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. 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. 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. 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. 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.

Due to their ability to solve complex problems, deep neural networks (DNNs) are becoming increasingly popular in medical applications. However, decision-making by such algorithms is essentially a black-box process that renders it difficult for physicians to judge whether the decisions are reliable. The use of explainable artificial intelligence (XAI) is often suggested as a solution to this problem. We investigate how XAI is used for skin cancer detection: how is it used during the development of new DNNs? What kinds of visualisations are commonly used? Are there systematic evaluations of XAI with dermatologists or dermatopathologists? Google Scholar, PubMed, IEEE Explore, Science Direct and Scopus were searched for peer-reviewed studies published between January 2017 and October 2021 applying XAI to dermatological images: the search terms histopathological image, whole-slide image, clinical image, dermoscopic image, skin, dermatology, explainable, interpretable and XAI were used in various combinations. Only studies concerned with skin cancer were included. 37 publications fulfilled our inclusion criteria. Most studies (19/37) simply applied existing XAI methods to their classifier to interpret its decision-making. Some studies (4/37) proposed new XAI methods or improved upon existing techniques. 14/37 studies addressed specific questions such as bias detection and impact of XAI on man-machine-interactions. However, only three of them evaluated the performance and confidence of humans using CAD systems with XAI. XAI is commonly applied during the development of DNNs for skin cancer detection. However, a systematic and rigorous evaluation of its usefulness in this scenario is lacking.

Background: Immune checkpoint inhibition and in particular anti-PD-1 immunotherapy have revolutionized the treatment of advanced melanoma. In this regard, higher tumoral PD-L1 protein (gene name: CD274) expression is associated with better clinical response and increased survival to anti-PD-1 therapy. Moreover, there is increasing evidence that tumor suppressor proteins are involved in immune regulation and are capable of modulating the expression of immune checkpoint proteins. Here, we determined the role of p53 protein (gene name: TP53) in the regulation of PD-L1 expression in melanoma. Methods: We analyzed publicly available mRNA and protein expression data from the cancer genome/proteome atlas and performed immunohistochemistry on tumors with known TP53 status. Constitutive and IFN-ɣ-induced PD-L1 expression upon p53 knockdown in wildtype, TP53-mutated or JAK2-overexpressing melanoma cells or in cells, in which p53 was rendered transcriptionally inactive by CRISPR/Cas9, was determined by immunoblot or flow cytometry. Similarly, PD-L1 expression was investigated after overexpression of a transcriptionally-impaired p53 (L22Q, W23S) in TP53-wt or a TP53-knockout melanoma cell line. Immunoblot was applied to analyze the IFN-ɣ signaling pathway. Results: For TP53-mutated tumors, an increased CD274 mRNA expression and a higher frequency of PD-L1 positivity was observed. Interestingly, positive correlations of IFNG mRNA and PD-L1 protein in both TP53-wt and -mutated samples and of p53 and PD-L1 protein suggest a non-transcriptional mode of action of p53. Indeed, cell line experiments revealed a diminished IFN-ɣ-induced PD-L1 expression upon p53 knockdown in both wildtype and TP53-mutated melanoma cells, which was not the case when p53 wildtype protein was rendered transcriptionally inactive or by ectopic expression of p53^L22Q,W23S, a transcriptionally-impaired variant, in TP53-wt cells. Accordingly, expression of p53^L22Q,W23S in a TP53-knockout melanoma cell line boosted IFN-ɣ-induced PD-L1 expression. The impaired PD-L1-inducibility after p53 knockdown was associated with a reduced JAK2 expression in the cells and was almost abrogated by JAK2 overexpression. Conclusions: While having only a small impact on basal PD-L1 expression, both wildtype and mutated p53 play an important positive role for IFN-ɣ-induced PD-L1 expression in melanoma cells by supporting JAK2 expression. Future studies should address, whether p53 expression levels might influence response to anti-PD-1 immunotherapy.

Hotspot mutations of the oncogenes BRAF and NRAS are the most common genetic alterations in cutaneous melanoma. Specific inhibitors of BRAF and MEK have shown significant survival benefits in large phase III trials. However, the prognostic significance of BRAF and NRAS mutations outside of clinical trials remains unclear. The mutational status of BRAF (exon 15) and NRAS (exon 2 and 3) was determined in melanoma samples of 217 patients with pyrosequencing and Sanger sequencing. The genotypes were correlated with clinical outcomes and pathologic features of the primary tumors. Time to disease progression was calculated with the cumulative incidence function. Survival analyses were performed with Kaplan-Meier estimates and Cox proportional hazards regression analysis. Relative survival was calculated with the Ederer-II method. Treatment with BRAF and MEK inhibitors and immune checkpoint blockade (ICB) was allowed. Mutations in BRAF and NRAS were identified in 40.1 and 24.4% of cases, respectively. Concurrent mutations in both genes were detected in further 2.3%. The remaining 33.2% were wild type for the investigated exons (WT). BRAF mutations were significantly associated with younger age at first diagnosis (p < 0.001) and truncal localization of the culprit primary (p = 0.002). The nodular subtype was most common in the NRAS cohort. In addition, NRAS-mutant melanoma patients showed a higher frequency of nodal relapse (p = 0.013) and development of metastatic disease (p = 0.021). The time to loco-regional nodal relapse was shortest in NRAS-mutant melanoma (p = 0.002). Presence of NRAS mutation was an independent risk factor for disease progression in multivariate analysis (HR 2.01; 95% CI 1.02 – 3.98). BRAF-mutant melanoma patients showed a tendency for better overall and relative survival. Genotype was not a consistent risk factor in multivariate analysis. Instead, positive sentinel lymph node status (HR 2.65; 95% CI 1.15 – 6.10) and treatment with ICB in stage IV disease (HR 0.17; 95% CI 0.06–0.48) were significant multivariate risk factors. NRAS-mutant tumors tended to behave more aggressively particularly in early stages of the disease in this high-risk melanoma population. Treatment with immune checkpoint blockade improved survival in stage IV disease in a real-world setting.

Clinicians and pathologists traditionally use patient data in addition to clinical examination to support their diagnoses. 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. 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. 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%. 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.