessels and melanocytes/melanoblasts. In the nail organ, the nail bed but not the nail matrix was immunoreactive for SOX9. In plantar skin, SOX9 specifically labels the evolving eccrine glands but not the interfollicular keratinocytes. CONCLUSIONS: The distinctive expression pattern of SOX9 during human cutaneous embryogenesis indicates a key role in skin homeostasis that includes but goes beyond its role in outer root sheath differentiation. Studying immunohistochemical markers in developing human skin has the potential to further our understanding of adult skin physiology and to deepen our concepts especially of the histogenesis of adnexal tumors (including those of the nail unit) and the relationship of the various adnexal structures to each other.

Blimp-1: a marker of terminal differentiation but not of sebocytic progenitor cells.

BACKGROUND: The role of stem cells in maintaining the sebaceous gland throughout the various stages of life is not satisfactorily resolved. In a recent article, the transcription factor B lymphocyte-induced maturation protein 1 (Blimp-1) was proposed as a marker of a population of unipotent progenitor cells that reside in the sebaceous gland, regulating its size and activity. METHODS: We used standard immunohistochemical methods to examine Blimp-1 expression in samples from embryonic, fetal and adult human skin and in 119 sebaceous lesions comprising all major categories of sebocytic lineage, including hamartomas, cysts and benign and malignant neoplasms. RESULTS: Blimp-1 is expressed late in embryonic development and is restricted to the evolving sebaceous gland, the terminally differentiating components of the hair follicle and nail organ and the granular layer. This pattern is preserved into adult life. In all sebaceous lesions, Blimp-1 labels only the most mature cellular constituents. CONCLUSIONS: The reported expression pattern is difficult to reconcile with a function of Blimp-1 as a marker for sebocytic progenitor cells but indicates a major role in terminal differentiation. Within the interfollicular epidermis, its exclusive localization to the granular layer suggests a central function in skin barrier homeostasis in the human.

Expression pattern of GATA-3 in embryonic and fetal human skin suggests a role in epidermal and follicular morphogenesis.

BACKGROUND: The transcription factor GATA-3 was recently identified as a master regulator in the specification of the inner root sheath. Additionally, it seems to play a role in skin barrier physiology. p63 binds and transactivates the GATA-3 promoter. While the expression profile of GATA-3 is delineated for the mouse, little is known about its expression in the adult human hair follicle and no studies are published about its distribution during human cutaneous embryogenesis. METHODS: We examined samples from embryonic, fetal and adult human skin for the expression of GATA-3 using immunohistochemistry. RESULTS: GATA-3 is expressed late during human skin development. Its expression pattern is comparable to the mouse and confined to the Huxley layer and inner root sheath cuticle but sparing the Henle layer. In addition, GATA-3 localizes to the spinous cell layer of the interfollicular epidermis. CONCLUSIONS: From the described expression pattern, it is highly probable that GATA-3 plays a role in follicular and epidermal morphogenesis. What the anatomically confined expression of GATA-3 to the spinous layer means biologically for the physiology of the skin is still unclear. Likewise, it still needs to be shown if GATA-3 could be exploited in the diagnosis of adnexal neoplasms.

A scanning microscopic clue to the diagnosis of arthropod assault reaction: alteration of interstitial tissue is more common than a wedge-shaped inflammatory infiltrate.

BACKGROUND: A scanning microscopic clue to the diagnosis of arthropod assault reactions is a wedge-shaped inflammatory infiltrate. However, to describe an inflammatory infiltrate as wedge-shaped or not involves a high degree of subjectivity. METHODS: We studied hematoxylin and eosin-stained sections of 137 biopsies of arthropod assault reactions for epidermal and dermal changes and for the composition, density and depth of the inflammatory infiltrate. RESULTS: We found a wedge-shaped inflammatory component in only 10.2% of the cases. A much more common feature is an alteration of the interstitial tissue present in 85.4% of the biopsies. It consisted of a narrowing of the spaces between the collagen bundles, which was readily observable on scanning magnification. On higher magnification, a loosely textured basophilic material was often noted within the dermis. CONCLUSIONS: The hitherto often emphasized wedge-shaped configuration of the inflammatory component of arthropod assault reactions is not of great diagnostic value. The altered interstitial tissue is easily recognizable by its diminished interstitial spaces at low power magnification and can serve as a scanning magnification clue to the diagnosis of arthropod assault reactions.

p16 expression differentiates between desmoplastic Spitz nevus and desmoplastic melanoma.

BACKGROUND: Loss of p16 in melanomas reflects worse outcomes for patients. It is associated with depth of invasion, ulceration, vascular invasion, lymph node metastases, metastases, recurrence of melanoma and decreased 5-year survival. Desmoplastic melanoma is an insidious malignant melanoma subtype that commonly occurs on sun-damaged skin of the head and neck area in elderly patients. The diagnostic conundrum occurs with confusion of desmoplastic melanoma with scars, hyalinizing blue nevi, desmoplastic Spitz nevi and diffuse neurofibromas. METHODS: The present study uses immunohistochemistry with a p16 antibody to differentiate desmoplastic Spitz nevi (n = 15 cases) from desmoplastic melanomas (n = 11). To date, no other studies have been published defining the expression pattern of p16 in desmoplastic melanoma. RESULTS: 81.8% of desmoplastic melanomas were negative for p16 and 18.2% were only weakly stained. In contrast, all desmoplastic Spitz nevi were moderately to strongly positive for p16. CONCLUSIONS: The staining pattern for p16 in desmoplastic melanomas and Spitz nevi in conjunction with the histopathologic features, S100 staining, Ki67 proliferation index and clinical scenario may aid in the difficult differential diagnosis between these two entities. Further confirmatory studies are indicated.