Explore the vast range of titles available.

An artificial intelligence trained to classify images of skin lesions as benign lesions or malignant skin cancers achieves the accuracy of board-certified dermatologists. Neural network identifies skin cancers Andre Esteva et al . used 129,450 clinical images of skin disease to train a deep convolutional neural network to classify skin lesions. The result is an algorithm that can classify lesions from photographic images similar to those taken with a mobile phone. The accuracy of the system in detecting malignant melanomas and carcinomas matched that of trained dermatologists. The authors suggest that the technique could be used outside the clinic as a visual screen for cancer. Skin cancer, the most common human malignancy 1 , 2 , 3 , is primarily diagnosed visually, beginning with an initial clinical screening and followed potentially by dermoscopic analysis, a biopsy and histopathological examination. Automated classification of skin lesions using images is a challenging task owing to the fine-grained variability in the appearance of skin lesions. Deep convolutional neural networks (CNNs) 4 , 5 show potential for general and highly variable tasks across many fine-grained object categories 6 , 7 , 8 , 9 , 10 , 11 . Here we demonstrate classification of skin lesions using a single CNN, trained end-to-end from images directly, using only pixels and disease labels as inputs. We train a CNN using a dataset of 129,450 clinical images—two orders of magnitude larger than previous datasets 12 —consisting of 2,032 different diseases. We test its performance against 21 board-certified dermatologists on biopsy-proven clinical images with two critical binary classification use cases: keratinocyte carcinomas versus benign seborrheic keratoses; and malignant melanomas versus benign nevi. The first case represents the identification of the most common cancers, the second represents the identification of the deadliest skin cancer. The CNN achieves performance on par with all tested experts across both tasks, demonstrating an artificial intelligence capable of classifying skin cancer with a level of competence comparable to dermatologists. Outfitted with deep neural networks, mobile devices can potentially extend the reach of dermatologists outside of the clinic. It is projected that 6.3 billion smartphone subscriptions will exist by the year 2021 (ref. 13 ) and can therefore potentially provide low-cost universal access to vital diagnostic care.

Skin cancer is one of most deadly diseases in humans. According to the high similarity between melanoma and nevus lesions, physicians take much more time to investigate these lesions. The automated classification of skin lesions will save effort, time and human life. The purpose of this paper is to present an automatic skin lesions classification system with higher classification rate using the theory of transfer learning and the pre-trained deep neural network. The transfer learning has been applied to the Alex-net in different ways, including fine-tuning the weights of the architecture, replacing the classification layer with a softmax layer that works with two or three kinds of skin lesions, and augmenting dataset by fixed and random rotation angles. The new softmax layer has the ability to classify the segmented color image lesions into melanoma and nevus or into melanoma, seborrheic keratosis, and nevus. The three well-known datasets, MED-NODE, Derm (IS & Quest) and ISIC, are used in testing and verifying the proposed method. The proposed DCNN weights have been fine-tuned using the training and testing dataset from ISIC in addition to 10-fold cross validation for MED-NODE and DermIS-DermQuest. The accuracy, sensitivity, specificity, and precision measures are used to evaluate the performance of the proposed method and the existing methods. For the datasets, MED-NODE, Derm (IS & Quest) and ISIC, the proposed method has achieved accuracy percentages of 96.86%, 97.70%, and 95.91% respectively. The performance of the proposed method has outperformed the performance of the existing classification methods of skin cancer.

Epidermal nevi are common congenital skin lesions with an incidence of 1 in 1,000 people; however, their genetic basis remains elusive. Germline mutations of the FGF receptor 3 (FGFR3) cause autosomal dominant skeletal disorders such as achondroplasia and thanatophoric dysplasia, which can be associated with acanthosis nigricans of the skin. Acanthosis nigricans and common epidermal nevi of the nonorganoid, nonepidermolytic type share some clinical and histological features. We used a SNaPshot multiplex assay to screen 39 epidermal nevi of this type of 33 patients for 11 activating FGFR3 point mutations. In addition, exon 19 of FGFR3 was directly sequenced. We identified activating FGFR3 mutations, almost exclusively at codon 248 (R248C), in 11 of 33 (33%) patients with nonorganoid, nonepidermolytic epidermal nevi. In 4 of these cases, samples from adjacent histologically normal skin could be analyzed, and FGFR3 mutations were found to be absent. Our results suggest that a large proportion of epidermal nevi are caused by a mosaicism of activating FGFR3 mutations in the human epidermis, secondary to a postzygotic mutation in early embryonic development. The R248C mutation appears to be a hot spot for FGFR3 mutations in epidermal nevi.

Recently, initial studies describing the use of multicolor fluorescence in situ hybridization (FISH) for classifying melanocytic skin lesions have been published demonstrating a high sensitivity and specificity in discriminating melanomas from nevi. However, the majority of these studies included neither histologically ambiguous lesions nor a clinical long-term follow up. This study was undertaken to validate a special multicolor FISH test in histologically ambiguous melanocytic skin lesions with known clinical long-term follow up. FISH was scored by three independent pathologists in a series of 22 melanocytic skin lesions, including 12 ambiguous cases using four probes targeting chromosome 6p25, centromere 6, 6q23, and 11q13. The FISH results were compared with array comparative genomic hybridization data and correlated to the clinical long-term follow up (mean: 65 months). Pair-wise comparison between the interpretations of the observers showed a moderate to substantial agreement ( κ 0.47–0.61). Comparing the FISH results with the clinical behavior reached an overall sensitivity of 60% and a specificity of 50% ( χ 2 =0.25; P =0.61) for later development of metastases. Comparison of array comparative genomic hybridization data with FISH analyses did not yield significant results but array comparative genomic hybridization data demonstrated that melanocytic skin lesions with the development of metastases showed significantly more chromosomal aberrations ( P <0.01) compared with melanocytic skin lesions without the development of metastases. The FISH technique with its present composition of locus-specific probes for RREB1/MYB and CCND1 did not achieve a clinically useful sensitivity and specificity. However, a reassessment of the probes and better standardization of the method may lead to a valuable diagnostic tool.

With rising melanoma incidence and mortality, early detection and surgical removal of primary lesions is essential. Multispectral imaging is a new, non-invasive technique that can facilitate skin cancer detection by measuring the reflectance spectra of biological tissues. Currently, incident illumination allows little light to be reflected from deeper skin layers due to high surface reflectance. A pilot study was conducted at the University Hospital Basel to evaluate, whether multispectral imaging with direct light coupling could extract more information from deeper skin layers for more accurate dignity classification of melanocytic lesions. 27 suspicious pigmented lesions from 23 patients were included (6 melanomas, 6 dysplastic nevi, 12 melanocytic nevi, 3 other). Lesions were imaged before excision using a prototype snapshot mosaic multispectral camera with incident and direct illumination with subsequent dignity classification by a pre-trained multispectral image analysis model. Using incident light, a sensitivity of 83.3% and a specificity of 58.8% were achieved compared to dignity as determined by histopathological examination. Direct light coupling resulted in a superior sensitivity of 100% and specificity of 82.4%. Convolutional neural network classification of corresponding red, green, and blue lesion images resulted in 16.7% lower sensitivity (83.3%, 5/6 malignant lesions detected) and 20.9% lower specificity (61.5%) compared to direct light coupling with multispectral image classification. Our results show that incorporating direct light multispectral imaging into the melanoma detection process could potentially increase the accuracy of dignity classification. This newly evaluated illumination method could improve multispectral applications in skin cancer detection. Further larger studies are needed to validate the camera prototype.

In daily clinical practice melanocytic nevi are commonly encountered. Traditionally, both benign and malignant melanocytic tumors have been sub-classified by their histopathologic characteristics with differing criteria for malignancy applied to each group. Recently, many of the mutations that initiate nevus formation have been identified and specific sets of mutations are found in different subtypes of nevi. Whereas a single mutation appears sufficient to initiate a nevus, but is not enough to result in melanoma, specific combinations of mutations have been identified in some melanocytic tumors that are regarded to be of low biologic potential. The term “melanocytoma” has recently been proposed by the World Health Organization to describe those tumors that demonstrate genetic progression beyond the single mutations that are found in nevi but are not frankly malignant. Melanocytomas occupy intermediate genetic stages between nevus and melanoma and likely have an increased risk of malignant transformation as compared to nevi. This review provides an update on the broad spectrum of melanocytic nevi and melanocytomas and outlines their key histopathologic and genetic features.

The potential role of stem cells in neoplasia is a subject of recent interest. Three markers of melanocytic stem cells have been described recently. CD166 is expressed on the surface of mesenchymal stem cells and has been found on human melanoma cell lines. CD133 is expressed on the surface of dermal-derived stem cells that are capable of differentiating into neural cells. Nestin is an intermediate filament expressed in the cytoplasm of neuroepithelial stem cells. In this study, we evaluate the expression of these markers and possible differences among banal nevi, primary melanoma, and metastastic melanoma. Tissue microarrays containing normal tissue and 226 melanocytic lesions (71 banal nevi, 71 in situ and invasive melanomas, and 84 metastatic melanomas) were studied by immunohistochemistry using monoclonal antibodies CD166, CD133, and nestin. A significantly greater percentage of melanomas (combined primary and metastatic) contained cells that expressed CD166 (P=0.005), CD133 (P=0.003), and nestin (P=0.03) than banal nevi. Only nestin showed a statistical difference when comparing primary and metastatic melanoma (P=0.05). A stepwise increase in the proportion of lesions expressing all three markers was observed from banal nevi (2/19) to primary melanomas (8/17) to metastatic melanoma (19/28), P=0.0005. All cases of metastatic melanoma expressed at least one stem cell marker. The increased expression of CD166, CD133, and nestin in melanoma suggests that progression to malignant melanoma likely involves genetic pathways instrumental to stem cell biology and normal tissue development. Further studies and characterization of these pathways may also reveal new prognostic markers for a disease whose prognosis in advanced stages is dismal.

We describe a fully automated system for the classification of acral volar melanomas. We used a total of 213 acral dermoscopy images (176 nevi and 37 melanomas). Our automatic tumor area extraction algorithm successfully extracted the tumor in 199 cases (169 nevi and 30 melanomas), and we developed a diagnostic classifier using these images. Our linear classifier achieved a sensitivity (SE) of 100%, a specificity (SP) of 95.9%, and an area under the receiver operating characteristic curve (AUC) of 0.993 using a leave-one-out cross-validation strategy (81.1% SE, 92.1% SP; considering 14 unsuccessful extraction cases as false classification). In addition, we developed three pattern detectors for typical dermoscopic structures such as parallel ridge, parallel furrow, and fibrillar patterns. These also achieved good detection accuracy as indicated by their AUC values: 0.985, 0.931, and 0.890, respectively. The features used in the melanoma–nevus classifier and the parallel ridge detector have significant overlap.

Although nevi are highly polymorphous, it has been suggested that each individual is characterized by only a few dominant patterns of nevi. Therefore, a nevus that does not fit in with these patterns, the “ugly duckling” nevus, is suspicious. Our objective was to study the intra-individual diversity of nevi, using human ability to build “perceived similarity clusters” (PSCs). Nine dermatologists had to cluster all the nevi of 80 patients into PSCs, at the clinical scale (CS) and at the dermoscopic scale (DS) (subset of 30 patients). Nine novices did the same in a subset of 11 patients. The experts identified a mean of 2.8 PSCs/patient at CS. Concordance was higher between experts than between novices at CS and at DS. Despite a trend for more PSCs at DS than at CS, the number of nevus patterns per patient remained low, regardless of the number of nevi. Inter-expert concordance permits a consensus representation of nevus diversity in each individual. Nevus diversity is limited in each patient and constitutes an individual reference system, which we can intuitively perceive. This reference is probably crucial for nevus analysis and melanoma detection and opens perspectives for computer-aided diagnostics.

While most melanomas can be distinguished from nevi by histopathology, the histology is ambiguous for some melanocytic tumors, contributing to diagnostic uncertainty. Therefore molecular assays, including FISH or SNP array, and more recently a gene expression test (myPath, Myriad Genetics) have been proposed to aid in the work-up of ambiguous tumors. Two hundred and sixty-eight prospectively submitted cases were gathered, with the goal of comparing the myPath assay to morphologic diagnosis in (1) morphologically unequivocal cases (198), and to morphologic diagnosis and FISH in (2) morphologically ambiguous cases (70). Melanoma FISH was performed using probes for 6p25, 6q23, 11q13, Cep6, 9p21, and Cep9 and scored according to established criteria. The myPath assay was scored by the manufacturer as benign, indeterminate, or malignant. In the unequivocal group, myPath assay showed 75% agreement with morphologic diagnosis, with 67% sensitivity and 81% specificity. In the ambiguous group, FISH and myPath showed 69% inter-test agreement. For these cases agreement with histopathologic interpretation was 84% for FISH and 74% for myPath. Sensitivity and specificity of FISH was 61 and 100%, 50 and 93% for myPath, respectively. Cases from both groups in which myPath was discordant with either morphologic diagnosis and/or FISH (81/268 cases), were submitted for evaluation by two experienced dermatopathologist and also by SNP-array. SNP-array results correlated better than FISH, which correlated better than myPath, with the morphologic interpretation. Our findings document that molecular diagnostics show good correlation with consensus diagnoses, but discordant results occur, and vary in level of correlation with consensus interpretations. Studies with long-term outcomes data within specific ambiguous lesion subsets are required to establish the accuracy of this test, as each molecular diagnostic technique has limitations based on both lack of clinical outcomes data in ambiguous melanocytic tumors and in terms of their sensitivity and specificity in melanocytic lesion subtypes.