Expert-Level Diagnosis of Nonpigmented Skin Cancer by Combined Convolutional Neural Networks.

Importance: Convolutional neural networks (CNNs) achieve expert-level accuracy in the diagnosis of pigmented melanocytic lesions. However, the most common types of skin cancer are nonpigmented and nonmelanocytic, and are more difficult to diagnose. Objective: To compare the accuracy of a CNN-based classifier with that of physicians with different levels of experience. Design, Setting, and Participants: A CNN-based classification model was trained on 7895 dermoscopic and 5829 close-up images of lesions excised at a primary skin cancer clinic between January 1, 2008, and July 13, 2017, for a combined evaluation of both imaging methods. The combined CNN (cCNN) was tested on a set of 2072 unknown cases and compared with results from 95 human raters who were medical personnel, including 62 board-certified dermatologists, with different experience in dermoscopy. Main Outcomes and Measures: The proportions of correct specific diagnoses and the accuracy to differentiate between benign and malignant lesions measured as an area under the receiver operating characteristic curve served as main outcome measures. Results: Among 95 human raters (51.6% female; mean age, 43.4 years; 95% CI, 41.0-45.7 years), the participants were divided into 3 groups (according to years of experience with dermoscopy): beginner raters (<3 years), intermediate raters (3-10 years), or expert raters (>10 years). The area under the receiver operating characteristic curve of the trained cCNN was higher than human ratings (0.742; 95% CI, 0.729-0.755 vs 0.695; 95% CI, 0.676-0.713; P < .001). The specificity was fixed at the mean level of human raters (51.3%), and therefore the sensitivity of the cCNN (80.5%; 95% CI, 79.0%-82.1%) was higher than that of human raters (77.6%; 95% CI, 74.7%-80.5%). The cCNN achieved a higher percentage of correct specific diagnoses compared with human raters (37.6%; 95% CI, 36.6%-38.4% vs 33.5%; 95% CI, 31.5%-35.6%; P = .001) but not compared with experts (37.3%; 95% CI, 35.7%-38.8% vs 40.0%; 95% CI, 37.0%-43.0%; P = .18). Conclusions and Relevance: Neural networks are able to classify dermoscopic and close-up images of nonpigmented lesions as accurately as human experts in an experimental setting.

Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists.

Background: Deep learning convolutional neural networks (CNN) may facilitate melanoma detection, but data comparing a CNN's diagnostic performance to larger groups of dermatologists are lacking. Methods: Google's Inception v4 CNN architecture was trained and validated using dermoscopic images and corresponding diagnoses. In a comparative cross-sectional reader study a 100-image test-set was used (level-I: dermoscopy only; level-II: dermoscopy plus clinical information and images). Main outcome measures were sensitivity, specificity and area under the curve (AUC) of receiver operating characteristics (ROC) for diagnostic classification (dichotomous) of lesions by the CNN versus an international group of 58 dermatologists during level-I or -II of the reader study. Secondary end points included the dermatologists' diagnostic performance in their management decisions and differences in the diagnostic performance of dermatologists during level-I and -II of the reader study. Additionally, the CNN's performance was compared with the top-five algorithms of the 2016 International Symposium on Biomedical Imaging (ISBI) challenge. Results: In level-I dermatologists achieved a mean (±standard deviation) sensitivity and specificity for lesion classification of 86.6% (±9.3%) and 71.3% (±11.2%), respectively. More clinical information (level-II) improved the sensitivity to 88.9% (±9.6%, P = 0.19) and specificity to 75.7% (±11.7%, P < 0.05). The CNN ROC curve revealed a higher specificity of 82.5% when compared with dermatologists in level-I (71.3%, P < 0.01) and level-II (75.7%, P < 0.01) at their sensitivities of 86.6% and 88.9%, respectively. The CNN ROC AUC was greater than the mean ROC area of dermatologists (0.86 versus 0.79, P < 0.01). The CNN scored results close to the top three algorithms of the ISBI 2016 challenge. Conclusions: For the first time we compared a CNN's diagnostic performance with a large international group of 58 dermatologists, including 30 experts. Most dermatologists were outperformed by the CNN. Irrespective of any physicians' experience, they may benefit from assistance by a CNN's image classification. Clinical trial number: This study was registered at the German Clinical Trial Register (DRKS-Study-ID: DRKS00013570; https://www.drks.de/drks_web/).

The status of dermoscopy in Germany - results of the cross-sectional Pan-Euro-Dermoscopy Study.

BACKGROUND: Survey on the current status of dermoscopy in Germany. METHODS: In the context of a pan-European internet-based study (n = 7,480) conducted by the International Dermoscopy Society, 880 German dermatologists were asked to answer questions with respect to their level of training as well as their use and perceived benefit of dermoscopy. RESULTS: Seven hundred and sixty-two (86.6 %) participants practiced dermatology in a publicly funded health care setting; 98.4 % used a dermoscope in routine clinical practice. About 93 % (n = 814) stated to have had more than five years of experience in the use of dermoscopy. Dermoscopy was considered useful in the diagnosis of melanoma by 93.6 % (n = 824); for pigmented skin tumors, by 92.4 % (n = 813); in the follow-up of melanocytic lesions, by 88.6 % (n = 780); for non-pigmented lesions, by 71.4 % (n = 628), in the follow-up of non-melanocytic lesions, by 52.7 % (n = 464); and for inflammatory skin lesions, by 28.5 % (n = 251). Overall, 86.5 % (n = 761) of participants felt that - compared to naked-eye examination - dermoscopy increased the number of melanomas diagnosed; 77,7 % (n = 684) considered the number of unnecessary excisions of benign lesions to be decreased. Participants who personally felt that dermoscopy improved their ability to diagnose melanoma were significantly i) younger, ii) had been practicing dermatology for a shorter period of time, iii) were less commonly employed by an university-affiliated dermatology department, iv) were more frequently working in an office-based public health care setting, and v) had more frequently been trained in dermoscopy during their dermatology residency. CONCLUSIONS: The findings presented herein ought to be integrated into future residency and continuing medical education programs with the challenge to improve dermato-oncological care and to expand the diagnostic spectrum of dermoscopy to include inflammatory skin diseases.

Collision skin lesions-results of a multicenter study of the International Dermoscopy Society (IDS).

BACKGROUND: Collision lesions as two independent and unrelated skin tumors often manifest an atypical morphology. OBJECTIVE: To determine the combinations of collision skin lesions (CSLs). METHODS: Twenty-one pigmented lesion clinics in nine countries included 77 histopathologically proven CSLs in this retrospective observational study. RESULTS: Seventy-seven CSLs from 75 patients (median age 59.8 years) were analyzed; 24.7% of CSLs were located on the head and neck area, 5.2% on the upper extremities, 48.1% on the trunk, and 11.7% on the lower extremities; 40.3% revealed a melanocytic component (median age 54.7 years), followed by 45.5% with a basal cell carcinoma (BCC) (median age 62.4 years) and 11.7% with a seborrheic keratosis (median age 64.7 years). CSLs with a BCC component were more often found on the head and neck area compared to tumors with a melanocytic component (34.3% versus 16.1%). Lesions with a melanocytic component were more often detected on the trunk compared to lesions with a BCC (64.5% versus 37.1%). Patients with CSLs with epidermal-epidermal cell combination were older than patients with epidermal-dermal cell combination (63 versus 55.2 years), were more often male than female (63% versus 43.3%), more often had the lesion on the head and neck area (32.6% versus 13.3%), and less often on the upper (2.2 % versus 10%) or lower extremities (8.7% versus 16.6%). CONCLUSIONS: CSLs consist of a heterogeneous group of lesions of varying cell types. They are associated with advancing age and cumulative UV-exposure. CSLs manifest a complex morphology making it challenging to diagnose correctly.

How to diagnose nonpigmented skin tumors: a review of vascular structures seen with dermoscopy: part I. Melanocytic skin tumors.

Dermoscopy is a noninvasive tool that can be helpful in the diagnosis of nonpigmented skin tumors. This is because dermoscopy permits the visualization of key vascular structures that are usually not visible to the naked eye. Much work has concentrated on the identification of specific morphologic types of vessels that allow a classification into melanocytic versus nonmelanocytic and benign versus malignant nonpigmented skin tumors. Among a broad spectrum of different types of vascular patterns, six main morphologies can be identified. These are comma-like, dotted, linear-irregular, hairpin, glomerular, and arborizing vessels. With some exceptions, comma, dotted, and linear irregular vessels are associated with melanocytic tumors, while the latter three vascular types are generally indicative of keratinocytic tumors. Aside from vascular morphology, the architectural arrangement of vessels within the tumor and the presence of additional dermoscopic clues are equally important for the diagnosis. This article provides a general overview of the dermoscopic evaluation of nonpigmented skin tumors and is divided into two parts. Part I discusses the dermoscopic vascular patterns of benign and malignant melanocytic skin tumors. Part II discusses the dermoscopic vascular patterns of benign and malignant nonmelanocytic nonpigmented skin tumors. In each part, additional special management guidelines for melanocytic and nonmelanocytic nonpigmented skin tumors, respectively, will be discussed.

How to diagnose nonpigmented skin tumors: a review of vascular structures seen with dermoscopy: part II. Nonmelanocytic skin tumors.

Nonmelanoma skin cancer refers to a broad class of tumors, including actinic keratosis, basal cell carcinoma, and squamous cell carcinoma, and as a group these are the most frequent cancers occurring in light skinned humans. In contrast to the rarity of amelanotic melanoma, nonmelanoma skin cancer commonly lacks pigmentation. Although these tumors rarely cause death related to metastases, they commonly destroy underlying tissues and should be removed at the earliest possible stage. Dermoscopy improves the clinical diagnosis of nonpigmented skin tumors by allowing the visualization of specific vascular structures that are usually not visible to the naked eye. Dermoscopic vascular patterns of several nonmelanocytic nonpigmented skin tumors, such as sebaceous hyperplasia, seborrheic keratosis, clear cell acanthoma, Bowen disease, or nodular cystic basal cell carcinoma are highly specific, allowing a ready diagnosis in most cases. Others, such as actinic keratosis, pyogenic granuloma, or uncommon adnexal tumors, may be difficult to differentiate even with the aid of dermoscopy. For this reason, general guidelines have been established to assist in making the most appropriate management decision. In the second part of this review of dermoscopic vascular structures of nonpigmented skin tumors, the dermoscopic patterns associated with benign and malignant nonmelanocytic skin tumors and recommendations for the management of these tumors will be discussed.

Dermoscopy of pigmented skin lesions: results of a consensus meeting via the Internet.

BACKGROUND: There is a need for better standardization of the dermoscopic terminology in assessing pigmented skin lesions. OBJECTIVE: The virtual Consensus Net Meeting on Dermoscopy was organized to investigate reproducibility and validity of the various features and diagnostic algorithms. METHODS: Dermoscopic images of 108 lesions were evaluated via the Internet by 40 experienced dermoscopists using a 2-step diagnostic procedure. The first-step algorithm distinguished melanocytic versus nonmelanocytic lesions. The second step in the diagnostic procedure used 4 algorithms (pattern analysis, ABCD rule, Menzies method, and 7-point checklist) to distinguish melanoma versus benign melanocytic lesions. kappa Values, log odds ratios, sensitivity, specificity, and positive likelihood ratios were estimated for all diagnostic algorithms and dermoscopic features. RESULTS: Interobserver agreement was fair to good for all diagnostic methods, but it was poor for the majority of dermoscopic criteria. Intraobserver agreement was good to excellent for all algorithms and features considered. Pattern analysis allowed the best diagnostic performance (positive likelihood ratio: 5.1), whereas alternative algorithms revealed comparable sensitivity but less specificity. Interobserver agreement on management decisions made by dermoscopy was fairly good (mean kappa value: 0.53). CONCLUSION: The virtual Consensus Net Meeting on Dermoscopy represents a valid tool for better standardization of the dermoscopic terminology and, moreover, opens up a new territory for diagnosing and managing pigmented skin lesions.

Dermoscopy of pigmented seborrheic keratosis: a morphological study.

OBJECTIVES: To describe morphological features of seborrheic keratosis as seen by dermoscopy and to investigate their prevalence. DESIGN: Prospective cohort study using macrophotography and dermoscopy for the documentation of seborrheic keratosis. SETTINGS: Seborrheic keratoses were prospectively collected in 2 sites: a private practice in Plantation, Fla (site 1), and the Department of Dermatology at the University Hospital Geneva in Switzerland (site 2). PATIENTS: A total of 203 pigmented seborrheic keratoses (from 192 patients) with complete documentation were collected (111 from site 1 and 93 from site 2). INTERVENTIONS: Screening for new morphological features of seborrheic keratosis and evaluation of all lesions for the prevalence of these criteria. MAIN OUTCOME MEASURES: Identification of new morphological criteria and evaluation of frequency. RESULTS: A total of 15 morphological dermoscopic criteria were identified. Standard criteria such as milialike cysts and comedolike openings were found in a high number of cases (135 and 144, respectively). We found network and networklike structures to be present in 94 lesions (46%). Using standard diagnostic criteria for seborrheic keratosis, 30 lesions would not have been diagnosed as such. CONCLUSIONS: The classic dermoscopic criteria for seborrheic keratosis (milialike cysts and comedolike openings) have a high prevalence but the use of additional dermoscopic criteria such as fissures, hairpin blood vessels, sharp demarcation, and moth-eaten borders improves the diagnostic accuracy. The proper identification of pigment network and networklike structures is important for the correct diagnosis.