Deep learning based histological classification of adnex tumors

BACKGROUND

Cutaneous adnexal tumors are a diverse group of tumors arising from structures of the hair appendages. Although often benign, malignant entities occur which can metastasize and lead to patients´ death. Correct diagnosis is critical to ensure optimal treatment and best possible patient outcome. Artificial intelligence (AI) in the form of deep neural networks has recently shown enormous potential in the field of medicine including pathology, where we and others have found common cutaneous tumors can be detected with high sensitivity and specificity. To become a widely applied tool, AI approaches will also need to reliably detect and distinguish less common tumor entities including the diverse group of cutaneous adnexal tumors.

METHODS

To assess the potential of AI to recognize cutaneous adnexal tumors, we selected a diverse set of these entities from five German centers. The algorithm was trained with samples from four centers and then tested on slides from the fifth center.

RESULTS

The neural network was able to differentiate 14 different cutaneous adnexal tumors and distinguish them from more common cutaneous tumors (i.e. basal cell carcinoma and seborrheic keratosis). The total accuracy on the test set for classifying 248 samples into these 16 diagnoses was 89.92 %. Our findings support AI can distinguish rare tumors, for morphologically distinct entities even with very limited case numbers (< 50) for training.

CONCLUSION

This study further underlines the enormous potential of AI in pathology which could become a standard tool to aid pathologists in routine diagnostics in the foreseeable future. The final diagnostic responsibility will remain with the pathologist.

Applying an artificial intelligence deep learning approach to routine dermatopathological diagnosis of basal cell carcinoma

BACKGROUND

Institutes of dermatopathology are faced with considerable challenges including a continuously rising numbers of submitted specimens and a shortage of specialized health care practitioners. Basal cell carcinoma (BCC) is one of the most common tumors in the fair-skinned western population and represents a major part of samples submitted for histological evaluation. Digitalizing glass slides has enabled the application of artificial intelligence (AI)-based procedures. To date, these methods have found only limited application in routine diagnostics. The aim of this study was to establish an AI-based model for automated BCC detection.

PATIENTS AND METHODS

In three dermatopathological centers, daily routine practice BCC cases were digitalized. The diagnosis was made both conventionally by analog microscope and digitally through an AI-supported algorithm based on a U-Net architecture neural network.

RESULTS

In routine practice, the model achieved a sensitivity of 98.23% (center 1) and a specificity of 98.51%. The model generalized successfully without additional training to samples from the other centers, achieving similarly high accuracies in BCC detection (sensitivities of 97.67% and 98.57% and specificities of 96.77% and 98.73% in centers 2 and 3, respectively). In addition, automated AI-based basal cell carcinoma subtyping and tumor thickness measurement were established.

CONCLUSIONS

AI-based methods can detect BCC with high accuracy in a routine clinical setting and significantly support dermatopathological work.

Einsatz künstlicher Intelligenz mittels Deep Learning in der dermatopathologischen Routinediagnostik des Basalzellkarzinoms: Applying an artificial intelligence deep learning approach to routine dermatopathological diagnosis of basal cell carcinoma

Hintergrund

Dermatopathologische Institute stehen aufgrund immer höherer Anforderungen bei andererseits schwindenden Ressourcen vor zunehmenden Herausforderungen. Basalzellkarzinome stellen einen Großteil des Einsendeguts mit entsprechendem Arbeitsaufwand dar. Gleichzeitig ermöglicht die Digitalisierung von Glasobjektträgern den Einsatz künstlicher Intelligenz (KI)‐basierter Verfahren in der Dermatopathologie. Bislang haben diese Verfahren keinen Einzug in die Routinediagnostik gefunden. Ziel dieser Studie war daher, den Einsatz eines KI‐basierten Modells zur automatisierten Basalzellkarzinom‐Erkennung zu etablieren.

Patienten und Methodik

In drei dermatopathologischen Zentren wurden während des täglichen Routinebetriebs Basalzellkarzinom‐Fälle digitalisiert und sowohl klassisch am Mikroskop als auch mittels KI‐basierter Methodik basierend auf neuronalen Netzen mit U‐Net‐Architektur befundet.

Ergebnisse

Im Routinebetrieb erzielte das Modell eine Sensitivität von 98,23 % und eine Spezifität von 98,51 % (Zentrum 1). Das Modell konnte übergangslos in den anderen Zentren Einsatz finden und erreichte ähnlich hohe Genauigkeiten in der Basalzellkarzinom‐Erkennung (Sensitivität von 97,67 % beziehungsweise 98,57 %, Spezifität von 96,77 % beziehungsweise 98,73 %). Zusätzlich wurden eine automatisierte, KI‐basierte Basalzellkarzinom‐Subtypisierung und Tumordickenmessung etabliert.

Schlussfolgerungen

KI‐basierte Verfahren können mit einer hohen Genauigkeit im Routinebetrieb Basalzellkarzinome erkennen und signifikant die dermatopathologische Arbeit unterstützen.

Deep learning detection of melanoma metastases in lymph nodes

BACKGROUND

In melanoma patients, surgical excision of the first draining lymph node, the sentinel lymph node (SLN), is a routine procedure to evaluate lymphogenic metastases. Metastasis detection by histopathological analysis assesses multiple tissue levels with hematoxylin and eosin and immunohistochemically stained glass slides. Considering the amount of tissue to analyze, the detection of metastasis can be highly time-consuming for pathologists. The application of artificial intelligence in the clinical routine has constantly increased over the past few years.

METHODS

In this multi-center study, a deep learning method was established on histological tissue sections of sentinel lymph nodes collected from the clinical routine. The algorithm was trained to highlight potential melanoma metastases for further review by pathologists, without relying on supplementary immunohistochemical stainings (e.g. anti-S100, anti-MelanA).

RESULTS

The established method was able to detect the existence of metastasis on individual tissue cuts with an area under the curve of 0.9630 and 0.9856 respectively on two test cohorts from different laboratories. The method was able to accurately identify tumour deposits>0.1 mm and, by automatic tumour diameter measurement, classify these into 0.1 mm to -1.0 mm and>1.0 mm groups, thus identifying and classifying metastasis currently relevant for assessing prognosis and stratifying treatment.

CONCLUSIONS

Our results demonstrate that AI-based SLN melanoma metastasis detection has great potential and could become a routinely applied aid for pathologists. Our current study focused on assessing established parameters; however, larger future AI-based studies could identify novel biomarkers potentially further improving SLN-based prognostic and therapeutic predictions for affected patients.

Evaluation of a Deep Learning Approach to Differentiate Bowen's Disease and Seborrheic Keratosis

Background: Some of the most common cutaneous neoplasms are Bowen’s disease and seborrheic keratosis, a malignant and a benign proliferation, respectively. These entities represent a significant fraction of a dermatopathologists’ workload, and in some cases, histological differentiation may be challenging. The potential of deep learning networks to distinguish these diseases is assessed. Methods: In total, 1935 whole-slide images from three institutions were scanned on two different slide scanners. A U-Net-based segmentation deep learning algorithm was trained on data from one of the centers to differentiate Bowen’s disease, seborrheic keratosis, and normal tissue, learning from annotations performed by dermatopathologists. Optimal thresholds for the class distinction of diagnoses were extracted and assessed on a test set with data from all three institutions. Results: We aimed to diagnose Bowen’s diseases with the highest sensitivity. A good performance was observed across all three centers, underlining the model’s robustness. In one of the centers, the distinction between Bowen’s disease and all other diagnoses was achieved with an AUC of 0.9858 and a sensitivity of 0.9511. Seborrheic keratosis was detected with an AUC of 0.9764 and a sensitivity of 0.9394. Nevertheless, distinguishing irritated seborrheic keratosis from Bowen’s disease remained challenging. Conclusions: Bowen’s disease and seborrheic keratosis could be correctly identified by the evaluated deep learning model on test sets from three different centers, two of which were not involved in training, and AUC scores > 0.97 were obtained. The method proved robust to changes in the staining solution and scanner model. We believe this demonstrates that deep learning algorithms can aid in clinical routine; however, the results should be confirmed by qualified histopathologists.

Checkpoint immunotherapy of cutaneous squamous cell carcinoma in patients suffering from chronic lymphocytic leukaemia: divergent outcomes in two men treated with PD-1 inhibitors

Cutaneous squamous cell carcinoma (cSCC) numbers among the most common types of skin cancer and is known as one of the cancer entities with the highest mutational burden among all solid tumours. Due to the positive correlation between mutational burden and response rate to inhibitors of the programmed cell death 1 (PD-1), those inhibitors are considered promising candidates for the systemic therapy of cSCC. Recently, the PD-1 inhibitors pembrolizumab, nivolumab and cemiplimab demonstrated efficacy in the systemic treatment of locally advanced or metastatic cSCC leading to the approval of cemiplimab by the FDA (U.S. Food and Drug Administration) in 2018 and the EMA (European Medicines Agency) in 2019. Patients with haematological malignancies tend to develop skin cancers of high aggressiveness, enhanced cumulative recurrence rate and higher rates of metastases with subsequent death. Chronic lymphocytic leukaemia (CLL) is the most frequent type of leukaemia in the United States and Europe with the majority of patients older than 50 years of age. This neoplasm predominantly originates from B -cells leading to an impaired immune system of the patient. Although CLL is a B-cell malignancy, studies have also described the involvement of T cells in the pathogenesis and progression of the disease with contradictory findings on the effects of PD-1 inhibitors in CLL. Due to their underlying hematologic malignancy, these patients have commonly no access to PD-1 inhibitor trials for treatment of advanced cSCC. We report on two patients with locally advanced or metastatic cSCC. Both patients had been suffering from a CLL for many years without indication for treatment. Despite a potential immunosuppressive state of the patients due to their CLL, both were treated with the PD-1 inhibitor pembrolizumab resulting in different therapy outcomes.

Management of partial and non-responding cutaneous squamous cell carcinoma

Cutaneous squamous cell carcinoma (cSCC) and basal cell carcinoma are the most common types of skin cancer. For patients with locally advanced and metastatic cSCC, the programmed cell death 1 (PD-1) inhibitor cemiplimab is approved for systemic treatment. Despite this revolutionary immunomodulatory therapeutic approach, tumours may fail to respond either completely or partially. In addition to the previously established local treatment with radiotherapy or systemic treatment with chemotherapy and epidermal growth factor receptor inhibitors, ongoing trials are currently focussed on re-stimulating the antitumour immune response in patients with advanced cSCC refractory to PD-1 inhibitors. In this review, ongoing and recently finished trials with different therapeutic approaches will be discussed.

[Immunotherapy for malignant melanoma]

Although cutaneous melanoma accounts for only about 4% of all skin cancers (including nonmelanocytic skin cancer), it is responsible for 80% of all deaths caused by skin cancer. The introduction of immune checkpoint inhibitors led to a significant improvement in long-term survival of patients in an advanced stage regardless of BRAF mutation status. In addition to targeted therapy for patients with BRAF-mutated melanoma, immunotherapies are the therapies of choice in advanced stages and, since 2018, also in the adjuvant setting. The effectiveness of combination therapies and sequences of targeted and immunotherapies are currently being tested.

GNAQ and GNA11 mutant nonuveal melanoma: a subtype distinct from both cutaneous and uveal melanoma

BACKGROUND

GNAQ and GNA11 mutant nonuveal melanoma represent a poorly characterized rare subgroup of melanoma with a gene mutation profile similar to uveal melanoma.

OBJECTIVES

To characterize these tumours in terms of clinical behaviour and genetic characteristics.

METHODS

Patients with nonuveal GNAQ/11 mutated melanoma were identified from the prospective multicentre tumour tissue registry ADOREG, Tissue Registry in Melanoma (TRIM) and additional cooperating skin cancer centres. Extensive data on patient, tumour and treatment characteristics were collected retrospectively. Targeted sequencing was used to determine tumour mutational burden. Immunohistochemistry staining was performed for programmed death-ligand 1 and BRCA1-associated protein (BAP)1. Existing whole-exome cutaneous and uveal melanoma data were analysed for mutation type and burden.

RESULTS

We identified 18 patients with metastatic GNAQ/11 mutant nonuveal melanoma. Tumours had a lower tumour mutational burden and fewer ultraviolet signature mutations than cutaneous melanomas. In addition to GNAQ and GNA11 mutations (nine each), six splicing factor 3b subunit 1 (SF3B1), three eukaryotic translation initiation factor 1A X-linked (EIF1AX) and four BAP1 mutations were detected. In contrast to uveal melanoma, GNAQ/11 mutant nonuveal melanomas frequently metastasized lymphatically and concurrent EIF1AX, SF3B1 and BAP1 mutations showed no apparent association with patient prognosis. Objective response to immunotherapy was poor with only one partial response observed in 10 treated patients (10%).

CONCLUSIONS

Our findings suggest that GNAQ/11 mutant nonuveal melanomas are a subtype of melanoma that is both clinically and genetically distinct from cutaneous and uveal melanoma. As they respond poorly to available treatment regimens, novel effective therapeutic approaches for affected patients are urgently needed. What is already known about this topic? The rare occurrence of GNAQ/11 mutations in nonuveal melanoma has been documented. GNAQ/11 mutant nonuveal melanomas also harbour genetic alterations in EIF1AX, SF3B1 and BAP1 that are of prognostic relevance in uveal melanoma. What does this study add? GNAQ/11 mutant nonuveal melanomas show metastatic spread reminiscent of cutaneous melanoma, but not uveal melanoma. GNAQ/11 mutant nonuveal melanomas have a low tumour mutational burden that is higher than uveal melanoma, but lower than cutaneous melanoma. What is the translational message? Primary GNAQ/11 mutant nonuveal melanomas are a subtype of melanoma that is clinically and genetically distinct from both cutaneous and uveal melanoma. As metastatic GNAQ/11 mutant nonuveal melanomas respond poorly to available systemic therapies, including immune checkpoint inhibition, novel therapeutic approaches for these tumours are urgently needed. Linked Comment: Rafei-Shamsabadi. Br J Dermatol 2020; 183:806-807.

The genetic basis for most patients with pustular skin disease remains elusive

BACKGROUND

Rare variants in the genes IL36RN, CARD14 and AP1S3 have been identified to cause or contribute to pustular skin diseases, primarily generalized pustular psoriasis (GPP).

OBJECTIVES

To better understand the disease relevance of these genes, we screened our cohorts of patients with pustular skin diseases [primarily GPP and palmoplantar pustular psoriasis (PPP)] for coding changes in these three genes. Carriers of single heterozygous IL36RN mutations were screened for a second mutation in IL36RN.

METHODS

Coding exons of IL36RN, CARD14 and AP1S3 were sequenced in 67 patients - 61 with GPP, two with acute generalized exanthematous pustulosis and four with acrodermatitis continua of Hallopeau. We screened IL36RN and AP1S3 for intragenic copy-number variants and 258 patients with PPP for coding changes in AP1S3. Eleven heterozygous IL36RN mutations carriers were analysed for a second noncoding IL36RN mutation. Genotype-phenotype correlations in carriers/noncarriers of IL36RN mutations were assessed within the GPP cohort.

RESULTS

The majority of patients (GPP, 64%) did not carry rare variants in any of the three genes. Biallelic and monoallelic IL36RN mutations were identified in 15 and five patients with GPP, respectively. Noncoding rare IL36RN variants were not identified in heterozygous carriers. The only significant genotype-phenotype correlation observed for IL36RN mutation carriers was early age at disease onset. Additional rare CARD14 or AP1S3 variants were identified in 15% of IL36RN mutation carriers.

CONCLUSIONS

The identification of IL36RN mutation carriers harbouring additional rare variants in CARD14 or AP1S3 indicates a more complex mode of inheritance of pustular psoriasis. Our results suggest that, in heterozygous IL36RN mutation carriers, there are additional disease-causing genetic factors outside IL36RN.

[Molecular and immunohistochemical diagnostics in melanoma]

To provide appropriate therapy and follow-up to patients with malignant melanoma, proper diagnostics are of critical importance. Targeted therapy of advanced melanoma is based on the molecular genetic analyses of tumor tissue. In addition, sequencing of genes and other genetic approaches can provide insight into the origin of melanocytic tumors and can aid in distinguishing benign from malignant lesions. In this regard, spizoid neoplasms remain a challenging entity. Aside from genetic analyses of tumor tissue, immunohistochemistry remains an essential tool in melanoma diagnostics and TNM classification. With new immunotherapies being approved for advanced melanoma, immunohistochemistry to determine PD-L1 expression has gained clinical interest. While PD-L1 expression is associated with response to PD-1 blockade, a substantial number of patients without PD-L1 expression can still experience tumor remission upon treatment. In this review, current and future developments in melanoma diagnostics with regard to molecular genetics and immunohistochemistry are summarized. The utilization of such analyses in clinical decision making is also discussed.

Differential influence of vemurafenib and dabrafenib on patients' lymphocytes despite similar clinical efficacy in melanoma

BACKGROUND

Since the majority of melanomas eventually become resistant and progress, combining selective BRAF inhibitors (BRAFi) with immunotherapies has been proposed to achieve more durable treatment responses. Here, we explored the impact of selective BRAFi on the hosts' immune system.

PATIENTS AND METHODS

Clinical data, whole blood counts (WBC) and serum lactate dehydrogenase (LDH) of 277 vemurafenib- and 65 dabrafenib-treated melanoma patients were evaluated. The frequency and phenotype of lymphocyte subpopulations were determined by flow cytometry while T cell cytokine secretion was measured by multiplex assays.

RESULTS

Progression-free survival (PFS) as well as overall survival (OS) were similar in patients treated with either BRAFi. High pretreatment LDH was associated with shorter PFS and OS in both groups. During therapy, peripheral lymphocytes decreased by 24.3% (median, P < 0.0001) in vemurafenib-treated patients but remained unchanged in dabrafenib-treated patients (+1.2%, P = 0.717). Differentiation of peripheral lymphocytes of vemurafenib-treated patients showed a significant decrease in CD4(+) T cells (P < 0.05). Within CD4(+) T cells obtained during treatment, an increase in CCR7(+)CD45RA(+) (naïve) and a decrease in CCR7(+)CD45RA(-) (central memory) populations were found (P < 0.01 for both). Furthermore, secretion of interferon-γ and interleukin-9 by CD4(+) T cells was significantly lower in samples obtained during vemurafenib treatment compared with baseline samples.

CONCLUSION

While both compounds have comparable clinical efficacy, vemurafenib but not dabrafenib decreases patients peripheral lymphocyte counts and alters CD4(+) T cell phenotype and function. Thus, selective BRAFi can significantly affect patients' peripheral lymphocyte populations. Fully understanding these effects could be critical for successfully implementing combinatorial therapies of BRAFi with immunomodulatory agents.

TERT promoter mutations in ocular melanoma distinguish between conjunctival and uveal tumours

BACKGROUND

Recently, activating mutations in the TERT promoter were identified in cutaneous melanoma. We tested a cohort of ocular melanoma samples for similar mutations.

METHODS

The TERT promoter region was analysed by Sanger sequencing in 47 uveal (ciliary body or choroidal) melanomas and 38 conjunctival melanomas.

RESULTS

Mutations of the TERT promoter were not identified in uveal melanomas, but were detected in 12 (32%) conjunctival melanomas. Mutations had a UV signature and were identical to those found in cutaneous melanoma.

CONCLUSION

Mutations of TERT promoter with UV signatures are frequent in conjunctival melanomas and favour a pathogenetic kinship with cutaneous melanomas. Absence of these mutations in uveal melanomas emphasises their genetic distinction from cutaneous and conjunctival melanomas.

Genetic and molecular characterization of uveal melanoma cell lines

The recent identification of frequent activating mutations in GNAQ or GNA11 in uveal melanoma provides an opportunity to better understand the pathogenesis of this melanoma subtype and to develop rational therapeutics to target the cellular effects mediated by these mutations. Cell lines from uveal melanoma tumors are an essential tool for these types of analyses. We report the mutation status of relevant melanoma genes, expression levels of proteins of interest, and DNA fingerprinting of a panel of uveal melanoma cell lines used in the research community.